METHOD OF MONITORING VEHICLE TIRES AND APPARATUS TO MONITORING VEHICLE TIRES

专利摘要:
predictive parity-based tire health monitoring apparatus and method a program method, apparatus, and product determine an health condition for tires on one or more vehicles by performing an individual self-comparison and parity-based comparison for each tire, where individual auto-comparison reduces variances between the plurality of tires on the vehicle, and parity-based comparison reduces variances within each tire. anomalies, such as severe leaks, and tire inflation, can be detected, and slow leaks can be predicted based on historical data associated with the health condition.
公开号:BR102014005772B1
申请号:R102014005772-2
申请日:2014-03-12
公开日:2021-06-01
发明作者:Peter J. Suh；Su Xu
申请人:The Goodyear Tire & Rubber Company；
IPC主号:

专利说明:

FIELD OF THE INVENTION
[001] The invention generally refers to monitoring and warning systems. BACKGROUND OF THE INVENTION
[002] It is well documented that maintaining correct tire pressure improves handling, increases mileage and extends vehicle tire life. Furthermore, while such factors are important for individual passenger vehicle owners, for commercial vehicle fleets such as tractor trailers, trucks, buses, vans, and other types of commercial vehicles, such factors can have a significant effect on profitability, both in terms of energy consumption costs and costs of retreading and/or tire replacement.
[003] Despite this irrefutable importance, tire pressure may not be monitored and maintained sufficiently frequently by many fleets, as well as a large part of the driving public in general. Furthermore, with the advent of “extended mobility tyres” (EMTs) and their increasingly widespread commercial presence, it can be difficult for the vehicle operator to detect a low pressure or leaky condition and take the appropriate action. As a result, prolonged use of a tire in a low pressure condition beyond the manufacturer's recommended limit may occur.
[004] Several legislative approaches requiring the communication of tire pressure information to the operator of a vehicle have been proposed, including a requirement that new vehicles be equipped with a tire pressure monitoring system. Conventional tire pressure monitoring systems (TPMSs) typically incorporate a sensor located on each tire on a vehicle to perform real-time monitoring of indoor air pressure and temperature. The information is wirelessly transmitted to the driver via radio frequency (RF) bands and displayed in the driver's compartment of the vehicle. The remote sensing module typically includes a temperature sensor, a pressure sensor, a signal processor, and an RF transmitter, and can be powered by a battery. Alternatively, a detection module can be “passive”, that is, power can be supplied to the detection module through magnetic coupling with a remote transmitter. The receiver can be unique for monitoring tire pressure or sharing other functions on the vehicle. For example, the receiver controller can be the existing panel controller or the bodywork controller. The receiver itself can be additionally shared with other systems using the same frequency range as a remote keyless entry system.
[005] The purpose of a tire monitoring system is to provide the driver with an alert if an anomaly occurs in one or more tires. In some cases, tire pressure and/or temperature can be reported and/or displayed, while in other cases a simple low pressure alert can be generated. To be useful, information must be communicated quickly and be reliable. However, displaying data derived from raw temperature and pressure sensor measurements is not always sufficient to accurately represent the status of a tire at any given time and under various loads and conditions. Interpretation of measured data with respect to temperature and pressure is therefore important, but so far has been problematic. Temperature and pressure readings by sensors communicating with a tire under actual wear conditions are influenced by several factors including the heat emitted by the brakes; thermal dissipation from the tire to the rim; load transfers that cause slight variations in tire volume; and heat build-up in the tire due to its hysteresis losses. Such factors can affect the accuracy of the information communicated to the driver, failing to alert the driver to marginal tire conditions under some circumstances and issuing false alarms to the driver in other cases.
[006] Opportunity is also a concern with conventional tire monitoring systems. Alerts to a driver of a low tire pressure condition can be based on simple thresholds, and when the driver is alerted due to pressure falling below a threshold, the tire has already reached a non-optimal state. Tire leaks can be slow or fast, and particularly for fast leaks, the driver alert can be too late to enable the driver to correct the situation without causing a tire failure or having to immediately stop the vehicle and change the tire or ask for assistance on the side of the road.
[007] Consequently there is a need in the art to process information in a tire pressure monitoring system in a timely and accurate manner.
[008] In addition, with respect to commercial vehicles, vehicles often have more tires, travel longer distances, and thus a greater probability of tire failure, as well as typically greater difficulty in solving tire problems while in transit. . For off-road tractor trailers, for example, the nearest service center can be tens of kilometers away and in many cases a service vehicle will need to be dispatched to provide roadside assistance. Furthermore, from a fleet perspective, coordinating the service and maintenance of multiple vehicles in the fleet increases these risks.
[009] Therefore, there is also a need in the art for a tire monitoring system capable of reducing vehicle downtime, improving fuel economy, and reducing tire costs associated with fleets and commercial vehicles used in this way. SUMMARY OF THE INVENTION
[010] The invention addresses these and other problems associated with the prior art by providing a method, apparatus and program product that monitors the integrity conditions of tires on a vehicle using a combination of individual self-comparison and comparison based on parity. Individual self-comparison reduces variances between tires on a vehicle, while parity-based comparison reduces variances within each tire. Anomalies, such as severe leaks and tire inflation, can also be detected, and slow leaks can be predicted based on historical data associated with the determined health condition.
[011] Therefore, consistent with an aspect of the invention, vehicle tires can be monitored by receiving tire data associated with a plurality of tires on a vehicle, and determining an integrity condition for each of the various tires upon performing an individual self-comparison and a parity-based comparison of tire data for each tyre. Individual self-comparison reduces variances between the plurality of tires on the vehicle, and parity-based comparison reduces variances within each tire.
[012] These and other advantages and characteristics, which characterize the invention, are presented in the attached claims and forming a part thereof. However, for a better understanding of the invention, and of the advantages and objectives obtained through its use, reference should be made to the Drawings, and the accompanying descriptive matter, in which exemplary embodiments of the invention are described. BRIEF DESCRIPTION OF THE DRAWINGS
[013] Figure 1 is a block diagram of a health monitoring system consistent with the invention.
[014] Figure 2 is a block diagram illustrating the data flow between components in the tire health monitoring system of Figure 1.
[015] Figure 3 is a block diagram of an exemplary hardware and software environment suitable for implementing the tire health monitoring system of claim 1.
[016] Figure 4 is a flowchart illustrating an exemplary sequence of steps performed by the tire integrity monitoring application referenced in Figure 3.
[017] Figure 5 is a flowchart illustrating an exemplary sequence of steps performed by the data pre-processing step referenced in Figure 4.
[018] Figure 6 is a graph of raw pressure for an exemplary set of data, and illustrating a strange, out-of-range data point.
[019] Figures, 7A and 7B, are graphs comparing raw pressure and compensated pressure for an exemplary set of data, and illustrating a data point, odd of impossible data combination.
[020] Figure 8 is a flowchart illustrating an exemplary sequence of steps performed by the integrity value generation step referenced in Figure 4.
[021] Figures, 9A and 9B, are graphs comparing individual integrity and compensated pressure values for an exemplary set of data.
[022] Figure 10 is a graph of individual health values for a tire in an exemplary set of data.
[023] Figure 11 is a graph of individual integrity values for four tires in an exemplary data set, illustrating the influence of a tire that leaks on an average value.
[024] Figure 12 is an exploded view of a graph of individual health values for four tires in an exemplary data set.
[025] Figure 13 is a graph of individual health values for four tires in an exemplary set of data, and including a graph of an average value.
[026] Figures 14A and 14B are respective graphs of individual integrity values and based on parity for a tire in an exemplary set of data.
[027] Figures 15A and 15B are respective graphs of two different anomaly detection characteristics for an exemplary set of data.
[028] Figures 16A and 16B are respective graphs of two different anomaly diagnostic characteristics for an exemplary set of data.
[029] Figure 17 is a flowchart illustrating an exemplary sequence of steps performed by a diagnostic and anomaly detection process capable of being performed by the tire integrity monitoring system of Figure 1.
[030] Figures 18A and 18B are respective graphs of two sets of exemplary data associated with two different leak events (designated respectively Event 1 and Event 2).
[031] Figures 19A and 19B are respective graphs illustrating application of an exemplary linear prediction model to the exemplary data sets of Figures 18A and 18B.
[032] Figures 20A and 20B are respective graphs illustrating application of an exemplary exponential prediction model with exponential extrapolation to the exemplary data sets of Figures 18A and 18B.
[033] Figures 21A and 21B are respective graphs illustrating application of an exemplary exponential prediction model with linear extrapolation to the exemplary data sets of Figures 18A and 18B.
[034] Figure 22 is a graph illustrating a limitation of linear and exponential prediction models with data sets having greater curvature.
[035] Figures 23A and 23B are respective graphs illustrating the application of an exemplary cubic routine prediction model to the exemplary data sets of Figures 18A and 18B.
[036] Figures 24A and 24B are respective graphs illustrating the application of a linear prediction model per exemplary piece to the exemplary data sets of Figures 18A and 18B.
[037] Figures 25A and 25B are respective graphs of two exemplary large data sets associated with the two leakage events (referred to respectively as Event 1 and Event 2) referenced in Figures 18A and 18B.
[038] Figures 26A and 26B are respective graphs illustrating the comparative performance of linear prediction models by part and cubic, exponential, linear routine using the exemplary data sets of Figures 18A and 18B.
[039] Figures 26C and 26D are respective graphs illustrating the comparative performance of linear prediction models by part and cubic routine, linear using the exemplary data sets of Figures 25A and 25B.
[040] Figure 27 is a graph of individual health values for ten tires on a vehicle in an exemplary dataset.
[041] Figure 28A is an enlarged view of a portion of the graph of Figure 27.
[042] Figure 28B is a graph of an anomaly detection characteristic corresponding to the portion of the graph of Figure 28A.
[043] Figures 29A and 29B are respective graphs of integrity values based on parity for ten tires on a vehicle in an exemplary set of data after retraining.
[044] Figure 30 is a pressure compensated graph for four tires on a vehicle in an exemplary dataset.
[045] Figure 31 is a graph of individual health values for four tires on a vehicle in an exemplary set of data.
[046] Figure 32 is a graph of health values based on parity for four tires on a vehicle in an exemplary data set. DETAILED DESCRIPTION
[047] Modalities consistent with the invention monitor the integrity conditions of tires on a vehicle using a combination of individual self-comparison and parity-based comparison.
[048] Individual self-comparison reduces the variances between tires on a vehicle, and in that regard, individual self-comparison can be considered to include various methodologies used to compare the current condition of a tire against what is considered an operating condition normal for the tyre. As will become more evident below, what is considered a normal operating condition for a tire is typically not tied to a fixed inflation level, for example, since a tire can be inflated to a pressure range, still be considered to be in normal operating condition. In one modality, for example, an individual self-comparison can compare a tire pressure with an average tire pressure calculated during training of a specific tire model for the tire, considered when the tire is supposed to be operating under normal operating conditions. .
[049] The parity-based comparison reduces the variances within each tire, and can be considered to include various methodologies used to compare the current condition of one tire with the current condition of other tires on the same vehicle, making an assumption that all tires on the vehicle are subject to approximately the same operating conditions. As such, the differences detected in a specific tire from the other tires on the vehicle, for example, as represented by an average of all tires or a subset of tires on the vehicle; they may be indicative of an insufficient integrity condition for the tyre, or other potential malfunction associated with the tyre.
[050] By combining individual self-comparison and parity-based comparison, the various tires on a vehicle can be effectively normalized to the same “normal” value during training based on current tire conditions during training, rather than any “normal”, specific predefined values. Then, during testing or monitoring, the combination of individual self-comparison and equally-based comparison can be used to isolate vehicle-specific tires that differ from other tires on the vehicle, such that despite the vehicle's current operating conditions , for example, due to ambient temperature and/or tire temperature, or if the vehicle is standing still or in operation, tires with insufficient integrity conditions can be identified.
[051] In additional modalities, anomalies, such as severe leakage and tire inflation, can also be detected, and can, for example, trigger retraining of specific tire models, for example, when inflating one or more tires is detected. Furthermore, in some modalities, a predictive algorithm can be used to predict a tire leak rate, for example, so that a time before tire re-inflation or repair is required can be calculated.
[052] Other variations of modifications will be evident to those of common knowledge in the art. HARDWARE AND SOFTWARE ENVIRONMENT
[053] Turning now to the drawings, in which like numbers denote parts throughout the various views, Figure 1 illustrates an exemplary tire health monitoring system 10 implemented as a tire health monitoring service 12 capable of monitoring a plurality of vehicles, eg tractor trailer 14 and a bus 16. It will be appreciated that service 12 may be capable of monitoring the tires of virtually any type of vehicle, including, for example, passenger vehicles, cars, trucks, vans, construction equipment, agricultural equipment, buses, etc., so the invention is not limited to the specific vehicles illustrated in Figure 1.
[054] Service 12 communicates wirelessly with vehicles 14, 16 through a network 18, for example, through a wireless carrier, which may be operated by the same entity that operates service 12, or by a completely separate entity, and may be public, private or patented in nature. Service 12 can be coupled to network 18 via wired and/or wireless communication means.
[055] Service 12 is coupled to a database 20 which is used to store the tire pressure monitoring system (TPMS) data retrieved from vehicles 14, 16, eg pressure, temperature, a vehicle identifier, a tire identifier, a wheel identifier, location data and/or a time stamp. In addition, as will be discussed in more detail below, service 12 can be accessed by various entities, including, for example, service agents 22 who are agents of service provider 12 or its authorized representatives, for example, authorized resellers and/ or service centers. Furthermore, in some modalities that monitor on behalf of vehicle fleets, fleet agents 24 can also be provided with access to service 12. Additional interfaces, eg for vehicle operators or owners, administrators, etc., can also be provided in some embodiments of the invention.
[056] Figure 2 illustrates in greater detail the components in system 10 used to retrieve, communicate, and process TPMS data. For example, on the vehicle 14, multiple TPMS sensors 28 can be installed on each vehicle tire/wheel, and configured to communicate the TPMS data to a receiver control unit (RCU) 30 disposed on the vehicle. Typical locations of these components are graphically illustrated by corresponding circles and inverted triangles in Figure 1, and it will be considered that multiple RCU 30 can be arranged at different locations on a vehicle for communication with nearby TPMS sensors 28.
[057] Each RCU 30 typically outputs the TPMS data to a panel display group 32 on vehicle 14, which can perform some processing of the TPMS data and can report such data to an operator, eg readings and pressure, readings temperature, and/or low pressure and/or temperature alerts. Cluster 32 may be a programmable computer or electronic device incorporating audio and/or visual indicators or displays, and may be integrated with other onboard electronic components. In some embodiments, for example, where no central monitoring service is used, a parity-based predictive algorithm as disclosed herein may be run locally on vehicle 14, for example, within cluster 32 or another electronic component. only one on board.
[058] In the illustrated embodiment that does not incorporate central monitoring, the vehicle 14 also includes a telematics/GPS unit 34 that communicates with the wireless carrier 18 to communicate the TPMS data to the service 12. The unit 34 can be configured to produce data location data generated by an integrated GPS receiver as well as additional data compiled by the sensors 28. It will be appreciated that the data communicated by the unit 34 may be pre-processed in some embodiments or may be raw data. Additionally, the protocol by which data is communicated to wireless carrier 18 can vary in different modalities. Additionally, in some modalities, GPS detection can be omitted. Furthermore, in some embodiments, bidirectional communication may be supported in such a way that, for example, service 12 may provide the operator of vehicle 14 with alerts or status information, and may provide a mechanism by which an operator may be placed in communication with a service agent, for example, via electronic message, voice and/or video communications to address any alert conditions or coordinate vehicle service.
[059] Wireless carrier 18 provides TPMS and other data provided by unit 34 to service 12, for example, by interfacing with an FTP server 38. Server 38 passes the arrival data to a database management system. data 40 to record the arrival data in the database 20. This data is then monitored and processed by a monitoring application 42, in the form discussed in more detail below.
[060] Turning now to Figure 3, an exemplary implementation of service hardware and software 12 is illustrated, within an apparatus 50. For the purposes of the invention, the apparatus 50 can represent virtually any type of computer, computer system. computer or other programmable electronic device, and will be referred to hereinafter as a computer for simplicity. It will be appreciated, however, that the apparatus 50 may be implemented using one or more networking computers, for example, in a cluster or other distributed computing system, or may be implemented within a single computer. or other programmable electronic device, for example, a desktop computer, a laptop computer, a handheld computer, a cell phone, a frequency converter, etc.
[061] Computer 50 typically includes a central processing unit 52 including at least one microprocessor coupled to memory 54 that may represent random access memory (RAM) devices comprising computer main storage medium 50 , as well as any additional levels of memory, eg cache memories, non-volatile or back-up memories (eg programmable or flash memories), read memories, etc. Furthermore, memory 54 can be considered to include memory storage means physically located elsewhere in computer 50, for example, any cache memory in a processor on CPU 52, as well as any storage capacity used as virtual memory, for example, as stored in a mass storage device 56 or another computer coupled to computer 50. Computer 50 also typically receives some inputs and outputs to communicate information externally. For interfacing with a user or operator, the computer 50 typically includes a user interface 58 incorporating one or more user input devices (e.g., a keyboard, a mouse, a track-ball, a joystick, a responsive element. touch, and/or a microphone, among others) and a display (for example, a CRT monitor, an LCD display panel, and/or a speaker among others). Otherwise, user input can be received via another computer or terminal.
[062] For additional storage, computer 50 may also include one or more mass storage devices 56, for example, a floppy disk or other removable disk drive, a hard disk drive, a direct access storage device (DASD) ), an optical drive (eg, a CD drive, a DVD drive, etc.), and/or a tape drive, among others. In addition, computer 50 may include an interface 60 with one or more networks 62 (e.g., a LAN, a WAN, a wireless network, and/or the Internet, among others) to allow communication of information with other computers and electronic devices, eg one or more client computers 64 (eg to interface with agents 22, 24) and one or more servers 66 (eg implementing other aspects of service 12). It should be appreciated that computer 50 typically includes suitable analog and/or digital interfaces between the CPU 52 and each of components 54, 56, 58 and 60 as is well known in the art. Other hardware environments are considered within the context of the invention.
[063] Computer 50 operates under the control of an operating system 68 and runs or otherwise relies on various computer software applications, components, programs, objects, modules, data structures, etc., for example , a call center application 70 (within which, for example, the monitoring application 42 can be implemented). In addition, several applications, components, programs, objects, modules, etc. they can also run on one or more processors in another computer coupled to computer 50 via network 62, e.g., in a distributed computing or server-client environment, whereby the processing required to implement the functions of a computer program can be allocated to multiple computers over a network.
[064] In general, the routines executed to implement the modalities of the invention, whether implemented as part of an operating system or a specific application, component, program, object, module or sequence of instructions, or even a subset of the themselves, will be referred to here as “computer program code” or simply “program code”. Program code typically comprises one or more instructions that are resident at various times in various storage and memory devices in a computer, and which, when read and executed by one or more processors in a computer, cause the computer to perform the steps necessary to carry out steps or elements embodying the various aspects of the invention. Furthermore, although the invention has been described and will be described below in the context of fully functioning computers and computer systems, those skilled in the art will appreciate that the various embodiments of the invention can be delivered as a program product in a variety of ways, and that the invention applies equally regardless of the specific type of computer-readable media used to actually carry out the distribution.
[065] Such computer readable media may include computer readable storage media and communication media. Computer readable storage media are non-transient in nature and may include non-removable and non-removable volatile and non-removable media implemented in any method or technology for storing information, such as computer readable instructions, data structures, modules program or other data. Computer readable storage media may further include RAM, ROM, erasable programmable read memory (EPROM), electrically erasable programmable read memory (EEPROM), flash memory or other solid state memory technology, CD-ROM, digital disks versatile (DVD), or other optical storage media, magnetic cassettes, magnetic tape, magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to store the desired information and which can be accessed by computer 50. The media may incorporate computer readable instructions, data structures or other program modules. By way of example, and not limitation, communication media may include wired media such as a wired network or wired-direct connection, and wireless media such as acoustic, RF, infrared and other wireless media. Combinations of any of the above may also be included in the scope of computer readable media.
[066] Several program codes described below can be identified based on the application in which it is implemented in a specific modality of the invention. However, it should be noted that any specific program nomenclature below is used for convenience only, and thus the invention should not be limited to the use only of any specific application identified and/or inferred by such nomenclature. Furthermore, given the number of typically endless ways in which computer programs can be organized into routines, procedures, methods, modules, objects, and the like, as well as the various ways in which program functionality can be allocated between various layers of software that are resident within a typical computer (eg operating systems, libraries, APIs, applications, small programs, etc.), it should be considered that the invention is not limited to the specific organization and allocation of functionality of program described here.
[067] Those skilled in the art will recognize that the exemplary embodiment illustrated in Figures 1-3 is not intended to limit the present invention. Indeed, those skilled in the art will recognize that other alternative hardware and/or software environments can be used without departing from the scope of the invention. TIRE INTEGRITY MONITORING BASED ON PREDICTIVE PARITY
[068] Embodiments consistent with the invention utilize predictive parity based tire health monitoring to monitor the health of tires on a specific vehicle. Such an implementation of tire health monitoring based on predictive parity; specifically in applying a centralized, tire health monitoring service suitable for monitoring the tire health of one or more fleets of commercial vehicles, eg trucks; is illustrated by process 100 in Figure 4.
[069] Process 100 includes six main steps: initialization 102, data acquisition 104, data pre-processing 106, integrity assessment 108, anomaly detection and diagnosis 110, and leak prediction 112, each of which will be discussed in more detail below. Steps 104-112, and in some cases, all or a portion of the operations performed in step 102, are typically implemented on a computer, for example, via software and/or hardware such as within the monitoring application 42 of Figures 2-3. In some embodiments, all or part of the functionality in these steps may be implemented in an on-board computer or programmable electronic device disposed in the vehicle, or in other ways that will be considered by those of ordinary skill in the art having the benefit of this revelation.
[070] Initialization step 102 is performed when the tire integrity monitoring system is first applied to the tires to ensure that all tires on the vehicle are not leaking. As will become more evident below, the system typically needs to be trained through data that is collected when tires are in good condition, and as such, any leak in a tire would violate this condition and lead to insufficient model parameters. and thus insufficient model accuracy. As such, at step 102, a tire check is performed at block 114 to ensure that all tires are in good condition, and based on whether all tires are good, block 116 or proceeds to block 118 to initiate a tire maintenance operation to replace or repair any defective tires and retry the initial tire check, or to step 104 to begin monitoring. It will be appreciated that tire checks and/or tire maintenance operations can be performed manually and/or automatically in various embodiments of the invention. For example, in one modality, a tire check can be performed by acquiring TPMS data for the tires over a period of time when the tire is known to be idle to test for any leaks.
[071] When the system is initialized and all tires for a vehicle have been verified to be good, monitoring can be established for the vehicle, and in such a way that TPMS data can be periodically sent to the service tire integrity monitoring, either only when the vehicle is operational, or alternatively at all times and regardless of the vehicle's operational status. Box 120 of Figure 1 illustrates the operations performed in response to each new TPMS data packet received by the tire monitoring service.
[072] In step 104, for example, a TPMS data packet, collected and transmitted by the onboard components discussed above in connection with Figure 2, can be acquired by the tire integrity monitoring service. Each package can include, for example, the current temperature and pressure of all tires from a vehicle, together with additional data such as a data compilation time stamp and GPS location. For a road tractor trailer, for example, a package might include pressure and temperature data for all 18 tires.
[073] Next, in step 106, the collected data are pre-processed. It was found that the collected data can sometimes include unexpected extraneous elements that are not in the normal range of values, or do not match the context values. These extraneous elements could potentially taint the accuracy of the last mentioned model if not removed, and as such, step 106 can preprocess all incoming data to ensure the data is clean. Then, further pre-processing in the form of temperature compensation, for example by means of a physical model, can be carried out. In one embodiment, for example, pre-processing of collected data can be used to generate a pressure compensated variable, representative of the tire pressure compensated for temperature effects.
[074] Next, as illustrated by blocks 108 and 110, the integrity assessment and the anomaly detection and diagnosis operations can be performed using the collected and pre-processed data. In block 108, each integrity assessment iteration initially checks a system mode to determine whether a model training process has been performed (block 122). Specifically, embodiments consistent with the invention utilize a specific vehicle tire health assessment model to monitor the tires of a specific vehicle, and the model typically must be trained prior to operational use. Thus, the model can switch between training and operational modes.
[075] Thus, if a training mode is detected in block 122, the control passes to block 124 to train a tire integrity assessment model 126 by updating the model parameters generated from historical data accordingly with the most recent data received. If an operational or test mode is detected, block 122 instead passes control to a health value generation block 128 to generate a health value (HV) for each tire as the output from step Integrity Assessment Score 108. Various operations, eg individual tire self-comparison, parity based comparison, etc. are performed at block 128 to generate a more accurate integrity value. These operations are discussed in more detail below.
[076] The anomaly detection and diagnosis step 110 receives the integrity values from the integrity assessment step 108 and determines in block 130 whether any anomalies have been detected. In the illustrated embodiment, anomalies to the system such as tire inflation are designed to be self-detectable. If no such anomaly is detected, control passes to block 132 to determine if all integrity values are “normal”, that is, within a range considered to represent a sufficiently pressurized and leak-free tire. Otherwise, control passes to block 134 to indicate a potential slow leak condition. Otherwise, if the integrity values are normal, control returns to step 104 to process any additional TPMS data packets waiting for processing.
[077] Returning to block 130, if a system anomaly is detected, the anomaly will be classified into several groups following different separation rules. Block 136, for example, determines whether an anomaly is caused by filling rather than severe leakage. If this is the case, control passes to block 138 to select the training mode so that the relevant integrity assessment model will again be trained based on the data received after filling. If instead a serious leak is the cause of an anomaly, control passes to block 140 to generate an urgent event to notify a service provider agent, vehicle operator, fleet agent, and/or other interested party .
[078] Returning to block 134, if a slow leak is detected in step 110, control passes to step 112 to initiate a prediction operation to predict based on historical data in the model the possible time the tire could operate until reach a predetermined threshold at which the tire is no longer suitable for operational use. Further details regarding the prediction operation will be discussed below.
[079] In addition, as illustrated in blocks 142, 144, a recovery check can be performed at each iteration in block 142 to determine if any tires that were previously determined to be leaking have been repaired from the trend data. If this is the case, control passes to block 144 to set the training mode for the tire. Otherwise, block 144 is ignored.
[080] Before further elaborating the aforementioned steps, some assumptions can be considered as a basis for the performance of tire integrity monitoring based on predictive parity consistent with the invention. First, it is assumed that during tire operation, the volume of a tire cavity does not change significantly, for example, the change is typically insignificant. Thus, the cavity volume can be considered as a constant number.
[081] Second, an assumption can be made that the data behaves similarly at any time as long as the data is considered from the same tire that is in good condition. Data-oriented statistical models are normally trained by the data considered from the system when its integrity condition is good, and future data is tested to determine if the new data is subjected to the same distribution so that if it is not, new data will be considered to represent poor system integrity. Therefore, if this assumption cannot be satisfied, then some of the test data is subjected to another distribution, and the model will automatically be treated as if it were an indication of negative integrity.
[082] Thirdly, after tire maintenances, such as inflating or fixing, it can be assumed that tire pressures are maintained in the normal range. It can also be assumed that prior to the first system start, none of the tires had any abnormal leaks (a condition that is tested in step 102 of Figure 4). In some modalities, such scenarios can also be detected simply by establishing thresholds in the pressure data.
[083] Fourth, an assumption can be made that all tires of a vehicle such as a tractor trailer share a similar operating condition in terms of road surface inequality, ambient temperature, and air pressure environment, normal rate of slow leakage, etc. This assumption facilitates the application of parity-based health condition comparison as will be discussed in more detail below.
[084] Fifth, it can be assumed that during normal operation, more than half of the tires from a vehicle are in good condition, ie, less than half of the tires will leak at any given time. This assumption makes it easy to select the most normal value for all tires to be the nominal value to which all other values can be compared.
[085] Sixthly, an assumption can be made that the possibility of two tires from a vehicle experiencing severe leakage simultaneously is very low.
[086] Additional details regarding the number of steps illustrated in Figure 4, and using the assumptions outlined above, are provided below. DATA PROCESSING
[087] Turning now to Figure 5, this figure illustrates the steps that can be performed in data processing step 106 in greater detail. Data acquisition methods were found to be often imperfect, resulting in problems such as out-of-range values and impossible data combinations, and data analysis without a proper data cleansing process that could cause misleading results. Modalities consistent with the invention, however, can preprocess the arrival data by firstly, in block 150, removing out-of-range foreign elements, secondly, in block 152, compensating the pressure data for temperature effects and, thirdly, in block 154, the removal of foreign elements of impossible combination.
[088] From the perspective of foreign elements, it has been found that there are two main types of foreign elements that can be observed from the raw TPMS data. The first is an out-of-range value, which can be thought of as a data point that contains a value that is outside the normal range. In one modality, for example, differences between two consecutive data points can be calculated, and if the difference value is beyond a predetermined threshold, then that data point can be removed. Figure 6, for example, illustrates at 155 a pressure reading of up to 160 psi that is far above what would be considered a normal range (for example, for a typical truck tire, the normal range could be 90- 130 psi), and which would be suitable for removal from the dataset.
[089] The second type of foreign element is defined as an “impossible data match” foreign element, which can be thought of as embedding where all variables are in their own normal ranges, respectively, but as a packet of data or a combination of all variables, the data value is unreasonable. It has been found, for example, that extraneous pressure and temperature values can be present in raw data but still fall within the normal ranges of pressure and temperature, so in some modalities a combination of these two variables can be used to assess the values together to detect extraneous elements of impossible combination.
[090] In the illustrated modalities, specifically, a compensated pressure is calculated to standardize the pressure measurements for a constant temperature.
[091] Traditionally, it has been recommended to take pressure readings for the tires when the tires are “cold”, ie not having been used for some time. If a tire has been running for even several minutes, the tire's internal temperature typically increases to a point where a measurable effect on pressure is experienced. However, TPMS sensors typically take measurements after the tires start to spin, so almost all pressure data collected from a TPMS sensor typically includes some temperature effect. Additionally, when temperature and pressure are plotted together, a high linear correlation is usually observed, for example, a correlation coefficient of approximately 0.96.
[092] The ideal gas law PV=nRT can be used to compensate pressure for temperature. A tire is assumed to be at a fixed, confined volume, with any change in tire size due to deflation considered to be negligible, so the regulation for ideal gas can be rewritten as:

[093] with all values on the right side of the equation being fixed, in which case the pressure is shown as being proportional to the temperature. Thus, by introducing two new values, C, compensated pressure, and nominal temperatureTO, the following relationship can be established:

[094] By selecting a predetermined fixed value such as 25 degrees Celsius for nominal temperature To, for each pressure and temperature pair collected from the TPMS system, a single corresponding compensated pressure can be generated corresponding to the temperature in 25 degrees Celsius. By making this transformation, ideally the temperature effect would be completely eliminated from the compensated pressure and the compensated pressure would be a flat line if a tire had absolutely no leakage. It is believed, however, that the compensated pressure can still fluctuate due to environmental factors such as where the TPMS sensors are mounted on the rims. The measured temperature is performed as cavity temperature whereas it is actually cavity close to wheel temperature, and a variance may exist between the actual cavity temperature and cavity close to wheel temperature. Furthermore, this variance can also be affected by ambient temperature. As such, some variance can still be expected in a pressure compensated value.
[095] It has been found that after temperature compensation, foreign elements of impossible combination can be identified more readily. Figures 7A and 7B, for example, respectively illustrate pressure and combined pressure values for a tire. As shown at 157 in Figure 7A, an extraneous pressure reading may not fall outside the normal readings, but as shown at 159 in Figure 7B, when that pressure value is compensated for the temperature, the resulting compensated pressure value it is more readily identified as a foreign element, and so can be discarded. INTEGRITY ASSESSMENT
[096] Returning now to Figure 4 and integrity assessment step 108, in the illustrated embodiment it may be desirable to generate a single value or set of values indicative of tire integrity to estimate the overall health of a vehicle. Since vehicles will typically maintain a regular health condition most of the time, it may be desirable to minimize the frequency with which failure diagnosis and prediction steps, which would otherwise incur substantial computational costs, are performed. Thus, by performing integrity assessment as a separate and preliminary step to failure diagnosis and prediction, such failure diagnosis and prediction could be delayed until an initial indication of a potential failure is detected. In other modalities, however, diagnosis and/or failure prediction can be performed more frequently.
[097] The 126 health assessment model, as noted above, operates in both training and testing modes. Training typically uses data from good integrity conditions so that a baseline of the system is described. Then, during testing, new data is compared to the trained baseline. If the test samples are similar to the baseline then a good system health condition can be completed, otherwise the system condition is not acceptable and some maintenance actions may be triggered including fault diagnosis and prediction.
[098] In the illustrated modality, the health assessment incorporates individual tire self-comparison and parity-based comparison. As shown in Figure 8, for example, the health value generation step 128 of Figure 4 may include a block 160 to generate an individual health value (IHV) for each tire on a vehicle using individual tire auto-comparison . Next, a parity-based comparison is performed in blocks 162-168. Depending on the algorithm used, an average of all a tire subset is calculated using one of blocks 162, 164 and 166. Then, at block 168, an equally based integrity value (PHV) is generated for each tire on the vehicle from from the average calculated on one of blocks 162, 164 or 166, typically by subtracting the average calculated from the IHV calculated for each tire.
[099] As will become more evident below, individual tire auto-comparison can be used to minimize or eliminate drift between different tires on a vehicle while equally-based comparison can be used to smooth out fluctuations within data from a tire. Additionally, it will be considered that although three blocks 162-166 are illustrated in Figure 8, in many embodiments, a parity-based comparison algorithm will use only one of the blocks when determining a PHV. INTEGRITY ASSESSMENT - INDIVIDUAL TIRE SELF COMPARISON
[0100] With respect to an individual tire self-comparison, if the tire maintenance is compared, for example, with a traditional machine tool maintenance process, the initial conditions for the two processes are similar, that is, a tire is properly inflated and tool is newly changed. However, for a machine tool, it starts to degrade from the first day it is used, whereas a tire will still be considered to be in good health condition as long as it maintains the same pressure level. The tire integrity condition starts to degrade only after the tire starts to leak. When a tire pressure reaches a certain level, which is analogous to that of a machine tool reaching a certain level indicative of excessive wear, component maintenance must be triggered. For a tire, a leak can be repaired while for machine tools, the tool would need to be replaced.
[0101] However, there is a significant difference between machine tool maintenance and maintenance. For tools that are newly installed, all typically have the same integrity conditions, as all tools must follow the same specification, whereas after maintenance the tires are typically inflated similarly, but not necessarily in the same pressure levels. Although tires can be inflated to different levels, they can still be considered to be in the same “good” integrity condition, ie they are not leaking. As such, in the illustrated embodiment, discrepancies in inflation levels between different instances of tire on a vehicle are conveniently compensated for to ensure that all tires are subjected to the same initial integrity condition after maintenance. Otherwise, such discrepancies can mask the actual tire integrity conditions and prevent a decay condition from being prematurely detected.
[0102] As such, in the illustrated modality, individual tire self-comparison is used to compare the current performance of a specific tire with its "normal" performance, that is, the condition in which the tire model is trained to reduce or eliminate discrepancies between individual tires. In one modality, for example, distance-based assessment (DBA) can be used to centralize the pressure compensated from all tires on a vehicle to the same level.
[0103] Using a DBA model, the first several data samples from each tire are used to “train” the model baseline for each tire. The data center, which is the average value of the data samples from each tire, is calculated to cancel any data acquisition errors from the TPMS system and establish the baseline of normal behavior. Subsequent data samples are then compared to this mean value to generate a “distance” or difference between the data sample and the mean value. This distance, which is referred to here as an individual integrity value (IHV), is then used to determine variations in individual tires. In one sport, for example, an average pressure value for each tire can be determined during training, and an IHV can be calculated for each tire based on the difference between the compensated pressure and the average value.
[0104] Figures 9A-9B, for example, illustrate the performance of DBA on an exemplary set of data points for four tires on a vehicle. The graph in Figure 9A illustrates the calculated offset pressures of four tires, while the graph in Figure 9B illustrates the IHV of the same four tires as generated from the DBA model. In this example, the Fisher Criterion value for the IHV is 0.1374, which is substantially greater than the corresponding value of 0.007 for the compensated pressure alone. Fisher's Criterion is an informatics test that assesses the ability to separate multiple variables when there are more than two classes in a data set. For a two-class dataset, the Fisher Criterion is given by:

[0105] here mi and m2 are respectively means of two classes, while si and S2 are standard deviations of two groups. The numerator represents the squared distance between two groups while the denominator represents the sum of variance within each group. Therefore, a higher Fisher Criterion value indicates a higher separation capability.
[0106] It will be considered that other individual self-comparison models and techniques may be used in other embodiments of the invention. For example, two self-comparison techniques, alternatives that can be used include logistic regression and self-organizing maps. As with DBA, to start testing as prematurely as possible to detect possible leaks, a basic training of at least a minimum of two samples may be desired, after which model parameters will continue to be updated as more data samples are received. In some modalities, testing and training can be carried out in parallel, and each time after receiving a new data package, the models can be updated until enough data samples have been collected to properly train the models, in which training may be discontinued while testing continues. The number of training samples can be chosen, for example, by leveraging the training data size and test performance so that the data samples in the training step can cover a normal performance range that is wide enough from all tires to form a baseline of overall integrity. A model is typically built from each tire, ie for a truck with 18 tires, 18 models can be built.
[0107] Logistic regression (LR), sometimes also known as logistic model or logit model, is a binomial regression, which is commonly used to predict the probability of a discrete event occurring by mapping multidimensional data to a value between 0 and 1. The data used by the model can be numerical and categorical, such as gender, presence, year and so on. For the purpose of tire health monitoring, it can be used to quantitatively assess current system health conditions as a supervised method.
[0108] LR is based on the logistic function:
where z is defined as:
and where is: is the intercept, pi, pz, ... are regression coefficients, whereas xi, X2, ... are independent variables.
[0109] From these equations, a significant advantage of LR can be seen that although a linear combination of all independent variables can be any value from negative infinity to positive infinity, the result is a limited value between 0 and 1.
[0110] When rewriting the aforementioned equations, the following relationships are then observed:

[0111] In practice, LR is typically used as a supervised statistical model, which means that during the training step, data from both acceptable and unacceptable conditions are required. A group of sample inputs {x1, x2, ...} can be observed from both system conditions, and the corresponding probabilities {p(z)} can be specified according to model needs (for example , 0.95 can be used for acceptable conditions and 0.05 for unacceptable conditions). Then, the model parameters βo, βi, β2,... can be obtained for maximum probability estimation (MLE), which exploits the best combination of parameters that maximizes the probability of the observed data through an iterative circuit.
[0112] After the model parameters are all trained, the test data can be fitted into the model to calculate current Individual System Health (IHV) values. To meet the training requirements of the LR model, LR models can be trained for data from both acceptable and unacceptable conditions. For tire inflation, 0 psi is the lowest pressure value the tire could obtain, and could be considered as data from bad condition. Thus, some pressure compensated samples can be used to represent an acceptable condition while some other values of 0 psi can be used for an unacceptable condition. Two corresponding HVs, 0.95 and 0.05, can be given respectively to each integrity condition. After training, all IHVs produced from LR will be limited to values between 0 and 1.
[0113] The Self-Organization Map (SOM) is a type of neural network model that is typically regarded as an unsupervised learning model because it is trained only by the topological structure of the data, and no information from data classification is required. SOM preserves the topological structure of the data by introducing a neighborhood function and is typically a suitable data visualization tool as it typically reduces the dimensionality of the data to a small number, usually two.
[0114] SOM consists of nodes of neurons, each of which is correlated with a weight vector. At first all nodes are arranged randomly on the map, then in an iterative training step they are updated based on learning rules such as:
[0115] When a sample vector is given as an input to the map, the distance between that sample vector and all nodes will be calculated, and the node with the shortest distance will be selected and considered as the Best Matching Unit ( BMU).
[0116] BMU neighbor nodes, which are predetermined by topological formats, are all updated according to the new sample vector, and the learning rate is respective to the distance between BMU and the sample value, which is monotonically decreasing.
[0117] At the end of the learning step, nodes are grouped into groups, and each group represents a class of data type.
[0118] Mean Quantization Error - Self-Organization Map (SOM-MQE) is an extended SOM model that can also be used for health variance. As the tire health range it is relatively easy to compile data during a health condition, whereas in contrast failure data is typically more precious and difficult to compile, as a semi-supervised learning method SOM-MQE has. advantages in terms of model training convenience. It trains the model based only on the healthy data and calculates the distance between the test samples and the trained baseline during testing to assess how far the test data is from the baseline. The further the test sample is from the baseline, the worse integrity is indicated for the current system.
[0119] The map is first trained with normal operating data, and then the MQE, which is defined as the distance between the test sample and its BMU, is calculated. As the BMU identification process is a search process that aims to find a trained node that has the minimum Euclidean distance for a test sample, even a sample that belongs to a space not covered by the training data can find a BMU and calculate its MQE, which is considered to be the health indicator of the SOM-MQE model.
[0120] Similar to the LR model, the SOM-MQE model can be trained by the first samples, which are supposed to be collected from an acceptable condition. The regular output of the SOM-MQE model is the MQE distance, which is a positive number. However, to make the result more intuitive, it may be desirable to nullify all MQE distances to better represent leaks.
[0121] Additional models and methodologies can also be used to generate the IHVs, for example, statistical pattern recognition, Gaussian blend models, neural networks, etc. Therefore, the invention is not limited to the specific models and methodologies disclosed herein. INTEGRITY ASSESSMENT - PARITY BASED COMPARISON
[0122] Although the aforementioned processing steps are found in many embodiments to minimize or reduce many of the variations between tires on a vehicle, it is desirable in many embodiments to additionally use a parity-based comparison to address the fluctuations that can occur inside each tire.
[0123] Again, with comparative reference to machine tool maintenance, for traditional machine tool monitoring, features that are used to better represent the original dataset and reduce the dataset size are typically extracted based on strokes (a duration in which a machine tool completes all of its repeatable movements). As the machine tool repeats the same series of movements over and over, assuming it has only one regime, the characteristics must remain constant if the machine tool is kept in good condition. On the other hand, tires do not have a fixed operating plan, for example a vehicle can be driven on different routes every day. Furthermore, even when a vehicle operates on the same route every day, environmental conditions, such as the ambient temperature, change from time to time. Therefore, it is much more difficult to define a fixed duty cycle to use as a basis for analysis. In other words, tire operation has been found to be a dynamic process that is greatly affected by environmental uncertainties.
[0124] As a result of these environmental uncertainties, even after temperature compensation, a theoretical flat line of pressure compensated that one would expect to see in a leak-free tire is not seen in practice. Instead, a flat trend with many fluctuations is typically observed. Fluctuations can be caused by many aspects such as ambient temperature variances, tire load, speed, roughness of road surfaces, ambient atmospheric pressure, etc. Furthermore, through individual tire self-comparison, the fluctuation effect is passed from the compensated pressure to the IHV, for example, as shown in Figure 10.
[0125] From a system monitoring standpoint, fluctuations reduce the accuracy of both imperfect detection and prediction. For example, for imperfect detection, since the trend can have a relatively wide variation, the threshold adjustment is faced with a great balance between sensitivity and false alarm.
[0126] It is known that valleys in HIV data are generally formed when a vehicle is parked when tire temperatures are low while high peaks are mainly caused by temperature developments. It can also be seen that even after temperature compensation, the compensated pressure still has a certain degree of impact on temperature, for example, due to ambient temperature. Although ambient temperature can form the basis for further modeling and/or compensation, in many modes it is desirable to avoid ambient temperature monitoring, as ambient temperature is typically not provided by many conventional TPMS systems.
[0127] Although ambient temperature can be used in some modalities, in the illustrated modality a comparison based on data-oriented parity is used to reduce the effects of fluctuations due to ambient temperature and other environmental effects.
[0128] The parity-based comparison in the illustrated mode is based on the aforementioned assumption that all tires on a given vehicle share a similar operating condition including environmental conditions, road conditions, and normal slow tire leaks. As all tires on a vehicle share a similar overall performance, in some sports an average value of IHVs from all tires can be calculated such that the differences between the individual tires and the average can be indicative of conditions of leakage, as illustrated by block 162 of Figure 8.
[0129] It has been found, however, that although the average of all tires may be sufficient when all tires are in integrity condition, when leaking tires are involved, especially when leakage is severe, the average value may be biased. for that condition of the tire that leaks and may not be sufficiently indicative of an overall good tire condition. Figure 11, for example, illustrates an exemplary dataset for a vehicle where the average value of the first half of the data (before data point 648) is heavily skewed towards tire 1, which has a leak. On the other hand, tire 1 is set at point 648 and the IHVs from four tires are centered back on the same level and more closely track the average value.
[0130] In other embodiments, an additional rule can be integrated to resolve the potential tendency of a tire to leak by attempting to calculate the average value of the tires determined to not have a leak. Thus, if during operation any tire is detected with a potential leak, the tire can be labeled such that when the average value is calculated, the data from that leaking tire is not included, and the average is based on only on a subset of tires, as illustrated in block 164 of Figure 8. Application of the rule has been found to at least partially address the trends that are attributed to the leaking tire. It was also found, however, that the bias problem may still be present in some instances because at the onset of a leak, while the pressure drop may initially be small, the pressure drop may have an impact on the mean value. . For example, in the same case illustrated in Figure 11, an average value could be calculated based on all four tires until the leak was detected at data point 260. As shown in Figure 12, for example, it can be seen that the average value represents the overall performance of all good tires well after leak detection. However, prior to detection, tire 1 already had the leak, and a small trend of the mean value can be better seen from the exploded graph.
[0131] Therefore, in still other modalities, it may be desirable to integrate an additional rule to further address the potential tendency of a tire to leak. Specifically, another assumption that can be made, as noted above, is that during regular operation, more than half of a vehicle's tires are in good condition. Put another way, at least six out of ten tires on a tractor are supposed to be not leaking at any given time, and such an assumption can be made that more than half of the tires that do not have a detectable leak do form a vehicle in good condition. For example, a tractor has ten tires, and two of them have detectable leaks. Thus, the additional rule assumes that at least five out of eight tires that have no detectable leakage are in good condition.
[0132] If the statement is well accepted, an average value of all or a subset of tires on the vehicle can be used to serve as an indication of overall good tire performance, as illustrated by block 166 of Figure 8. In addition, an average value calculation may only require at least one or two values that are right in the middle of all values, which refers to the value of good tires according to the aforementioned assumption of more than half of a vehicle's tires being in good conditions. Figure 13, for example, illustrates the IHVs of four tires and an average value calculated from them to represent the overall healthy tire performance of all good tires under a dynamic operating condition and even with some tires that leak. It should be noted, although in some modalities it may be desirable to exclude known or predicted leaking tires from the averaging, in many modalities the inclusion of these leaky tires in the averaging will not significantly alter the resulting average, so that it may not be desirable to try to exclude leaking tires from the calculation.
[0133] Thus, in some embodiments, the overall healthy tire performance based on the average calculated from a subset of tires can be subtracted from all individual tire performances, ie, IHVs to reduce fluctuation caused by dynamic operating conditions. This value obtained after performing the parity-based comparison is referred to here as the Parity-based Integrity Value (PHV), and as illustrated by Figures 14A-14B, it can be seen that the PHV value calculated in the form described here (Figure 14B) has a comparatively smaller variance than the corresponding IHV value for the same tire (Figure 14A), and also that environmental factors including temperature effects have substantially less impact on the data.
[0134] It will be appreciated that averaging using a subset of tires from a vehicle may be implemented in several ways consistent with the invention, and may include various numbers and combinations of tires from a vehicle. For example, an average can be taken from each axle, from each tire type (eg steering/drive/trailer), or from the steering, internal drive, drive tires. external, internal towing and/or external towing. Alternatively, an average can be taken from the high and low, or the second high and the second low from any of the aforementioned tire combinations. The invention, however, is not limited to the specific determinations disclosed herein. ANOMALY DETECTION AND DIAGNOSIS
[0135] Again, returning to Figure 4, and specifically to the anomaly detection and diagnosis step 110, the PHV calculated for each tire on a vehicle is analyzed to detect failures and perform diagnostics in response to the detected failures.
[0136] It can be assumed that all tires have normal leaks, and are inflated regularly to keep their pressures in the normal range. Furthermore, there are two possible results that can be found from the data after each inflation: 1. the pressure differences between the tires remained the same; 2. the pressure differences between the tires have changed. In most cases, the pressure differences change after fillings. However, the tire integrity assessment models for the vehicle discussed above are all based on these pressure differences, and as such, after tire inflation, the tire integrity assessment models typically need to be retrained to determine the new pressure differences between the tyres. Although in some modalities, a manual trigger of the new model training may be supported, in other modalities it may not be possible or desirable to support the new manual training, and as such, it is usually desirable to provide an algorithm for detecting data-driven anomaly to automatically identify tire inflations and thereby trigger retraining of tire integrity assessment models.
[0137] Anomaly detection conveniently provides satisfactory accuracy providing a relatively low incidence of false alarms while maintaining relatively good sensitivity. Furthermore, anomalies must be conveniently detected in a relatively short time after their occurrence, as the opportunity is typically highly valued by real-time systems.
[0138] To address these normally competing issues, different characteristics can be extracted from the processed and/or raw data, including, for example: mean, mean number of a series, difference, standard deviation, maximum, minimum , etc., both within a single tire and between the tires. “Inside a single tyre” means that characteristics are extracted from a tire over a period of time (or alternatively, using a moving window technique). “Between tires”, on the other hand, means that characteristics are extracted from data collected from multiple tires at the same time stamp. It was found that although many tested features can indicate filling events, many require a relatively long delay after fillings occur to collect enough data to extract suitable features.
[0139] In some embodiments, two different types of features can be used: the first, referred to here as an anomaly detection feature, can be used to locate abnormal points; and a second, referred to herein as an anomaly diagnostic feature, can be used to diagnose what causes an abnormal spot.
[0140] Figures 15A and 15B, for example, illustrate two candidate characteristics that can be used as an anomaly detection characteristic, PHV standard deviation between healthy tires (Figure 15A), and the mean of temporal differences (Figure 15B). As an example, to average temporal differences, the difference between two consecutive measurements from the same tire position for all tires can be calculated, and the average of all differences from healthy tires can be calculated. So, for example, if there are 10 healthy tires, the readings from the first data packet would be [100, 100, ... ... 100, 105] and the readings from the second data packet would be [100, 100, 100, 100]; then the time difference would be [0.0, ... 0,-5] leading to an average of -0.5).
[0141] From the two figures, two clusters are illustrated, with the initial cluster changing to form a new cluster when the anomaly occurs (illustrated by lines 170, 172). The anomaly detection feature is used to detect system level, ie vehicle level anomalies, and as such information from all tires on the vehicle is considered for each feature. It will be assumed that Figures 15A-15B provide a view of the respective characteristics over a long period, however, in a real-time system, the characteristic values are calculated one by one, and as such, when an anomaly occurs, only one point would be observed in the second grouping. As such, in a real-time system, it may be desirable to provide a delay of one or more extraordinary data packets to ensure that a cluster change is not caused by a single spike as a foreign element, as it would be. the case with the data value represented at 174 in Figure 15B.
[0142] After the trouble point is detected, it is desirable to provide a diagnostic process to try to understand the reason behind the trouble point. Such a diagnostic process utilizes one or more fault diagnostic features, which in contrast to a fault detection feature are intended for use in understanding individual tire performances when a fault is detected, and such features may accordingly be considered. from individual tyres. For example, Figures 16A-16B respectively illustrate the use of two different characteristics, the difference between IHV and its minimum (Figure 16A) and standard deviation (Figure 16B) for each of ten tires.
[0143] Figure 17 illustrates an anomaly detection and diagnosis process 200, exemplary consistent with the invention. Process 200 can be performed for each new set of data points, after the PHV and IHV values have been calculated for each tire. Process 200 starts at block 202 by calculating one or more fault detection characteristics, which can be based, for example, on the IHVs from the tires in good condition. Anomaly detection characteristics can include, for example, standard deviation of PHV, IHV, or compensated pressure, mean of time differences, etc.
[0144] Next, in block 204, a dynamic or moving threshold can be calculated based on the historical trends of the characteristics in question. In one modality, for example, a 5-sigma distance from an average can be considered as a threshold for deciding whether a point is an anomaly. In alternative modalities, a fixed threshold can be used.
[0145] Next, blocks 206 and 208 determine respectively whether the characteristics exceed the calculated thresholds, and if so, whether the thresholds are exceeded for a predetermined number of consecutive sample periods. In one modality, for example, if two consecutive anomaly detection characteristic values have exceeded the moving thresholds, anomaly can be detected. In other modalities, only a single threshold may need to be exceeded to detect an anomaly, while in still other modalities, more than two thresholds may need to be exceeded.
[0146] If the condition of block 206 or block 208 is not satisfied, no anomaly has been detected, and process 200 is complete. Otherwise, control passes to block 210 to identify the first point that is out of threshold as the anomaly point.
[0147] Block 212 then initiates anomaly diagnostics to attempt to determine whether anomaly is due to a leak event or a filling event. In the illustrated mode, four possible events can be indicated: (1) a severe leak occurs in a tire; (2) a single tire was inflated; (3) two tires were inflated; and (4) more than two tires were inflated.
[0148] Block 212 specifically recalculates one or more fault detection characteristics, but excluding the tire having the minimum IHV from the calculation. For example, if out of ten tires, tires 1 to 9 are considered good, and at the time when a malfunction occurs, tire 1 has the minimum IHV between tires 1 to 9, then the malfunction detection characteristics will again be cal- calculated based on tires 2 to 9. In doing so, if the anomaly is caused by a severe leak, the leaking tire will then have the lowest IHV, thus the anomaly detection characteristics with and without the lowest IHV can be compared in block 214, so any non-trivial difference (ie, a difference above a threshold) between the results will typically indicate that the tire with the minimum IHV caused the fault point, and as such , has a serious leak, while the small difference indicates that the cause of the anomaly point is not a serious leak.
[0149] Therefore, in the event of a severe leak, control passes to block 216 to signal an event indicating severe leak from the minimum IHV tire. Otherwise, control passes to block 218 to calculate one or more diagnostic anomaly characteristics.
[0150] As noted above, one assumption on which to base the illustrated modalities is that the chance of two tires starting a serious leak at the same time is very low, so that if an anomaly is not driven by a serious leak in a single tire, it can be considered to be caused by the inflation of one or more tyres. As such, when an ano-malia is detected but a serious leak is not indicated, a check can be conducted on all good tires for potential inflation, tire replacements or tire rotations. Also, since the new model training will readjust the tire integrity values, and return these values to approximately a value of 0, a slow tire leak trend can, in some cases, be disguised as a result. of the new training. Therefore, to prevent potentially unnecessary model retraining, it may be desirable in some modalities to detect the inflation of only a subset of the tires so that only those tires determined to have been inflated will be retrained. In the illustrated mode, for example, events of "inflation of one tyre" and "inflation of two tires" can be detected together with an event of "more than two tires are inflated", so that only one, two, or all tire models will be retrained as needed. It will be appreciated that in some disciplines all tires may be retrained in response to any one tire being inflated, whereas in other disciplines three, four or more tires being inflated may be checked separately.
[0151] Fault diagnostic features are used in system change detection, for example, inflation check, tire replacement, tire rotation and/or any other deliberate change of tire pressure, and feature values extracted after the occurrence of an anomaly are tested with thresholds generated by their historical characteristic values. The number of values that are beyond the thresholds can then be used to categorize the root cause of the anomaly. Thus, for example, block 218 calculates one or more diagnostic anomaly characteristics such as the difference between IHV and its minimum standard deviation, etc.
[0152] Next, in block 220, one or more dynamic or moving thresholds can be calculated based on the historical trends of the characteristics in question. In one embodiment, for example, the mean and standard deviation of the last 100 fault diagnostic characteristic values from each individual tire can be calculated. In one example, two thresholds can be used, with a first threshold being the mean plus/minus three times the standard deviation, and a second threshold being the mean plus/minus four times the standard deviation. In alternative modalities, fixed thresholds can be used.
[0153] Block 222 then determines the number of tires that exceed the threshold(s) and thus determined to have been deliberately changed. In one embodiment the at least two diagnostic fault characteristic values from the same tire can be compared with the two thresholds respectively, and if the two diagnostic fault characteristic values for a tire are outside the second threshold, then the tire can be labeled as an inflated/replaced/rotated tire. In addition, the number of malfunction diagnostic characteristic values that are beyond the first threshold can be accumulated in relation to all tires, and if the total number of malfunction diagnostic characteristic values beyond the first threshold is greater than a preset number (eg five) all tires can be retrained together.
[0154] If a single tire is determined to be inflated, control passes to block 224 to signal an event indicating that the tire has been inflated so that the model for that tire can be retrained. If two tires have been determined to be inflated, control passes to block 226 to signal an event indicating that both tires have been inflated so that models for those tires can be retrained. Otherwise, control passes to block 228 to signal an event indicating that models for all tires must be retrained. SLOW LEAKAGE CHECK
[0155] Again, going back to Figure 4, and specifically the slow leak check in step 134, if the PHV for a tire is normal then a slow leak will be checked. In one modality, a threshold criterion; for example, a PHV dropping below -3 psi; can be set (eg by a user) to decide if a tire has a slow leak. All PHVs from all wheel positions can be compared to a threshold criterion in turns for each iteration, and if a PHV is much lower than threshold (eg more than 1 psi lower than threshold), two points can be accumulated for an indicator or slow leak count associated with that wheel position. If a PHV is not much lower than the threshold (for example, less than 1 psi), a point can be added to the indicator or slow leak count, but if a PHV is not lower than the threshold, the indicator or slow leak count , accumulated, can be reset back to zero and restart accumulating points.
[0156] When the slow leak indicator satisfies another threshold criterion, for example reaching a predetermined value such as 3, the average value of the last two PHVs from the same wheel position can be calculated and stored as an initial slow leak level. As long as the slow leak indicator is not reset to zero, the same initial slow leak level can be used for that wheel position. A slow leak event can later be triggered in response to another threshold criterion, for example where the average values of the last two PHVs are compared to an initial slow leak level from the same wheel position, of such that a slow leak event can be triggered and tire health predictions can be initiated if the average value of the last two PHVs is less than the initial slow leak level by a certain amount (eg 1.5 psi) . TIRE RECOVERY CHECK
[0157] Again returning to Figure 4, and specifically to the recovery check in step 142, this step checks if any of the leaking tires detected have been repaired based on the trend of the data. It is assumed that when a leaking tire is repaired or rebuilt, the tire will be re-inflated, and thus cause an increase in the tire's PHV. In one modality, for each iteration, a PHV delta of the leaking tire (the difference between the current PHV and the last PHV) can be compared to a threshold, for example, 5 psi. If the current PHV delta is greater than the threshold, a next PHV delta (from the next iteration) can be compared to a second threshold, eg 10 psi. If the two delta PHVs are greater than the thresholds, the tire can be considered as recovered from the leak (a healthy tire again), causing the tire recovery event to trigger and retraining to be performed for that tire (block 144). TIRE INTEGRITY PROGNOSTICS
[0158] Returning yet again to Figure 4, an additional aspect of the illustrated modalities is the tire integrity prediction or prediction, which is implemented in block 112. The prediction is used to predict the future trend of tire integrity a system based on its historical conditions, and on the illustrated modalities, the historical PHVs from the low-leak tires are adjusted in a prediction model to predict the future performance of the tire.
[0159] A prediction model consistent with the invention can be implemented using a few different types of predictive techniques. Four such predictive techniques, designated here as linear, exponential, cubic routine, and piecewise linear, are discussed in more detail below. Within the context of this discussion, an exemplary dataset of 10 tractor/trailer combinations (ie 10-18 wheel vehicles), based on TPMS collected every 16 minutes, is used to generate the PHV values at intervals of 16 minutes, with the PHV values in turn used as the input data for the respective models, and with the output of each model being a future trend of leakage for approximately 5 hours (20 intervals) to reach a low threshold pre-set pressure.
[0160] Each model can be configured, for example, to be updated repeatedly starting on detection of an event until the end of the event has been reached (for example, when a tire is fixed and inflated again). An initial training set can include the data collected between the last tire inflation event and the leak detection event. In addition, prediction with a prediction model can cover N steps (eg 20 steps) ahead, with a mean squared prediction error (RMSE) calculated for each iteration using 20 steps prediction.
[0161] For the purpose of comparing the different techniques, two typical events were used. A first event (Event 1) is relatively simpler, and is more linear, while a second event (Event 2) has more than one slope change. Figures 18A and 18B respectively illustrate the PHV values corresponding to these events.
[0162] Figures 19A and 19B respectively illustrate the application of a linear regression model to data points for Events 1 and 2, respectively. Linear regression prediction was found to have a relatively low vulnerability to local trends, and can be represented as - i.:.: = o + . where is is the slope and ■ is the intercept, and where the slope is bound to be less than or equal to zero.
[0163] Figures 20A and 20B respectively illustrate the application of an exponential regression model to data points for Events 1 and 2, respectively, using exponential extrapolation, while Figures 21A and 21B respectively illustrate the application of another exponential regression model to data points for Events 1 and 2, respectively, using linear extrapolation. The exponential regression prediction can be represented in the form ■ = , with a slope limitation ab not being positive. With exponential extrapolation (Figures 20A and 20B), the fitted exponential equation is used for extrapolation, whereas with linear extrapolation (Figures 21A and 21B), the position/slope of the last training point is used for extrapolation.
[0164] Both linear and exponential regression models may have limitations in some applications. For example, as illustrated in Figure 22, if additional training data (eg 10 days) is used for Event 2, the two models will provide relatively poor prediction results. In some applications, it may be desirable to improve the fit by using only a portion of the data points as training data to reduce the probability of a large curvature being in the training dataset (eg, by using only the latter 50, 150, 250, etc. data points as training data).
[0165] Another predictive technique that can be used is piecewise regression, which is often used to model systems that have sudden changes of condition within the data. With part regression, different conditions are modeled across different parts of a model, and typically an overall least square estimate is used to optimize the model's fit performance across all parts. Connection points between two parts and/or the leftmost/rightmost points on a curve are considered nodes, and the number and location of nodes are typically positioned to maximize model performance. Both can be decided directly or computer using optimization models.
[0166] Two piece regression models, examples discussed below (cubic and linear piece routine) use 6 as the maximum number of nodes, such that training data can be truncated to 5 pieces and mo. - fingers separately. In addition, the exact number of nodes can be optimized in some modalities using general cross validation, which is discussed in more detail below.
[0167] Node locations can be optimized by the model itself, in such a way that the curvatures in the training dataset are better modeled, and additionally a better prediction can be generated.
[0168] Among other techniques, a Multivariate Adaptive Regression Routines technique (also known as MARS) can be used to build the piecewise regression models.
[0169] With the MARS technique, the final model typically takes the form
where a plurality (n) of basic functions Bi(x) is individually weighted by a corresponding weight wi. Each basic function can typically be implemented as a constant term, a critical point function, or a multiplication of multiple critical point functions. A critical point function is a per-piece function that typically takes the form max(0,cx) or max(0,xc), where c is a constant corresponding to a node.
[0170] The model building process is similar to a step regression that typically has two phases: advance selection and retreat clearing. During selection and advancement, the training dataset is segmented into multiple separate pieces until an acceptable minimized model error is obtained. The model is initialized by having only one term, which is the average of all training data. Then, one employs a voracious algorithm and keeps searching for the next node (the breaking point between two pieces) to segment a next piece and add a next pair of basic functions that provide the greatest reduction for global model training error in terms of squared error sum. Basic functions are typically added to the model until a preset criterion is satisfied, for example, where the training error is less than a preset value, where before and after adding the next basic function, the reduction in model error is smaller than a preset value, or where the number of basic functions reaches a preset limit.
[0171] In the back-off erasure phase, the at least effective adjacent pieces of model from the advance selection are combined to increase the generalizability of the model. A General Cross Validation (GCV) value can be used to consider both the model's fit performance and complexity, with an objective to minimize the GCV across all parts:
where k is the number of basic functions, d is a penalty for each node (eg 2), and n is the number of training data points.
[0173] A cubic routine model is a third-order polynomial piece regression model, where each piece can be represented by an equation taking the form fi(t)=ai + bit + cit2+dit3, and bounded by such that everywhere on the line, the line/slope/curvature is continuous and the slope is not positive. In this example, 6 nodes are used, and the prediction is based on a linear extrapolation using the slope and location of the last (rightmost) training point. Figures 23A and 23B, for example, illustrate the application of such a cubic routine model to data points for Events 1 and 2, respectively.
[0174] A piece linear model, in comparison, is a first polynomial piece regression model, where each piece can be represented by an equation taking the form fi(t)=ai + bit, and limited in such a way that in anywhere on the line, the line is continuous and the slope is not positive. In this example, 6 knots are used, and the prediction is based on a linear extrapolation using the slope and location of the last training point (far right). Figures 24A and 24B, for example, illustrate the application of such a piecewise linear model to data points for Events 1 and 2, respectively.
[0175] Based on the foregoing, the performance of the five models (linear, exponential with exponential extrapolation, exponential with linear extrapolation, cubic routine and linear per piece) can be compared using a moving window-size Mean Square Error value fixed (20 steps ahead) based on the predicted results for both Events 1 and 2. The prediction can be started right after a leak is detected and can be updated whenever a new measurement is received (that is, whenever a new measurement is received. is added to the training dataset). Therefore, an array of RMSE values can be generated for each model between detection and fixation (end) of a leak. The training dataset can include all PHV values that are generated in the form revealed above since the last time the prediction model was trained again (re-training model can be triggered, for example, upon tire inflation/replacement). As the duration from the last, retraining to the onset of a leak varies, the main composition of the training data may vary. If a leak starts shortly after retraining the model, most of the training data is collected after the leak starts, whereas if a leak starts several days after retraining the last model, the set Training data may include more data collected before starting the leak. For the purposes of this discussion, small (ie most data after the leak starts) and large (ie most data before the leak starts) training datasets are used respectively for both events. previously mentioned leaks. Figures 18A and 18B previously discussed, for example, illustrate small training datasets for Events 1 and 2, respectively, while Figures 25A and 25B illustrate exemplary large training datasets for Events 1 and 2.
[0176] Figures 26A-26B illustrate the comparative performance (in terms of RMSE) of the five models mentioned above for the small training set for Events 1 and 2, respectively. Figure 26A illustrates that for Event 1 and the small training set, the exponential with exponential extrapolation model lags behind the performance of the other four models. Figure 26B illustrates that for Event 2 and the small training set, where the data forms a trend with more curvature, the linear, cubic routine, and linear per piece models perform better than the exponential models.
[0177] Figures 26C-26D illustrate the comparative performance (in terms of RMSE) of only linear, cubic routine, and linear per piece models for Events 1 and 2, respectively. In general, for the large training set, the cubic and linear per piece routine models perform better than a linear model. Also, as the tire leak trend is usually relatively noisy, it is believed that the cubic routine model can in some cases be very flexible (compared to the linear piece model), and may take a longer time. to train (since it is third order vs. first order), so in some modalities the piecewise linear regression model may be desirable for use in slow leakage prediction.
[0178] It will be considered, however, that alternative regression or extrapolation algorithms can be used to predict future inflation levels of a tire, including, for example, Autoregressive Moving Average (ARMA), linear regression, regression polynomial, quadratic regression, exponential regression, exponential regression with linear extrapolation, routine regression, piecewise linear regression, neural networks, Kalman filters, particle filters, similarly based prediction, etc. It will also be considered that in some modalities, it may be desirable to apply one or more noise removal techniques to improve model accuracy, for example, by top-down sampling (eg, taking an average of several data points) and/or ascending sampling (eg, by interpolating between data points) of a dataset.
[0179] Furthermore, in some modalities, it may be desirable to use multiple regression models in cooperation with each other, for example, merging the linear regression together with the quadratic and exponential regression, with the regression linear responsible for straight line prediction and quadratic and exponential regression used to change the slope curve fits. Fusion rules can be based on knowledge of the shapes of the training dataset and predicted samples, for example, based on the knowledge that the curve shape of a degrading tire should not continue to climb, and can be used to combine and/or select the various types of regressions at different points.
[0180] Based on the determined predictions, various modalities can provide fins and prediction-related information to a vehicle operator, a vehicle or fleet owner, a service or maintenance provider, and/or third parties so that, for example, tire maintenance operations can be scheduled automatically or manually. For example, tires can be scheduled for repair or replacement. In addition, information related to prediction can be used, for example, to determine how long a tire will be able to be used before needing inflation, repair and/or replacement.
[0181] It will also be considered that although the illustrated modalities use tire data such as pressure, temperature and time stamps, additional data such as ambient temperature, tread depth, additional temperatures performed at different points in a tire, rim or wheel, vibration, etc. they can also be used in a tire monitoring system consistent with the invention. Therefore, the invention is not limited to the specific embodiments discussed herein. WORK EXAMPLE
[0182] For further explanation of the prediction and monitoring functionality described above, a real-world dataset, considered from 10 tractor-trailer combinations (ie 18 tires for each truck) uses in the truck cargo transport (LTL) business (that is, all the loads on a truck come from multiple locations and are also distributed to multiple locations) were analyzed. Each truck was provided with a TPM system that acquired data from all 18 tires, with measurements taken approximately every 30 seconds when the tires were rotating (detected by accelerometers), and collected approximately every 16 minutes. The collected data were manually downloaded, and from these data, a relative time stamp, temperature, and pressure for each tire at each collection interval was extracted for analysis. Overall, approximately 7 months of data were collected for each truck.
[0183] To simulate real-time conditions using off-line data, an iterative algorithm was used, such that a data packet was loaded into the system and owned in a moment. If an unhealthy tire were detected, an alarm would be triggered, and otherwise the other iteration would begin.
[0184] From the point of view of anomaly detection in intact tires, the input variables for the entire system were temperature, pressure and time. It was found, however, that due to large gaps in data, whether due to nights or weekends when trucks were not operating, data visualization could be improved by graphically plotting data points against their higher indices. properly than against time.
[0185] It was found, for example, that the use of an anomaly detection feature allowed an anomaly, particularly the inflation of multiple tires on a vehicle, to be detected. Figure 27, for example, illustrates an exemplary IHV chart for a set of 10 tires on a vehicle. Also, this figure shows that the IHVs of different tires before point 365 are close to each other, but are separated after passing point 365.
[0186] From a real-time system perspective, it is desirable that this event be detected relatively early after its occurrence. So it makes more sense to look at the IHV before point 366 (including 366, a data point after filling takes place) which is shown in Figure 28A. Although it may be difficult to visualize the change of point 365 from its previous points by applying the anomaly detection feature discussed above, for example, by calculating the PHV standard deviation among healthy tires (for example, according to shown in Figure 15A), the change is easier to visualize. The anomaly detection feature is shown in Figure 28B, and illustrates where two consecutive feature data points (points 365, 366) were above the threshold indicating that a system anomaly was detected.
[0187] Subsequently, by recalculating the anomaly detection characteristic without the tire with the minimum IHV, as discussed above, similar characteristic values that were above the threshold were also found at points 365 and 366, and thus the cause of the system anomaly was categorized as filling. In addition, the fault diagnosis feature showed that more than two tires were inflated, and as such, the health assessment models on all tires were retrained. After retraining, including tire self-comparison and parity-based comparison, the IHVs of all tires were merged together and the PHVs returned to a value of approximately zero, indicating that the health condition remained good. Figures 29A and 29B respectively show the IHVs and PHVs after retraining, and it should be noted that although after inflation more air was pumped into the tires, the tires were still indicated to be in substantially the same condition as integrity as before. As such, the tire integrity conditions did not change based on the amount of air in each tire. The PHVs in Figure 29B were all at approximately 0, which indicated that the tires were in equally good condition before and after the inflation event. Therefore, after the new model training, the PHV continued to accurately reflect the integrity conditions of all tires.
[0188] Therefore, it was found that reducing the variance within a group of good conditions by bringing tires from different inflation levels to the same integrity condition and extracting their similar duty regime effect from the data, facilitated detection of a possible bad tire condition. Additionally, system changes such as filling events were accurately detected.
[0189] Next, from a leak detection and prediction point of view, the same data set was used to try to detect a slow leak and predict the leak rate of the leaking tire. Figure 30, for example, illustrates a pressure compensated plot of four tires from a truck in the dataset. To detect leaking tires while not generating too many false alarms, thresholds below the lowest normal tire pressures were established. In addition, some of the four tires were known to have been inflated at data point 1189 at two different thresholds that needed to be adjusted to detect tire leakage. Before filling a threshold 260 of 97 psi was set while a threshold 262 after filling was set at 103 psi. Using compensated pressure, a leak from tire 3, shown in graph 264, was initially detected at point 1387, where its pressure had the first major drop.
[0190] However, by applying the aforementioned prediction algorithm, premature leak detection was obtained. First, through tire auto-comparison, inflation detection, and model retraining, it was found that only a single threshold was needed for any tire at any one time, as shown in Figure 31. Also, a slow leak that could not be observed from the compensated pressure is shown at 266 on the IHV graph, and the leaking tire detection time could be changed to an earlier time, for example around point 300. Then after a comparison based on parity being conducted, a smaller data range within each tire is obtained, as shown in Figure 32, and the leak detection time can be further advanced to approximately point 120. Also, as the leak was not severe when it was initially detected, a prediction process was subsequently triggered to infer the future health condition of the tyre.
[0191] Several additional modifications can be made without departing from the essence and scope of the invention. Therefore, the invention is based on the claims appended below.

权利要求:
Claims (10)
[0001]
1. A method of monitoring vehicle tires (14, 16), the method comprising: receiving tire condition data associated with a plurality of tires on a vehicle (14, 16); and determining an integrity condition for each of the plurality of tires, wherein determining the integrity condition for each of the plurality of tires includes: reducing the tire condition data variances among the plurality of tires on the vehicle (14, 16 ), performing an individual self-comparison of the tire condition data for each tire, where the individual self-comparison compares the current condition of a tire with what is considered a normal operating condition of the tire; and CHARACTERIZED by the fact that: reducing the variances of the tire condition data within each tire by performing a parity based comparison of the tire condition data for each tire, where the parity based comparison compares the tire condition data tire for a first tire with those of at least a subset of the plurality of tires.
[0002]
2. Method, according to claim 1, CHARACTERIZED by the fact that performing the individual self-comparison includes: applying an assessment model to the tire data, in which the assessment model comprises a distance-based assessment model (DBA), a logistic regression model (LR) or a self-organizing map model (SOM); and generate an individual integrity value (IHV) for each tire based on the evaluation model (126), where the parity integrity value (PHV) for each tire compares a current performance of such tire with a normal performance of such. tire determined during training of the evaluation model (126).
[0003]
3. Method according to claim 1 or 2, CHARACTERIZED by the fact that performing the comparison based on parity includes one or more comparing the tire data for a first tire with those of at least a subset of the plurality of tires, comparing the tire data for the first tire with an average determined from the tire data for at least a subset of the plurality of tires, and comparing the tire data for the first tire with an average determined from the tire data for at least a subset of the plurality of tires.
[0004]
4. Method according to any one of claims 1 to 3, CHARACTERIZED by the fact that the method further includes: training a specific tire model for each tire using a first-ratio of the tire data to establish a normal operating condition for each tire; and applying a second portion of the tire data to the specific tire model for each tire to determine the health condition of each tire, wherein training the specific tire model for each tire includes determining an average pressure value for each tire. , and wherein applying the second portion of the tire data includes: determining an individual integrity value (IHV) for each tire based on the individual self-comparison, wherein determining the parity integrity value (PHV) for each tire includes determining a difference between a pressure compensated value and the average pressure value; and determining a parity integrity value (PHV) for each tire based on the parity based comparison, wherein determining the parity integrity value (PHV) for each tire includes determining a difference between the integrity value of parity (PHV) for each tire and an average of the integrity of parity (PHV) values for at least a subset of the plurality of tires.
[0005]
5. Method according to claim 4, CHARACTERIZED by the fact that the training of the specific tire model (124) is carried out in a training mode (122) for the specific tire model and the application of the second portion of the tire data is performed in a test mode for the specific tire model, and that method further includes, in response to detecting an anomaly for a tire, returning the specific tire model to the training mode (122) to again train the specific tire model of the tire.
[0006]
6. Method, according to any one of claims 1 to 5, CHARACTERIZED by the fact that the method further includes: determining an anomaly detection characteristic; detect an anomaly based on the anomaly detection feature; determine a dynamic threshold based on a historical trend of the anomaly detection characteristic, where detecting the anomaly is based on the anomaly detection characteristic exceeding the dynamic threshold; in response to detection of an anomaly: determining an anomaly diagnostic feature; and determining a malfunction type based on the malfunction diagnostic feature, where the malfunction type is selected from the group consisting of a severe leak, a single tire inflation, a subset tire inflation, and an all-inflation the tires, and in that for the inflation of a single tyre, a specific tire model for the single tire is again trained, for the inflation of a subset of tires, specific tire models for the subset of tires are again trained, and for the inflation of all tires, specific tire models for all tires are trained again.
[0007]
7. Method according to any one of claims 1 to 6, CHARACTERIZED by the fact that the method further includes predicting leakage (112) in a tire based on historical parity integrity values (PHVs) of the tire, where predicting leakage (112) includes one or more of performing at least one linear, exponential, routine cubic and linear regression per part, and building a prediction model using Multivariate Adaptive Regression Routines (MARS) regression analysis to minimize mean square error (RMSE) in the prediction model.
[0008]
8. Method according to any one of claims 1 to 7, CHARACTERIZED by the fact that the method further includes detecting a slow leak (134) in a tire based on the parity integrity values (PHVs) of the tire through a the plurality of iterations, wherein detecting the slow leak (134) includes, during each iteration: comparing a tire parity integrity (PHV) value to a first threshold criterion; accumulating a count for the tire if the tire's parity integrity (PHV) value satisfies a first threshold criterion; and resetting the count for the tire if the tire parity integrity (PHV) value does not satisfy the first threshold criterion.
[0009]
9. Method according to any one of claims 1 to 8, CHARACTERIZED by the fact that the method further includes detecting the recovery (142) of a tire based on historical parity integrity values (PHVs) of the tire, where detecting the recovery (142) of the tire comprises: comparing a current PHV of the tire with a threshold criterion; and triggering a tire recovery event (142) in response to the tire's current parity integrity (PHV) value satisfying the threshold criterion.
[0010]
10. Apparatus for monitoring vehicle tires (14, 16) on a vehicle (14, 16) comprising: at least one processor; and set of instructions defined by executing at least one processor to monitor vehicle tires (14, 16), receiving tire condition data associated with a plurality of tires on the vehicle (14, 16) and determining an integrity condition for each of the plurality of tires, wherein determining the health condition for each of the plurality of tires includes: reducing the tire condition data variances among the plurality of tires on the vehicle (14, 16), performing an individual self-comparison of the tire condition data for each tire, where individual self-comparison compares the current condition of a tire with what is considered a normal operating condition of the tire; and CHARACTERIZED by the fact that: reducing the variances of the tire condition data within each tire by performing a parity based comparison of the tire condition data for each tire, where the parity based comparison compares the tire condition data tire for a first tire with those of at least a subset of the plurality of tires.

类似技术:

---

公开号 | 公开日 | 专利标题

---

BR102014005772B1|2021-06-01|METHOD OF MONITORING VEHICLE TIRES AND APPARATUS TO MONITORING VEHICLE TIRES

---

US6868358B2|2005-03-15|Method for processing information in a tire pressure monitoring system

---

US20160167446A1|2016-06-16|Tire service life monitoring

---

JP2017509103A|2017-03-30|Method, apparatus, and system for estimating battery degradation state during operation of electric or hybrid vehicle, and method for constructing model for estimation

---

US9376118B2|2016-06-28|Assessment of tire condition based on a tire health parameter

---

EP3017972B1|2021-12-01|Wheel diagnostic monitoring

---

JP2020129233A|2020-08-27|Abnormality detection device

---

CN105658937A|2016-06-08|Method for monitoring operation of sensor

---

Egaji et al.2019|An innovative decision rule approach to tyre pressure monitoring

---

CN105480030B|2017-07-21|Judge the abnormal method of vehicle load

---

US20210217256A1|2021-07-15|Method and Apparatus for Diagnosing and Monitoring Vehicles, Vehicle Components and Routes

---

US20200231010A1|2020-07-23|Tire health monitoring systems and methods thereto

---

US10933705B2|2021-03-02|Apparatus and method for monitoring tire pressure of vehicle

---

JP2019191122A|2019-10-31|Tire state detector

---

JP6346280B2|2018-06-20|Decision support method based on Bayesian method in automobile tire pressure monitoring system

---

Carrascal et al.2009|Unsupervised methods for anomalies detection through intelligent monitoring systems

---

US20200134639A1|2020-04-30|Homogeneous model of hetergeneous product lifecycle data

---

Xu2011|Techniques for real-time tire health assessment and prognostics under dynamic operating conditions

---

JP2004322821A|2004-11-18|Tire air pressure lowering alarm device and method, and tire air pressure lowering alarm program

---

BR102020025719A2|2021-07-27|TIRE CONDITIONS ESTIMATION METHOD

---

US20210074080A1|2021-03-11|Part maintenance and value estimation system

---

US20210125428A1|2021-04-29|Apparatus and Method for Tire Separation Monitoring

---

CN114103562A|2022-03-01|Tire pressure monitoring system

---

CN112440629A|2021-03-05|Method for extracting variations in tyre characteristics

---

KR102089675B1|2020-03-16|Fuzzy-logic-based decision support method in a system for monitoring motor vehicle tire pressure

同族专利:

---

公开号 | 公开日

---

US9079461B2|2015-07-14|

---

CN104044413B|2017-11-07|

---

US20140277910A1|2014-09-18|

---

JP6289946B2|2018-03-07|

---

EP2777957A3|2014-12-10|

---

CN104044413A|2014-09-17|

---

EP2777957B1|2017-04-19|

---

EP2777957A2|2014-09-17|

---

JP2014177273A|2014-09-25|

---

BR102014005772A2|2015-06-23|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US4773011A|1986-01-27|1988-09-20|The Goodyear Tire & Rubber Company|Method of surveying, selecting, evaluating, or servicing the tires of vehicles|

---

FR2622845B1|1987-11-09|1990-02-09|Michelin & Cie|PROCESS FOR THE EXPLOITATION OF PRESSURE AND TEMPERATURE MEASUREMENTS IN A TIRE MONITORING DEVICE|

---

US20080046149A1|1995-06-07|2008-02-21|Automotive Technologies International, Inc.|Vehicle Component Control Methods and Systems Based on Vehicle Stability|

---

US5886624A|1991-08-08|1999-03-23|Compagnie General Des Etablissements Michelin-Michelin & Cie|Method for processing pressure and temperature measurements in a tire monitoring system|

---

US5900809A|1991-08-08|1999-05-04|Michelin Recherche Et Technique, S.A.|Method for processing pressure and temperature measurements in a tire monitoring system|

---

US7103460B1|1994-05-09|2006-09-05|Automotive Technologies International, Inc.|System and method for vehicle diagnostics|

---

US7082359B2|1995-06-07|2006-07-25|Automotive Technologies International, Inc.|Vehicular information and monitoring system and methods|

---

US7421321B2|1995-06-07|2008-09-02|Automotive Technologies International, Inc.|System for obtaining vehicular information|

---

US7313467B2|2000-09-08|2007-12-25|Automotive Technologies International Inc.|System and method for in-vehicle communications|

---

US6738697B2|1995-06-07|2004-05-18|Automotive Technologies International Inc.|Telematics system for vehicle diagnostics|

---

US7786864B1|2000-09-08|2010-08-31|Automotive Technologies International, Inc.|Vehicular RFID and sensor assemblies|

---

US20050192727A1|1994-05-09|2005-09-01|Automotive Technologies International Inc.|Sensor Assemblies|

---

US5463374A|1994-03-10|1995-10-31|Delco Electronics Corporation|Method and apparatus for tire pressure monitoring and for shared keyless entry control|

---

US7089099B2|2004-07-30|2006-08-08|Automotive Technologies International, Inc.|Sensor assemblies|

---

US7630802B2|1995-06-07|2009-12-08|Automotive Technologies International, Inc.|Information management and monitoring system and method|

---

WO1996006747A2|1994-08-31|1996-03-07|Otter Controls Limited|A tyre condition monitoring system|

---

US5656993A|1995-05-08|1997-08-12|Semisystems, Inc.|Vehicle wheel condition monitor and data storage system|

---

US6484080B2|1995-06-07|2002-11-19|Automotive Technologies International Inc.|Method and apparatus for controlling a vehicular component|

---

US5749984A|1995-12-29|1998-05-12|Michelin Recherche Et Technique S.A.|Tire monitoring system and method|

---

DE69709417T2|1996-01-26|2002-08-29|Michelin & Cie|Method for evaluating signals from a tire pressure monitoring device of a moving vehicle|

---

FR2745755A1|1996-03-06|1997-09-12|Michelin & Cie|METHOD FOR EXPLOITING PRESSURE AND TEMPERATURE MEASUREMENTS IN A TIRE MONITORING SYSTEM|

---

DE69717183T2|1996-03-20|2003-07-24|Michelin & Cie|Method for evaluating pressure measurement in a tire pressure monitoring device|

---

US6246317B1|1998-02-27|2001-06-12|William Pickornik|Target pressure learn strategy for vehicular tire pressure systems|

---

US6118369A|1998-08-17|2000-09-12|Ford Motor Company|Tire diagnostic system|

---

US7783507B2|1999-08-23|2010-08-24|General Electric Company|System and method for managing a fleet of remote assets|

---

US6809637B1|1999-09-03|2004-10-26|The Goodyear Tire & Rubber Company|Monitoring a condition of a pneumatic tire|

---

US7167796B2|2000-03-09|2007-01-23|Donnelly Corporation|Vehicle navigation system for use with a telematics system|

---

JP4357074B2|2000-03-17|2009-11-04|株式会社ブリヂストン|Tire wear prediction method and tire wear prediction apparatus|

---

WO2001086576A1|2000-05-08|2001-11-15|Systech International, L.L.C.|Monitoring of vehicle health based on historical information|

---

FR2812394A1|2000-07-25|2002-02-01|Michelin Soc Tech|METHOD FOR EVALUATING THE AUTONOMY OF A FLAT ROLLING SYSTEM|

---

US6636790B1|2000-07-25|2003-10-21|Reynolds And Reynolds Holdings, Inc.|Wireless diagnostic system and method for monitoring vehicles|

---

US7161476B2|2000-07-26|2007-01-09|Bridgestone Firestone North American Tire, Llc|Electronic tire management system|

---

US20020130771A1|2000-08-17|2002-09-19|Osborne Corwin K.|Vehicle wheel monitoring system|

---

US6662642B2|2000-09-08|2003-12-16|Automotive Technologies International, Inc.|Vehicle wireless sensing and communication system|

---

JP3769459B2|2000-10-13|2006-04-26|株式会社豊田中央研究所|Tire burst prediction device|

---

CA2426357A1|2000-10-23|2002-05-02|Michelin Recherche Et Technique S.A.|Tyre blown-out detector tor|

---

US7729823B2|2000-11-30|2010-06-01|Pirelli Pneumatici S.P.A.|Method and system for monitoring tyres|

---

US6571617B2|2001-01-17|2003-06-03|Microchip Technology Incorporated|Method and apparatus using directional antenna or learning modes for tire inflation pressure monitoring and location determination|

---

US6738748B2|2001-04-03|2004-05-18|Accenture Llp|Performing predictive maintenance on equipment|

---

US6973824B2|2001-06-12|2005-12-13|Continental Aktiengesellschaft|Tire status detection system|

---

DE10130640A1|2001-06-26|2003-01-02|Opel Adam Ag|Method and device for monitoring tire pressures|

---

US6759952B2|2001-07-06|2004-07-06|Trw Inc.|Tire and suspension warning and monitoring system|

---

US6741169B2|2001-07-06|2004-05-25|Trw Inc.|Tire tread integrity monitoring system and method|

---

US7103760B1|2001-07-16|2006-09-05|Billington Corey A|Embedded electronic device connectivity system|

---

EP1285786B1|2001-08-22|2016-02-03|Sumitomo Rubber Industries Limited|Method and apparatus for detecting abnormalities of tire, and program for detecting abnormalities of tire|

---

DE10148214C2|2001-09-28|2003-07-31|Daimler Chrysler Ag|Method for providing a maintenance algorithm|

---

FR2831107B1|2001-10-18|2004-10-15|Michelin Soc Tech|METHOD FOR ESTIMATING THE AIR TEMPERATURE OF THE INTERNAL CAVITY OF A TIRE AND APPLICATION TO THE DETECTION OF AN ABNORMAL OPERATION OF A FLAT ROLLING SYSTEM|

---

DE10218781A1|2002-04-26|2003-11-13|Tuev Automotive Gmbh|Pneumatic tire mountable on a rim, sensor network, revolution measuring unit and vehicle monitoring system|

---

US20030214394A1|2002-05-17|2003-11-20|Behrendsen Paul Alfred|Method and apparatus for sensing tire performance and wear|

---

US7075421B1|2002-07-19|2006-07-11|Tuttle John R|Tire temperature and pressure monitoring sensors and systems|

---

US6868358B2|2002-07-24|2005-03-15|The Goodyear Tire & Rubber Company|Method for processing information in a tire pressure monitoring system|

---

JP4397327B2|2002-08-12|2010-01-13|株式会社ブリヂストン|Method and apparatus for determining the life expectancy of a run flat tire that continuously runs in a run flat state|

---

US6879893B2|2002-09-30|2005-04-12|The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration|Tributary analysis monitoring system|

---

JP3938354B2|2002-10-28|2007-06-27|横浜ゴム株式会社|Mounting structure of tire internal information collection device|

---

DE10360122C5|2003-12-20|2016-01-07|Patentportfolio S. à. r. l.|Method for pressure monitoring of motor vehicle tires|

---

CA2553023C|2004-01-09|2013-03-19|United Parcel Service Of America, Inc.|System, method and apparatus for capturing telematics data with an active rfid tag|

---

DE102004010010B3|2004-03-01|2005-10-13|Siemens Ag|Method and monitoring device for monitoring a wheel characteristic of a wheel|

---

DE102004010665B4|2004-03-04|2014-09-18|Infineon Technologies Ag|Device and method for determining a state parameter of an object to be monitored|

---

JP4382562B2|2004-04-12|2009-12-16|株式会社ブリヂストン|Method and apparatus for determining end of run-flat running life of run-flat tire|

---

JP2005313846A|2004-04-30|2005-11-10|Yokohama Rubber Co Ltd:The|Designing method of support body for emergency travel, computer program for design of support body for emergency travel, manufacturing method of support body for emergency travel, designing method of support body / tire assembly for emergency travel, and manufacturing method of support body / tire assembly for emergency travel|

---

US20060010961A1|2004-07-19|2006-01-19|Alex Gibson|Method and apparatus for detecting leakage rate in a tire pressure monitoring system|

---

US8565959B2|2004-12-03|2013-10-22|The Goodyear Tire & Rubber Company|Method for detection of low leak rates in a tire|

---

US7301445B2|2004-12-28|2007-11-27|Caterpillar Inc.|Tire maintenance system|

---

FR2884455B1|2005-04-13|2010-02-26|Michelin Soc Tech|METHOD AND DEVICE FOR MONITORING THE PRESSURE OF A TIRE|

---

US7323975B2|2005-04-19|2008-01-29|General Motors Corporation|Tire pressure monitoring system and method therefor|

---

US7272536B2|2005-05-23|2007-09-18|Potts Gerald R|Failure warning for a tire among a plurality of tires|

---

KR20070008988A|2005-07-14|2007-01-18|주식회사 유컴테크놀러지|Tire pressure monitoring apparatus using radio frequency identification system|

---

JP5276778B2|2005-08-31|2013-08-28|株式会社ブリヂストン|Tire information management system|

---

FR2895703B1|2006-01-05|2008-02-15|Michelin Soc Tech|METHOD FOR PROCESSING DATA OF A TIRE MONITORING SYSTEM OF A VEHICLE|

---

US20070222568A1|2006-03-21|2007-09-27|International Business Machines Corporation|System and method for vehicle tire usage information|

---

FR2900099B1|2006-04-21|2008-07-04|Michelin Soc Tech|METHOD OF CONFIRMING A PRESS ALERT IN A PULSE MONITORING SYSTEM|

---

US8009027B2|2006-05-31|2011-08-30|Infineon Technologies Ag|Contactless sensor systems and methods|

---

US7751955B2|2006-06-30|2010-07-06|Spx Corporation|Diagnostics data collection and analysis method and apparatus to diagnose vehicle component failures|

---

US20080018441A1|2006-07-19|2008-01-24|John Robert Orrell|Tire failure detection|

---

US7657504B2|2006-10-10|2010-02-02|Microsoft Corporation|User interface for displaying images of sights|

---

US7908928B2|2006-10-31|2011-03-22|Caterpillar Inc.|Monitoring system|

---

JP5018116B2|2007-02-15|2012-09-05|横浜ゴム株式会社|Tire design method and tire design apparatus|

---

US8494704B2|2007-03-16|2013-07-23|Nira Dynamics Ab|Tire pressure classification based tire pressure monitoring|

---

WO2009070063A1|2007-11-30|2009-06-04|Volvo Lastvagnar Ab|Wheel-monitoring module|

---

WO2009073945A1|2007-12-12|2009-06-18|Tirestamp Inc.|Tire asset tracking system|

---

US9026304B2|2008-04-07|2015-05-05|United Parcel Service Of America, Inc.|Vehicle maintenance systems and methods|

---

FR2933034B1|2008-06-30|2011-08-26|Michelin Soc Tech|INFLATION ASSISTING DEVICE AND METHOD|

---

JP2010165242A|2009-01-16|2010-07-29|Hitachi Cable Ltd|Method and system for detecting abnormality of mobile body|

---

US8659412B2|2009-12-10|2014-02-25|Continental Automotive Systems, Inc.|Tire pressure monitoring apparatus and method|

---

FI20105845A|2010-08-11|2012-02-12|Secure Oy W|Vehicle wheel control system|US7592903B2|2006-10-30|2009-09-22|Spx Corporation|Tire pressure monitor system tool with re-learn and diagnostic procedures|

---

NO2833907T3|2012-11-02|2018-07-28|

---

US20160200153A1|2013-08-20|2016-07-14|Mobile Awareness, Llc|Tire wellness system|

---

US10245905B2|2013-09-20|2019-04-02|Komatsu Ltd.|Tire abnormality management system and tire abnormality management method|

---

WO2015164763A1|2014-04-25|2015-10-29|Oceaneering International, Inc.|Remotely operated vehicle power management system and method of use|

---

US9376118B2|2014-07-08|2016-06-28|The Goodyear Tire & Rubber Company|Assessment of tire condition based on a tire health parameter|

---

US9881428B2|2014-07-30|2018-01-30|Verizon Patent And Licensing Inc.|Analysis of vehicle data to predict component failure|

---

DE102014220814A1|2014-10-14|2016-04-14|Continental Automotive Gmbh|Driver assistance system for a vehicle and method for operating a driver assistance system|

---

US9636956B2|2014-11-04|2017-05-02|The Goodyear Tire & Rubber Company|Wheel diagnostic monitoring|

---

US20160129734A1|2014-11-06|2016-05-12|The Goodyear Tire & Rubber Company|Mobile-phone based data processing system|

---

US9418492B2|2014-11-06|2016-08-16|The Goodyear Tire & Rubber Company|Mobile phone enabled data processing system|

---

IN2015CH02899A|2015-06-10|2015-06-19|Tymtix Technologies Pvt Ltd|

---

TWI585382B|2016-03-02|2017-06-01|緯創資通股份有限公司|Method, system, and computer-readable medium for detecting depth of tire tread pattern|

---

EP3225429B1|2016-03-28|2019-11-27|Dana Heavy Vehicle Systems Group, LLC|Tire condition telemetrics system|

---

KR101843256B1|2016-05-11|2018-03-28|한국타이어 주식회사|A method of evaluating the state of tire and providing an alarm and a smart tire therewith it|

---

US10916129B2|2017-01-30|2021-02-09|International Business Machines Corporation|Roadway condition predictive models|

---

US10809742B2|2017-03-06|2020-10-20|The Goodyear Tire & Rubber Company|System and method for tire sensor-based autonomous vehicle fleet management|

---

CN107089102B|2017-04-21|2019-03-22|重庆长安汽车股份有限公司|A kind of detection method of the slow gas leakage of tire|

---

CN107539038B|2017-06-06|2020-10-13|新华三技术有限公司|Vehicle tire pressure state monitoring method and device|

---

CN107862375A|2017-10-30|2018-03-30|北京计算机技术及应用研究所|A kind of two stage equipment fault diagnosis method|

---

US11151883B2|2017-11-03|2021-10-19|International Business Machines Corporation|Empathic autonomous vehicle|

---

JP2019119170A|2018-01-10|2019-07-22|セイコーエプソン株式会社|Abnormality warning method and abnormality warning system|

---

US10579054B2|2018-01-29|2020-03-03|Uatc, Llc|Systems and methods for on-site recovery of autonomous vehicles|

---

US10596865B2|2018-01-31|2020-03-24|Ford Global Technologies, Llc|Systems and methods for controlling vehicle TPMS sensor localization|

---

IT201800005600A1|2018-05-22|2019-11-22|SYSTEM AND METHOD FOR MANAGING DATA RELEVANT TO A WHEEL SERVICE|

---

US11243957B2|2018-07-10|2022-02-08|Verizon Patent And Licensing Inc.|Self-organizing maps for adaptive individualized user preference determination for recommendation systems|

---

WO2020020437A1|2018-07-24|2020-01-30|Bayerische Motoren Werke Aktiengesellschaft|Method indicating unexpected behaviour and vehicle, system, and storage medium comprising the same|

---

US11200141B2|2018-07-25|2021-12-14|Dell Products L.P.|Apparatus and method for troubleshooting poor part life zones in a datacenter|

---

IT201900001585A1|2019-02-04|2020-08-04|Tn Service Srl|REMOTE CONTROL SYSTEM OF THE TIRE STATUS OF ANY VEHICLE|

---

WO2021261610A1|2020-06-23|2021-12-30|엘지전자 주식회사|Method for monitoring, by vehicle, control message in automated vehicle & highway systems|

---

CN113561713B|2021-09-24|2021-11-30|深圳市沃易科技有限公司|Tire pressure abnormity troubleshooting method and system based on Internet of things|

法律状态:
2015-06-23| B03A| Publication of a patent application or of a certificate of addition of invention [chapter 3.1 patent gazette]|

---

2018-11-13| B06F| Objections, documents and/or translations needed after an examination request according [chapter 6.6 patent gazette]|

---

2020-03-31| B06U| Preliminary requirement: requests with searches performed by other patent offices: procedure suspended [chapter 6.21 patent gazette]|

---

2021-05-11| B09A| Decision: intention to grant [chapter 9.1 patent gazette]|

---

2021-06-01| B16A| Patent or certificate of addition of invention granted|Free format text: PRAZO DE VALIDADE: 20 (VINTE) ANOS CONTADOS A PARTIR DE 12/03/2014, OBSERVADAS AS CONDICOES LEGAIS. |

优先权:

---

申请号 | 申请日 | 专利标题

---

US13/828.124|2013-03-14|

---

US13/828,124|US9079461B2|2013-03-14|2013-03-14|Predictive peer-based tire health monitoring|

---