device and system including multiple devices for supervision and control of machines in industrial i

专利摘要:
it is a system to supervise the operation of at least one machine in an industrial installation or to supervise such operation and act on at least one machine based on such supervision. the system comprises a network comprising a server and a plurality of devices that form a computing cluster. at least some devices among the plurality of devices are connectable to a machine in the industrial installation. each device among at least some devices comprises: a first processor configured to compute in real time, with data obtainable from the machine to which the device is connected, a first processing task to solve a first query; and a second processor configured to share its processing power with the network and to compute, when assigned by the server, at least one chain of a second processing task to solve a second query. the server is configured to: control the computing cluster; partition the second processing task into a plurality of threads; and assigning one or more threads among the plurality of threads to the second processor of at least some devices among the plurality of devices. a device for supervising the operation of at least one machine in an industrial facility is also revealed. also revealed is an industrial installation comprising: a plurality of machines; and a network to supervise the operation of at least one machine among the plurality of machines or to supervise and operate at least one machine, wherein the network comprises a server and a plurality of devices that form a computing cluster.
公开号:BR112019018116A2
申请号:R112019018116
申请日:2018-03-01
公开日:2020-04-07
发明作者:Díaz Javier；Juan Gabilondo José
申请人:Plethora Iiot S L；
IPC主号:

专利说明:

DEVICE AND SYSTEM INCLUDING MULTIPLE DEVICES FOR SUPERVISION AND CONTROL OF MACHINES IN INDUSTRIAL INSTALLATION
TECHNICAL FIELD [0001] The present invention relates to the field of devices for supervision and control in industrial installations. More particularly, it refers to high availability devices to capture and process data in real time in industrial environments that generate large amounts of data. These devices can be particularly advantageous for implementing intelligent services, such as energy efficiency and predictive maintenance on industrial machinery. The present invention also relates to networks, systems and industrial installations that include such devices.
STATE OF THE TECHNIQUE [0002] In the nineties (1990), the concept of a network composed of smart devices was developed and a term that defines such a concept was enshrined: the Internet of Things (IoT). LoT refers to physical objects or devices that can be connected to the Internet and that have the ability to identify themselves to other objects or devices with so-called machine-machine or M2M communications. IoT can be defined as the network of physical objects integrated with electronic products, software, sensors and the network connectivity that allows these objects to collect and exchange data. Objects that belong to the IoT can be captured and controlled remotely across the entire network infrastructure
Petition 870190089699, of 10/09/2019, p. 6/104
2/87 existing. Each Thing is uniquely identifiable through its integrated computing system and can be operated through the existing Internet infrastructure. LoT currently unites multiple technologies, covering both wireless and wired communication and networks, sensors and both integrated systems and microelectromechanical systems (MEMS). LoT is applicable to many different environments, including home (smart homes), public infrastructure (such as smart cities, including, for example, environmental supervision) and industry (or in the industry-specific sense, such as smart grids, or in the broad sense industry, covering some specific infrastructures, such as intelligent transport).
[0003] When loT is applied to industrial infrastructures (in a broader sense), the specific term Industrial Internet of Things (IIoT), also called Industry 4.0, is used. When industrial equipment is integrated into IIoT, machines are able to report on their state of work to higher-level supervisory systems, thereby enabling intelligent predictive services such as real-time quality control (RT), as named by Acatech (National Academy of Science and Engineering), Smart Service Welt, March 2014. This, in turn, will enable manufacturers of industrial equipment to become suppliers of advanced services.
[0004] The main difference between IIoT and loT is the fact that the devices and machines that comprise IIoT can be (or be part of) systems whose failure or defect can
Petition 870190089699, of 10/09/2019, p. 7/104
3/87 result in at least one loss of or serious damage to equipment / property, can harm the environment and / or severe loss of production, among others. In this sense, the systems of an industrial installation require solutions that provide real-time response and synchronization between different machines so that they are actuated when there is any potential defect that could cause any previous events. Non-limiting examples of such systems are power generation plants such as nuclear power plants, transmission and transportation of electricity, oil refineries, natural gas processing plants, chemical treatment plants and manufacturing production and assembly lines.
[0005] Referring, in particular, to the example of manufacturing production lines, an exemplary defect that causes a critical event may be the failure of a ball in one of the bearings included in a machine used in a machining process. The machine can comprise many elements and / or mechanical assemblies of high complexity, however, the failure of the ball (ie the damaged ball) in a bearing can take the machine out of operation, thereby causing the interruption of the entire line of production. In this sense, a single worn out sphere can trigger a chain of events that ends up ruining the entire machine. In other cases, the machine is still operative even after the ball has been damaged, however, on the other hand, this event causes the failure of additional elements within the same machine, which can eventually break the machine. In these situations, the time required from the exact moment of the component failure (for example, sphere) to the instruction
Petition 870190089699, of 10/09/2019, p. 8/104
4/87 to stop the machine being supplied and carried out is the most crucial aspect to avoid failures or even disasters. Another example of a crucial event in the field of machining processes is the breaking of a single cutting edge of a tool (or even the same damaged sphere), which can cause serious quality problems and involve production losses: the processed goods by the machine they are processed incorrectly and therefore become defective and do not meet the minimum quality requirements; it can happen that no one detects that the goods are defective, or they can be detected, but long after they have been delivered to distributors or end customers. Therefore, such faults need to be detected in real time, so that quick detection and a reaction (depending on the nature of the machine and the fault, this can take a few milliseconds) to component failure are a must in industrial environments. In addition, it is preferred that industrial systems can somehow predict when the components inside the machines are about to fail so that they can be replaced before the failure.
[0006] When machines from industrial installations encompass IIoT, they tend to generate large amounts of data (a single machine can include tens, hundreds or even thousands of sensors), thereby creating challenging new business opportunities. Such amounts of data provide the consequent need for fast data processing time, and an increase in the need to quickly index, store and process such data since otherwise generated data can
Petition 870190089699, of 10/09/2019, p. 9/104
5/87 become insignificant. For example, a machine tool that can emit thousands of variables, that is, thousands of different pieces of potential information provided by sensors and / or devices that control the operation of the machine tool, can be stopped before failure or some moments after the failure has occurred if these variables (or a subset of those variables) are processed quickly and a meaning is extracted from the combination of some or all of them. Considering the particular case of a machine tool that has 3 servo motors and a machining spindle, the number of variables emitted is 78. If the sample yield is 2.8 MB / cycle, the amount of data captured per year is approximately 1 ALSO. The delay in processing the variables and in reacting to the subsequent extraction of knowledge from the variables determines whether the failure is effectively prevented or not, in this case, it determines what the cost of the failure occurring will be. As described in the examples above, the later the failure is detected, the greater the cost incurred in solving the resulting problems.
[0007] The current approach to handling large amounts of data is based on a 190 pyramid computing stack, as depicted in Figure 1. At a first level, sensors continuously capture data from a device or machine in an industrial facility (system critical, in general) under supervision / quality control. A second level is the level of local control implemented, for example, with PLCs (Programmable Logic Controllers) or Control tools
Petition 870190089699, of 10/09/2019, p. 10/104
6/87
Computer Numeric (CNC), in which the sensor signals are converted into digital data. The data obtained from the sensors are communicated to an Industrial Control System (ICS), such as SCADA (Supervision Control and Data Collection), through the PLC / CNC level. SCADA is a centralized computer-based system for remote monitoring and control of industrial processes (which can also be disseminated in large areas), such as manufacturing, power generation, infrastructure processes, transmission and transportation of electricity, among others. Local control actions are performed by the PLC / CNC, while SCADA is in charge of global control actions that affect the entire system. The SCADA layer deals with information, while the lower layer (PLC / CNC) deals directly with data. In this context, the term data refers to raw values (such as a voltage or other value), as provided by their corresponding source (for example, sensor, PLC / CNC or device in general), although the term information also refers to refer to data, but in a contextualized way. For example, the voltage resulting from a thermistor is data (particularly raw data), and the voltage, since it receives a meaning (that is, what is the temperature for that particular voltage) is also information.
[0008] The upper levels of the pyramid are MES (Manufacturing Execution System) and ERP (Enterprise Resource Planning). MES are systems responsible for managing product definitions, managing resources, scheduling, dispatching and executing production orders, collecting data from
Petition 870190089699, of 10/09/2019, p. 10/114
7/87 production and perform production analysis, among others, while ERP refers to business management software that provides an integrated view of core business processes. They are used, for example, to collect, store, manage and interpret data from different business activities, such as product planning, marketing and sales, inventory management and shipping and payment. The upper layers (MES and ERP) no longer deal with this type of information, but instead work with business information (also known as knowledge) that can be extracted from such information when it is processed.
[0009] SOADA systems in the prior art have to manage low amounts of information that reflect the current status of supervised machines (eg active / inactive devices in general, open / closed valve states, pressure level in tanks, etc.). ); these situation values can be recoverable and a PLC / CNC, for example. When tasks other than simple supervision are intended (for example, performing predictive maintenance or evaluating whether a device or machine is operating correctly despite evident overall corrective performance), it is necessary to consider all or most of the data provided by related sensors and devices of the industrial machine. A single machine can include hundreds or even thousands of sensors, among which each can supply more than one variable (for example, a photo from a digital camera has as many variables as the number of pixels the photo presents). Consequently, a great deal of
Petition 870190089699, of 10/09/2019, p. 10/124
8/87 amount of data (ie data, which can then provide information and knowledge). In the prior art, data has to be stored in the cloud where it can be processed through so-called cloud computing. Cloud computing is a model that enables ubiquitous on-demand network access typically via the Internet to a shared grouping of configurable computing / processing resources, such as servers, networks and databases. Due to the nature of the Internet, problems such as network latency, security and privacy are inherent in cloud systems. In general, security and privacy are problematic issues when sensitive information and data is stored in the cloud. However, even if these two issues should not be a problem, cloud computing cannot provide the appropriate answer to the demand conditions for controlling industrial environments. First, simple data storage is generally not feasible: there is high data productivity for each machine in an industrial installation, so the bandwidth required to transfer data to the cloud is generally insufficient; In addition, the system is not limited to the communication channel but also to all devices that make it possible to transmit data to the cloud, for example, switches, routers, etc. This is specifically problematic in systems such as those exemplified above, which needs to be controlled in response to abnormal behavior that can be detected only by using the raw data provided by the machine itself and performing the detection in an extremely short period of time, and in
Petition 870190089699, of 10/09/2019, p. 10/13
9/87 secondly, due to the fact that to detect such abnormal behavior, it is necessary to process all stored data and, despite the cloud computing capabilities, this may not be sufficient to provide an answer, let alone in short periods of time. time.
[0010] In order to alleviate these disadvantages, a new paradigm commonly known as fog computing or edge computing was developed with the intention of bringing the cloud closer to the industrial facility. Cloud / edge computing extends cloud computing to the physical location of machines that belong to the network, ensuring proximity to end users and grouping of local resources, thereby reducing network latency, as reported, for example, by Pranali More in Review of implementing Fog computing, IJT: International Journal of Research in Engineering and Technology, Volume 04, Issue 06, June 2015. Fog / edge computing addresses some of the security and privacy issues or concerns associated with computing in to the point that data or information may not have been stored in the cloud. However, fog / edge computing requires implementing a large infrastructure that can be extremely expensive and that, in any case, does not solve all the problems of cloud computing. In particular, the same bandwidth and computing problems are also present in fog / edge computing. However, it is true that network latency may be lower for cloud computing, yet the bandwidth of the communication channel may prove not to be sufficient for the productivity generated in an industrial installation. If
Petition 870190089699, of 10/09/2019, p. 10/144
10/87 network can, in some way, cooperate with the amount of data generated by the entire industrial installation, again, that amount of data would have to be processed in time in order to detect any abnormal behavior of the machines.
[0011] By way of example, Figure 2 shows a schematic of the current architecture that connects industrial networks and the cloud. By way of example, four end points are shown, each representing a different example of industrial networks: a wind turbine park, a locomotive, a machine tool and a cell to insert components into a workpiece using a robot. Each end point is connected to a Ciber-Fisico production system (CPPS) responsible for capturing data from the end points and performing local pre-processing of the captured data. Examples of preprocessing performed by conventional CPPSs are to eliminate erroneous data to save space and simple sample transformations (application of signal processing, for example, mean values, median values, standard deviations, domain changes such as, for example, Transform Fast Fourier time domain to frequency domain, FFT). Secure communication is provided between the CPPS and the cloud in order to send pre-processed data to the cloud for additional storage and processing. In many applications, multiple CPPSs are required to be connected together, for example, over an Ethernet network. These are currently companies that offer devices working according to the illustrated CPPS and with the ability to connect to neighboring devices. Examples of such
Petition 870190089699, of 10/09/2019, p. 10/154
11/87 devices are ^DE Switch Akro 6/0 TSN developed by TTTech, CPPS-Gate40 developed by SoCe and eWON Flexy developed by eWON.
[0012] Armando Astarloa et. there. proposed a CPPS that integrates the IEEE 1588 high availability protocol, composed of intelligent nodes that can capture and process data from multiple sensors, where the nodes are in a high availability ring network to support submicrosecond synchronization (1588-aware HighAvailability Cyber-Physical Production Systems, Precision Clock Synchronization for Measurement, Control, and Communication (ISPCS), 2015 IEEE International Symposium on, Beijing, 2015, pp. 25-30). The network meets extremely high demand synchronization requirements. Each smart node sends a large amount of sensor data combined with timestamp values to a server in the cloud, where they are processed and managed.
[0013] The problem with such a system is the fact that the amount of data generated within each smart node (CPPS) can cause the communication link between smart nodes and the server to not function as expected. Referring again to the example of an industrial machine represented by thousands of variables per second (samples per submisecond), even after preprocessing the data associated with these variables in a CPPS, the server may not be able to process this data within the premises demand time in industrial environments, thus making certain services intelligent, such as services to improve overall efficiency and predictive maintenance of equipment,
Petition 870190089699, of 10/09/2019, p. 10/164
12/87 possible.
[0014] This problem can be solved by implementing a vast infrastructure of communication channels and switches providing enough capacity to handle the huge amount of data to be delivered to the cloud (including data obtainable from sensors, as shown in Figure 1) and ensuring - short times to process data and extract knowledge from the data (this means that sufficient computational resources must always be available and that the network latency is low and stable). However, in addition to the large economic investment required in equipment and maintenance, this is only a short-term solution that is both cost-effective and the demands on channel bandwidth and processing power continuously grow as that new industrial machines are added to the industrial installation, therefore, the system is not scalable.
DESCRIPTION OF THE INVENTION [0015] The device, network and system described in the present disclosure are intended to address deficiencies in prior art smart devices and networks therefrom. The device, network and system is a device, network and system for supervising the operation of at least one machine in an industrial facility. The device, network and system of the invention can also predict the behavior of at least one machine or act on at least one machine, for example, by providing an instruction or command to at least one machine or a component of it in order to react to any defect that may have been detected or diagnosed. Each
Petition 870190089699, of 10/09/2019, p. 10/174
13/87 one of the device, network and system are particularly applicable for use in industrial environments that require intelligent services, such as services to improve the overall efficiency of the equipment and predictive maintenance, in a non-expensive manner, as will be evident from the present description .
[0016] In the context of this disclosure, the term supervise the operation should be interpreted broadly, covering not only passive surveillance of one or more machines in an industrial facility, but also reaction, activation, prediction, control or active management in one or more machines .
[0017] In the context of the present disclosure, the term machine refers to any machine, machinery, station, cell, component or peripheral equipment connected to a machine, machinery, station or cell in an industrial environment. In this text, the term sensor is intended to cover not only sensors in a literal sense but also equipment to interact with a plurality of sensors, such as PLC and CNC providing any data that may be emitted by these machines, including sensor data that can be captured by the equipment for sensor integration. In other words, the sensors belong to a stage of obtaining data in the field and cover any means of obtaining data. In the context of the present disclosure, the term data refers to raw values (such as a voltage or other value), as provided by the corresponding source (for example, sensor, PLC / CNC or device in general), although the term information also refers to
Petition 870190089699, of 10/09/2019, p. 10/184
14/87 data, but in a contextualized form. For example, the voltage resulting from a thermistor is data (particularly raw data), and the voltage, since it receives a meaning (that is, what is the temperature for that particular voltage) is also information. Of course, the term data also includes a sample or a set of samples obtainable from sensors when referring to digitized pieces of information, in the form of bits or bytes, for example.
[0018] A first aspect of the invention relates to a device to supervise the operation of at least one machine in an industrial installation or to supervise such operation and act on at least one machine based on such supervision. The device comprises a first processor configured to compute a first processing task in real time with data obtainable from a machine to which the device is connectable, where the first processing task is to solve a first query. The device further comprises a second processor configured to share its processing power with a network to which the device is connectable and to compute at least one chain of a second processing task when assigned by another device (called a servo or master in the present disclosure. ) of the network, where the second processing task is to solve a second query.
[0019] The device has the capacity to solve queries regarding the operation of at least one machine or at least one component of it. In modalities
Petition 870190089699, of 10/09/2019, p. 10/194
Preferred 15/87, the machine to which the device is connectable is at least one device. The network to which the device is connected is preferably a network of computing clusters in a high-performance computing context. [0020] A second aspect of the invention relates to a system to supervise the operation of at least one machine from an industrial installation or to supervise such operation and act on at least one machine based on such supervision. The system comprises a network comprising, in turn, a server and a plurality of devices that form a computing cluster. At least some devices among the plurality of devices are devices according to the first aspect of the invention. In this sense, each device among the at least some devices among the plurality of
devices is connectable to an machine installation industrial and comprises: one first processor configured to compute an first task of
real-time processing with data obtainable from the machine to which the device is connectable, where the first processing task is to solve the first query; and a second processor configured to share its processing power with the network and to compute at least one thread of a second processing task when assigned by the server, where the second processing task is to solve a second query. The network server is configured to control the computing cluster in order to partition the second processing task into a plurality of threads; and assign at least one thread to the
Petition 870190089699, of 10/09/2019, p. 10/204
16/87 second processor of at least some devices among the plurality of devices. In this sense, the server comprises a processor configured to manage the computing cluster, in order to partition the second processing task into a plurality of threads and assign and transmit at least one thread to the second processor of at least some devices among the plurality of threads. devices. The processor comprised on the server can also be configured to receive output from at least some devices among the plurality of devices; process the outputs of the threads to compute the second processing task; and provide a solution to the second consultation.
[0021] A third aspect of the invention concerns an industrial installation comprising: a plurality of machines; and a network to supervise the operation of at least one machine among the plurality of machines or to supervise such an operation and act on at least one machine based on such supervision, wherein the network comprises a server and a plurality of devices that form a cluster computing. At least some devices among the plurality of devices are connectable to one machine among the plurality of machines, in which each device among at least some devices comprises: a first processor configured to compute, in real time with data obtainable from the machine to which the device is connectable, a first processing task to solve a first query; and a second processor configured to share its
Petition 870190089699, of 10/09/2019, p. 10/21
17/87 processing power with the network and to compute, when assigned by the server, at least one chain of a second processing task to solve a second query. The network server is configured to: control the computing cluster; partition the second processing task into a plurality of threads; and assigning one or more threads among the plurality of threads to the second processor of at least some devices among the plurality of devices. The server can also be configured to receive output from at least some devices among the plurality of devices; process the outputs of the threads to compute the second processing task; and provide a solution to the second consultation. [0022] In preferential modalities, at least some devices comprise each device among the plurality of devices. In addition, in some of these preferred embodiments, each device among the plurality of network devices (of the system or the industrial installation) is a device according to the first aspect of the invention. [0023] The device can resolve queries in real time while obtaining data from the machine in real time, something that, in turn, makes it possible to react in real time or in near real time, for example, to prevent a critical event happen or prevent a critical event that has already occurred from causing damage to the industrial facility or to operators within the industrial facility; In that sense, the device can derive an instruction for the machine from such queries (for example,
Petition 870190089699, of 10/09/2019, p. 10/22
18/87 adjust parameters of a component, machine shutdown etc.). The device can also be integrated within a network of computing clusters in order to contribute to the solution of queries that, due to the complexity or computational load of the same, may not be solved in real time, however, due to the network of clusters of computing, can be solved in time in process, that is, at the same time that an industrial process is taking place (a process of a machine of the industrial installation), thus, enabling a reaction in process. Queries can be queries to supervise the operation of at least one machine (or at least one component of a machine) in an industrial facility or queries to predict the behavior of at least one machine / component or to act on at least one machine / component or queries whose response is used by the device or by the server that controls the devices to react to any defect that may have been detected or diagnosed or to act on the machine / component or to prescribe any action on the machine / component. In addition, since data from the industrial facility can be stored within the network of computing clusters, the device can also contribute to solving queries that do not require a real-time or in-process solution, but which can provide degradation information , possible failures or potential improvements to be made within a given time horizon.
[0024] In the context of the present disclosure, the term real time refers to a time within a range
Petition 870190089699, of 10/09/2019, p. 10/23
19/87 variable between a minimum value V _m in and a value greater than 100 ms (milliseconds), as a variable range between a minimum value V _m in and a value greater than 50 ms, a variable range between a minimum value V _m in and a higher value of 5 ms or a variable range between a minimum value V _m in and a higher value of 1 ms. Considering the current technology, the minimum value V _m in can be, for example, but without limitation, 1 ps (microseconds). However, an element skilled in the art will understand that the evolution of technology may make it possible to reduce the minimum value V _m in the range to a minimum value less than 1 ps, as a minimum value of 500 ns, a minimum value of 100 ns, a minimum value of 20 ns or a minimum value of 1 ns.
[0025] In the context of the present disclosure, queries that can be resolved in real time by a single device (preferably the device that obtains the data required to resolve queries from the machine that can be connected to them) are called first queries, at the whereas queries that can be resolved in time in process are called second consultations.
[0026] Second queries have a task associated with them that can be parallelized by partitioning them into one or more parts (that is, threads), which means that one or more threads of the task to solve the second query can be provided and assigned to different devices in a network of computing clusters. Each of these threads can be computed in a distributed manner (for example, on different devices). When a second processor
Petition 870190089699, of 10/09/2019, p. 10/24
20/87 a device computes a thread, the result of the thread can be integrated with other threads (for example, on a device or on the server of the network of computing clusters) of the same task so that the query can be solved. The time it takes to solve a query by parallelizing the computation of its task may be less than solving it without any parallelization.
[0027] The second processor of the device is configured to compute at least one thread from a second processing task when the server (from a network to which the device is connectable and / or connected) assigns such a thread to the device. The assignment of threads by the server is performed considering the available computational capacity of the devices that form the network. Since the second processor contributes to the solution of a second query by computing a part of the corresponding second processing task, the second processor is configured to share its processing power with the network of computing clusters to which the device is connectable and / or to which the device is connected. If the computing cluster network has, for example, nodes in the form of the device of the first aspect of the invention, the processing power of the second processor can be shared with the second processors of similar devices within the computing cluster network. Although the computing cluster network preferably comprises the number of nodes equal to the number of devices and a server, the computing cluster network can
Petition 870190089699, of 10/09/2019, p. 10/254
21/87 understand additional nodes in the form, for example, but without limitation, Ethernet switches, routers, data storage devices, devices with processing power not shared with the computing cluster network, etc.
[0028] The integration in the same device of obtaining data from multiple sensors with local processing capabilities (in the first processor) and distributed processing capabilities (in the second processor) generates a new pyramidal computing concept for high availability data processing that replaces the traditional pyramid: Although a large amount of business decisions (knowledge) can continue to be stored and processed in the cloud and / or certain information (for example, as a result of processing raw data as collected from machinery interfaces in order to generate a view global industrial installation under quality control) can continue to be stored and processed in the fog / edge, a new computational layer was developed and defined: ground computing.
[0029] Regarding the system server, this includes means to connect to the network (for example, an input / output port or a network interface controller) through which the server is connectable and / or connected to the network. The server and the processor thereof can be further configured to optimize the performance of the computing cluster by balancing the load of the devices within the computing cluster. In this sense, each device can transmit information regarding their computational load to the
Petition 870190089699, of 10/09/2019, p. 10/26
22/87 so that the server processor can know the computational load of each device and, thus, assign any threads to the devices depending on their computational load. The computational capabilities of the computing cluster are then optimized, due to the fact that it may happen that some devices are almost idle while others are performing tasks close to their computational capacity. The processing or computational capacity of the network increases when additional devices are incorporated into the network as new devices and machines are implemented in the industrial facility. This, in turn, can enable the solution of more complex queries due to the fact that the data does not need to be removed from the network (the data remains at ground level).
[0030] The device and system of the present disclosure can provide a new computational scheme formed by a local computational level in which the sensor data (I understand sensors in the broad sense already specified, and data as comprehensive samples) are obtained first and, in then used to solve one or more first queries locally, and a distributed computational level in which there is a distributed solution of one or more second queries. This new computational scheme can reduce the response time to any query and the reaction time to resolve a query (and to derive an instruction from it) within the industrial facility to which the network is associated. In other words, devices and distributed computing (considered high-end computing)
Petition 870190089699, of 10/09/2019, p. 10/274
23/87 performance, that is, HPC) realized by the devices that work in the network of computing clusters increase the overall computational performance derived from the predictive control / supervision applied to the industrial installation. Consequently, a new computational layer is added to the current pyramidal computational stack 190 in Figure 1. This computational layer is called ground computing and covers the processing of the second processed of each device on the network. In other words, ground computing includes the high performance computing performed by the devices included in the network to resolve seconds within the ground computing layer. The ground computing layer is built on top of the lowest layer, hereinafter referred to as the liquid computing layer, which covers the processing of the first processor of each device on the network. The liquid computing layer computes tasks to solve the first queries in real time with data that can be obtained from the machine to which the device is connectable. In order to meet the requirement to resolve the first queries in real time, a first processor that has programmable electronic hardware / components can be used advantageously. For example, but without limitation, a field programmable integrated circuit such as an FPGA (ie, field programmable port arrangement), as a hardware accelerator, can be used. This makes it possible to compute tasks associated with the first queries in the processing cycle times, such as between 1 / 1.5 GHz and 1 / 0.8 GHz, that is, between 0.67 ns (nanoseconds) and 1.25 ns. In this way, the supervision of a machine or a
Petition 870190089699, of 10/09/2019, p. 10/284
24/87 component can be obtained in real time.
[0031] Furthermore, due to the fact that large amounts of data produced in complex industrial installations are processed within the computing cluster (which are closer to the machine than the cloud / fog computing) in two different layers (liquid computing) , solves queries that require real-time processing and, therefore, enables the reaction to an event in near real time, for example, in the order of milliseconds; and ground computing, resolving queries without parallel computing are solved in an insufficient period for quick reaction and, also in some cases, if necessary, to resolve queries with less time-consuming responses, which allows a reaction in the process, that is, during the time necessary to carry out a process), the amount of data delivered to the equipment that performs edge / fog / cloud computing (to be handled offline for an out-of-process reaction, that is, in a time longer than the time required io to perform a process) can be reduced. In fact, there is no need for raw data to be transmitted to the cloud / fog / edge for processing, therefore, the telecommunications infrastructure between the different layers may have a lower capacity and / or higher latency. In contrast, information that is only substantially contextualized leaves the computing cluster (ground computing) for further processing or computing at the edge / fog / cloud. The new computational scheme described can therefore coexist with existing paradigms, such as cloud computing and
Petition 870190089699, of 10/09/2019, p. 10/29
25/87 fog / edge computing, however these are dedicated to refined processing (ie, advanced processing and information in order to gain knowledge of the industrial system under analysis).
[0032] In some embodiments of the invention, the first processing task includes: pre-processing the first data to form a data set to solve the first query. That is, the first processor of the device is additionally configured to pre-process the first data in order to form a data set to solve the first query. In some embodiments of the invention, the first processing task includes: pre-processing the first data to form a data set; and select a subset of data, from the data set, to solve the first query. That is, the first processor of the device is additionally configured to pre-process the first data to form a data set and select a subset of data, from the data sets, to solve the first query.
[0033] The device obtains the necessary data (usually sensor data), pre-processes the data obtained, thereby reducing the size of the data, and solves the queries using the pre-processed data (that is, the data with reduced size). In addition, some or all of the pre-processed data can also be delivered to fog / cloud computing equipment so that, with fog and / or cloud computing, additional processing can be provided for queries or operations of a business nature,
Petition 870190089699, of 10/09/2019, p. 10/30
26/87 for example. In some embodiments of the invention, the second processor of the device is additionally configured to locally compute a third processing task to solve a query (a query not linked to parallel computing, which is also called third queries in the present disclosure), that is, the second processor is not limited to solving tasks in relation to a second query of a parallel nature.
[0034] In some of these modalities, the third processing task includes pre-processing the data obtainable from any device on the network (including the server or the same device) to form a data set to solve the query (not linked to parallel computing) , that is, third consultation); that is, the second processor of the device can be additionally configured to process the data obtainable from any device on the network to form a data set to solve the query not linked to parallel computing.
[0035] In some embodiments of the invention, the first processor of the device is additionally configured to derive an instruction after solving the first query. The first processor of the device is additionally configured to transmit the instruction either to the machine and to the device, to the network, or to a machine connectable to another device among the plurality of devices (on the network). In some cases, the first processor may first transmit the instruction to the second processor which, in turn, is additionally configured to transmit the instruction to the machine connectable to the device and / or to another device among the
Petition 870190089699, of 10/09/2019, p. 10/314
27/87 plurality of devices.
[0036] In some embodiments of the invention, the second processor of the device is additionally configured to provide a solution to the first receptive query of the first processor of the same device to the network or server thereof.
[0037] In some embodiments of the invention, the device additionally includes a data storage device that includes at least one non-volatile memory, such as a hard disk drive (HDD) or, preferably, a solid state drive (SSD) . The second processor of the device can store data and retrieve data from non-volatile memory; For example, the second processor can temporarily store any data that has to be transmitted to the network when the communications channel does not have available free bandwidth and can store any data that is received from the network before processing the data.
[0038] In some embodiments of the invention, the second processor is additionally configured to send data to the server to generate the second processing task, in which the data sent to the server is data obtainable from the machine to which the device is connectable and / or a solution for the first consultation.
[0039] In some embodiments of the invention, the system server comprises or is connected to a data storage device including at least one non-volatile memory. In these modalities, any data produced within the network devices can be stored in the server's non-volatile memory.
Petition 870190089699, of 10/09/2019, p. 10/32
28/87 [0040] In some embodiments of the invention, each device on the network is configured to obtain data (from the machine to which the device is connectable) synchronized with the data obtained by other devices among the plurality of devices (from machine to which each of the other devices are connectable). In order to work synchronously, each device on the network implements a protocol (for example, on the second processor or in the middle of connecting to a network) to synchronize the clocks across the network and to obtain data simultaneously so that each device get data from a machine in the same instant as other devices on the network. For example, this protocol can be Precision Time Protocol (PTP) (IEEE 1588), Coelho Branco or Time Sensitive Network (TSN). In particular modes, each device includes a synchronized timer for synchronization, preferably with sub-second precision, with the server and with other devices on the network.
[0041] In some embodiments of the invention, the server includes a first processor and a second processor. In some embodiments, the server is a device according to the first aspect of the invention.
[0042] In some embodiments, the server includes a processor configured to execute a first instruction set architecture different from a second instruction set architecture executed by the second processor of each device of at least some devices or all devices among the plurality of devices. 0 server includes
Petition 870190089699, of 10/09/2019, p. 10/334
29/87 additionally a network interface that can be connected to the network (for example, to a device within the computing cluster), where the network interface is configured to convert instructions from the first instruction set architecture to the second instruction set architecture and vice versa. The network interface allows the plurality of devices on the network to execute an instruction set architecture that is different from the server's instruction set architecture; this is advantageous since the server can be interconnected with other equipment (for example
example, computing equipment in computation in mist and / or cloud) that runs an architecture in set instruction (ubiquitous) as x86. Therefore,
Regardless of the instruction set architecture executed on the network, the server can communicate any data generated within the network to other devices outside the network by converting (with the network interface) the instructions into the corresponding instruction set architecture. Non-limiting examples of possible instruction set architectures are ARM, Arduino, Raspberry PI, x86, PowerPC, SoC devices etc.
[0043] In some embodiments of the invention, the first processor of the device includes one of: a central processing unit or at least one core thereof, a graphics processing unit, a field programmable integrated circuit, such as an FPGA (ie , field programmable port arrangement), such as a hardware accelerator or an integrated circuit (for example, a system-on-chip, a system-on-chip of multiple processors) - for example, Zynq, MPSoC by Xilinx- and a combination of
Petition 870190089699, of 10/09/2019, p. 10/34
30/87 same. In some embodiments of the invention, the second processor of the device includes one of: a central processing unit or at least one core thereof, a graphics processing unit, a field programmable integrated circuit, such as an FPGA or an integrated circuit (for example, example, a system-on-chip, a system-on-chip of multiple processors) and a combination of them. In some embodiments of the invention, the device additionally includes a multi-core central processing unit and the first processor includes at least one core of the multi-core central processing unit, and the second processor includes at least one other core within the processing unit multiple-core system.
[0044] In some embodiments of the invention, the system network includes a network device (for example, a router, a switch, etc.) to transmit data obtainable as a result of performing the tasks already described on the first processor (net computing) and the second processor (ground computing) from one or more devices (on the network) to an external network or a server on it. In this sense, data that is within the network can be transmitted to a computing device external to the network, for example, a computing device configured to perform foggy or cloud computing. The liquid computing layer and the ground computing layer can then coexist, intensify and use synergies with fog / cloud computing. The external network can be in the cloud or / and in the fog. The network device can be connected to the
Petition 870190089699, of 10/09/2019, p. 10/35
31/87 system network, in which case the server transmits data to the external network through the network device.
[0045] In some embodiments of the invention, the second processor of the device is additionally configured to dedicate part of the processing power to compute at least one thread of the second processing task, that is, part of the processing power of the second processor is dedicated to HPC. In these modalities, the part of the second processor not dedicated to HPC can be dedicated or to compute locally a third processing task that includes solving a query (not linked to parallel computing) or, with the first processor, to pre-process data and / or samples obtainable by the device. Such preprocessing may be applicable especially to critical systems where devices receive very large amounts of data in very short periods of time and, therefore, a relatively large amount of local resources is required to capture and pre-process data.
[0046] The network can be deterministic, meaning that all data generated within the network includes timestamps that allow the identification of the exact time at which the data originated. In a particular mode, the network can be a real-time deterministic network of the Ethernet type, such as Coelho Branco. In an alternative mode, the network can be a TSN network, usually implemented on an Industrial Ethernet network. [0047] The second processor of each device that forms the network's computing cluster works asynchronously. However, in some modalities of
Petition 870190089699, of 10/09/2019, p. 10/36
32/87 invention, the second processor processes data synchronously in relation to the second processor of other devices that form the network of computing clusters.
[0048] Although the device is suitable for IIoT, it can support the connection of machines and / or sensors not equipped with IP, that is, machines and / or sensors that do not produce IP packets. Therefore, the device can be connected to machines and / or sensors with IP or other protocols such as Modbus, PROFINET or 10-Link communication. This feature makes it possible for the device to be deployed in a wide range of industrial facilities such as manufacturing, generation of chemicals and energy, transmission transmission, for example, nuclear power plants, wind power plants, oil refineries, power plants. natural gas processing, chemical treatment plants and manufacturing production lines.
[0049] A fourth aspect of the invention relates to a device to supervise the operation of at least one machine from an industrial installation or to supervise such operation and act on at least one machine based on such supervision. The device comprises a first processor configured to compute a first processing task in real time with data obtainable from a machine to which the device is connectable, where the first processing task is to solve a first query. The device further comprises a second processor configured to locally compute a second processing task to solve another query.
Petition 870190089699, of 10/09/2019, p. 37/104
33/87 [0050]
The device can solve the first query and the other query locally with the first processor and the second processor, respectively, that is, neither the first task nor the second task computed by the processor are linked to parallel computing.
[0051] device of the fourth aspect of the invention can be connectable to (other) similar devices to form a network. Devices can be configured to exchange data over the network. The network can comprise a server. The server and at least some of the devices can form a computing cluster.
[0052] In some embodiments, the first processing task and / or the second processing task includes processing the data obtainable from any device on the network (including the server or the same device) to form a data set to solve the first query and / or other consultation. That is, both the first processor and the second processor are configured to process the data obtainable from any device on the network to form a data set to solve a query (not dealt with in parallel computing).
[0053] The device has the capacity to solve queries in relation to the operation of at least one machine or at least one component of it, in this sense, the first consultation and the other consultation can refer to the operation of at least one machine or at least a component of it.
[0054] A fifth aspect of the invention concerns a system to supervise the operation of at least one
Petition 870190089699, of 10/09/2019, p. 38/104
34/87 machine of an industrial installation or to supervise such operation and act on at least one machine based on such supervision. The system comprises a network comprising a plurality of devices. At least some devices among the plurality of devices are devices according to the fourth aspect of the invention. Each device among at least some devices among the plurality of devices is connectable to a machine in the industrial installation and comprises: a first processor configured to compute a first processing task in real time with data obtainable from the machine to which the device is connectable, where the first processing task is to solve the first query; and a second processor configured to compute a second processing task locally to solve another query (i.e., the second processing task is not linked to parallel computing).
[0055] A sixth aspect of the invention relates to an industrial installation comprising: a plurality of machines; and a network to supervise the operation of at least one machine among the plurality of machines or to supervise such an operation and act on at least one machine based on such supervision. The network comprises a plurality of devices. At least some devices among the plurality of devices are devices according to the fourth aspect of the invention. Each device among at least some devices among the plurality of devices is connectable to a machine in the industrial installation and comprises: a first processor
Petition 870190089699, of 10/09/2019, p. 10/39
35/87 configured to compute a first processing task in real time with data obtainable from the machine to which the device is connectable, where the first processing task is to solve the first query; and a second processor configured to compute a second processing task locally to solve another query (i.e., the second processing task is not linked to parallel computing).
[0056] In some embodiments of the invention, the second processor of a device is additionally configured to share its processing power with the network to which the device is connected and to compute at least one thread of a third processing task when assigned by another device (that is, the server) of the network, where the third processing task is to solve the second query. In these modalities, the network forms a cluster of computers with at least some devices among the plurality of devices. The network additionally comprises a server configured to control the computing cluster; partition the third processing task into a plurality of threads; and assigning one or more threads among the plurality of threads to the second processor of at least some devices among the plurality of devices. In this sense, the server comprises a processor configured to manage the computing cluster, in order to partition the third processing task into a plurality of threads and assign and transmit at least one
Petition 870190089699, of 10/09/2019, p. 10/40
36/87 chaining to the second processor of at least some devices among the plurality of devices. A seventh aspect of the invention relates to a device for supervising and controlling one or more machines of industrial application that have sensors, in which the device comprises: input / output ports for connection to other similar devices; a data acquisition block with a sensor interface connected to an industrial application machine to receive data from multiple sensors of the machine; at least one memory; a first processor that receives data from multiple sensors and uses the data to compute a first processing task in real time to solve a first query regarding the operation of the machine to which the device is connected; and a second processor that shares the processing power of the same with a network of computing clusters, composed of similar devices, to which the device is connected, computing, when assigned by a device in the network, at least one chain of a second task processing to resolve the second query.
[0057] In some embodiments of the invention, the second processor sends a solution of the second processing task to the network device to solve the second query.
[0058] In some embodiments of the invention, the first processor provides a solution for the first query to the second processor and / or derives a first instruction from a solution for the first query and provides the first instruction to the machine to which it is connected.
Petition 870190089699, of 10/09/2019, p. 41/104
[0059] In some embodiments of the invention, the first processor additionally receives a second instruction from the computer cluster network or the second processor and supplies the second instruction to the machine to which it is connected.
[0060] An eighth aspect of the invention relates to a system for supervising and controlling one or more industrial application machines that have sensors, using a plurality of separate computing devices, in which the system comprises: a server connected to communication with the plurality of devices that form a computing cluster, in which the server controls the operation of the devices; where each of the devices has input / output ports for connection to other similar devices; each device has a data acquisition block with a sensor interface connected to a machine to receive data from multiple sensors on the machine; each device has at least one memory; each device has a first processor that receives data from multiple sensors and uses the data to compute a first processing task in real time to solve a first query regarding the operation of the machine to which the device is connected; and the server has at least one memory and at least one processor to solve a second query by partitioning the second processing task into threads to solve the second query and assigning the threads to devices for cooperative computing of the second processing task; and each of the devices that has a
Petition 870190089699, of 10/09/2019, p. 42/104
38/87 second processor that shares the processing power of the same with the computing cluster, computing at least one thread assigned by the server and sending a solution of the same to the server to solve the second query.
[0061] In some embodiments of the invention, the first processor of each device provides a solution for the first query to the second processor of each device and / or derives a first instruction from a solution for the first query and provides the first instruction to the machine to the which it is connected.
[0062] In some embodiments of the invention, at least one server processor derives a second instruction from a solution for the second query and supplies the second instruction to a machine through the device that is connected to said machine.
[0063] In some embodiments of the invention, the at least one server processor additionally receives outputs from the threads of the devices, processing the outputs to compute the second processing task and provides the solution for the second query.
[0064] A new aspect of the invention relates to a system to supervise and control one or more machines of industrial application that have sensors, using a plurality of separate computing devices, in which the system comprises: a server connected to communication with the plurality of devices that form a computing cluster, in which the server controls the operation of the devices; each of the devices has input / output ports for connection to
Petition 870190089699, of 10/09/2019, p. 43/104
39/87 other similar devices; each device has a data acquisition block with a sensor interface connected to a machine to receive data from multiple sensors on the machine; each device has at least one memory; each device has a first processor that receives data from multiple sensors and uses the data to compute a first processing task in real time to solve a first query regarding the operation of the machine to which the device is connected; and the server has at least one memory and at least one processor to solve a second query by partitioning the second processing task in threads to solve the second query and assigning the threads to some devices among the plurality of devices for cooperative computing the second processing task; and each of the devices that has a second processor that shares its processing power with the computing cluster, computing at least one thread assigned by the server and sending a solution to the server to resolve the second query.
[0065] In some embodiments of the invention, the first processor of each device derives a first instruction from a solution for the first consultation and supplies the first instruction to the machine to which it is connected.
[0066] In some embodiments of the invention, at least one server processor derives a second instruction from a solution for the second query and supplies the second instruction to a machine through the device that is
Petition 870190089699, of 10/09/2019, p. 44/104
40/87 connected to said machine.
[0067] In some embodiments of the invention, the at least one processor of the server additionally receives outputs from the one or more threads of the devices, process the outputs to compute the next processing task and provide a solution for the next query.
[0068] A tenth aspect of the invention relates to a method to supervise the operation of at least one machine in an industrial installation or to supervise such operation and act on at least one machine based on such supervision, through a network that comprises a server and a plurality of devices that form a computing cluster, in which at least some devices among the plurality of devices are connectable to a machine in the industrial installation, in which the method comprises: obtaining, each device from at least some devices, sensor data of the machine to which the device is connectable; compute in real time, on a first processor of each device among at least some devices, a first processing task to solve a first query with the obtained data; receiving, each device among the at least some devices, at least one thread of a sequential processing task to solve a sequential query when assigned by the server; compute, in a processor sequence for each device of at least some devices, the at least one received thread, in which the processor sequence is configured to share its processing power with the network; on the server, control the computing cluster, partition
Petition 870190089699, of 10/09/2019, p. 45/104
41/87 the second processing task in a plurality of threads and assigning one or more threads among the plurality of threads to the second processor of at least some devices among the plurality of devices.
[0069] In some embodiments of the invention, at least some devices comprise each device among the plurality of devices.
[0070] In some embodiments of the invention, the computation of the first processing task comprises: preprocessing the data to form a data set and selecting a subset of data, from the data set, to solve the first query; or preprocess the data to form a data set to solve the first query.
[0071] In some embodiments of the invention, the method additionally comprises: in the first processor, deriving an instruction after solving the first query and transmitting the instruction to the second processor of the same device or the machine connectable to the device.
[0072] In some embodiments of the invention, the method additionally comprises, on the second processor, locally computing a third processing task to resolve a query. In some embodiments of the invention, the computation of the third processing task comprises pre-processing data obtainable from the server or any device among the plurality of devices to form a data set to solve the query.
[0073] In some embodiments of the invention, the method
Petition 870190089699, of 10/09/2019, p. 46/104
42/87 further comprises, on the server: receiving outputs from one or more threads from at least some devices among the plurality of devices; process the outputs to compute the second processing task; and provide a solution to the second consultation.
[0074] In some embodiments of the invention, each device among at least some devices performs data acquisition, from the machine to which the device is connectable, synchronized with the data obtained by other devices among at least some devices, of the machine to which each of the devices is connectable.
[0075] In some embodiments of the invention, the method further comprises, by means of a network device, transmitting data within the network to a computing device external to the network.
[0076] Furthermore, similar advantages, as described for the first, second and third aspects of the invention can also be applicable to the fourth, fifth and sixth aspects of the invention.
BRIEF DESCRIPTION OF THE DRAWINGS [0077] To complete the description and to provide a better understanding of the invention, a set of drawings is available. Said drawings form an integral part of the description and illustrate modalities of the invention that should not be interpreted as restrictions on the scope of the invention and only as examples of how the invention can be realized. The drawings include the following figures:
[0078] Figure 1 shows the pyramidal model that illustrates
Petition 870190089699, of 10/09/2019, p. 47/104
43/87 the paradigm of how data are processed in industrial networks based on loT and / or IIoT according to the prior art.
[0079] Figure 2 shows a schematic of a prior art architecture for the supervision of industrial installations.
[0080] Figure 3 shows schematically a system formed by a plurality of nodes according to an embodiment of the invention.
[0081] Figures 4 and 5 show schematically systems formed by a plurality of nodes according to different modalities of the invention, with the machines to which the nodes that form respective networks are connected, and also the interconnection of respective networks to the cloud / fog . [0082] Figure 6A illustrates a schematic of a machine from an industrial installation to which the device of the present disclosure is connectable. Figure 6B illustrates a schematic of a company's worldwide industrial infrastructure; the infrastructure can be controlled by the devices and system of the present disclosure in cooperation with existing cloud / fog facilities.
[0083] Figure 7A schematically shows the interfaces and processors of a device according to one embodiment of the invention, and Figure 7B schematically shows several functionalities that such a device can be provided for its operation.
[0084] Figure 8 illustrates in diagrams a query solvable by a device according to a modality of the invention, either locally or in parallel with other devices.
Petition 870190089699, of 10/09/2019, p. 48/104
44/87 [0085] Figure 9A shows a pyramidal model similar to that of Figure 1, but which illustrates the paradigm of how data can be processed with a system according to a modality of the invention. Figure 9B shows a computational pyramid with net computing, ground computing, fog computing and cloud computing.
[0086] Figure 10 shows in diagrams an example of how the system can supervise the operation of a machine in an industrial installation.
[0087] Figure 11 shows in diagrams another example of how the system can supervise the operation of a machine in an industrial installation.
[0088] Figure 12 illustrates a rotating component whose operation can be supervised with a device and / or a system according to the modalities of the invention.
DESCRIPTION OF A WAY TO CARRY OUT THE INVENTION [0089] Figures 3 to 5 show different modalities of a system that forms a network according to the present invention that can be particularly suitable for supervising and analyzing the behavior, operation and / or performance of one or more machines in an industrial installation and eventually to act on one or more machines and to predict their behavior, operation and / or performance and prescribe a reaction to such behavior, operation and / or performance.
[0090] The network includes a plurality of devices and a server that controls the network and its devices. Throughout the present disclosure, devices and servers are also referred to as network nodes. Without limitation, the industrial installation may belong to one of the
Petition 870190089699, of 10/09/2019, p. 49/104
45/87 following fields: automotive industry, transportation industry, including marine and aerospace, air traffic control industry, energy industry, medical equipment industry, cyber defense industry and other industries for manufacturing purposes among others. In the context of the present disclosure, a machine in an industrial facility refers to any device or machine used to carry out an industrial process.
[0091] In Figure 3, the system includes a network 100 which includes, in turn, a server 10 and a plurality of devices 21-26 configured as a computing cluster. Each device 21-26 has a first processor and a second processor, as described later, and is connectable to either a machine (or a component thereof) in an industrial facility (through the interfaces or input / output ports 29) and to a network (through interfaces or input / output ports 28) such as network 100. The illustrated network 100 is implemented in a ring topology, but any other topology that allows a cluster configuration is also possible, such as a star topology. However, due to the nature of industrial facilities, the use of a ring topology can be highly convenient since the network remains fully operational even after the failure of a communication link between two adjacent nodes; at least, this topology facilitates the cabling of the network since when a new machine is added to the industrial installation, two adjacent nodes can be disconnected from each other so that an additional device can be added between the
Petition 870190089699, of 10/09/2019, p. 50/104
46/87 same. Similarly, since the system is scalable and the processing power of the system depends on the number of devices on the network, it may be convenient to add more nodes to the network in the form of devices, even if new machines are not incorporated into the industrial installation. Devices 21-26, since they are interconnected with links 110, form a computing cluster managed by server 10. Server 10 is configured to assign task threads to be computed to at least some 21-26 devices for high computing. performance, as will be explained later. Links 110 preferably include optical fibers. Server 10 may comprise a device, such as devices 2126, or may comprise a different device, for example, that has enhanced or more powerful processing capabilities.
[0092] In relation to Figure 4, each of the devices 21-26 is connectable to one or more machines 121-125 (for example, as shown in Figure 4) through one or more interfaces 29. The connection between a device and a machine can be a direct connection or an indirect connection (that is, through an interconnect device, such as an Ethernet switch); in addition, this connection can be wired or wireless. A device is connectable to a machine so that it can obtain or receive data from multiple sensors on the machine, for example, by means of sensors and / or by means of control devices (ie controllers) such as a PLC or a CNC that can influence the machine. In this sense, since the device is directly or indirectly connectable to the machine, when the
Petition 870190089699, of 10/09/2019, p. 51/104
47/87 device is connected to the machine, the device can also send data to the machine when necessary (for example, to adjust parameters of a component, to turn off the machine etc.).
[0093] Since each machine can produce different volumes of data (understood as covering samples), in some modalities, it may be necessary that more than one device can be connected to the machine in order to cooperate with the amount of data that it generates. This is illustrated in Figures 4 and 5, in which devices 25, 26 of the respective networks 101, 102 are connected to machine 125, while each of the remaining devices 21-24 of network 101, 102 is connected respectively to a among machines 121-124. Due to the fact that devices 2126 and network 100, and similarly devices 21-27 and respective networks 101, 102, can reduce the reaction time to any inquiry or potential / real defect in the machines associated with networks 100-102 , 100-102 nets are especially useful in industrial applications where the consequence of a machine failure can be, for example, a major economic loss.
[0094] As shown in the modalities illustrated in Figures 4 and 5, networks 101, 102 can be connected to the cloud and / or fog / edge 150 by means of a network device 200 (for example, Ethernet modem / router) of systems that allow communication with the cloud and / or fog / edge 150. Such a connection to the cloud and / or fog can also be implemented in the system of Figure 3 (not illustrated in it). Particularly, in Figure 4, a server 11 is connected via a communication link 18 to a
Petition 870190089699, of 10/09/2019, p. 52/104
48/87 device 27 and server 11 is then connected to network device 200. Device 27 also forms part of network 101. In this way, server 11 also acts as a communications port that makes it possible to connect the network 101 to other networks outside the network 101, such as networks 150 for cloud or fog / edge computing. [0095] Alternatively, as shown in Figure 5, a server 12 can include a network interface 19 through which it is connectable to device 27 and therefore to network 102. The processing power of server 12 can be added to the high-performance computing functionality of the network 102. Processors for devices 21-27 (for example, second processors for them) can run an instruction set architecture different from an instruction set architecture performed by server 12. When this is in this case, the network interface 19 can be configured to convert instructions from the first instruction set architecture to the second instruction set architecture and vice versa. Server 12 is connected to network device 200 so that network 102 can communicate with the cloud and / or fog / edge 150.
[0096] In any of the modalities shown in Figures 3-5, the server 10,11,12 can include data storage means 15. Data storage means 15 are preferably non-volatile memory means, such as a hard disk drive (HDD) or a solid state drive (SSD). The data storage means 15 can store any data, including data associated with queries, tasks and threads.
Petition 870190089699, of 10/09/2019, p. 53/104
49/87 [0097] Figure 6A shows a scheme in the form of functional blocks of how different units / subsystems of an exemplary machine 500 for machining parts are interrelated and how they can be connected to a device to share the data / sample / information generated in the subsystems / units. The machine outlined 500 is a typical machine in the automotive industry, but similar functional blocks can represent other machines in different fields.
[0098] Machine 500 in this example consists of the following units, subsystems and / or components:
[0099] The 511 machining unit: Its function is to ensure that the tool for machining has sufficient precision, strength and torque to carry out the machining task with the necessary performance (quality, production rate ...). For this purpose, the 511 machining unit has four servo-controlled geometry axes: X, Y and Z geometry axes for linear tool movement in space and the S axis to control tool rotation. Each geometry axis is driven by a motor and controlled by a CNC 581 (usually covered by the PLC / CNC 516 block). A CNC 501 bus communicates the motors with the CNC 581, sharing a high amount of information regarding the situation of the motors (energy consumption, temperature, commanded paths etc.) and the sensor readings 561 - 564 (one for each axis) geometric) that control movement with high precision (encoders, temperature sensors, etc.). In addition, an I / O module 521 connected to a 502 fieldbus allows you to connect additional sensors 531 (accelerometers, switches
Petition 870190089699, of 10/09/2019, p. 54/104
50/87 inductive, pressure switches etc.) to the network and control other 541 actuators (counterweight cylinders etc.) of the 511 machining unit.
[0100] A 512 clamping unit: This 512 unit fixes the part to be machined. It needs to locate the part precisely and needs to be able to absorb the cutting forces produced in the machining process. The actuators used by the mechanisms that fix the part need to be commanded and controlled by the 532 (limit switches, analog position control switches, etc.). Sensors 532 and actuators 542 are connected to an I / O module 522 that is connected to a fieldbus 502.
[0101] A 513 hydraulic / pneumatic unit: Many of the machine 500's mechanisms are driven by hydraulic or pneumatic cylinders. This 513 unit provides adequate flow and pressure to the mechanisms. Some 533 sensors (pressure, flow, temperature, level, etc.) control this function so that it is performed properly at the same time that the 543 solenoid valves manage the circuits. Sensors 533 and valves 543 are connected to an I / O module 523 that is connected to field bus 502.
[0102] A 514 cooling unit: The 514 cooling unit provides cooling fluid to these systems of the machine 500 that needs to be cooled. There are some 534 sensors to control this function so that it is performed properly. Similar to hydraulic unit 513, some 544 valves manage the circuit. Sensors 534 and valves 544 are connected to an I / O module 524 that is connected to the fieldbus
Petition 870190089699, of 10/09/2019, p. 55/104
51/87
502.
[0103] A loading / unloading unit 515: It is the system that loads and unloads the part to be processed automatically on machine 500. For this purpose, unit 515 has two geometrical servo-controlled axes 565-566 connected to CNC 581 like those in the machining unit 511. These geometrical axes 565-566 have to carry out the movements of unit 515. In addition, an I / O module 525 connected to field bus 502 allows additional sensors 535 to be connected (inductive switches, pressure switches etc. .) to the network and control other 545 actuators (fasteners, etc.) of the loading / unloading unit 515.
[0104] A 517 quality control unit: Considering the high production rates of these types of production lines in the automotive industry, it is mandatory
control the quality of parts produced from so that are under specification. In that example, this function is performed by a camera 583 and a probe touch of measurement 584, switch 570 among the wed is both are connected to one [0105] One 51 wattmeter 8: This lets you know how the
machine 500 works in terms of energy consumption in order to optimize it. The 518 wattmeter is connected to another switch 571.
[0106] A human-machine interface (HMI) 519: This provides an interface through which a user can interact with machine 500. HMI 519 is also connected to switch 571.
[0107] 0 switch 57 0 has fieldbus 502
Petition 870190089699, of 10/09/2019, p. 56/104
52/87 as an input, and switch 570 is connected to PLC 582 of PLC / CNC 516. CNC 581 is also connected to PLC 582 which, in turn, is connected to switch 571. A device, as described in the present disclosure, it can be, for example, connectable to machine 500 of Figure 6A by means of switch 571, that is, the device can connect to switch 571 in order to obtain data from machine 500.
[0108] In the scheme in Figure 6A, switch 571 can communicate with the connectable device of the same using IP (Internet Protocol), and data can be transferred using a protocol for reliable transmission, such as TCP (Transmission Control Protocol); communication can be established through a physical or wireless connection. In alternative implementations, the device can connect directly to the I / O modules on the machine with the corresponding device interfaces / ports. For this purpose, the device is equipped with I / O ports that allow the connection of physical interfaces; the ports can be adapted to connect different physical interfaces, for example, without limitation, RJ-45. In short, each device (the devices in Figures 3-5) receives data and / or samples from a plurality of sensors supplied in different elements or components of the machine, through interfaces directly or indirectly connectable to the sensors and / or other sources of processing and communication media such as machine controllers (eg PLC, CNC etc.). Non-limiting examples of sensors are temperature sensors, vibration sensors, pressure sensors, position sensors, speed sensors, CCD cameras
Petition 870190089699, of 10/09/2019, p. 57/104
53/87 and / or CMOS, microphones among others. Alternative ways of connecting or interconnecting the I / O modules to a control unit can be implemented instead. Although some machines in an industrial installation are equipped with alarm systems that trigger an alarm when one or more measured values are beyond threshold values, these systems are not able to analyze all data emitted by the machine, its controllers and / or its sensors; the emitted data can be indicative of possible failures even when the measured values are within a range that corresponds to normal operation. Thus, in order to verify the operation and a machine and react to it or predict its performance or operation (for example, possible failure, efficiency or reduced productivity, quality assurance etc.) in order to anticipate a failure, for example , it may be necessary to resolve queries with the data provided by the machine and its associated devices (for example, sensors, actuators, controllers, etc.), as will be explained in detail below in the present disclosure.
[0109] With reference to Figure 6B, it is explained how the device, network and system of the present disclosure enables the control of an industrial environment and enables the provision of an adequate response to the demand conditions of the control of an industrial environment. It is also applied as the network coexists, intensifies and uses synergies with fog / cloud computing.
[0110] Figure 6B shows schematically the levels at which a company's global industrial infrastructure can be divided. At a higher level 2000, a customer or
Petition 870190089699, of 10/09/2019, p. 58/104
54/87 final company is represented. The company can operate, for example, in the automotive industry. Company Management 2000 can have different industrial plants 2100, 2200 spread around the world, represented at a second level in Figure 6B. For example, a first plant may be located in Europe, a second plant may be located in America and a third plant may be located in Asia. Each plant is composed of several industrial lines, forming a third level. By way of example, in Figure 6B, three lines 2110, 2120, 2130 from the first plant 2100 are shown and a line 2210 from the second plant 2200 is shown. A fourth level development is made up of machines that form part of each industrial line. For example, in Figure 6B, two machines 2111, 2112 that belong to line 2110 are schematic. One such machine can be, for example, the machine 500 shown in Figure 6A. As in Figure 6A, machines 2111, 2112 belonging to the 2110 line comprise several components. The 2111 machine can be a machining tool comprising a 2111a machining unit, a 2111b clamping unit, etc. Each component can, in turn, have several elements; For example, the machining unit can comprise servo motors to move the machining unit along 3 geometry axes (geometry X axis, Y geometry axis and Z geometry axis) and a spindle. The spindle, in turn, comprises different sub-elements to be controlled, such as an engine, a front bearing and a rear bearing. Sensors, like the accelerometers shown in Figure 6B, can be used to control the many sub-elements of a machine. a
Petition 870190089699, of 10/09/2019, p. 59/104
55/87 system to analyze the behavior, operation and / or performance of an industrial installation, as revealed, according to Figures 3-5 can be used, for example, to control the operation of each line 2110, 2120, 2130, 2210 shown in Figure 6B, for example, by connecting at least one device (such as device 20 of Figures 7A7B) to each machine on each line.
[0111] Now, business control schematically represented in Figure 6B involves different levels of control or supervision. In a business management world, this control is applied by imposing queries whose response reflects the operation and / or performance of all or part of the business. For example, in order to check the behavior of a component and react to it, or to predict the performance of an installation (for example, possible failure, reduced efficiency or productivity, quality assurance, etc.) in order to anticipate a failure , for example, it may be necessary to resolve queries with / from the data provided by the machine and its associated devices (eg sensors, controller actuators, etc.). For example, a query can refer to the operation or situation of a component of a machine or to the operation or situation of a machine or even to the operation or situation of an entire industrial line or installation to the evolution of all business worldwide. Queries can be defined on a network device, on the server of the same or even outside the network formed by the devices, for example, on a server located in the cloud / fog / edge, in which case the queries can be transmitted to nodes and processors of them, through
Petition 870190089699, of 10/09/2019, p. 60/104
56/87 of communication links.
[0112] As will be explained with reference to Figure 8, in order to solve a query, a task associated with the query needs to be computed. In other words, a task comprises or involves all the processing necessary to provide an answer to the query. Some non-limiting examples of consultations to oversee different levels of an industrial business are as follows. In a heat heat treatment, in which a high-speed thermal camera is used to control the treatment process (such a camera is, for example, included in the machine shown in Figure 6A and indicated with 583), different queries can be imposed: 0 Has the thermal process started , Is the heat source working , Is the temperature distribution adequate to obtain the required surface treatment , Is there any surface at risk of reaching the melting temperature , The temperature distribution is constant for each work product . In a rotating component that has ball bearings, different queries can be imposed: Is the ball passing frequency of the inner ring below its maximum limit , What is the instantaneous acceleration amplitude for the ball turning frequency , When is the remaining service life of a ball bearing component , the ball bearing that is undergoing abnormal degradation , does the ball bearing need to be replaced or serviced . Previous queries are queries regarding the operation of a machine or a component of it. These consultations are also called first consultations and second
Petition 870190089699, of 10/09/2019, p. 61/104
57/87 consultations, as explained with reference to Figures 7A-7B. For example, another second query may be to determine the operation of one machine compared to the operation of other machines of the same type. That is, according to the operation of all machines of the same type, a working model is produced, comprising any small variations between machines of the same type. Then, the state of a machine can be calculated and compared constantly to the normal model.
[0113] Other queries that refer to a higher level of abstraction, such as queries regarding the knowledge (contextualized information) of the industrial installation (ie an industrial line, a plant or even a group of plants), such as business nature, are those treated outside the network of computing clusters. These queries are preferably handled in the cloud and / or in the fog / edge. The non-limiting examples of these consultations are: How many human resources does the industrial facility need to ensure a General Equipment Efficiency of 90% Or how many parts can this facility produce if General Equipment Efficiency is 90% These queries normally have as inputs the solutions to queries treated at the level of computing on the ground (that is, within the computing cluster), as a consequence of this the volume of data sent and treated in fog / cloud / edge computing is reduced in relation to the volume of data processed in ground computing. The level of computing on the ground is described in detail with reference to Figures 7A-7B. Figure 7A shows schematically a device
Petition 870190089699, of 10/09/2019, p. 62/104
58/87 according to an embodiment of the invention in the form of a block diagram. The device 20 can be, for example, any of the devices 21-26 shown in Figures 3 or any of the devices 21-27 shown in Figures 4-5.
[0114] The device 20 comprises a data acquisition block of multiple sensors 60 through which the device 20 is connectable to a machine (for example, the machine 500). The multi sensor data acquisition block 60 includes data acquisition software and using the sensor as an interface. In other words, the multi-sensor data acquisition block acts as an interface with one or more ports of a machine that belongs to the industrial system, installation or infrastructure under control. This interface is also illustrated in Figures 3-4 as ports 29; ports 29 can directly or indirectly provide an interface with sensors. Some non-limiting examples of sensors are temperature sensors, vibration sensors, image capture sensors, among others.
[0115] The connection between a machine and a respective device can be wired or wireless; the device can be connected to a controller (for example, PLC, CNC) controlling the machine, or to a field bus through which data originated in different sensors that belong to a given machine are transmitted. This connection can be direct (direct connection between the device and the machine) or indirect (through a hierarchical chain, for example).
[0116] The device 20 additionally comprises the
Petition 870190089699, of 10/09/2019, p. 63/104
59/87 first processor 61 that is configured to perform net computing 71. In particular, net computing 71 refers to computing a processing task to resolve a query. To compute the processing task, in liquid computing 71, the first processor 61 processes the data from multiple sensors obtained through the multi sensor data acquisition block 60 (from a machine); processing of data from multiple sensors can include pre-processing of data and / or selecting variables from data from multiple sensors. When the first processor 61 performs net computing 71, it performs data processing and resolves a query locally and in real time (also called a first query) so that the system can detect and react to sudden abnormal behavior or a machine operation.
[0117] In some embodiments, the first processor 61 may additionally comprise a central processing unit that has programmable hardware / electronic components, such as, but not limited to, a field programmable integrated circuit, such as an FPGA (ie, port arrangement field programmable), then configured to run a real-time operating system that manages the field programmable integrated circuit or SoC and liquid computing 71. This makes it possible to compute tasks associated with first queries at processing cycle times, such as between 1 / 1.5 GHz and 1 / 0.8 GHz, that is, between 0.67 ns (nanoseconds) and 1.25 ns. In this way, supervision of a machine or component can be achieved in real time. Beyond
Petition 870190089699, of 10/09/2019, p. 64/104
60/87 In addition, the modalities in which the first processor 61 includes a field programmable integrated circuit are especially advantageous in applications where fast query programming / reprogramming is desired, due to the fact that these circuits allow for such rapid programming / reprogramming.
[0118] For example, referring again to Figure 6B, a query regarding any machine or component in any 2100, 2200 plant can be reprogrammed by an offline operator. In other words, queries - or first queries, or second queries or queries from a higher abstraction level - can be defined on a device 20 of the network 100, 101, 102, on the server 10, 11, 12 of the same or even even outside the network formed by the devices, for example, in one located in the cloud / fog. This makes it possible for the remote reprogramming of a query from any remote physical location that provides access to the cloud / fog is available.
[0119] In addition, device 20 also comprises a second processor 62 that is configured to perform computing on the floor 72. During the execution of computing on the floor 72, the second processor 62 can solve threads in which a task associated with a query is divided, as explained later with reference to Figure 8, when device 20 forms part of a network in a computing cluster configuration (such as network 100, 101, 102 in Figures 3-5), that is, in a configuration of HPC. Therefore, in this case, an answer or solution to these queries can be obtained by computing a task in a distributed manner. Consequently, the
Petition 870190089699, of 10/09/2019, p. 65/104
61/87 second processor 62 of device 20 cooperates with the second processor of other devices within the same network to resolve a query (also called a second query). The second processor 62 is designed to solve the threads (portions of a task associated with the query) while performing a computation on soil 72. The threads are parallel parts of the task associated with the query to be solved and that the network server distributes by all devices on the network. Thus, second consultations are consultations associated with tasks of a parallel nature, that is, tasks solved with HPC. A second processor 62 of a device 20 locally resolves the thread assigned to it by the network server. The second processor 62 can also compute tasks independent of the HPC, that is, it can perform data processing that does not involve cooperation with other devices on the network.
[0120] Figure 8 diagrammatically illustrates a query 1500. A first processor 61 or a second processor 62 of a device according to an embodiment of the invention can solve query 1500, which, for example, can provide information regarding a behavior, operation or situation of a machine (or a component thereof) of an industrial installation. In order to resolve query 1500, an associated task 1501 needs to be computed. Task 1501 comprises all the processing required to resolve query 1500. In some cases, the first processor 61 can resolve query 1500 by computing task 1501 locally.
Petition 870190089699, of 10/09/2019, p. 66/104
62/87
Particularly, in some of these cases, task 1501 comprises processing data, thus solving query 1500; the processing of the data can result in the pre-processing of data and / or the selection of variables of the data when pre-processed, and then, in the processing of the data when pre-processed and / or one when specific variables are selected. In some other cases, query 1500 is resolved by more than one device, that is, query 1500 is resolved in a distributed manner. Consequently, a server (for example, as shown in Figures 3-5) on a network can parallelize the computation of task 1501 by dividing it into several 1510a-1510n threads (shown with dashed arrow lines for illustration only) in order to perform the HPC. Each 1510a-1510n thread can be sent to one or more devices so that the second processor of the same can compute. Computing all 1510a-1510n threads, task 1501 can then be computed in order to resolve query 1500. In addition, in some cases, once all 1510a1510n threads have been computed, before solving the query 1500 the additional processing of the result of task 1501 may be necessary to solve the query 1500. By partitioning a task in threads, the time to solve it can be reduced to a greater or lesser degree depending on the number of devices (and the power second processor) available to perform solo computing 72.
[0121] In solo computing 72, the second processor 62 of device 20 can use any data or
Petition 870190089699, of 10/09/2019, p. 67/104
63/87 information obtained from any device on the network or the server itself via network connectivity block 63 and / or the first processor 61 of the same device 20. When the data or information to be used in computing in solo 72 are obtained from other devices or from the server, packets will reach device 20 from one of the two nodes (or even from both nodes, in this case a set of packets is discarded) adjacent to it when the network has a topology However, this does not mean that the system is limited to communications between adjacent nodes: a node retransmits the packets so that they travel from node to node until they reach the destination node.
[0122] The network connectivity block 63 can comprise one or more interfaces through which the device 20 is connectable to a network. In the preferred embodiment in which the network provides a ring topology, the network connectivity block 63 includes at least two interfaces and, preferably, comprises a low-latency network switch for frame forwarding as a 3-port switch: two of the ports being dedicated to ring connectivity and an internal Ethernet port for communication with the multi sensor data acquisition block 60.
[0123] Figure 7B shows in a block diagram several functionalities with which the device 20 is provided in an embodiment of the invention. Device 20 is organized hierarchically into two parts: a first part (the bottom half of device 20, as shown in Figure 7B, for illustration only)
Petition 870190089699, of 10/09/2019, p. 68/104
64/87 dedicated to net computing 71, that is, to process data locally and solve tasks (for example, to answer first queries) 81 with the data as processed; and a second part (the upper half of device 20, as shown in Figure 7B for illustration only) dedicated to computing on the ground 72.
[0124] In relation to the lower half of device 20, device 20 obtains data from one or more machines to which it is connectable. After the data acquisition stage 80, the first processor 61 of the device 20 processes the data 81. Data processing 81 can encompass preprocessing, whereby data from multiple sensors is processed to form a first data set of a smaller size than the original data. Examples of preprocessing performed at this stage may include, for example, clearing nonsignificant data (ie, those outside the operating range of a sensor), defining the accuracy of data collection (for example, truncating decimals in order to convert the variable of a type of numeric data in other types of data with fewer bits, if truncated decimals can be neglected), data transformations (for example, average computation values, median heat, standard deviations, entropies, domain changes such as time domain for frequency domain with the Fast Fourier Transform, for example), and the application of models already trained for variable selection to save data transmission bandwidth (for example, selecting points or pixels of interest in a high resolution digital image
Petition 870190089699, of 10/09/2019, p. 69/104
65/87 or extracting background information from a digital image). A reduction in the amount of data remaining to be processed is thus achieved by minimizing additional general computations and energy consumption. Data processing 81 can also encompass a selection of variables. Thus, from a first data set (for example, generated in a preprocessing stage), the first processor 61 of the device 20 can carry a selection of variables in order to reduce the number of variables that are redundant or that have low significance, according to the additional processing, to be performed by the first processor 61 to solve queries, while some other variables may contain significant information for the first query or queries (or even for other purposes, for example, to solve another first query or queries, to share with other devices within the network of computing clusters that may need them, for foggy / cloud computing and / or for the purpose of logging). The selection of variable allows or contributes to the reduction of overfitting within the computations carried out later. It should be noted that, in some modalities, a single device can receive data from hundreds of sensors that lead to high productivity; from the large volumes of data that are processed, a fraction of the data can be sufficient to solve the queries, thus, the variable solution can reduce the amount of data that will be used in the task computation without affecting its result, that is, without influencing the solution to the consultation
Petition 870190089699, of 10/09/2019, p. 70/104
66/87 associated with the task that is computed task. In other words, due to a variable selection stage, efficiency can be increased by reducing the computational load on the task. An example of a variable selection can refer to a digital image with thousands of pixels from which only a few pixels have been extracted in a pre-processing stage; in the variable selection stage, a subset of pixels from the few pixels already extracted is selected since this subset, together with several different ones of this particular digital image, may be sufficient to compute a given task. Another example can refer to the vibration of determined components within a device in which the processing of local data 81 can produce the frequency spectrum of the vibration; part of the spectrum can be discarded since it does not contain any useful information. Then, the variable selection can only extract particular values from the rest of the spectrum that are relevant to solve some queries in data processing 81 of net computing 71.
[0125] With the data set obtained after the variable selection stage, the first processor 61 solves a task (associated with a first query). The task can be solved locally (on the first processor 61) due to the fact that the original data set may have been reduced by applying pre-processing and variable selection. In this way, the first processor 61 can operate with a reduced amount of data and computes a task with that data; the first processor 61 performs computations using part or all of the
Petition 870190089699, of 10/09/2019, p. 71/104
67/87 computational capacity that the first processor 61 has available.
[0126] The upper half of device 20 is dedicated to computing in soil 72. In computing in soil 72, device 20 performs HPC 93 in such a way that queries (second queries) that require great processing power, that is, associated queries to tasks whose computation requires great processing power, can be solved faster when distributed among different devices through chains of the parallel task associated with them. The second processor 62 can also compute tasks independently of the HPC by performing local data processing 91, that is, data processing that does not involve cooperation with other second device processors on the network. Such processing 91 may include pre-processing and / or selection of variables.
[0127] Although the second processor 62 is configured to work in a computing cluster configuration, the way that the second processor 62 works to solve threads of a parallelizable task (the task that represents the computational work to be computed to answer a query ) can be similar to the operation of the first processor 61. The second processor 62 can process data (generally different from the data obtained in the net computing part) that it obtained either from device 20 alone or from the network.
[0128] During operation in HPC mode, the second processor 62 provides a solution to the chain whose
Petition 870190089699, of 10/09/2019, p. 72/104
68/87 computation was requested by the server. The chaining output is generally transmitted to the network server where all the different outputs produced within the network (that is, from other devices on the network) are collected for their integration in the task. By computing all threads (by the server alone or by a device assigned by the server), the task can be computed in order to solve a query. In some cases, once the task has been computed, prior to resolving the query further processing of the task result may be necessary to provide an answer to the query. The data communication block 94 represents that the second processor 62 can send the solution to the chain and / or other data or with a deterministic data transmission protocol (which can be particularly convenient when it is necessary to ensure that the data reaches its destination without being altered by a party that may have gained non-legitimate access to the network; Examples of such protocols are DATA Distribution Service, ie DDS, for Real Time Systems and Time Sensitive Network, ie TSN) or a non-deterministic data transmission protocol (which can be particularly convenient when the data to be transmitted is not crucial - for example, for logging or for non-immediate actions - since such protocols will not access the communication channel if there is no available free bandwidth; an example of such protocols is Communications in an Open Platform architecture (ie, OPC-UA) and which, similarly, can r and receive data from the network to which it is
Petition 870190089699, of 10/09/2019, p. 73/104
69/87 data are either with a deterministic data transmission protocol or a non-deterministic data transmission protocol. Communications can be managed by the second processor 62 and / or by means of connection to the network. [0129] Although not shown in Figures 3-5 or 7A, device 20 may additionally include data storage means (not shown) accessible by second processor 62. Data storage means are preferably memory means non-volatile, such as a hard disk drive (HDD) or a solid state drive (SSD). Data storage media can store data (for example, samples, solutions for threads, solutions for tasks, queries and solutions for them, etc.) receivable by device 20 and that the second processor 62 can process and / or use for the solution of threads, tasks and / or queries. The data storage means can also store or temporarily store data (for example, samples, solutions for threads, solutions to tasks, queries and solutions to them, etc.) so that the second processor 62 can transmit data to another device on the network 100, 101, 102 and / or to server 10, 11, 12 thereof once the communications channel has available bandwidth, thus not losing data when device 20 cannot transmit due to the fact that all bandwidth is occupied. Device 20 may also include volatile memory means, such as RAM (random access memory), through which data or solutions to tasks can be transferred between the first and second processors 61, 62. The first processor 61 of the
Petition 870190089699, of 10/09/2019, p. 74/104
70/87 device 20 can store data and retrieve data from volatile memory media; As an example, the first processor 61 can temporarily store sensor data before preprocessing the sensor data and can temporarily store the preprocessed data and or data resulting from computing a processing task before transmitting it or to the device's second processor or to the network.
[0130] Still referring to Figure 7B, the upper half of device 20 can also be equipped with cybersecurity mechanisms (not shown) aimed at preventing hacker attacks from outside the network and protecting any data transferred across the network from being read by an unauthorized person or party if data packets are captured by the person or party. Consequently, device 20 may comprise a software-deployable firewall and whose purpose is to block any communications from outside the network that appear to lack the necessary privileges or permissions to establish communications with device 20; the firewall can be run on the second processor 62. In addition, other cybersecurity mechanisms implantable in the device 20 is encryption and decryption of the data in the data packets to be transmitted to or received from the network in order to provide an additional layer of security.
[0131] Device 20 additionally includes mechanisms for device synchronization 90, for example, using the Precision Time Protocol (PTP), so that each device on the network can obtain data on the same devices.
Petition 870190089699, of 10/09/2019, p. 75/104
71/87 moments.
[0132] The two levels of computing (net computing on the first processor 61 and ground computing on the second processor 62) can coexist with other existing computing paradigms, such as cloud / fog / edge computing, to provide different types of responses of consultations, for example, consultations involving different aspects or levels of industrial business as outlined in Figure 6B. The queries to be solved by device 20 or a network comprising device 20 can be defined by operators who monitor the situation and the performance of the industrial installation, namely, to control the correct operation of the machinery within the industrial installation. Queries can be defined on a device 20, or on the server 10, 11, 12, or in addition to the system that includes network 100, 101, 102, for example, in the cloud / fog / edge. An example of a very simple query associated with a machine for machining parts can be to check if the vibrations are submitted by a bearing are within a particular range (for example, in terms of frequency and / or amplitude) considered as normal or correct. This query is preferably resolved in real time on the device's first processor. This type of information can be extracted only from the sensors that capture the bearing while it is in operation, for example, from the accelerometers connected to that bearing. An example of a more complex query could be to verify that all the mechanisms of a robotic arm are working as expected and that the remaining life of each is at least
Petition 870190089699, of 10/09/2019, p. 76/104
72/87 minus two weeks. In that case, there may be up to ten or hundreds of sensors that refer to mechanisms whose sensor and / or sample data can provide the answer to the query only if the data is processed and combined properly. This query does not need to be resolved in real time and may require the processing power of the second processor from several devices (ie, HPC). In this sense, it is necessary that the data collection be carried out in a synchronized manner and provide the same timestamp so that any problem that occurs in a component or machine can be identified and diagnosed with the data (if the data is obtained at different times, it may not be possible to trace the causes that generate the problem); this does not mean that all sensors need to produce data at the same rate, a sensor can pick up more or less frequently depending on the time evolution of a magnitude (for example, the ambient temperature is not expected to change significantly every second while the The energy of a laser can vary significantly many times in a second) as long as the sensor data corresponding to a particular moment can be identified for processing.
[0133] The nature of the sensors can be very variable: whereas an ambient temperature sensor can emit data at a rate of a few bytes per second, an accelerometer in a bearing can emit data at a rate of kilobytes per second (for example , 20-30 kB / s), or a digital camera can output data at a rate of megabytes per second. It is clear that the data add up quickly and the
Petition 870190089699, of 10/09/2019, p. 77/104
73/87 aggregate productivity for a single device can be in the order of units or tens of megabytes per second. To solve a query by computing a task, the data then needs to be processed and the solution to the task needs to be extracted.
[0134] As an example, if there should be movement of a single spindle of a machining tool, the data to monitor the spindle can include variables such as the vibrations of each one between the geometric axis, the bearings, the cage and the rails which can be captured at 20 kilohertz and also variables such as power, torque, temperature and angular spindle speed, in which these variables are captured at 10 hertz for example. Aggregating all this data, productivity can be 0.5 megabytes per second. The time it can take to provide a solution to a query associated with data outside the industrial facility (that is, in the cloud or in the fog) can be in the order of several seconds or even minutes, if the communication channel has bandwidth enough, latency is low and sufficient computing resources are readily available. In addition, it must be considered that although the data is transferred and processed either in the fog / edge or in the cloud, additional data is generated by the same sensors and controllers that must also be analyzed due to the fact that the spindle behavior may already have changed. The first processor (liquid layer) of the device can obtain the data involved and perform, for example, a fast Fourier Transform (FFT). Then, if, for example, the amplitude of a frequency of interest
Petition 870190089699, of 10/09/2019, p. 78/104
74/87 (included in the FFT) is above a threshold, an alarm is sent to a supervisory system for further action.
[0135] As already explained, the two levels of computing (net computing on the first processor 61 and ground computing on the second processor 62) can coexist with other existing computing paradigms, such as cloud / fog / edge computing. For example, by analyzing the instantaneous energy consumption of each machine, improvements to the entire installation can be planned. This query belongs to the business world, typically dealt with in the cloud / fog / edge. Another example consultation of this type is: How many human resources does the industrial installation need to guarantee a 90% General Equipment Efficiency
[0136] Whenever a new device (such as device 20) is integrated into the network as illustrated, for example, in the form of Figures 3-5, the demand for processing / computational resources within the network may increase due to the addition to the network multiple sensor data collected by the new device. However, the processing / computational resources of the network are simultaneously increased due to the incorporation into the network of the first and second processors of the new device and, especially due to the second processor as it is added to HPC distributed processing. The network server monitors the status of each device on the network in terms of load and progress, which solves a particular task or portion of a task (that is, chaining). The server communicates with the devices to do this and so on, for example, using
Petition 870190089699, of 10/09/2019, p. 79/104
75/87 of message transmission protocols such as MPI (Message Transmission Interface). The network can be configured to allow scalable expansion by adding new devices to the network and having the server configured to assign new tasks or existing processing tasks to the new devices. For example, by connecting a new device, a synchronization process can be performed to synchronize the processing of the second processor for all devices within the network. Distributed processing (HPC) is carried out while maintaining network synchronization requirements. Therefore, the workload is spread across multiple devices.
[0137] The first processor of devices in a network 100, 101, 102 works synchronously to process all data received from machines 121-125 and compute tasks, whereas the second processor comprised in each device can work synchronously or asynchronous with respect to the second processor of corresponding devices within the computing cluster when they are computing tasks or threads into which a task is divided.
[0138] The devices and the server of the network 100, 101, 102 can be configured to execute different protocols and processes that allow a correct distributed computing. The second processor of the devices can communicate with other devices and second processors of the same through message transmission communication protocols (for example, MPI); the devices and the server send and receive messages with
Petition 870190089699, of 10/09/2019, p. 80/104
76/87 the use of this type of protocols to perform distributed computing. The message transmission protocols can be used within structures or programming models that can be leveraged towards asynchronous and heterogeneous parallel computing, that is, structures or programming models that support solving the threads in an asynchronous way and by processors. different natures and by processors of different natures (for example, one or several cores of a central processing unit, one or several processing units in a parallel configuration, field programmable integrated circuits, etc.).
[0139] The server 10, 11, 12 manages the HPC within the computing cluster. In particular, it partitions tasks to be solved in parallel to form a plurality of task threads for high performance computing. It also transmits the parts (threads) of such a task to some devices for HPC; once server 10, 11, 12 is aware of the load situation of each device, server 10, 11, 12 can transmit the threads of a task to selected devices on the network that have enough processing power free to compute the themselves. The server 10, 11, 12 receives solutions to the corresponding device threads and reassembles them in order to compute the task and eventually provide an answer to a query.
[0140] A cluster workload manager on the server distributes threads across the entire network 100, 101, 102 focusing on the following: threads for
Petition 870190089699, of 10/09/2019, p. 81/104
77/87 be resolved, information regarding the processing power (available) of each device, and in some cases requests from devices to solve a task in a distributed manner. The cluster workload manager decides which thread is assigned to each device (only a few devices on the network or all devices on the network can, since only a subset of the devices on the network can be used to resolve the threads) based on device status and then the threads using a message transmission protocol. In this sense, a monitoring module or library can dynamically determine the workload of each device and the associated available capacity for the HPC of the device. Another module or library (for example, Load Balancing or DLB) can dynamically balance the load on the devices by adjusting the number of active threads (on the second processor of the same) used in a given process. With the cooperation between the different modules and processes responsible for HPC, the server 10, 11, 12 knows at all times what is the situation of the network 100, 101, 102 and the devices in it so that, for example, the network manager cluster workload can adjust the operation of each device by changing the urgency in which a task to resolve a query needs to be resolved or a new urgent query is declared, thereby speeding up the solution for a particular query.
[0141] The first processor 61 and the second processor 62 are coupled communicatively in order to share
Petition 870190089699, of 10/09/2019, p. 82/104
78/87 data, usually through random access memory (RAM). The data to be shared can be task output (for example, the output of a task solved by the first processor 61 can be transmitted to the second processor 62), data from multiple sensors processed (for example, the first processor 61 can process the data of multiple sensors in order to form a smaller data set that is transmitted to the second processor 62 to perform HPC), instructions for the machine and / or data in relation to instructions submitted to the machine for logging etc.
[0142] Figures 9A and 9B show pyramidal models
600, 610 which illustrate the paradigm of how data is processed with devices and systems in accordance with the invention.
[0143] The level of local computing confined within the system network comprises the aforementioned soil layer and liquid layer, shown schematically together in Figure 9A with respect to the traditional sensors, PLC & SCADA layers of the traditional pyramid 190. The device and system innovators can reduce the reaction time to any inquiry, need or potential / real defect within a system associated with the network. For this reason, the device and system are especially applicable to industrial installations where this aspect is crucial. In addition, due to the fact that data is processed locally within the computing cluster (layers in the ground & liquid), the amount of delivered to edge / fog / cloud computing equipment can be reduced. In fact, the data delivered to computing in
Petition 870190089699, of 10/09/2019, p. 83/104
79/87 edge / fog / cloud should be reduced mainly to information regarding the business / knowledge of the installation. In addition, devices that function as a computing cluster allow you to balance your performance in order to optimize the computing capabilities of the computing cluster. Last but not least, the processing / computational capacity of the network increases at the same time as new devices are added to the network.
[0144] Next, the examples are discussed in order to illustrate the advantages of a system in accordance with a modality of the invention both in terms of efficiency of processing time and reduction of data delivered to fog / cloud computing.
[0145] An example of controlling the operation of a machine in an industrial installation by means of the revealed system is illustrated in Figure 10. Three devices 921, 922, 923 of a network formed by a plurality of devices, in addition to the 910 server, are shown. In order to monitor a heat treatment using a laser applied to a machine, a 901 high-speed thermal camera is used. The device 921 collects data from that camera 901. Camera 901 has a resolution of 1024 pixels per frame, that is, the camera covers 1024 variables (1 pixel is equivalent to 1 variable). Each variable has a word length of 10 bits. The sample rate is 1000 frames per second (1 kHz). Therefore, this camera produces data at a speed of 1280000 bytes / s (1.28 MB / s, 1.28 megabytes per second).
[0146] Regarding the treatment that is applied in
Petition 870190089699, of 10/09/2019, p. 84/104
80/87 machine, it is possible to make first consultations, for example: has the thermal process started (Or similarly, is the heat source (laser) working ) In order to resolve this query, a task associated with that query and based on it is created. Computing this task will provide the answer to the query. In this case, the task to be computed is to obtain a region of interest (ROI) and to process it.
[0147] In this way, the data obtained from camera 901 at a rate of 1.28 MB / s is sent to the first processor 921a of the device 921 in which this task is computed by applying a region of interest (ROI) algorithm to eliminate background pixels from each frame image and only work with pixels that contain information. In a particular example, the ROI is a bit greater than 70%: 729 pixels per frame are selected, which corresponds to 911,000 bytes / s (911 Kbytes / s). In Figure 10, an example of a region of interest obtained after applying an ROI algorithm is shown. If the ROI is the appropriate size, then it can be concluded that the thermal process has started (that is, the first consultation is answered). Then, on the first 921a processor, the data that comes from the ROI is used to generate a connectivity matrix, in this case, the matrix with dimensions 729 x 729, which define for each pixel the neighboring pixels that follow a data structure. The cellular result of this connectivity matrix provides an output of 66.4 kB / s, which, added to 911 kB / s, results in 977.4 kB / s as an output of net computing (first 921 a processor). This connectivity matrix is useful for detecting the spatial configuration
Petition 870190089699, of 10/09/2019, p. 85/104
81/87 pixels. In other words, the ROI result is further processed in net computing.
[0148] In addition, second consultations can be made. For example: Is the temperature distribution adequate to obtain the required surface treatment Or is there any surface at risk of reaching the melting temperature Or is the temperature distribution constant for each work product In order to solve this query, a task associated with it and based on it is created. Computing this task will provide the answer to the query. In this case, the task to be computed is to obtain the temperature distribution on the surface. The solution to this task involves analyzing the frames captured by the 901 camera at a rate of 1000 frames / s.
[0149] In order to perform this processing, the ROI and the connectivity matrix are sent to the temporary memory storage 921c of the second processor 921b of the device 921 in which the first task was computed. The data stored in temporary storage 921c is sent to server 910. Then, the task is divided into threads by server 910 (in programmer 910a) to be sent to different second processors 921 b, 922b, 923b from corresponding devices 921, 922, 923. In particular, each thread runs an algorithm to process the data associated with a different picture frame. Each chain can be running a cumulative grouping algorithm on the pixels of an image frame with the obtained ROI and with common characteristics in terms of time and space (space-time) resulting from the matrix of
Petition 870190089699, of 10/09/2019, p. 86/104
82/87 connectivity. The server 910 assigns threads to various devices 921, 922, 923, particularly, to the second processors 921b, 922b, 923b thereof. In other words, the task is parallelized in order to be able to cooperate with the large amount of data from the camera (1000 frames / s). Each second processor 921b, 922b, 923b therefore processes different image frames.
[0150] The result of each chaining is the median, minimum and maximum temperatures and the standard deviation for each grouped frame. In this particular example, the number of clusters is 9. The output data transmission is 144 kB / s (144000 bytes / s). This data (solution of all threads) is sent to the server (in programmer 910b) in order to generalize the values along the cycle time of the heat treatment process. This combination is assigned by the server 910 to the second processor 921b of a device 921. The temperature distribution over the task surface provides the answer to the second query. [0151] Another example of controlling the operation of a machine in an industrial installation through the revealed system is illustrated in Figure 11. Three devices 921, 922, 923 of a network formed by a plurality of devices, in addition to the 910 server, are shown. In order to perform condition monitoring on a rotating component 130, an accelerometer 1201 is used.
[0152] Figure 12 shows two views of a rotating component 130 (side view on the left and a 3D section view on the right). The 1201 accelerometer has a word length of 24 bit. The sample rate is 30 kHz to monitor phenomena of a maximum of 10 kHz.
Petition 870190089699, of 10/09/2019, p. 87/104
83/87
Therefore, the 1201 accelerometer produces data at a speed of 90 kB / s (90000 bytes / s). From accelerometer 1201, data is sent to the first processor 921a of device 921 in a fast Fourier Transform (FFT) is applied to move from one time domain to a frequency domain. This transformation reduces the amount of data transmitted by half, providing 45 kB / s (45000 bytes / s). From the FFT, the frequencies of interest for ball bearings are selected: FTF fundamental training frequency, inner ring ball pass frequency ΒΡΕΙ, BPFO outer ring ball pass frequency, BSP ball turn frequency and frequency of SRF stem rotation, as shown in Figure 12 (right). With these frequencies, it is possible to make first consultations, for example: is the ball passing frequency of the inner ring below its maximum limit What is the instantaneous amplitude of acceleration for the ball turning frequency
[0153] Then, the amplitude for each of the 5 frequencies of interest is sent to the temporary storage of temporal memory 921c of the second processor 921b of the device 921. The 5 frequencies and their corresponding amplitudes imply 10 variables, among which each requires 4 bytes. Due to the fact that in this example 2 FFT are performed per second, the output of the liquid stage provides data at 80 B / s. Data stored in temporary storage 921c is sent to server 910 (in programmer 910a), which assigns threads to other second processors 921b, 922b, 923b of devices
Petition 870190089699, of 10/09/2019, p. 88/104
84/87 respective 921, 922, 923. In this case, the parallelization is not due to a large amount of data to be processed (as was the case with the data collected by the camera in the example illustrated in Figure 10); In contrast, parallelization is required due to the fact that an agglomeration algorithm of k means needs to be initialized several times in order not to fall into local minimums (due to the fact that the algorithm is launched from random points that can lead to a local minimum). Therefore, in this case, each thread sent to other second processors comprises executing a clustering algorithm of k means to group amplitudes with common characteristics in terms of time. The algorithm is then initialized several times (once per thread) with the same input data (historical data stored in temporary storage 921c) in order to later select the best (the most separate centroides). The result of each chain is centroid (5 variables, one per frequency) and a diameter for each cluster (that is, 6 variables). There are 4 bytes / variables. In this particular example, the number of clusters is 3. Therefore, each half k provides 72 bytes. Due to the fact that, in this example, it was established that 10000 FFT are used for each k media clustering algorithm and 2 FFT are performed per second, and considering that each k media provides 72 bytes, at this stage, data transmission The resulting output is about 0.0144 bytes / s. Parallelization of the execution of each k media clustering algorithm is necessary due to the fact that each execution can take several seconds. This data that starts from each thread is sent back to the
Petition 870190089699, of 10/09/2019, p. 89/104
85/87 server 910 (in programmer 910b) in order to perform a statistical test to detect whether the groupings are involving (change of concept). The computation of this test is assigned by the server to another second 921b processor. With the concept change values (new centroid positions), a task to solve the second queries is completed. Examples of such second queries are: what is the remaining service life of the ball bearing component Does the ball bearing show abnormal degradation Does the ball bearing need to be replaced or serviced As can be seen, although the second processor 923b of the device 923 is computer-based, according to which a k means clustering algorithm is executed in order to solve the task associated with a rotating element of the machine to which the device 921 is connected, the first 923a processor of device 923 is computing, in net computing, another task to solve a query associated with a rotating element of the machine to which device 923 is connected.
[0154] As can be deduced from the previous examples, the volume of data obtained in the net computing output (first processor of a device) in relation to the data processed by the first processor (that is, in relation to the data in the computing input net) is reduced. In the first example, a reduction of 1280 / 977.4 times is approximately achieved. In the second example, a reduction of 90,000 / 80 times is obtained in net computing. In relation to ground computing, in the first example, a reduction of 977.4 / 144 times is obtained,
Petition 870190089699, of 10/09/2019, p. 90/104
86/87 whereas in the second example, a reduction of 80 / 0.0144 times is obtained. This implies that the volume of data is provided for further processing in fog / cloud computing is also reduced. In the embodiments of the invention, the volume of data provided for further processing in edge / fog / cloud computing is preferably at least 10 times less than the volume of data arriving at ground computing, more preferably, 20 times less , even more preferably, 100 times smaller and, in some cases, is even up to 10 ⁶ times smaller. How often it is less depends on the consultations that are treated.
[0155] The first, second and third queries as described in the present disclosure can be queries for at least one of the following: supervising the operation of at least one machine (or at least one machine component) in an industrial facility; predict the behavior of at least one machine / component; act on at least one machine / component; control the devices (with the response to queries that are used by a device or the network server) in order to react to any defect that may have been detected or diagnosed; and prescribe any action on the machine / component.
[0156] In this text, the term understands and includes and its derivations (as it understands, that includes etc.) should not be interpreted in an excluding sense, that is, these terms should not be interpreted as excluding the possibility that what is described and defined can include elements, additional steps etc. At the
Petition 870190089699, of 10/09/2019, p. 91/104
87/87 In this text, the terms multiplicity and plurality were used interchangeably.
[0157] The invention is obviously not limited to the modality (or modalities) described in this document, but also covers any variations that may be considered by any technician in the subject (for example, in relation to the choice of materials, dimensions, components, configuration , etc.), within the general scope of the invention as defined in the claims.

权利要求:
Claims (29)
[1]
1. System to supervise the operation of at least one machine (121-125, 500) of an industrial installation or to supervise such operation and act on at least one machine (121-125, 500) based on such supervision, the system being characterized by understanding:
a network (100-102) comprising a server (10-12) and a plurality of devices (20-27) that form a computing cluster, in which at least some devices among the plurality of devices (20-27) are connectable to a machine (121-125, 500) of the industrial facility to receive sensor data from the machine, where each device (20-27) out of at least some devices comprises:
a first processor (61) configured to compute in real time, with data obtainable from the machine (121-125, 500) to which the device (20-27) is connectable, a first processing task (1501) to solve a first consultation (1500) regarding the operation of a machine or a component of the machine to which a device can be connected; and a second processor (62) configured to share its processing power with the network (100-102) and to compute, when assigned by the server (10-12), at least one thread (1510a1510n) of a second processing task ( 1501) to solve a second query (1500) in relation to the operation of at least one machine or a component of it to which at least one device is
Petition 870190085137, of 30/08/2019, p. 49/73
[2]
2/11 connectable;
wherein the server (10-12) is configured to: control the computing cluster;
partition the second processing task (1501) into a plurality of threads (1510a1510n); and assigning one or more threads (1510a-1510n) among the plurality of threads to the second processor (62) of at least some devices among the plurality of devices (20-27).
2. System according to claim 1, characterized by the fact that at least some devices comprise each device (20-27) among the plurality of devices (20-27).
[3]
3. System, according to any previous claim, characterized by the fact that the first processing task (1501) comprises:
pre-process the data to form a data set and select a subset of data, from the data set, to solve the first query (1500); or pre-process the data to form a data set to solve the first query (1500).
[4]
4. System, according to any previous claim, characterized by the fact that the first processor (61) is additionally configured to derive an instruction after solving the first query (1500) and transmitting the instruction to the second processor (62) from the same device or machine (121-125, 500) connectable to the device (20-27).
Petition 870190085137, of 30/08/2019, p. 50/73
3/11
[5]
5. System, according to any previous claim, characterized by the fact that the second processor (62) is additionally configured to locally compute a third processing task (1501) to solve a query (1500).
[6]
6. System according to claim 5, characterized by the fact that the third processing task (1501) comprises pre-processing data obtainable from the server (10-12) or any device among the plurality of devices (20-27) to form a data set to solve the query (1500).
[7]
7. System, according to any previous claim, characterized by the fact that the server (10-12) is additionally configured for:
receiving outputs from one or more threads (1510a1510n) from at least some devices among the plurality of devices (20-27);
process the outputs to compute the second processing task (1501); and provide a solution to the second consultation (1500).
[8]
8. System, according to any previous claim, characterized by the fact that the second processor (62) is additionally configured to send data to the server (10-12) to generate the second processing task (1501), in which the data sent to the server (10-12) are data obtainable from the machine (121-125, 500) to which the device (20-27) is connectable (1501) and / or a solution for the first consultation (1500).
[9]
9. System, according to any previous claim, characterized by the fact that each device
Petition 870190085137, of 30/08/2019, p. 51/73
4/11 among at least some devices is configured to obtain data, from the machine (121-125, 500) to which the device is connectable, synchronized with data obtained by other devices from at least some devices, from the machine (121-125, 500) to which each of the other devices is connectable.
[10]
System according to any preceding claim, characterized in that it further comprises a network device (200) for transmitting data within the network (100-102) to a computing device external to the network (100-102), wherein the computing device is preferably configured to perform fog computing or cloud computing.
[11]
11. System, according to any previous claim, characterized by the fact that the server (12) comprises:
a processor configured to execute a first instruction set architecture other than a second instruction set architecture performed by the second processor (62) of each device among at least some devices; and a network interface (19) connectable to a device (27) out of the plurality of devices (20-27), where the network interface (19) is configured to convert instructions from the first instruction set architecture to the second architecture instruction set and vice versa.
[12]
12. Device (20-27) to supervise the operation of at least one machine (121-125, 500) from an industrial installation or to supervise such operation and act at
Petition 870190085137, of 30/08/2019, p. 52/73
5/11 at least one machine (121-125, 500) based on such supervision, the device (20-27) being characterized by comprising:
a first processor (61) configured to compute in real time, with data obtainable from at least one machine (121-125, 500) to which the device (20-27) is connectable, a first processing task (1501) to solve a first query (1500) regarding the operation of a machine or a component of the machine to which a device is connectable; and a second processor (62) configured to share its processing power with a network of computing clusters (100-102) to which the device (20-27) is connectable and for computing, when assigned by a device on the network (100 -102), at least one thread (1510a-1510n) of a second processing task (1501) to solve a second query (1500) in relation to the operation of at least one machine or a component of it to which at least one device it is connectable.
[13]
13. Device according to claim 12, characterized by the fact that the second processor (62) is additionally configured to send data to the server (10-12) to generate the second processing task (1501), in which the data sent to the server (10-12) data is obtained from the machine (121-125, 500) to which the device (20-27) is connectable (1501) and / or a solution for the first consultation (1500).
[14]
14. Device according to any one of claims 12-13, characterized by the fact that the
Petition 870190085137, of 30/08/2019, p. 53/73
6/11 the first processor (61) and / or the second processor (62) comprise a multi-processor system-on-chip.
[15]
Device according to any one of claims 12-14, characterized in that the second processor (62) is additionally configured to locally compute a third processing task (1501) to resolve a query (1500).
[16]
16. Device according to any one of claims 12-15, characterized by the fact that the first processing task (1501) comprises:
pre-process the data to form a data set and select a subset of data, from the data set, to solve the first query (1500); or pre-process the data to form a data set to solve the first query (1500).
[17]
17. Device according to any one of the preceding claims 12 to 16, characterized in that the first processor (61) is additionally configured to derive an instruction after solving the first query (1500) and transmitting the instruction to the second processor ( 62) from the same device or from the machine (121-125, 500) connectable to the device (20-27).
[18]
18. Industrial installation characterized by comprising:
a plurality of machines (121-125, 500); and a network (100-102) to supervise the operation of at least one machine (121-125, 500) of the plurality of machines or to supervise such an operation and act on at least one machine (121-125, 500) based on such supervision, where
Petition 870190085137, of 30/08/2019, p. 54/73
The network (100-102) comprises a server (10-12) and a plurality of devices (20-27) that form a computing cluster;
at least some devices among the plurality of devices (20-27) are connectable to one machine (121— 125, 500) among the plurality of machines to receive sensor data from the machine, where each device (20-27) among the at least some devices include:
a first processor (61) configured to compute in real time, with data obtainable from the machine (121-125, 500) to which the device (20-27) is connectable, a first processing task (1501) to solve a first consultation (1500) regarding the operation of a machine or a component of the machine to which a device can be connected; and a second processor (62) configured to share its processing power with the network (100-102) and to compute, when assigned by the server (10-12), at least one thread (1510a1510n) of a second processing task ( 1501) to solve a second query (1500) regarding the operation of at least one machine or a component of the same machine to which at least one device is connectable;
wherein the server (10-12) is configured to: control the computing cluster;
partition the second processing task (1501) into a plurality of threads (1510a1510n); and assign one or more threads (1510a-1510n)
Petition 870190085137, of 30/08/2019, p. 55/73
8/11 among the plurality of threads to the second processor (61) of at least some devices among the plurality of devices (20-27).
[19]
19. Industrial installation according to claim 18, characterized by the fact that at least some devices comprise each device (20—27) from among the plurality of devices (20-27).
[20]
20. Industrial installation according to any one of claims 18-19, characterized by the fact that the server (10-12) is additionally configured for:
receiving outputs from one or more threads (1510a1510n) from at least some devices among the plurality of devices (20-27);
process the outputs to compute the second processing task (1501); and provide a solution to the second consultation (1500).
[21]
21. Method for supervising the operation of at least one machine (121-125, 500) from an industrial facility or for supervising such operation and acting on at least one machine (121-125, 500) based on such supervision, by means of a network (100-102) comprising a server (10-12) and a plurality of devices (20-27) that form a computing cluster, in which at least some devices among the plurality of devices (20-27) are connectable to a machine (121-125, 500) of the industrial installation, the method being characterized by understanding:
obtain, through each device from at least some devices, sensor data of the machine to which the device is connectable;
compute in real time on a first processor
Petition 870190085137, of 30/08/2019, p. 56/73
9/11 each device (20-27) among at least some devices, a first processing task to solve a first query (1500) with the obtained data;
receive, each device among at least some devices, at least one thread of a second processing task to solve a second query when assigned by the server;
compute, in a second processor of each device of at least some devices, the at least one received thread, in which the second processor is configured to share its processing power with the network;
on the server (10-12), control the computing cluster, partition the second processing task (1501) into a plurality of threads (1510a-1510n) and assign one or more threads (1510a-1510n) among the plurality of threads to the second processor (62) of at least some devices among the plurality of devices (20-27).
[22]
22. Method according to claim 21, characterized by the fact that at least some devices comprise each device (20-27) among the plurality of devices (20-27).
[23]
23. Method according to any one of claims 21-22, characterized in that the computation of the first processing task (1501) comprises: pre-processing the data to form a data set and selecting a subset of data, from the data set, to solve the first query
Petition 870190085137, of 30/08/2019, p. 57/73
10/11 (1500); or pre-process the data to form a data set to solve the first query (1500).
[24]
24. Method according to any of claims 21-23, characterized in that it further comprises, in the first processor (61), deriving an instruction after solving the first query (1500) and transmitting the instruction to the second processor (62) of a device or the machine (121-125, 500) connectable to the device (20-27).
[25]
25. Method according to any one of claims 21-24, characterized in that it comprises, on the second processor (62), locally computing a third processing task (1501) to solve a query (1500).
[26]
26. Method according to claim 25, characterized by the fact that the computation of the third processing task (1501) comprises pre-processing data obtainable from the server (10-12) or any device among the plurality of devices (20- 27) to form a data set to solve the query (1500).
[27]
27. Method according to any one of claims 21-26, characterized in that it further comprises, on the server (10-12): receiving outputs from one or more threads (1510a-1510n) from at least some devices among the plurality of devices (20-27); process the outputs to compute the second processing task (1501); and provide a solution to the second consultation (1500).
Petition 870190085137, of 30/08/2019, p. 58/73
11/11
[28]
28. Method according to any one of claims 21-27, characterized by the fact that each device among at least some devices obtains data, from the machine (121-125, 500) to which the device is connectable, synchronized with data obtained by other devices from at least some devices, the machine (121-125, 500) to which each of the other devices is connectable.
[29]
29. Method according to any one of claims 21-28, characterized in that it further comprises transmitting, via a network device (200), data within the network (100-102) to a computing device external to the network ( 100-102).

类似技术:

---

公开号 | 公开日 | 专利标题

---

US10860006B2|2020-12-08|Device and system including multiple devices for supervision and control of machines in industrial installation

---

BR112019018116A2|2020-04-07|device and system including multiple devices for supervision and control of machines in industrial installation

---

Wu et al.2017|A fog computing-based framework for process monitoring and prognosis in cyber-manufacturing

---

RU2667545C1|2018-09-21|Drilling rig control system

---

Liu et al.2016|A cyber-physical system architecture in shop floor for intelligent manufacturing

---

US20180173205A1|2018-06-21|Industrial internet connected control system

---

CN110719210B|2020-08-28|Industrial equipment predictive maintenance method based on cloud edge cooperation

---

BR102013028304A2|2015-09-22|cloud based engine monitoring solution

---

CN110609512B|2021-10-15|Internet of things platform and Internet of things equipment monitoring method

---

Faltinski et al.2012|Detecting anomalous energy consumptions in distributed manufacturing systems

---

Wu et al.2016|Fog-enabled architecture for data-driven cyber-manufacturing systems

---

Isaja et al.2018|Distributed ledger architecture for automation, analytics and simulation in industrial environments

---

Jiang et al.2018|Cyber-physical system based factory monitoring and fault diagnosis framework with plant-wide performance optimization

---

Vater et al.2019|A reference architecture based on edge and cloud computing for smart manufacturing

---

Bonnard et al.2019|A big data/analytics platform for industry 4.0 implementation in SMEs

---

CN105956118B|2019-06-04|A method of realizing real time information quality stamp distribution global data base

---

Wöstmann et al.2019|A retrofit approach for predictive maintenance

---

Syfert et al.2013|DiaSter-Intelligent systems for diagnostics and automatic control support of industrial processes

---

Villalonga et al.2019|Visual analytics framework for condition monitoring in cyber-physical systems

---

Ning et al.2016|A cloud based framework of prognostics and health management for manufacturing industry

---

Możaryn et al.2020|Design and development of industrial cyber-physical system testbed

---

Stankovski et al.2020|Edge Computing for Fault Detection in Smart Systems

---

Shroff et al.2011|MIMOSA and OPC UA applied in Wapice Remote Management system

---

Rubio et al.2019|Smart gateways for IOT-Factory integration: trends and use case

---

Jeong et al.2013|Monitoring system with wireless sensor network: a survey

同族专利:

---

公开号 | 公开日

---

GB201912333D0|2019-10-09|

---

GB2573962A|2019-11-20|

---

WO2018158404A1|2018-09-07|

---

EP3590013A1|2020-01-08|

---

CA3054637A1|2018-09-07|

---

JP6983805B2|2021-12-17|

---

JP2021015614A|2021-02-12|

---

EP3370125A1|2018-09-05|

---

DE112018001085T5|2019-11-14|

---

SG11201809889PA|2018-12-28|

---

JP2020502596A|2020-01-23|

---

KR20190122125A|2019-10-29|

---

CN109313423A|2019-02-05|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US7778713B2|2007-02-27|2010-08-17|Rockwell Automation Technologies, Inc.|Construction of an industrial control system using multiple instances of industrial control engines|

---

US7797060B2|2007-02-27|2010-09-14|Rockwell Automation Technologies, Inc.|Prioritization associated with controller engine instances|

---

US8881157B2|2009-09-11|2014-11-04|Empire Technology Development Llc|Allocating threads to cores based on threads falling behind thread completion target deadline|

---

CN105531668B|2013-08-08|2019-04-23|英派尔科技开发有限公司|Method, transference apparatus and the computer-readable medium of migration for executive process|

---

US9665088B2|2014-01-31|2017-05-30|Fisher-Rosemount Systems, Inc.|Managing big data in process control systems|

---

JP6563187B2|2014-11-12|2019-08-21|株式会社東芝|Distributed control system, control device, and control method|

---

JP6540166B2|2015-03-31|2019-07-10|オムロン株式会社|Control device|ES2782299A1|2019-03-11|2020-09-11|Univ Valladolid|PARTS AND PRODUCTS COLLECTION SYSTEM FOR ASSEMBLY OR ORDER PREPARATION |

---

CN110311946A|2019-05-10|2019-10-08|国网浙江省电力有限公司宁波供电公司|Business datum security processing, the apparatus and system calculated based on cloud and mist|

---

JP2020194435A|2019-05-29|2020-12-03|オムロン株式会社|Control system, support device and support program|

---

EP3770704A1|2019-07-24|2021-01-27|Siemens Aktiengesellschaft|Cloud-based decentralized automation system|

---

CN110765288A|2019-09-04|2020-02-07|北京旷视科技有限公司|Image information synchronization method, device and system and storage medium|

---

KR102244815B1|2019-12-26|2021-04-27|한국브로치주식회사|Control Method of active broaching machine that can apply IoT-based preventive maintenance technology|

---

TWI741540B|2020-03-25|2021-10-01|王志民|Plant and mushroom cultivation monitoring system|

---

EP3929682A4|2020-06-25|2021-12-29|Volkswagen Ag|Estimating a mechanical degradation of a machine|

---

CN111835849B|2020-07-13|2021-12-07|中国联合网络通信集团有限公司|Method and device for enhancing service capability of access network|

---

CN112094959B|2020-09-22|2021-09-21|江苏坤泰机械有限公司|Full-automatic leather cutting machine|

---

KR102294851B1|2021-02-22|2021-08-27|강민석|System for measurement and control, control method of edge computing terminal unit, and computer program|

法律状态:
2021-10-19| B350| Update of information on the portal [chapter 15.35 patent gazette]|

优先权:

---

申请号 | 申请日 | 专利标题

---

EP17382107.5A|EP3370125A1|2017-03-01|2017-03-01|Device and system including multiple devices for supervision and control of machines in industrial installation|

---

PCT/EP2018/055115|WO2018158404A1|2017-03-01|2018-03-01|Device and system including multiple devices for supervision and control of machines in industrial installation|

---