Machine learning model for electric vehicle component health monitoring

Abstract

Methods, computing systems, and techniques for machine-learned vehicle component health monitoring are presented. An example method may include obtaining operational data describing one or more operational characteristics of a component of a subsystem on a vehicle. The example method may include generating a component life value for the component using the component life model and based on the operational data. The example method may include generating, using a vehicle usage model and based on a component life value, a pre-decision of the component. In an example method, a vehicle usage model may be configured to evaluate component life values based on usage patterns associated with a vehicle. The example method may include initiating a corrective action to mitigate degradation of the component based on the anticipation.

Description

Technical Field

[0001] This disclosure relates in general to the use of artificial intelligence (including machine learning models) to monitor the operational health of components of a vehicle, such as an electric vehicle. Background Technology

[0002] Research on climate change has identified emissions from internal combustion engine vehicles as a driving force behind global warming. Replacing internal combustion engine vehicles with electric vehicles can reduce net greenhouse gas emissions. Compared to internal combustion engine vehicles, improving electric vehicles can increase their use and adoption. Therefore, improving electric vehicles can help mitigate climate change.

[0003] Vehicles, such as automobiles, are complex electromechanical assemblies of components. Some components may experience wear and tear during use. The operational health of some components may decline over time. Many vehicle operators may be familiar with recognizing wear patterns in conventional (e.g., internal combustion) vehicle components. For example, engine noise or other indicators can provide early warning of component wear.

[0004] However, electric vehicles may be unfamiliar to many vehicle operators. Therefore, concerns about the maintenance of electric vehicles may inhibit their adoption. Improved maintenance techniques for electric vehicles could increase their adoption, thereby contributing to climate change mitigation. Summary of the Invention

[0005] The specific aspects and advantages of this disclosure will be set forth in part in the description which follows, or may be learned from the description or by practice of its implementation.

[0006] In one example aspect, this disclosure provides an example computational system for monitoring the operational health of a vehicle subsystem. The example computational system may include control circuitry. The control circuitry may be configured to acquire operational data describing one or more operational characteristics of components of the subsystem on the vehicle. The control circuitry may be configured to use a component lifetime model and generate component lifetime values ​​based on the operational data. The control circuitry may also be configured to use a vehicle usage model and generate predictions about the components based on the component lifetime values, wherein the vehicle usage model is configured to evaluate the component lifetime values ​​based on usage patterns associated with the vehicle. The control circuitry may be configured to initiate corrective actions based on the predictions to mitigate component degradation.

[0007] In some specific implementations of the example computing system, predictions may include the effective lifetime value of a component.

[0008] In some specific implementations of the example computing system, the control circuitry can be configured to obtain an initial effective lifetime value.

[0009] In some specific implementations of the example computing system, the control circuitry may be configured to use a component lifetime model and generate a component lifetime value based on operating data and an initial effective lifetime value.

[0010] In some specific implementations of the example computing system, the control circuitry can be configured to iteratively update the storage location for the stored lifetime value.

[0011] In some specific implementations of the example computing system, the component lifetime model may include a physics-based model.

[0012] In some specific implementations of the example computing system, the component lifetime model may include a machine learning neural network trained to output a set of associations between input operational characteristics and one or more failure modes of the component.

[0013] In some specific implementations of the example computing system, the components may be active electrical components. In some specific implementations of the example computing system, one or more failure modes may include at least one of the following: (i) bond wire lift-off / detachment, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

[0014] In some specific implementations of the example computing system, the control circuitry may be configured to, for each of a plurality of components of the subsystem,: obtain corresponding operational data describing one or more corresponding operational characteristics of the corresponding component; and generate a corresponding component lifetime value using a corresponding component lifetime model and based on the corresponding operational data. In some specific implementations of the example computing system, the control circuitry may be configured to generate a prediction of the subsystem using a vehicle usage model and based on multiple corresponding component lifetime values.

[0015] In some specific implementations of the example computing system, the vehicle usage model may include a machine learning neural network trained to generate effective lifespan values ​​based on latent embeddings of usage patterns.

[0016] In some specific implementations of the example computing system, the control circuitry may be configured to obtain usage patterns from a usage pattern database, which is stored in the database in association with a specific user of the vehicle.

[0017] In some specific implementations of the example computing system, the vehicle usage model may include a machine learning neural network trained to identify anomalous usage patterns of the vehicle based on one or more operational characteristics. In some specific implementations of the example computing system, control circuitry may be configured to use the vehicle usage model to identify anomalous usage patterns outside the domain of the component lifetime model. In some specific implementations of the example computing system, control circuitry may be configured to utilize predictive overlay component lifetime values ​​based on the identified anomalous usage patterns.

[0018] In some specific implementations of the example computing system, initiating corrective actions based on prediction may include: initiating a control signal configured to cause the vehicle to present a warning message to the occupants of the vehicle.

[0019] In some specific implementations of the example computing system, initiating corrective actions based on predictions may include sending a message instructing the prediction to a remote server.

[0020] In one example aspect, this disclosure provides an example method for monitoring the operational health of a vehicle subsystem. This example method may be a computer-implemented method. The example method may include obtaining operational data describing one or more operational characteristics of components of the subsystem on the vehicle. The example method may include using a component lifetime model and generating a component lifetime value based on the operational data. The example method may include using a vehicle usage model and generating a prediction of the component based on the component lifetime value, wherein the vehicle usage model is configured to assess the component lifetime value based on usage patterns associated with the vehicle. The example method may include initiating corrective actions based on the prediction to mitigate component degradation.

[0021] In some specific implementations of the example method, the prediction may include the effective lifetime value of the component.

[0022] In some specific implementations of the example method, the example method may include obtaining an initial effective lifetime value.

[0023] In some specific implementations of the example method, the example method may include using a component lifetime model and generating a component lifetime value based on operational data and an initial effective lifetime value.

[0024] In some specific implementations of the example method, the example method may include iteratively updating the storage location of the stored lifetime value.

[0025] In some specific implementations of the example method, the component lifetime model may include a physics-based model.

[0026] In some specific implementations of the example method, the component lifetime model may include a machine learning neural network trained to output a set of associations between input operational characteristics and one or more failure modes of the component.

[0027] In some specific implementations of the example method, the component may be an active electrical component. In some specific implementations of the example method, one or more failure modes may include at least one of the following: (i) bond wire detachment, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

[0028] In some specific implementations of the example method, the example method may include, for each of a plurality of components of a subsystem: obtaining corresponding operational data describing one or more corresponding operational characteristics of the corresponding component among the plurality of components; and using a corresponding component lifetime model and based on the corresponding operational data to generate a corresponding component lifetime value. In some specific implementations of the example method, the example method may include using a vehicle usage model and based on multiple corresponding component lifetime values ​​to generate a prediction of the subsystem.

[0029] In some specific implementations of the example method, the vehicle usage model may include a machine learning neural network trained to generate effective lifespan values ​​based on latent embeddings of usage patterns.

[0030] In some specific implementations of the example method, the example method may include obtaining usage patterns from a usage pattern database, which is stored in the database in association with a specific user of the vehicle.

[0031] In some specific implementations of the example method, the vehicle usage model may include a machine learning neural network trained to identify anomalous usage patterns of the vehicle based on one or more operational characteristics. In some specific implementations of the example method, the method may include using the vehicle usage model to identify anomalous usage patterns outside the domain of the component lifetime model. In some specific implementations of the example method, the method may include leveraging predictive overlay component lifetime values ​​based on the identified anomalous usage patterns.

[0032] In some specific implementations of the example method, initiating a corrective action based on a prediction may include: initiating a control signal configured to cause the vehicle to present a warning message to the occupants of the vehicle.

[0033] In some specific implementations of the example method, initiating a corrective action based on a prediction may include sending a message instructing the prediction to a remote server.

[0034] In one example aspect, this disclosure provides one or more example non-transitory computer-readable media storing instructions. The instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to obtain operational data describing one or more operational characteristics of components of a subsystem on a vehicle. The instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to generate a component lifetime value for a component using a component lifetime model and based on the operational data. The instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to generate a prediction of a component using a vehicle usage model and based on the component lifetime value, wherein the vehicle usage model is configured to evaluate the component lifetime value based on usage patterns associated with the vehicle. The instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to initiate corrective actions based on the prediction to mitigate component degradation.

[0035] In some specific implementations of one or more example nontransitory computer-readable media, the prediction may include the effective lifetime value of a component.

[0036] In some specific implementations of one or more example nontransitory computer-readable media, the instructions of one or more example nontransitory computer-readable media may be executable by control circuitry to obtain an initial effective lifetime value.

[0037] In some specific implementations of one or more example nontransitory computer-readable media, the instructions of one or more example nontransitory computer-readable media may be executable by control circuitry to generate a component lifetime value for the component using a component lifetime model and based on operating data and an initial effective lifetime value.

[0038] In some specific implementations of one or more example nontransitory computer-readable media, the instructions of one or more example nontransitory computer-readable media may be executable by control circuitry to iteratively update the storage location of the stored lifetime value.

[0039] In some specific implementations of one or more example nontransitory computer-readable media, the component lifetime model may include a physical-based model.

[0040] In some specific implementations of one or more example nontransitory computer-readable media, the component lifetime model may include a machine learning neural network trained to output a set of associations between input operational characteristics and one or more failure modes of the component.

[0041] In some specific embodiments of one or more example nontransitory computer-readable media, the component may be an active electrical component. In some specific embodiments of one or more example nontransitory computer-readable media, one or more failure modes may include at least one of: (i) bond wire detachment, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

[0042] In some specific embodiments of the example one or more example non-transitory computer-readable media, the instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to: obtain corresponding operational data describing one or more corresponding operational characteristics of the corresponding component among the plurality of components; and generate a corresponding component lifetime value for the corresponding component using a corresponding component lifetime model and based on the corresponding operational data. In some specific embodiments of the example one or more example non-transitory computer-readable media, the instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to generate a prediction of the subsystem using a vehicle usage model and based on the lifetime values ​​of multiple corresponding components.

[0043] In some specific implementations of one or more example nontransitory computer-readable media, the vehicle usage model may include a machine learning neural network trained to generate effective lifetime values ​​based on latent embeddings of usage patterns.

[0044] In some specific implementations of one or more example nontransitory computer-readable media, the instructions of one or more example nontransitory computer-readable media may be executable by control circuitry to obtain usage patterns from a usage pattern database, which is stored in the database in association with a specific user of the vehicle.

[0045] In some specific embodiments of one or more example non-transitory computer-readable media, the vehicle usage model may include a machine learning neural network trained to identify anomalous usage patterns of the vehicle based on one or more operational characteristics. In some specific embodiments of one or more example non-transitory computer-readable media, the instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to use the vehicle usage model to identify anomalous usage patterns outside the domain of a component lifetime model. In some specific embodiments of one or more example non-transitory computer-readable media, the instructions of the one or more example non-transitory computer-readable media may be executable by control circuitry to utilize a predictive overlay component lifetime value based on the identified anomalous usage patterns.

[0046] In some specific implementations of one or more example nontransitory computer-readable media, initiating a corrective action based on a prediction may include: initiating a control signal configured to cause the vehicle to present a warning message to the occupants of the vehicle.

[0047] In some specific implementations of one or more example nontransitory computer-readable media, initiating a corrective action based on a prediction may include sending a message instructing the prediction to a remote server.

[0048] Other exemplary aspects of this disclosure relate to other systems, methods, vehicles, apparatuses, tangible non-transitory computer-readable media, and devices for improving the operation of a vehicle and the computational efficiency associated with the vehicle.

[0049] These and other features, aspects, and advantages of the various embodiments will become more readily understood with reference to the following description and the appended claims. The accompanying drawings, which are incorporated in and form a part of this specification, illustrate specific embodiments of the disclosure and, together with the description, serve to explain the relevant principles. Attached Figure Description

[0050] The specification provides a detailed discussion of specific implementations applicable to those skilled in the art, with reference to the accompanying drawings, in which:

[0051] Figure 1 An example computing ecosystem is illustrated in the example implementation of this disclosure;

[0052] Figure 2 A schematic diagram illustrating an example data processing pipeline specifically implemented according to the present disclosure is shown;

[0053] Figure 3A A schematic diagram illustrating an example model configuration specifically implemented according to the present disclosure is shown;

[0054] Figure 3B A schematic diagram illustrating an example model configuration specifically implemented according to the present disclosure is shown;

[0055] Figure 4 A schematic diagram illustrating an example data processing pipeline specifically implemented according to the present disclosure is shown;

[0056] Figure 5 A schematic diagram illustrating an example data processing pipeline specifically implemented according to the present disclosure is shown;

[0057] Figure 6 A schematic diagram illustrating an example data processing pipeline specifically implemented according to the present disclosure is shown;

[0058] Figure 7A flowchart illustrating an example method for implementing an example specific embodiment of this disclosure is shown; and

[0059] Figure 8 A block diagram illustrating an example computing system specifically implemented according to the present disclosure is shown. Detailed Implementation

[0060] One aspect of this disclosure relates to machine learning-based vehicle component health monitoring. The operational health status of vehicle components can be a useful metric for assessing the current condition of the vehicle, planning future maintenance needs, or diagnosing past component failures. Advantageously, example embodiments of this disclosure can use a machine learning data processing pipeline to analyze real-time operating characteristics of the vehicle to estimate the current operational health status for determining when maintenance may be necessary. For example, embodiments of this disclosure can analyze the vehicle's lifespan during driving or control states when the control unit is active and current / voltage is present in the system. The machine learning data processing pipeline can leverage vehicle usage data to customize predicted maintenance intervals based on historical usage patterns. The machine learning data processing pipeline can utilize a physics-based component model that provides initial estimates that can be improved using machine learning models.

[0061] For example, a vehicle may include electrical equipment or subsystems, such as an inverter. The inverter may contain numerous components. Recorded operational data can characterize the components. Operational data can characterize the components locally (e.g., recording measurements of the component itself, the circuit in which the component resides, etc.), globally (e.g., recording measurements of the vehicle as a whole, a sub-assembly of the vehicle, or a larger subsystem containing the inverter, etc.), or at any other scale. For example, operational data may include voltage measurements across the component, current measurements within the component, acceleration experienced by the component, cycle counts of operations performed by the component, or other operational characteristics.

[0062] To estimate the health status of an inverter, an example machine learning data processing pipeline can process operational data to evaluate each component. A component lifetime model can process the operational data and output a component lifetime value. A component lifetime model can be configured to estimate the health status (e.g., remaining useful life) of a specific component or component type. A component lifetime model can be or include machine learning models, physics-based analysis models, or combinations thereof. Different component lifetime models can correspond to different component types. For example, a capacitor component lifetime model can process operational data monitoring capacitors to evaluate the capacitor's health status, such as its remaining useful life. A semiconductor component lifetime model can process operational data monitoring semiconductors to evaluate the semiconductor's health status, such as its remaining useful life. In this way, for example, a component lifetime model can assess the health status of the inverter's components.

[0063] A data processing pipeline can improve the output of a component lifetime model to obtain a more accurate inverter health estimate. The pipeline can use a vehicle usage model to process the component lifetime model's output to estimate the effective lifetime achievable if the vehicle is used according to its intended usage patterns. For example, the vehicle usage model can process vehicle usage data (e.g., associated with the vehicle, associated with the vehicle's user account, etc.) indicating driving style, time, duration, trip type, etc., to predict the inverter's degradation trajectory. For example, a particular driver might suddenly accelerate and brake, or frequently carry heavy loads. Various demands on the vehicle due to additional electrical or mechanical loads on the inverter can shorten the component's lifespan. Furthermore, a given remaining lifetime estimate for a component may be implemented differently in actual usage scenarios. For example, the inverter (or its components) might withstand N trips for the first light-load user or 0.5N trips for the second heavy-load user. Therefore, the vehicle usage model can improve the component lifetime values ​​applied in different usage scenarios to obtain predictions of the inverter's application lifetime.

[0064] Based on this prediction, the data processing pipeline can initiate corrective actions to mitigate inverter degradation. Corrective actions may include adjusting driving modes or other driving parameters to reduce the load placed on the inverter (e.g., electrical load, mechanical load, etc.). Corrective actions may also include triggering notifications or alarms regarding the health status of components. In this way, for example, wear and tear can be mitigated by promoting reduced demand and timely maintenance.

[0065] The technology disclosed herein provides numerous technical benefits and improvements to transportation vehicles and computing technologies. For example, computing systems can utilize machine learning data processing pipelines to increase overall vehicle system uptime, reduce vehicle system downtime, detect and mitigate natural component wear and degradation, etc. Systems (such as computing systems) can utilize machine learning data processing pipelines to detect symptoms leading to component failure or damage and mitigate such degradation before it occurs. In this way, for example, the exemplary implementations can improve vehicle performance and reliability.

[0066] In some specific implementations, the computing system can utilize both physics-based analytical models and machine learning-based empirical models, thereby leveraging the advantages of each. In this way, for example, a vehicle computing system can provide improved monitoring capabilities. As vehicles become increasingly complex machines with advanced electromechanical systems, the improved vehicle computing systems according to exemplary embodiments of this disclosure can facilitate further development in the long-term care of such vehicles. In this way, for example, the exemplary embodiments can increase vehicle lifespan and avoid the depletion of natural resources in premature vehicle replacement. In this way, for example, the exemplary embodiments can advance the field of vehicles and vehicle computing systems as a whole.

[0067] For example, exemplary embodiments of this disclosure can help avoid unnecessary maintenance by more intelligently predicting the remaining useful life of components. By avoiding unnecessary maintenance, fewer components can be used, thereby reducing the carbon footprint associated with the use of the vehicle. For example, exemplary embodiments of this disclosure can help avoid catastrophic component failures by more intelligently monitoring the health of other components and alerting users to address maintenance needs before such catastrophic failures occur. By avoiding catastrophic component failures, fewer components can be used, and fewer vehicles can be deemed total loss, scrapped, or withdrawn from service, thereby increasing the overall lifespan of the vehicle. This, in turn, can spread the vehicle's carbon footprint over a longer period, thereby effectively reducing emissions per mile per hour.

[0068] Furthermore, the example implementation can facilitate increased efficiency for vehicles under service. The machine learning data processing pipeline according to this disclosure can generate advance notifications of component wear and degradation. This notification can be sent to a remote computing system associated with a maintenance network (e.g., a dealer network). The remote computing system can use this notification to pre-order components necessary for repair or otherwise addressing component degradation. Therefore, the remote computing system can have sufficient notifications to time service access for vehicles, thereby optimizing the use of service center resources. Consequently, service center resources (e.g., site space, equipment, workshop consumables, utilities, etc.) can be deployed more efficiently, reducing waste. In this way, for example, the example implementation can advance the field of vehicle repair and maintenance.

[0069] Reference will now be made in detail to the embodiments, one or more examples of which are illustrated in the accompanying drawings. The exemplary features are described to illustrate various possible embodiments and are not intended to limit this disclosure. Indeed, it will be apparent to those skilled in the art that various modifications and variations can be made to the embodiments without departing from the scope or spirit of this disclosure. For example, a function illustrated or described as part of one embodiment may be used with another embodiment to produce yet another embodiment. Therefore, aspects of this disclosure are intended to cover such modifications and variations.

[0070] The techniques disclosed herein may include such collection where the user explicitly authorizes the collection of user-associated data. Such authorization may be provided by the user in response to a prompt explicitly requesting such authorization, via explicit user input to a user interface. The collected data may be anonymized, pseudonymous, encrypted, noise-added, securely stored, or otherwise protected. The user may opt out of such data collection at any time.

[0071] Figure 1 An example computing ecosystem 100 according to one embodiment of this disclosure is illustrated. Ecosystem 100 may include a vehicle 105, a remote computing platform 110 (also referred to herein as computing platform 110), and a user device 115 associated with a user 120. User 120 may be a driver of the vehicle. In some specific embodiments, user 120 may be a passenger of the vehicle. Vehicle 105, computing platform 110, and user device 115 may be configured to communicate with each other via one or more networks 125.

[0072] Systems / devices within ecosystem 100 can communicate using one or more application programming interfaces (APIs). This can include externally-facing APIs to transfer data from one system / device to another. Externally-facing APIs allow systems / devices to establish secure communication channels via secure access channels on network 125 using any number of methods, such as network-based methods, programmatic access via RESTful APIs, Simple Object Access Protocol (SOAP), Remote Procedure Call (RPC), script access, etc.

[0073] Computing platform 110 may include a computing system located remotely from vehicle 105. In one embodiment, computing platform 110 may include a cloud-based server system. Computing platform 110 may include one or more backend services for supporting vehicle 105. These services may include, for example, remote collaboration services, navigation / route planning services, performance monitoring services, etc. Computing platform 110 may host or otherwise include one or more APIs for communicating data to / from computing system 130 or user equipment 115 of vehicle 105.

[0074] The computing platform 110 may include one or more computing devices. For example, the computing platform 110 may include control circuitry 185 and non-transitory computer-readable medium 190 (e.g., memory). The control circuitry 185 of the computing platform 110 may be configured to perform the various operations and functions described herein. In one embodiment, the control circuitry 185 may include one or more processors (e.g., microprocessors), one or more processing cores, programmable logic circuitry (PLC) or programmable logic / gate array (PLA / PGA), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), or any other control circuitry.

[0075] In one embodiment, control circuitry 185 may be programmed by one or more computer-readable or computer-executable instructions stored on non-transitory computer-readable medium 190. In one embodiment, non-transitory computer-readable medium 190 may be a memory device (also referred to as a data storage device), which may include electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. Non-transitory computer-readable medium 190 may be formed, for example, a hard disk drive (HDD), a solid-state drive (SDD) or solid-state integrated memory, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), dynamic random access memory (DRAM), portable optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD), and / or memory stick. In some cases, non-transitory computer-readable medium 190 may store computer-executable or computer-readable instructions, such as instructions for performing the operations and methods described herein.

[0076] In various implementations, the terms "computer-readable instructions" and "computer-executable instructions" are used to describe software instructions or computer code configured to perform various tasks and operations. In various implementations, if the computer-readable instructions or computer-executable instructions form a module, the term "module" broadly refers to a collection of software instructions or code configured to cause control circuitry 185 to perform one or more functional tasks. When control circuitry or other hardware components are executing a module or computer-readable instructions, the module and computer-readable / executable instructions can be described as performing various operations or tasks.

[0077] User equipment 115 may include computing devices owned or otherwise accessible to user 120. For example, user equipment 115 may include a telephone, laptop computer, tablet computer, wearable device (e.g., smartwatch, smart glasses, headset), personal digital assistant, gaming system, personal desktop device, other handheld device, or other types of mobile or non-mobile user equipment. As further described herein, user equipment 115 may include one or more input components, such as buttons, touchscreens, joysticks or other cursor controls, styluses, microphones, cameras or other imaging devices, motion sensors, etc. User equipment 115 may include one or more output components, such as display devices (e.g., displays), speakers, etc. In one embodiment, user equipment 115 may include components (such as, for example, touchscreens) configured to perform input and output functions to receive user input and present information to user 120. User equipment 115 may execute one or more instructions to run instances of software applications and present user interfaces associated with them. Launching a software application on a suitable transport platform can initiate a user network session with computing platform 110.

[0078] Network 125 can be any type of network or combination of networks that enables communication between devices. In some implementations, network 125 may include one or more of a local area network, wide area network, Internet, secure network, cellular network, mesh network, point-to-point communication link, or some combination thereof, and may include any number of wired or wireless links. Communication over network 125 may be accomplished via a network interface, for example, using any type of protocol, protection scheme, encoding, format, encapsulation, etc. Communication between computing system 130 and user equipment 115 may be facilitated by near field communication or short-range communication technologies (e.g., Bluetooth Low Energy protocol, radio frequency signaling, NFC protocol).

[0079] Vehicle 105 can be a vehicle operable by user 120. In one embodiment, vehicle 105 can be a car or another type of land-based vehicle manually driven by user 120. For example, vehicle 105 could be a Mercedes-Benz. ® A car or van. In some embodiments, vehicle 105 may be an aircraft (e.g., a private plane) or a waterborne vehicle (e.g., a boat). Vehicle 105 may include operator assistance features such as cruise control, advanced driver assistance systems, etc. In some embodiments, vehicle 105 may be a fully autonomous vehicle or a semi-autonomous vehicle.

[0080] Vehicle 105 may include a powertrain and one or more power sources. The powertrain may include motors (e.g., internal combustion engines, electric motors, or hybrids thereof), electric motors (e.g., electric motors), transmissions (e.g., automatic transmissions, manual transmissions, continuously variable transmissions), drive shafts, axles, differentials, electronic components, gears, etc. Power sources may include one or more types of power sources. For example, vehicle 105 may be a fully electric vehicle (EV) capable of using batteries to operate the powertrain (e.g., for propulsion) and onboard functions of vehicle 105. In one embodiment, vehicle 105 may use combustible fuel. In one embodiment, vehicle 105 may include a hybrid power source, such as, for example, a combination of combustible fuel and electricity.

[0081] Vehicle 105 may include a vehicle interior. The vehicle interior may include areas within the vehicle body of vehicle 105, including, for example, the user compartment of vehicle 105. The vehicle interior may include seats for users, steering mechanisms, accelerator interfaces, brake interfaces, etc. The vehicle interior may include display devices, such as displays associated with an infotainment system. Such components may be referred to as display devices for an infotainment system, or may be considered as devices for implementing embodiments including the use of an infotainment system. For illustrative and exemplary purposes, such components may herein be referred to as head unit display devices (e.g., located in the front area / dashboard area of ​​the vehicle interior), rear unit display devices (e.g., located in the rear passenger area of ​​the vehicle interior), infotainment head unit, or rear unit, etc.

[0082] The display device can display various content to user 120, including information about vehicle 105, prompts for user input, etc. The display device may include a touchscreen through which user 120 provides user input to the user interface. The display device may be associated with an audio input device (e.g., a microphone) for receiving audio input from user 120. In some embodiments, the display device may be used as a dashboard for vehicle 105.

[0083] The interior of the vehicle 105 may include one or more lighting elements. The lighting elements may be configured to emit light in various colors, brightness levels, etc.

[0084] Vehicle 105 may include a vehicle exterior. The vehicle exterior may include the outer surface of vehicle 105. The vehicle exterior may include one or more lighting elements (e.g., headlights, brake lights, high beams). Vehicle 105 may include one or more doors for accessing the vehicle interior by, for example, operating a door handle on the vehicle exterior. Vehicle 105 may include one or more windows, including windshields, door windows, passenger windows, rear windows, sunroofs, etc.

[0085] For the sake of brevity, certain routines and conventional components of vehicle 105 (e.g., engine) are not illustrated and / or discussed herein. Those skilled in the art will understand the operation of conventional vehicle components in vehicle 105.

[0086] Vehicle 105 may include a computing system 130 on vehicle 105. The computing system 130 may be on vehicle 105 because it is included on or within vehicle 105. The computing system 130 may include one or more computing devices, which may include various computing hardware components. For example, the computing system 130 may include control circuitry 135 and a non-transitory computer-readable medium 140 (e.g., memory). The control circuitry 135 may be configured to perform various operations and functions for implementing the techniques described herein.

[0087] In one embodiment, control circuitry 135 may include one or more processors (e.g., microprocessors), one or more processing cores, programmable logic circuitry (PLC) or programmable logic / gate array (PLA / PGA), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), or any other control circuitry. In some specific embodiments, control circuitry 135 and / or computing system 130 may be part of, or may form part of, a vehicle control unit (also referred to as a vehicle controller), which is embedded in or otherwise disposed in vehicle 105 (e.g., Mercedes-Benz). ® In a car or van. For example, the vehicle controller may be or may include an infotainment system controller (e.g., infotainment head unit), telematics control unit (TCU), electronic control unit (ECU), central powertrain controller (CPC), charging controller, central external and internal controller (CEIC), zone controller or any other controller (the terms “or” and “and / or” are used interchangeably herein).

[0088] In one embodiment, control circuitry 135 may be programmed by one or more computer-readable or computer-executable instructions stored on non-transitory computer-readable medium 140. In one embodiment, non-transitory computer-readable medium 140 may be a memory device (also referred to as a data storage device), which may include electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. Non-transitory computer-readable medium 140 may be formed, for example, a hard disk drive (HDD), a solid-state drive (SDD) or solid-state integrated memory, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), dynamic random access memory (DRAM), portable optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD), and / or memory stick. In some cases, non-transitory computer-readable medium 140 may store computer-executable or computer-readable instructions, such as instructions for performing the methods of Figures 10A through 10B and 11. Additionally or alternatively, similar instructions may be stored in computing platform 110 (e.g., non-transitory computer-readable medium 190) and made available via network 125.

[0089] The computing system 130 (e.g., control circuitry 135) may be configured to communicate with other components of the vehicle 105 via a communication channel. The communication channel may include one or more data buses (e.g., Controller Area Network (CAN)), on-board diagnostic connectors (e.g., OBD-II), or a combination of wired or wireless communication links. The on-board systems may transmit or receive data, messages, signals, etc., from each other via the communication channel.

[0090] In one implementation, the communication channel may include a direct connection, such as a connection provided via a dedicated wired communication interface (such as an RS-232 interface, a Universal Serial Bus (USB) interface) or via a local computer bus (such as a Peripheral Component Interconnect (PCI) bus). In one implementation, the communication channel may be provided via a network. The network may be any type or form of network, such as a Personal Area Network (PAN), Local Area Network (LAN), Intranet, Metropolitan Area Network (MAN), Wide Area Network (WAN), or the Internet. The network may utilize different technologies and protocol layers or protocol stacks, including, for example, Ethernet protocol, Internet Protocol Suite (TCP / IP), ATM (Asynchronous Transfer Mode) technology, SONET (Synchronous Optical Networking) protocol, or SDH (Synchronous Digital Hierarchy) protocol.

[0091] In one implementation, the systems / equipment of vehicle 105 may communicate via an intermediate storage device or, more generally, via an intermediate non-transitory computer-readable medium. For example, a non-transitory computer-readable medium 140, which may be located outside the computing system 130, may act as an external buffer or repository for storing information. In such an example, the computing system 130 may retrieve or otherwise receive information from the non-transitory computer-readable medium 140.

[0092] Vehicle 105 may include one or more human-machine interfaces (HMIs) 145. HMI 145 may include display devices as described herein. The display devices (e.g., touchscreens) may be viewable by users of vehicle 105 located at the front of vehicle 105 (e.g., driver's seat, front passenger seat) (e.g., user 120, second user 175). Additionally or alternatively, the display devices (e.g., rear unit) may be viewable by users located at the rear of vehicle 105 (e.g., rear passenger seat).

[0093] Vehicle 105 may include one or more sensors 150. Example sensors 150 may include, for example, oxygen sensors, air flow sensors, air pressure sensors, air temperature sensors, coolant temperature sensors, oil temperature sensors, oil pressure sensors, crankshaft position sensors, camshaft position sensors, knock sensors, transmission temperature sensors, motor torque sensors, motor speed sensors, throttle position sensors, brake pedal sensors, brake pressure sensors, wheel speed sensors, steering angle sensors, steering torque sensors, rain sensors, light sensors, tire pressure sensors, exhaust temperature sensors, suspension position sensors, airbag sensors, seatbelt sensors, occupancy sensors, inertial measurement units, battery charging sensors, battery temperature sensors, and other temperature sensors (e.g., junction temperature), voltage probes (e.g., input / output voltage), current probes (e.g., input / output current), strain gauges, optical sensors, etc. Any type of data source can be used. Electrical components may report input voltage (AC or DC), input current (AC or DC), output voltage (AC or DC), output current (AC or DC), etc.

[0094] Sensor 150 may be configured to acquire sensor data. This may include sensor data associated with the surrounding environment of vehicle 105, sensor data associated with the interior of vehicle 105, or sensor data associated with a specific vehicle function. Sensor data may indicate conditions observed inside, outside, or in the surrounding environment of the vehicle. For example, sensor data may acquire image data, internal / external temperature data, weather data, data indicating the position of a user / object inside vehicle 105, weight data, motion / gesture data, audio data, or other types of data. Sensor 150 may include one or more of the following: a camera (e.g., a visible spectrum camera, an infrared camera), a motion sensor, an audio sensor (e.g., a microphone), a weight sensor (e.g., for a vehicle seat), a temperature sensor, a humidity sensor, a light detection and ranging (LIDAR) system, a radio detection and ranging (RADAR) system, or other types of sensors. Vehicle 105 may also include other sensors configured to acquire data associated with vehicle 105. For example, vehicle 105 may include an inertial measurement unit, a tire odometer, or other sensors.

[0095] Vehicle 105 may include a positioning system 155. Positioning system 155 may be configured to generate location data (also referred to as location data) indicating the position (also referred to as location) of vehicle 105. For example, positioning system 155 may determine location using one or more of the following methods: using inertial sensors (e.g., inertial measurement units, etc.), satellite positioning systems; based on IP addresses; using triangulation and / or proximity to network access points or other network components (e.g., cell towers, WiFi access points, etc.); or other suitable technologies. Positioning system 155 may determine the current location of vehicle 105. Location may be represented as a set of coordinates (e.g., latitude, longitude), addresses, semantic locations (e.g., "at work"), etc.

[0096] In one implementation, positioning system 155 may be configured to locate vehicle 105 within the environment in which it is located. For example, vehicle 105 may access map data that provides detailed information about its surrounding environment. The map data may provide information about: the identification and location of different roads, road segments, buildings, or other objects; the location and direction of driving lanes (e.g., parking lanes, turning lanes, bicycle lanes, or other lanes within a specific road); traffic control data (e.g., the location, timing, or instructions of signs (e.g., stop signs, yield signs), traffic lights (e.g., stop lights) or other traffic signals or control devices / markers (e.g., pedestrian crossings); or any other data. Positioning system 155 may locate vehicle 105 within the environment (e.g., across multiple axes) based on map data. For example, positioning system 155 may process sensor data (e.g., LiDAR data, camera data, etc.) to match it with a map of the surrounding environment to determine the vehicle's location within that environment. The determined location of vehicle 105 can be used by various systems of computing system 130 or provided to computing platform 110.

[0097] Vehicle 105 may include a communication system 160 configured to allow vehicle 105 (and its computing system 130) to communicate with other computing devices. Computing system 130 may use communication system 160 to communicate with computing platform 110 or one or more other remote computing devices via network 125 (e.g., via one or more wireless signal connections). In some implementations, communication system 160 may allow communication between one or more systems on vehicle 105.

[0098] In one embodiment, the communication system 160 may be configured to allow the vehicle 105 to communicate with or otherwise receive data from the user equipment 115. The communication system 160 may utilize various communication technologies, such as, for example, Bluetooth Low Energy protocol, radio frequency signaling, or other short-range or near-field communication technologies. The communication system 160 may include any suitable components for interfacing with one or more networks, including, for example, transmitters, receivers, ports, controllers, antennas, or other suitable components that may facilitate communication.

[0099] Figure 2An example data processing pipeline 200 for vehicle component health monitoring, specifically implemented according to an example of this disclosure, is illustrated. The data processing pipeline 200 can process operational data 202 using a component lifetime model 210 (e.g., optionally including a physics-based model 212, a machine learning model 214, or both). The output of the component lifetime model 210 can provide an initial estimate of the lifetime value of one or more components. The data processing pipeline 200 can input vehicle usage data 220 into a vehicle usage model 230 to improve the output of the component lifetime model 210. The vehicle usage model 230 can generate predictive data 240 to describe the effective lifetime of components in the application context of anticipated vehicle use.

[0100] Operational data 202 may include virtually any measured, inferred, or generated quantity, quality, or other value describing the operation of the system containing the monitored components or describing its environment. For example, operational data 202 may describe the operational characteristics of the external environment of the vehicle (e.g., weather conditions). Operational data 202 may also describe the operational characteristics of the vehicle itself.

[0101] Operational data 202 can describe the global operational characteristics of the vehicle. Global operational characteristics can be those shared throughout the entire vehicle. For example, global operational characteristics may include user account data associated with the user driving the vehicle. Global operational characteristics may include speed or acceleration or other motion measurements associated with a reference point of the vehicle (e.g., a rigid part of the chassis). Global operational characteristics may include the vehicle's power status (e.g., on, off, auxiliary, charging, etc.). Global operational characteristics may include passenger count, load weight, fuel / charging status, mileage, etc.

[0102] Operational data 202 can describe local operating characteristics of various systems, subsystems, or components. Local operating characteristics may include absolute or relative measurements. For example, local operating characteristics may include absolute or relative motion measurements (e.g., defined relative to a reference point of a vehicle). Local operating characteristics may include local temperatures measured near the component of interest. Local operating characteristics may include electrical parameters measured for the component of interest, such as voltage across the component or current through the component, or spectral analysis of electrical signals. Local operating characteristics may include cycle counts or uptime measurements for individual systems, subsystems, or components.

[0103] The computing system 130 can directly collect operational data 202 from the monitored components. The computing system 130 can record signals received from various systems, subsystems, and components of the vehicle. The computing system 130 can record signals received from sensors or probes deployed across the vehicle. These sensors or probes can be dedicated to monitoring system health or can be used in other ways for the normal operation of the vehicle.

[0104] Operational data 202 may include raw data or preprocessed data. Operational data 202 may include filtered data processed to remove spurious or noisy signals. Operational data 202 may include data output from a preprocessing stage (not shown) of the data processing pipeline 200. The preprocessing stage may implement digital or analog filters. The preprocessing stage may implement one or more machine learning models trained to extract relevant parameters of interest from a set of underlying raw data.

[0105] Operational data 202 may include subsampled data. Operational data 202 may be subsampled from a large set of log data. The sampling resolution of operational data 202 can be configured to optimize the processing latency of the real-time implementation of the data processing pipeline 200.

[0106] Example data sources for operational data 202 may include, for example, oxygen sensors, air flow sensors, air pressure sensors, air temperature sensors, coolant temperature sensors, oil temperature sensors, oil pressure sensors, crankshaft position sensors, camshaft position sensors, knock sensors, transmission temperature sensors, motor torque sensors, motor speed sensors, throttle position sensors, brake pedal sensors, brake pressure sensors, wheel speed sensors, steering angle sensors, steering torque sensors, rain sensors, light sensors, tire pressure sensors, exhaust temperature sensors, suspension position sensors, airbag sensors, seat belt sensors, occupancy sensors, inertial measurement units, battery charging sensors, battery temperature sensors, and other temperature sensors (e.g., junction temperature), voltage probes (e.g., input / output voltage), current probes (e.g., input / output current), strain gauges, optical sensors, etc. Any type of data source can be used. Electrical components may report input voltage (AC or DC), input current (AC or DC), output voltage (AC or DC), output current (AC or DC), etc.

[0107] Component lifetime model 210 can process operational data 202 to generate initial lifetime values ​​for one or more components. Component lifetime model 210 can process operational data 202 in a streaming manner in real time, in batches at predetermined intervals, or on demand. Component lifetime model 210 can be executed on a vehicle (e.g., vehicles 105, 180) or remotely.

[0108] The physics-based model 212 may be or include analytical or empirical models for modeling component behavior. For example, the physics-based model 212 may include a fault model based on the Arrhenius model. The physics-based model 212 may include a closed-form analytical thermodynamic model of heat transfer through the component. For example, the physics-based model 212 may include a heat transfer model used to determine local temperatures (e.g., the junction temperature of a transistor) based on temperatures located at temperature sensors located remotely from the junction itself (e.g., on a heatsink, etc.). The physics-based model 212 may include an analytical thermomechanical stress model to calculate stress, strain, and fatigue in the component or its fixtures.

[0109] Example physics-based model 212 may include, for example, semiconductor models for estimating bond wire delamination, time-dependent dielectric breakdown, threshold voltage instability, bias temperature instability, etc. Example physics-based model 212 may include, for example, capacitor models for estimating capacitance drift, parasitic component drift, temperature-driven, leakage current, electrolytic degradation, etc.

[0110] Figure 3A An example configuration of a physics-based model 212 is illustrated. The physics-based model 212 may include multiple models. These multiple models may be associated with corresponding components or component types. For example, the physics-based model 212 may include component type A model 312-A, component type B model 312-B, etc., up to component type N model 312-N for N different component types. These multiple models may be the same or different types of models. Multiple models within the multiple models may be of the same model type (e.g., based on an Arrhenius model), although they may have optionally different parameters. These multiple models may process the same input data or different input data. The corresponding model within the multiple models may optionally handle local operational characteristics associated with the corresponding component being analyzed, taking into account other global operational characteristics.

[0111] Refer again Figure 2 The data processing pipeline 200 can use a physics-based model 212 to obtain various outputs. Example outputs may include component lifetime values. Component lifetime values ​​can be any value, flag, or other indicator of a component's health status. Component lifetime values ​​may include or otherwise indicate an estimate of remaining useful life. Remaining useful life can refer to the estimated amount of time a system or component can continue operating before reaching the end of its useful life. Remaining useful life can be a prediction of the remaining operating time of a system or component based on its current condition and historical performance. Remaining useful life can be output based on time (e.g., hours) or usage (e.g., cycle counts, such as vehicle power cycles, mileage, trip counts, etc.).

[0112] Machine learning model 214 can be or includes any kind of machine learning model trained to process input data to assess component health. Machine learning model 214 can be or includes models trained using supervised or unsupervised techniques. Machine learning model 214 can be or includes sequence-based models (e.g., transformer networks, long short-term memory networks, recurrent neural networks, etc.) for processing the sequence of operational data 202. Machine learning model 214 can be or includes convolutional neural networks for generating feature maps of operational data 202 (e.g., feature maps of rasterized waveforms, etc.). Machine learning model 214 can operate in conjunction with or instead of the physics-based model 212. Machine learning model 214 can infer component health status from operational data 202. Machine learning model 214 can regress estimated lifetime values ​​or generate categorical outputs indicating healthy, warning, or unhealthy states. (See below for reference.) Figure 8 Describe further details of the machine learning model and the training process.

[0113] Figure 3B An example configuration of machine learning model 214 is illustrated. Machine learning model 214 may include multiple models. These multiple models may be associated with corresponding multiple components or component types. For example, machine learning model 214 may include component type A model 314-A, component type B model 314-B, etc., up to component type N model 314-N for N different component types. These multiple models may be models of the same or different types. Multiple models among the multiple models may be of the same model type (e.g., token sequence based, image based, etc.), although they may have parameters that are optionally trained independently. These multiple models may process the same input data or different input data. The corresponding model among the multiple models may optionally process local operational characteristics associated with the corresponding component being analyzed, taking into account other global operational characteristics.

[0114] Refer again Figure 2 Vehicle usage data 220 can describe historical usage patterns of the vehicle. Historical patterns can be extracted from recorded operational data (e.g., previously collected operational data 202). Vehicle usage data 220 can contain any or all of the same data as operational data 202. Vehicle usage data 220 can contain operational data 202 plotted over time. Vehicle usage data 220 can contain scores or other evaluation metrics output by a rating model configured to evaluate operational data 202. Vehicle usage data 220 can be associated with user accounts. For example, vehicle usage data 220 can include data describing usage patterns associated with a specific user of the vehicle.

[0115] The user can be a driver / operator or a passenger. Vehicle usage data 220 may include data describing the use of the vehicle as a whole or the use of any of its components or subsystems. For example, vehicle usage data 220 may include data describing how user A (e.g., user 120) has driven the vehicle in the past (e.g., gently, aggressively, as indicated by acceleration and braking), the time and frequency at which user A tends to drive the vehicle, etc. Vehicle usage data 220 may include data describing how user B (e.g., user 175) has driven the vehicle in the past (e.g., gently, aggressively, as indicated by acceleration and braking), the time and frequency at which user B tends to drive the vehicle, etc. Vehicle usage data 220 may include data describing how passenger user A (e.g., user 120) has adjusted the passenger seat in the past. Vehicle usage data 220 may include data describing how passenger user B (e.g., user 175) has adjusted the passenger seat in the past.

[0116] Vehicle usage data 220 can be embedded in a latent space. Operational data 202 can be processed using a machine learning embedding model to generate multidimensional embeddings representing usage associated with a specific user account. Such embeddings can be stored in association with the user account (e.g., stored on the vehicle, at a remote computing system, etc.). Vehicle usage data 220 can be encrypted to protect the security of user account data.

[0117] Data processing pipeline 200 can obtain vehicle usage data 220 from data storage devices on the vehicle. Data processing pipeline 200 can also obtain vehicle usage data 220 from data storage devices on a remote computing system. For example, when a user uses the vehicle for the first time, the vehicle can check the registry on the remote computing system to obtain any pre-existing vehicle usage data associated with the account. The vehicle can cache any obtained vehicle usage data locally for processing by data processing pipeline 200. The vehicle can periodically update the registry when recording or generating vehicle usage data 220.

[0118] Vehicle usage model 230 can improve the output of component lifespan model 210 based on vehicle usage data 220. For example, component lifespan model 210 can provide an estimate of component lifespan based on a physical model and learning trends associated with the entire category or type of component. Vehicle usage model 230 can refine such estimates to adapt and tailor them to the specific use experienced by the component. For example, component lifespan model 210 can output a measure of remaining useful life in a given unit (e.g., hours, cycles, etc.). However, a given number of hours or cycles can translate differently into meaningful constraints for different usage scenarios. For example, a sample component might be projected by component lifespan model 210 to have 100 cycles remaining before a potential failure. For a usage pattern of only 1 cycle per day (e.g., light use), the prediction might not be as urgent as for a usage pattern of 50 cycles per day (e.g., heavy use). Initiating replacement for the user under light use scenarios could lead to unnecessary or premature component replacement, resulting in unnecessary costs and potential waste of resources. Conversely, under heavy usage, if the user fails to replace the component within the next two days, it may push the component towards failure, so timely replacement may be appropriate. In this way, for example, vehicle model 230 can assess the component's lifespan and determine personalized predictions.

[0119] The vehicle usage model 230 can receive multiple outputs from the component lifetime model 210. For example, the vehicle usage model 230 can receive multiple outputs from the component lifetime model 210, which is associated with multiple components of a subsystem, respectively. The vehicle usage model 230 can refine the outputs of the component lifetime model 210 by calculating the total lifetime value of the subsystem based on the corresponding component lifetime values. For example, for necessary components, the minimum component lifetime value (e.g., minimum remaining useful life) can be selected as the total lifetime value of the subsystem including the corresponding necessary component. For non-necessary components, the total lifetime value can be calculated based on a weighted combination of lifetime values. For example, component lifetime values ​​can be or include the probability of degradation or failure. The corresponding component can be associated with relative cost or other metrics associated with failure. The total lifetime value can be calculated based on a weighted average of cost or other failure metrics according to the failure probability.

[0120] Vehicle usage model 230 can determine effective component life values. Effective component life values ​​may include component life values ​​that have been increased, decreased, or otherwise altered based on vehicle usage data 220.

[0121] The transportation use model 230 may be or include one or more machine learning models. The transportation use model 230 may include a decoder architecture configured to receive latent embeddings of the transportation use data 220 to output personalized predictions. The transportation use model 230 may be or include statistical models, heuristic-based models, or other decision-making algorithms.

[0122] Vehicle usage model 230 may be or include one or more machine learning models trained to recognize anomalous usage patterns. For example, vehicle usage model 230 may be configured to identify usage patterns outside the valid input domain of component lifespan model 210. For example, vehicle usage model 230 may process operational data 202, vehicle usage data 220, or both to identify anomalous patterns. Based on the identified anomalous patterns, vehicle usage model 230 may override component lifespan model 210. For example, detecting an anomalous pattern may indicate that a failure has occurred or is beginning to occur.

[0123] Predictive data 240 may be or include effective lifetime values. Predictive data 240 may be or include the effective remaining lifetime of a subsystem component when it is used by a specific user.

[0124] The predicted data 240 may include mitigation instructions for implementing corrective actions on the vehicle. Mitigation instructions provide instructions to be executed by the vehicle computing system to mitigate potential component failures. For example, mitigation instructions may be configured to cause the vehicle to change its driving mode (e.g., locking "Sport" mode to ensure gentler acceleration, etc.). Mitigation instructions may be configured to cause the vehicle to display warning signals (e.g., warning lights on the dashboard, warning messages transmitted to the user's device, audible warning messages output by the speaker driver, etc.). Mitigation instructions may be imperceptible to the user, such as instructions to change the operating conditions of the components being analyzed by the vehicle computing system.

[0125] Mitigation commands can trigger the uploading of component health status to a remote server. For example, the remote server can manage the inventory and logistics of replacement components and service appointments. The remote server can receive notifications from the vehicle based on mitigation commands. The remote server can update inventory or logistics accordingly based on anticipated demand for the repaired components. The remote server can initiate contact with one or more users associated with the vehicle (e.g., driver, fleet manager, etc.) to alert them to the component's condition. For example, an alert describing predictive data 240 can be issued electronically via mobile application, email, SMS message, etc. Predictive data 240 may include recommended maintenance intervals. Predictive data 240 may include recommended maintenance times or locations based on users' anticipated usage or location of the vehicle.

[0126] Figure 4An example data processing pipeline for iterative health monitoring of vehicle components, specifically implemented according to an example of this disclosure, is illustrated. As an example, the data processing pipeline 200 may operate in a cyclic data stream. Upon initialization (e.g., when the component or vehicle is new), initial state data 302 may provide a starting point for the initial health state (e.g., a lifespan value, such as remaining useful life). This initial state data 302 may be processed by a component lifespan model 210 and a vehicle usage model 230 to obtain updated state data 304. The updated state data 304 may be or include a reduced remaining useful life estimate.

[0127] Component lifetime model 210 and vehicle usage model 230 can be configured to receive initial lifetime values ​​and generate new, reduced lifetime values. For example, in a first phase, component lifetime model 210 can be configured to generate an initial updated lifetime value. In a second phase, vehicle usage model 230 can be configured to generate an effective lifetime value based on the initial updated lifetime value. This effective lifetime value can be based on the initial updated lifetime value applied in the context of a specific usage pattern.

[0128] The updated state data 304 can be passed to the decision box 306. If the updated state data 304 of any particular component remains within a predetermined setpoint (e.g., threshold remaining lifetime), the updated state data 304 can be stored or cached as cached state data 308 for processing in subsequent iterations.

[0129] Decision box 306 can be configured to trigger operations outside of the monitoring loop. If the updated status data 304 is outside a predetermined setpoint, decision box 306 can trigger a correction action instruction 310.

[0130] Corrective action instructions 310 may include mitigation instructions. Mitigation instructions provide instructions to be executed by the vehicle computing system to mitigate potential component failure. For example, mitigation instructions may be configured to cause the vehicle to change its driving mode (e.g., locking "Sport" mode to ensure gentler acceleration, etc.). Mitigation instructions may be configured to cause the vehicle to display warning signals (e.g., warning lights on the dashboard, warning messages transmitted to the user's device, audible warning messages output by the speaker driver, etc.). Mitigation instructions may be imperceptible to the user, such as instructions to cause the vehicle computing system to change the operating conditions of the component being analyzed.

[0131] Figure 5An example data processing pipeline for iterative vehicle component health monitoring, specifically implemented according to an example of this disclosure, is illustrated. For example, decision box 306 can operate on the output of component lifetime model 210. If the initially updated component lifetime value is within a predetermined setpoint, monitoring can continue. Valid component lifetime values ​​can be output by vehicle usage model 230 and stored in a cache for future monitoring iterations. If the initially updated component lifetime value is outside the predetermined setpoint, decision box 306 can initiate a correction action instruction 310.

[0132] Figure 6 An example configuration of component lifetime model 210 is illustrated. One or more machine learning models 214 can monitor operational data for anomalies. If an anomaly is detected outside the validity domain of component lifetime model 210, decision box 602 can directly trigger corrective action instruction 310.

[0133] Similarly, refer again Figures 2 to 5 The data processing pipeline 200 can use the vehicle usage model 230 to identify abnormal usage patterns outside the domain of the component lifetime model 210. Based on the identified abnormal usage patterns, the data processing pipeline 200 can utilize predictive overlay component lifetime values.

[0134] For example, an abnormal condition might be associated with a decrease in confidence or an increase in uncertainty regarding the current lifetime value of a given component. Such a decrease in confidence or increase in uncertainty can reach a point where the component is preferably considered to have a lifetime value below a predetermined setpoint (e.g., as in decision box 306). In the example, corrective action instruction 310 may include instructions configured to maintain operational data 202 within a valid domain. For example, corrective action instruction 310 could cause the vehicle to have a reduced maximum speed, a reduced available motor torque, increased cooling applied to the component, etc.

[0135] Figure 7 A flowchart illustrating an example method 700 for health monitoring of vehicle components according to an example aspect of this disclosure is provided. Method 700 may be a computer-implemented method. Method 700 may be performed by a computing system described with reference to any of the other accompanying drawings. In one embodiment, method 700 may be performed by… Figure 1 The control circuit 135 of the computing system 130 executes the algorithm. One or more portions of method 700 may be implemented as an algorithm on the hardware components of the device described herein (e.g., such as...). Figures 1 to 5 B. Figure 8 (etc.). For example, each step of method 700 can be implemented as an operation / instruction that can be executed by computing hardware.

[0136] Figure 7Elements are illustrated in a particular order for illustrative and discussion purposes. Those skilled in the art will understand, when using the disclosure provided herein, that elements of any of the methods discussed herein can be adapted, rearranged, extended, omitted, combined, or modified in various ways without departing from the scope of this disclosure. Figure 7 The description is based on elements / terms described with reference to other systems and figures for illustrative purposes only and is not intended to be limiting. One or more portions of method 700 may be performed additionally or alternatively by other systems. For example, method 700 may be performed by control circuitry 185 of computing platform 110.

[0137] In one embodiment, method 700 may include step 702, in which the computing system obtains operational data describing one or more operational characteristics of components of a subsystem on a vehicle. For example, the computing system may obtain operational data 202. The operational data may include substantially any measured, inferred, or generated data describing the operational characteristics of the vehicle.

[0138] In one implementation, method 700 may include step 704, in which the computing system uses a component lifetime model and generates a component lifetime value for the component based on operational data. For example, the computing system may input operational data 202 into one or more component lifetime models 210 to generate a component lifetime value for the component. Component lifetime model 210 may include a physics-based model 212. Component lifetime model 210 may include a machine learning model 214. For example, machine learning model 214 may include a machine learning neural network trained to output a set of associations between input operational characteristics and one or more failure modes of the component. For example, failure modes of an example active electrical component may include at least one of the following: (i) bond wire delamination, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability. Each such failure mode can be identified from patterns in operational data 202.

[0139] In one implementation, method 700 may include step 706, in which the computing system uses a vehicle usage model and generates a prediction of the component based on the component lifetime value. The vehicle usage model may be configured to evaluate the component lifetime value based on usage patterns associated with the vehicle. For example, the computing system may input the output of component lifetime model 210 into vehicle usage model 230. Vehicle usage model 230 may generate a prediction of the component based on the received output. Vehicle usage model 230 may generate a prediction of the component based on the received output and vehicle usage data 220. In some specific implementations, the prediction may be the effective lifetime value of a subsystem.

[0140] In some implementations, the vehicle usage model may include a machine learning neural network trained to identify anomalous usage patterns of the vehicle based on one or more operational characteristics (e.g., operational data 202, vehicle usage data 220, etc.). The computing system may use the vehicle usage model 230 to identify anomalous usage patterns outside the domain of the component lifetime model 210. The computing system may then leverage predicted overlay component lifetime values ​​based on the identified anomalous usage patterns.

[0141] In some implementations, the vehicle usage model may include a machine learning neural network trained to generate effective lifespan values ​​based on latent embeddings of usage patterns. The computing system can obtain usage patterns from a usage pattern database associated with specific users of the vehicle.

[0142] In one implementation, method 700 may include step 708, in which the computing system initiates corrective actions based on a prediction to mitigate component degradation. Step 708 may include initiating a control signal configured to cause the vehicle to present a warning message to the vehicle's occupants. Step 708 may include initiating a control signal configured to cause the vehicle to present a warning message to the vehicle's user (e.g., a fleet manager not using the vehicle). Step 708 may include initiating a control signal configured to cause the vehicle to change its driving mode. Step 708 may include initiating a control signal configured to maintain the vehicle within an adjusted operating range, such as by reducing drive torque, engine RPM, maximum speed, etc. Step 708 may include initiating a control signal configured to cause the computing system to send a message indicating the prediction to a remote server.

[0143] In some specific implementations of method 700, the computing system can obtain an initial lifetime value and generate a component lifetime value for the component using component lifetime model 210 and based on operational data 202 and the initial lifetime value. The computing system can iteratively update the storage location of the stored lifetime value (e.g., cached state data 308).

[0144] In some specific implementations of method 700, the computing system can obtain corresponding operational data for each corresponding component among multiple components of the subsystem. This corresponding operational data describes one or more corresponding operational characteristics of the corresponding component among the multiple components. The computing system can generate a corresponding component lifetime value for each corresponding component among the multiple components of the subsystem using a corresponding component lifetime model and based on the corresponding operational data. The computing system can use a vehicle usage model and generate a prediction of the subsystem based on multiple corresponding component lifetime values.

[0145] Figure 8 A block diagram of an example computing system 1 according to one embodiment of the present disclosure is illustrated. System 1 includes a computing system 2 (e.g., a computing system on a vehicle), a server computing system 22 (e.g., a remote computing system, a cloud computing platform), and a training computing system 38, which are communicatively coupled via one or more networks 99.

[0146] The computing system 2 may include one or more computing devices 4 or circuitry. For example, the computing system 4 may include control circuitry 6 and a non-transitory computer-readable medium 8 (also referred to herein as memory). In one embodiment, the control circuitry 6 may include one or more processors (e.g., microprocessors), one or more processing cores, programmable logic circuitry (PLC) or programmable logic / gate array (PLA / PGA), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), or any other control circuitry. In some specific embodiments, the control circuitry 6 may be part of, or may form part of, a vehicle control unit (also referred to as a vehicle controller), which is embedded in or otherwise disposed in a vehicle (e.g., a Mercedes-Benz). ® In a car or van. For example, the vehicle controller may be or may include an infotainment system controller (e.g., an infotainment head unit), a telematics control unit (TCU), an electronic control unit (ECU), a central powertrain controller (CPC), a charging controller, a central external and internal controller (CEIC), a zone controller, or any other controller. In one embodiment, the control circuitry 6 may be programmed by one or more computer-readable or computer-executable instructions stored on a non-transitory computer-readable medium 8.

[0147] In one embodiment, the non-transitory computer-readable medium 8 may be a memory device (also referred to as a data storage device), which may include electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. The non-transitory computer-readable medium 8 may be formed, for example, a hard disk drive (HDD), a solid-state drive (SDD) or solid-state integrated memory, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), dynamic random access memory (DRAM), portable optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD), and / or memory stick.

[0148] Non-transitory computer-readable medium 8 may store information accessible by control circuitry 6. For example, non-transitory computer-readable medium 8 (e.g., a memory device) may store data 10 that can be acquired, received, accessed, written, manipulated, created, and / or stored. Data 10 may include any data or information, such as that described herein. In some embodiments, computing system 2 may acquire data from one or more memories located remotely from computing system 2.

[0149] The non-transitory computer-readable medium 8 may also store computer-readable instructions 12 executable by the control circuitry 6. Instructions 12 may be software written in any suitable programming language or may be implemented in hardware. Instructions may include computer-readable instructions, computer-executable instructions, etc. As described herein, in various embodiments, the terms "computer-readable instructions" and "computer-executable instructions" are used to describe software instructions or computer code configured to perform various tasks and operations. In various embodiments, if the computer-readable instructions or computer-executable instructions form a module, the term "module" broadly refers to a collection of software instructions or code configured to cause the control circuitry 6 to perform one or more functional tasks. When the control circuitry 6 or other hardware components are executing modules or computer-readable instructions, these modules and computer-readable / computer-executable instructions can be described as performing various operations or tasks.

[0150] Instruction 12 may be executed in a logically and / or virtually separate thread on control circuitry 6. For example, non-transitory computer-readable medium 8 may store instruction 12, which, when executed by control circuitry 6, causes control circuitry 6 to perform any of the operations, methods, and / or procedures described herein. In some cases, non-transitory computer-readable medium 8 may store computer-executable instructions or computer-readable instructions, such as those for performing... Figure 7 The instructions of at least a portion of the method.

[0151] In one embodiment, computing system 2 may store or include one or more machine learning models 14. For example, machine learning model 14 may be, or may otherwise include, various machine learning models, including machine learning model 214 or machine learning components of vehicle-use model 230. In one embodiment, machine learning model 14 may include (e.g., for generating data clusters) unsupervised learning models. In one embodiment, machine learning model 14 may include neural networks (e.g., deep neural networks) or other types of machine learning models (including nonlinear and / or linear models). Neural networks may include feedforward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or other forms of neural networks. Some example machine learning models may utilize attention mechanisms such as self-attention. For example, some example machine learning models may include multi-head self-attention models (e.g., transformer models).

[0152] In one implementation, one or more machine learning models 14 may be received from server computing system 22 via network 99, stored in computing system 2 (e.g., non-transitory computer-readable medium 8), and then used by control circuitry 6 or otherwise implemented. In one implementation, computing system 2 may implement multiple parallel instances of a single model.

[0153] Additionally or alternatively, one or more machine learning models 14 may be included in or otherwise stored and implemented by server computing system 22, which communicates with computing system 2 according to a client-server relationship. For example, machine learning model 14 may be implemented by server computing system 22 as part of a web service. Thus, one or more models 14 may be stored and implemented at computing system 2, and / or one or more models 14 may be stored and implemented at server computing system 22.

[0154] The computing system 2 may include one or more communication interfaces 16. Communication interfaces 16 can be used to communicate with one or more other systems. Communication interfaces 16 may include any circuitry, components, software, etc., for communicating via one or more networks (e.g., network 99). In some implementations, communication interfaces 16 may include one or more of the following: a communication controller, receiver, transceiver, transmitter, port, conductor, software, and / or hardware for conveying data / information.

[0155] The computing system 2 may also include one or more user input components 18 for receiving user input. For example, the user input component 18 may be a touch-sensitive component (e.g., a touch-sensitive display or touchpad) that is sensitive to the touch of a user input object (e.g., a finger or stylus). The touch-sensitive component may be used to implement a virtual keyboard. Other example user input components include a microphone, a traditional keyboard, a cursor device, a joystick, or other devices through which the user may provide user input.

[0156] The computing system 2 may include one or more output components 20. Output components 20 may include hardware and / or software for generating content audibly or visually. For example, output components 20 may include one or more speakers, handsets, headphones, mobile phones, etc. Output components 20 may include display devices, which may include hardware for displaying user interfaces and / or messages to a user. As examples, output components 20 may include displays, CRTs, LCDs, plasma screens, touchscreens, TVs, projectors, tablets, and / or other suitable display components.

[0157] Server computing system 22 may include one or more computing devices 24. In one embodiment, server computing system 22 may include one or more server computing devices or otherwise be implemented by one or more server computing devices. In instances where server computing system 22 includes multiple server computing devices, such server computing devices may operate according to a sequential computing architecture, a parallel computing architecture, or some combination thereof.

[0158] Server computing system 22 may include control circuitry 26 and nontransitory computer-readable medium 28 (also referred to herein as memory 28). In one embodiment, control circuitry 26 may include one or more processors (e.g., microprocessors), one or more processing cores, programmable logic circuitry (PLC) or programmable logic / gate array (PLA / PGA), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), or any other control circuitry. In one embodiment, control circuitry 26 may be programmed by one or more computer-readable or computer-executable instructions stored on nontransitory computer-readable medium 28.

[0159] In one embodiment, the non-transitory computer-readable medium 28 may be a memory device (also referred to as a data storage device), which may include electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. The non-transitory computer-readable medium may be formed, for example, a hard disk drive (HDD), a solid-state drive (SDD) or solid-state integrated memory, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), dynamic random access memory (DRAM), portable optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD), and / or memory stick.

[0160] Non-transitory computer-readable medium 28 may store information accessible by control circuitry 26. For example, non-transitory computer-readable medium 28 (e.g., a memory device) may store data 30 that can be acquired, received, accessed, written, manipulated, created, and / or stored. Data 30 may include any data or information, such as that described herein. In some embodiments, server computing system 22 may acquire data from one or more memories located remotely from server computing system 22.

[0161] The non-transitory computer-readable medium 28 may also store computer-readable instructions 32 that can be executed by the control circuitry 26. Instructions 32 may be software written in any suitable programming language or may be implemented in hardware. Instructions may include computer-readable instructions, computer-executable instructions, etc. As described herein, in various embodiments, the terms "computer-readable instructions" and "computer-executable instructions" are used to describe software instructions or computer code configured to perform various tasks and operations. In various embodiments, if the computer-readable instructions or computer-executable instructions form a module, the term "module" broadly refers to a collection of software instructions or code configured to cause the control circuitry 26 to perform one or more functional tasks. When the control circuitry 26 or other hardware components are executing modules or computer-readable instructions, these modules and computer-readable / computer-executable instructions can be described as performing various operations or tasks.

[0162] Instruction 32 may be executed in a logically and / or virtually separate thread on control circuitry 26. For example, non-transitory computer-readable medium 28 may store instruction 32, which, when executed by control circuitry 26, causes control circuitry 26 to perform any of the operations, methods, and / or procedures described herein. In some cases, non-transitory computer-readable medium 28 may store computer-executable instructions or computer-readable instructions, such as those for performing... Figure 7 The instructions of at least a portion of the method.

[0163] Server computing system 22 may store or otherwise include one or more machine learning models 34, including machine learning model 214 or any machine learning component of vehicle-used model 230. Machine learning model 34 may include model 14 stored in computing system 2 or the same model. In one embodiment, machine learning model 34 may include an unsupervised learning model. In one embodiment, machine learning model 34 may include a neural network (e.g., a deep neural network) or other types of machine learning models (including nonlinear and / or linear models). Neural networks may include feedforward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, or other forms of neural networks. Some example machine learning models may utilize attention mechanisms such as self-attention. For example, some example machine learning models may include multi-head self-attention models (e.g., transformer models).

[0164] The machine learning models described in this specification may have various types of input data and / or combinations thereof, representing data that can be used by sensors and / or other systems on a vehicle. Input data may include, for example, latently encoded data (e.g., latent spatial representations of inputs, etc.), statistical data (e.g., data calculated and / or processed from another data source), sensor data (e.g., raw and / or processed data collected by sensors on the vehicle), or other types of data.

[0165] Server computing system 22 may include one or more communication interfaces 36. Communication interfaces 36 can be used to communicate with one or more other systems. Communication interfaces 36 may include any circuitry, components, software, etc., for communicating via one or more networks (e.g., network 99). In some implementations, communication interfaces 36 may include one or more of the following: a communication controller, receiver, transceiver, transmitter, port, conductor, software, and / or hardware for conveying data / information.

[0166] Computing system 2 and / or server computing system 22 can directly train models 14 and 34. Computing system 2 and / or server computing system 22 can train models 14 and 34 via interaction with training computing system 38, which is communicatively coupled through network 99. Training computing system 38 may be separate from server computing system 22, or it may be part of or implemented by server computing system 22. Training computing system 38 may be separate from computing system 2, or it may be part of or implemented by computing system 2.

[0167] The training computing system 38 may include one or more computing devices 40. In one embodiment, the training computing system 38 may include one or more server computing devices or be implemented by one or more server computing devices. In instances where the training computing system 38 includes multiple server computing devices, such server computing devices may operate according to a sequential computing architecture, a parallel computing architecture, or some combination thereof.

[0168] The training computing system 38 may include control circuitry 42 and a non-transitory computer-readable medium 44 (also referred to herein as memory 44). In one embodiment, the control circuitry 42 may include one or more processors (e.g., microprocessors), one or more processing cores, programmable logic circuitry (PLC) or programmable logic / gate array (PLA / PGA), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), or any other control circuitry. In one embodiment, the control circuitry 42 may be programmed by one or more computer-readable or computer-executable instructions stored on the non-transitory computer-readable medium 44.

[0169] In one embodiment, the non-transitory computer-readable medium 44 may be a memory device (also referred to as a data storage device), which may include electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. The non-transitory computer-readable medium may be formed, for example, a hard disk drive (HDD), a solid-state drive (SDD) or solid-state integrated memory, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), dynamic random access memory (DRAM), portable optical disc read-only memory (CD-ROM), digital versatile optical disc (DVD), and / or memory stick.

[0170] Non-transitory computer-readable medium 44 may store information accessible by control circuitry 42. For example, non-transitory computer-readable medium 44 (e.g., a memory device) may store data 46 that can be acquired, received, accessed, written, manipulated, created, and / or stored. Data 46 may include any data or information, such as that described herein. In some embodiments, training computing system 38 may acquire data from one or more memories located remotely from training computing system 38.

[0171] The non-transitory computer-readable medium 44 may also store computer-readable instructions 48 executable by the control circuitry 42. Instructions 48 may be software written in any suitable programming language or may be implemented in hardware. Instructions may include computer-readable instructions, computer-executable instructions, etc. As described herein, in various embodiments, the terms "computer-readable instructions" and "computer-executable instructions" are used to describe software instructions or computer code configured to perform various tasks and operations. In various embodiments, if the computer-readable instructions or computer-executable instructions form a module, the term "module" broadly refers to a collection of software instructions or code configured to cause the control circuitry 42 to perform one or more functional tasks. When the control circuitry 42 or other hardware components are executing modules or computer-readable instructions, these modules and computer-readable / computer-executable instructions can be described as performing various operations or tasks.

[0172] Instruction 48 may be executed in a logically or virtually separate thread on control circuitry 42. For example, non-transitory computer-readable medium 44 may store instruction 48 that, when executed by control circuitry 42, causes control circuitry 42 to perform any of the operations, methods, and / or procedures described herein. In some cases, non-transitory computer-readable medium 44 may store computer-executable instructions or computer-readable instructions, such as those for performing... Figure 7 The instructions of at least a portion of the method.

[0173] The training computing system 38 may include a model trainer 50, which uses various training or learning techniques to train machine learning models 14, 34 stored at the computing system 2 and / or the server computing system 22.

[0174] Model trainer 50 can train models 14 and 34 continuously online. Models 14 and 34 can learn continuously as they are deployed on computing system 2 and server computing system 22. For example, models 14 and 34 may include machine learning model 214 or vehicle usage model 230, which can continuously learn from operational data 202 to detect component degradation, abnormal conditions, etc.

[0175] Model trainer 50 can train models 14 and 34 in batches or offline. For example, model trainer 50 can initially train models 14 and 34 in batches for later online fine-tuning.

[0176] Model trainer 50 can train models 14 and 34 in an unsupervised manner. Therefore, unlabeled data specific to an application or problem domain can be used to effectively train the models (e.g., detecting anomalous usage or operational patterns), which improves the model's performance and adaptability. Model trainer 50 can also train models 14 and 34 in a supervised manner. The model trainer can use reinforcement learning (e.g., providing rewards based on improved reliability, uptime, minimized operational disruption, etc.) to train models 14 and 34.

[0177] Model trainer 50 can train machine learning models 14, 34 in a supervised manner. For example, example machine learning models 14, 34 can be models configured to process voltage waveforms measured at transistor pins (e.g., a sliding window of the waveform's time history) and output labels (e.g., "healthy", "deteriorating", "deteriorated", etc.). Training data used to train example models may include measured waveforms known to be associated with components that are healthy, deteriorating, deteriorated, etc. Comparisons of the example model's outputs when given training inputs can indicate errors or losses. The computational system can update one or more parameters of the example model to train the model to better predict the health status of the device being tested.

[0178] Model trainer 50 can train machine learning models 14 and 34 in an unsupervised manner. For example, example machine learning models 14 and 34 can be models configured to learn to identify anomalous behavior of components. Example machine learning model 214 can be a clustering model configured to continuously learn groupings of observed data points. Data points located at a threshold distance from cluster points can be identified as anomalous.

[0179] In the example, model trainer 50 can backpropagate the loss or reward function through the model being trained to modify the model's parameters (e.g., weights). Whether or not the model's parameters (e.g., weights) are modified, model trainer 50 can continue backpropagating the loss or reward function through the machine learning model. For example, model trainer 50 can perform gradient descent, in which the parameters of the machine learning model can be modified in the direction of the negative gradient of the loss or reward function.

[0180] The model trainer 50 can utilize training techniques such as backpropagation of errors. For example, a loss function can be backpropagated through the model to update one or more parameters of the model (e.g., based on the gradient of the loss function). Various loss functions can be used, such as mean squared error, likelihood loss, cross-entropy loss, hinge loss, and / or various other loss functions. Gradient descent can be used to iteratively update the parameters over multiple training iterations.

[0181] In one implementation, backpropagation of the error may include performing truncated backpropagation over time. The model trainer 50 may perform various generalization techniques (e.g., weight decay, dropout, etc.) to improve the generalization ability of the model being trained. Specifically, the model trainer 50 may train machine learning models 14, 34 based on a set of training data 52.

[0182] Training data 52 may include unlabeled training data for unsupervised training. In the example, training data 52 may include multiple sets of labeled or unlabeled operational data indicating measurement characteristics of components of a vehicle.

[0183] In one implementation, if the user has provided consent / authorization, the training examples can be provided by computing system 2 (e.g., the user's vehicle). Therefore, in such a specific implementation, the model 14 provided to computing system 2 can be trained by training computing system 38 in a personalized manner for model 14.

[0184] Model trainer 50 may include computer logic for providing the desired functionality. Model trainer 50 may be implemented in hardware, firmware, and / or software that controls a general-purpose processor. For example, in one embodiment, model trainer 50 may include a program file stored on a storage device, loaded into memory, and executed by one or more processors. In other specific embodiments, model trainer 50 may include one or more sets of computer-executable instructions stored in a tangible computer-readable storage medium, such as RAM, a hard disk, or an optical or magnetic medium.

[0185] The training computing system 38 may include one or more communication interfaces 54. Communication interfaces 54 can be used to communicate with one or more other systems. Communication interfaces 54 may include any circuitry, components, software, etc., for communication via one or more networks (e.g., network 99). In some implementations, communication interfaces 54 may include one or more of the following: a communication controller, receiver, transceiver, transmitter, port, conductor, software, and / or hardware for conveying data / information.

[0186] One or more networks 99 can be any type of communication network, such as a local area network (e.g., an intranet), a wide area network (e.g., the Internet), or some combination thereof, and may include any number of wired or wireless links. Typically, communication over network 99 may be carried over any type of wired and / or wireless connection using various communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, Secure HTTP, SSL).

[0187] Figure 8An example computing system that can be used to implement this disclosure is illustrated. Other computing systems may also be used. For example, in one embodiment, computing system 2 may include model trainer 50 and training data 52. In such embodiments, models 14, 34 may be trained and used locally at computing system 2. In some embodiments of such embodiments, computing system 2 may implement model trainer 50 to personalize models 14, 34.

[0188] Additional discussion of various implementation schemes

[0189] Implementation scheme 1 relates to a computing system. The computing system may include control circuitry. The control circuitry may be configured to acquire operational data describing one or more operational characteristics of components of a subsystem on a vehicle. The control circuitry may be configured to use a component lifetime model and generate a component lifetime value for the component based on the operational data. The control circuitry may be configured to use a vehicle usage model and generate a prediction of the component based on the component lifetime value, wherein the vehicle usage model is configured to evaluate the component lifetime value based on usage patterns associated with the vehicle. The control circuitry may be configured to initiate corrective actions based on the prediction to mitigate component degradation.

[0190] Implementation scheme 2 includes the computing system described in implementation scheme 1. In this implementation scheme, the prediction includes the effective lifetime value of the component.

[0191] Implementation scheme 3 includes the computing system according to implementation scheme 2. In this implementation scheme, the control circuitry may be configured to: obtain an initial effective lifetime value; and generate the component lifetime value of the component using the component lifetime model and based on the operating data and the initial effective lifetime value.

[0192] Implementation scheme 4 includes the computing system described in implementation scheme 3. In this implementation scheme, the control circuitry can be configured to iteratively update the storage location storing the effective lifetime value.

[0193] Implementation scheme 5 includes a computing system according to any one of implementation schemes 1 to 4. In this implementation scheme, the component lifetime model includes a physics-based model.

[0194] Implementation scheme 6 includes a computing system according to any one of embodiments 1 to 5. In this embodiment, the component lifetime model includes a machine learning neural network trained to output a set of associations between input operational characteristics and one or more failure modes of the component.

[0195] Implementation 7 includes the computing system according to implementation 6. In this implementation, the component is an active electrical component, and the one or more failure modes include at least one of the following: (i) bond wire detachment, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

[0196] Implementation scheme 8 includes a computing system according to any one of implementation schemes 1 to 7. In this implementation scheme, the control circuitry may be configured to: obtain corresponding operational data for each corresponding component among a plurality of components of the subsystem, the corresponding operational data describing one or more corresponding operational characteristics of the corresponding component among the plurality of components; generate a corresponding component lifetime value for the corresponding component using a corresponding component lifetime model and based on the corresponding operational data; and generate a prediction of the subsystem using the vehicle usage model and based on the plurality of corresponding component lifetime values.

[0197] Implementation scheme 9 includes a computing system according to any one of implementation schemes 1 to 8. In this implementation scheme, the vehicle usage model includes a machine learning neural network trained to generate the effective lifespan value based on the latent embeddings of the usage pattern.

[0198] Implementation scheme 10 includes a computing system according to any one of embodiments 1 to 9. In this embodiment, the control circuitry may be configured to obtain the usage patterns from a usage pattern database, the usage patterns being stored in the database in association with a specific user of the vehicle.

[0199] Implementation scheme 11 includes a computing system according to any one of embodiments 1 to 10. In this embodiment, the vehicle usage model includes a machine learning neural network trained to identify abnormal usage patterns of the vehicle based on the one or more operational characteristics, and the control circuitry is configured to: use the vehicle usage model to identify abnormal usage patterns outside the domain of the component lifetime model; and based on the identified abnormal usage patterns, utilize the prediction to overlay the component lifetime value.

[0200] Implementation scheme 12 includes a computing system according to any one of implementation schemes 1 to 11. In this implementation scheme, initiating the correction action based on the prediction includes: initiating a control signal configured to cause the vehicle to present a warning message to the occupants of the vehicle.

[0201] Implementation scheme 13 includes a computing system according to any one of implementation schemes 1 to 12. In this implementation scheme, initiating the correction action based on the prediction includes: sending a message indicating the prediction to a remote server.

[0202] Implementation scheme 14 relates to a method for monitoring the operational health of a vehicle subsystem. The method may be a computer-implemented method. The method may include obtaining operational data describing one or more operational characteristics of components of a subsystem on the vehicle. The method may include using a component lifetime model and generating a component lifetime value based on the operational data. The method may include using a vehicle usage model and generating a prediction of the component based on the component lifetime value, wherein the vehicle usage model is configured to assess the component lifetime value based on usage patterns associated with the vehicle. The method may include initiating corrective actions based on the prediction to mitigate component degradation.

[0203] Implementation scheme 15 includes the method according to implementation scheme 14. In this implementation scheme, the prediction includes the effective lifetime value of the component.

[0204] Implementation scheme 16 includes the method according to implementation scheme 15. In this implementation scheme, the method may include: obtaining an initial effective lifetime value; and generating the component lifetime value of the component using the component lifetime model and based on the operating data and the initial effective lifetime value.

[0205] Implementation scheme 17 includes the method according to implementation scheme 16. In this implementation scheme, the method may include iteratively updating the storage location where the effective lifetime value is stored.

[0206] Implementation scheme 18 includes the method according to any one of implementation schemes 14 to 17. In this implementation scheme, the component lifetime model includes a physics-based model.

[0207] Implementation scheme 19 includes the method according to any one of implementation schemes 14 to 18. In this implementation, the component lifetime model includes a machine learning neural network trained to output a set of associations between input operating characteristics and one or more failure modes of the component.

[0208] Implementation 20 includes the method according to any one of Implementations 14 to 19. In this implementation, the component is an active electrical component, and the one or more failure modes said therein may include at least one of the following: (i) bond wire detachment, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

[0209] Implementation scheme 21 includes the method according to any one of implementation schemes 14 to 20. In this implementation scheme, the method may include: for each corresponding component of a plurality of components of the subsystem, obtaining corresponding operational data, the corresponding operational data describing one or more corresponding operational characteristics of the corresponding component among the plurality of components; and generating a corresponding component lifetime value for the corresponding component using a corresponding component lifetime model and based on the corresponding operational data; and generating a prediction of the subsystem using the vehicle usage model and based on the plurality of corresponding component lifetime values.

[0210] Implementation scheme 22 includes the method according to any one of implementation schemes 14 to 21. In this implementation, the vehicle usage model includes a machine learning neural network trained to generate the effective lifespan value based on the latent embeddings of the usage pattern.

[0211] Implementation scheme 23 includes the method according to any one of implementation schemes 14 to 22. In this implementation scheme, the method may include obtaining the usage pattern from a usage pattern database, the usage pattern being stored in the database in association with a specific user of the vehicle.

[0212] Implementation scheme 24 includes the method according to any one of implementation schemes 14 to 23. In this implementation scheme, the vehicle usage model includes a machine learning neural network trained to identify abnormal usage patterns of the vehicle based on the one or more operational characteristics, and the method may include: using the vehicle usage model to identify abnormal usage patterns outside the domain of the component lifetime model; and using the prediction to overlay the component lifetime value based on the identified abnormal usage patterns.

[0213] Implementation scheme 25 includes the method according to any one of implementation schemes 14 to 24. In this implementation scheme, initiating the corrective action based on the prediction includes: initiating a control signal configured to cause the vehicle to present a warning message to the occupants of the vehicle.

[0214] Implementation scheme 26 includes the method according to any one of implementation schemes 14 to 25. In this implementation scheme, initiating the correction action based on the prediction includes: sending a message indicating the prediction to a remote server.

[0215] Implementation scheme 27 relates to one or more non-transitory computer-readable media storing instructions executable by control circuitry to perform operations. In this implementation scheme, the one or more non-transitory computer-readable media stores instructions executable by control circuitry to obtain operational data describing one or more operational characteristics of components of a subsystem on a vehicle. The one or more non-transitory computer-readable media stores instructions executable by control circuitry to generate a component lifetime value for the component using a component lifetime model and based on the operational data. The one or more non-transitory computer-readable media stores instructions executable by control circuitry to generate a prediction of the component using a vehicle usage model and based on the component lifetime value, wherein the vehicle usage model is configured to evaluate the component lifetime value based on usage patterns associated with the vehicle. The one or more non-transitory computer-readable media stores instructions executable by control circuitry to initiate corrective actions based on the prediction to mitigate component degradation.

[0216] Implementation scheme 28 includes one or more non-transitory computer-readable media as described in implementation scheme 27. In this implementation scheme, the prediction includes the effective lifetime value of the component.

[0217] Implementation scheme 29 includes one or more non-transitory computer-readable media according to implementation scheme 28. In this implementation scheme, the one or more non-transitory computer-readable media stores instructions executable by control circuitry to obtain an initial effective lifetime value; and generates the component lifetime value of the component using the component lifetime model and based on the operating data and the initial effective lifetime value.

[0218] Implementation scheme 30 includes one or more non-transitory computer-readable media as described in implementation scheme 29. In this implementation, the one or more non-transitory computer-readable media stores instructions executable by control circuitry to iteratively update the storage location storing the effective lifetime value.

[0219] Implementation scheme 31 includes one or more non-transitory computer-readable media according to any one of embodiments 27 to 30. In this embodiment, the component lifetime model includes a physics-based model.

[0220] Implementation scheme 32 includes one or more non-transitory computer-readable media according to any one of embodiments 27 to 31. In this embodiment, the component lifetime model includes a machine learning neural network trained to output a set of associations between input operational characteristics and one or more failure modes of the component.

[0221] Implementation scheme 33 includes one or more non-transitory computer-readable media according to implementation scheme 32. In this implementation scheme, the component is an active electrical component, and said one or more failure modes may include at least one of the following: (i) bond wire detachment, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

[0222] Implementation scheme 34 includes one or more non-transitory computer-readable media according to any one of implementation schemes 27 to 33. In this implementation scheme, the one or more non-transitory computer-readable media stores instructions executable by control circuitry to: obtain corresponding operational data for each of a plurality of components of the subsystem, the corresponding operational data describing one or more corresponding operational characteristics of the corresponding component among the plurality of components; and generate a corresponding component lifetime value for the corresponding component using a corresponding component lifetime model and based on the corresponding operational data; and generate a prediction of the subsystem using the vehicle usage model and based on the plurality of corresponding component lifetime values.

[0223] Implementation scheme 35 includes one or more non-transitory computer-readable media according to any one of embodiments 27 to 34. In this embodiment, the vehicle usage model includes a machine learning neural network trained to generate the effective lifetime value based on the latent embeddings of the usage pattern.

[0224] Implementation scheme 36 includes one or more non-transitory computer-readable media according to any one of embodiments 27 to 25. In this embodiment, the one or more non-transitory computer-readable media stores instructions executable by control circuitry to obtain a usage pattern from a usage pattern database associated with a specific user of the vehicle.

[0225] Implementation scheme 37 includes one or more non-transitory computer-readable media according to any one of embodiments 27 to 36. In this embodiment, the vehicle usage model includes a machine learning neural network trained to identify abnormal usage patterns of the vehicle based on the one or more operating characteristics, and the one or more non-transitory computer-readable media stores instructions executable by control circuitry to: use the vehicle usage model to identify abnormal usage patterns outside the domain of the component lifetime model; and based on the identified abnormal usage patterns, utilize the prediction to overlay the component lifetime value.

[0226] Implementation scheme 38 includes one or more non-transitory computer-readable media according to any one of embodiments 27 to 37. In this embodiment, initiating the corrective action based on the prediction includes: initiating a control signal configured to cause the vehicle to present a warning message to the occupants of the vehicle.

[0227] Implementation scheme 39 includes one or more non-transitory computer-readable media according to any one of implementation schemes 27 to 38. In this implementation scheme, initiating the correction action based on the prediction includes sending a message indicating the prediction to a remote server.

[0228] Additional Public Content

[0229] As used herein, adjectives and their possessive forms are intended to be used interchangeably unless the context explicitly states and / or clearly indicates otherwise. For example, where appropriate, “component of a vehicle” and “vehicle component” are used interchangeably. Similarly, words, phrases and other disclosures herein are intended to cover obvious variations and synonyms, even if such variations and synonyms are not explicitly listed.

[0230] This paper discusses technical reference servers, databases, software applications, and other computer-based systems, as well as the actions taken and the information transmitted to and from these systems. The inherent flexibility of computer-based systems allows for a wide variety of possibilities in the configuration, combination, and task and functional division of components. For example, the processes discussed herein can be implemented using a single device or component, or a combination of multiple devices or components. Databases and applications can be implemented on a single system or distributed across multiple systems. Distributed components can operate sequentially or in parallel.

[0231] While the subject matter has been described in detail with respect to various specific example embodiments, each example is provided by way of explanation rather than limitation. Those skilled in the art, upon understanding the foregoing, will readily make changes, modifications, or equivalent treatments to such embodiments. Therefore, this disclosure does not exclude the inclusion of such modifications, modifications, and / or additions to this disclosure, which will be apparent to those of ordinary skill in the art. For example, functionality illustrated or described as part of one embodiment may be used with another embodiment to produce yet another embodiment. Therefore, this disclosure is intended to cover such changes, modifications, and equivalent treatments.

[0232] Various aspects of this disclosure have been described with reference to exemplary embodiments thereof. Many other embodiments, modifications, or variations within the scope and spirit of the appended claims will be apparent to those skilled in the art upon review of this disclosure. Any and all functions of the following claims may be combined or rearranged in any possible manner. Therefore, the scope of this disclosure is by way of example rather than limitation, and this disclosure does not exclude the inclusion of such modifications, variations, or additions to this disclosure, which will be apparent to those skilled in the art. Furthermore, terms are described herein using lists of example elements connected by conjunctions such as “and,” “or,” and “but.” It should be understood that such conjunctions are provided for illustrative purposes only. The terms “or” and “and / or” are used interchangeably herein. A list connected by a particular conjunction such as “or” may, for example, refer to “at least one” or “any combination” of the example elements listed in that list, where “or” is understood to mean “and / or” unless otherwise indicated. Furthermore, terms such as “based on” should be understood to mean “at least partially based on.”

[0233] When using the disclosure provided herein, those skilled in the art will understand that elements of any claim, operation, or process discussed herein can be adapted, rearranged, extended, omitted, combined, or modified in various ways without departing from the scope of this disclosure. Sometimes, letter designations may be used to list elements in the specification or claims for illustrative purposes, and this is not intended to limit them. If letter designations are used, they do not imply a particular order of operations or a particular importance of the listed elements. For example, letter identifiers (such as (a), (b), (c), ..., (i), (ii), (iii), ..., may be used to exemplify operations or different elements in a list. Such identifiers are provided for the reader's convenience and do not indicate a particular order, importance, or priority of steps, operations, or elements. For example, an operation exemplified by list identifiers such as (a), (i), etc., may be performed before, after, or concurrently with another operation exemplified by list identifiers such as (b), (ii).

Claims

1. A computing system for monitoring operational health of a vehicle subsystem, the computing system comprising: a control circuit configured to: obtain operational data describing one or more operational characteristics of a component of a subsystem on a vehicle; generate a component life value for the component using a component life model and based on the operational data; generate a prognosis for the component using a vehicle usage model and based on the component life value, wherein the vehicle usage model is configured to evaluate the component life value based on a usage pattern associated with the vehicle; and initiate a corrective action to mitigate degradation of the component based on the prognosis.

2. The computing system of claim 1, wherein the prognosis comprises an effective life value for the component.

3. The computing system of claim 2, wherein the control circuit is configured to: obtain an initial effective life value; and generate the component life value for the component using the component life model and based on the operational data and the initial effective life value.

4. The computing system of claim 3, wherein the control circuit is configured to: iteratively update a storage location storing the effective life value.

5. The computing system of claim 1, wherein the component life model comprises a physics-based model.

6. The computing system of claim 1, wherein the component life model comprises a machine learning neural network trained to output an association between a set of input operational characteristics and one or more failure modes of the component.

7. The computing system of claim 6, wherein the component is an active electrical component, and wherein the one or more failure modes comprise at least one of: (i) bond wire disconnection, (ii) dielectric breakdown, (iii) threshold voltage instability, or (iv) bias temperature instability.

8. The computing system of claim 1, wherein the control circuit is configured to: for each respective component of a plurality of components of the subsystem: obtaining corresponding operational data, the corresponding operational data describing one or more corresponding operational characteristics of the corresponding one of the plurality of components; and generate a respective component life value for the respective component using a respective component life model and based on the respective operational data; and generate a prognosis for the subsystem using the vehicle usage model and based on the plurality of respective component life values.

9. The computing system of claim 2, wherein the vehicle usage model comprises a machine learning neural network trained to generate the effective life value based on a latent embedding of the usage pattern.

10. The computing system of claim 1, wherein the control circuit is configured to: obtain the usage pattern from a usage pattern database, the usage pattern stored in the database in association with a particular user of the vehicle.

11. The computing system of claim 1, wherein the vehicle usage model comprises a machine learning neural network trained to identify anomalous usage patterns of the vehicle based on the one or more operational characteristics, and wherein the control circuit is configured to: identify, using the vehicle usage model, an anomalous usage pattern outside of a domain of the component life model; and override the component life value with the pre-empt based on identifying the anomalous usage pattern.

12. The computing system of claim 1, wherein initiating the corrective action based on the pre-empt comprises: initiating a control signal configured to cause the vehicle to present a warning message to an occupant of the vehicle.

13. The computing system of claim 1, wherein initiating the corrective action based on the pre-empt comprises: sending a message indicating the pre-empt to a remote server.

14. A method for monitoring operational health of a vehicle subsystem, the method comprising: obtaining operational data describing one or more operational characteristics of a component of a subsystem on a vehicle; generating a component life value for the component using a component life model and based on the operational data; generating a pre-empt for the component using a vehicle usage model and based on the component life value, wherein the vehicle usage model is configured to evaluate the component life value based on usage patterns associated with the vehicle; and initiating a corrective action to mitigate degradation of the component based on the pre-empt.

15. The method of claim 14, wherein the pre-empt comprises an effective life value for the component, and the method comprises: obtaining an initial effective life value; and generating the component life value for the component using the component life model and based on the operational data and the initial effective life value.

16. The method of claim 14, wherein the component life model comprises a machine learning neural network trained to output associations between a set of input operational characteristics and one or more failure modes of the component.

17. The method of claim 14, the method comprising: for each respective component of a plurality of components of the subsystem: obtaining respective operational data describing one or more respective operational characteristics of the respective component of the plurality of components; and generating a respective component life value for the respective component using a respective component life model and based on the respective operational data; and generating a pre-empt for the subsystem using the vehicle usage model and based on the plurality of respective component life values.

18. The method of claim 14, the method comprising: obtaining the usage pattern from a usage pattern database, the usage pattern stored in the database in association with a particular user of the vehicle.

19. The method of claim 14, the method comprising: identify, using the vehicle usage model, an abnormal usage pattern outside of a domain of the component life model, wherein the vehicle usage model comprises a machine learning neural network trained to identify abnormal usage patterns of the vehicle based on the one or more operational characteristics; and utilize the pre-empt based on the identification of the abnormal usage pattern to override the component life value.

20. One or more non-transitory computer-readable media storing instructions executable by control circuitry to: obtain operational data describing one or more operational characteristics of a component of a subsystem on a vehicle; generate, using a component life model and based on the operational data, a component life value for the component; generate, using a vehicle usage model and based on the component life value, a pre-empt for the component, wherein the vehicle usage model is configured to evaluate the component life value based on usage patterns associated with the vehicle; and initiate a corrective action to mitigate degradation of the component based on the pre-empt.

Images