Expected signal range for a vehicle

Abstract

A system and a method for indicating the failure of a vehicle component by identifying anomalies in a vehicle signal are provided. These methods include a model trained on training data, resulting in a trained artificial intelligence model. The trained model is executed by an electronic processor designed to: receive a plurality of input signals correlated with the vehicle signal; analyze the input signals; determine a predicted vehicle signal from the input signals; determine a tolerance band from the input signals; and, if the model is valid, generate a model validity signal.A comparison device is designed to: receive the predicted vehicle signal, the tolerance band, and the model validity signal; receive the measured vehicle signal; and determine whether the measured vehicle signal is an anomaly, based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal.

Description

background

[0001] The arrangements, examples and aspects described herein relate, among other things, to a system and a procedure for determining whether a vehicle signal is within an expected range, using artificial intelligence.

[0002] Detecting anomalies in measurements is a common problem in engine calibration. A direct comparison of measurements is not possible because operating conditions are constantly changing. These conditions include ambient temperature, altitude, driving style, and other vehicle characteristics.

[0003] Current approaches involve the manual inspection of vehicle data, which is prone to false positives. A false positive occurs when an evaluation returns a positive result indicating that no anomaly was detected, when in fact an anomaly did occur. Another approach involves rule-based algorithms, which are also prone to false negatives. A false negative occurs when an evaluation returns a negative result indicating that an anomaly was detected, when in fact no anomaly did occur. Brief description

[0004] The arrangement is designed to detect a failure of a vehicle component by identifying anomalies in measurements using a trained artificial intelligence model.

[0005] An arrangement is designed as a system for indicating the failure of a vehicle component by identifying anomalies in a measured vehicle signal. The system comprises: a memory that stores a trained artificial intelligence model, and an electronic processor coupled to the memory that executes the trained artificial intelligence model. The electronic processor is designed to: retrieve a plurality of input signals correlated with the measured vehicle signal; determine a predicted vehicle signal from the plurality of input signals; determine a tolerance band from the plurality of input signals; determine whether the model is valid, and if so, generate a model validity signal.The system includes a comparator designed to: retrieve the predicted vehicle signal, tolerance band, and model validity signal; retrieve the measured vehicle signal; and determine whether the measured vehicle signal is an anomaly based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal. The system includes an aggregator coupled to the comparator designed to: count and store the times and occurrences of anomalies; and indicate a vehicle component failure based on the number of anomalies over a period of time.

[0006] In another arrangement, a method for indicating a failure of a vehicle component by determining anomalies for a measured vehicle signal involves running a trained artificial intelligence model with an electronic processor.The electronic processor is designed to: retrieve a plurality of input signals correlated with the measured vehicle signal; determine a predicted vehicle signal from the plurality of input signals; determine a tolerance band from the plurality of input signals; determine whether the model is valid, and if so, generate a model validity signal; retrieve the measured vehicle signal; determine whether the measured vehicle signal is an anomaly based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal; count and store the times and occurrences of the anomalies; and indicate a vehicle component failure depending on a number of anomalies over a period of time.

[0007] An arrangement is directed towards a procedure for indicating a vehicle component failure. The procedure includes: retrieving a plurality of input signals correlated with a vehicle signal; determining a predicted vehicle signal from the plurality of input signals using a trained artificial intelligence model; determining a tolerance band from the plurality of input signals; determining whether the model is valid, and if so, generating a model validity signal; retrieving a measured vehicle signal; determining whether the measured vehicle signal is an anomaly based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal; counting and storing the times of occurrences of the anomalies; and indicating a vehicle component failure as a function of the number of anomalies over a period of time. Brief description of the drawings Fig. Figure 1 is a block illustration of a model that receives inputs and provides a predicted vehicle signal, a tolerance band, and a model validity signal, according to some aspects. Fig. Figure 2 illustrates a graph of a predicted vehicle signal for the model of Fig. 1 according to some aspects. Fig. 2A is a block diagram of an electronic operating device 40 for the trained KL model 20 of Fig. 1. Fig. Figure 3 is a block diagram of a system that represents the model of Fig. 1 includes a device connected to a comparator to evaluate for an anomaly in a vehicle signal. Fig. Figure 4 is a block diagram of an electronic vehicle operating unit and sensors connected to it via a bus to provide information regarding exhaust gas recirculation, according to some aspects. Fig. Figure 5 is a block diagram of an electronic vehicle operating unit and sensors connected to it via a bus to provide information regarding boost pressure, according to some aspects. Fig. 6 is a graph that represents the boost pressure arrangement for the arrangement of Fig. The 5 data received corresponds, according to some aspects. Fig. Figure 7 is a block diagram of a model that receives inputs and training data to provide a predicted EGR rate signal, according to some aspects. Fig. Figure 8 is a block diagram of an electronic vehicle operating unit and sensors connected to it via a bus to provide and store information regarding the EGR rate, according to some aspects. Fig. 9 is a graph that represents the EGR rate arrangement for the arrangement of Fig. The 8 data received corresponds, according to some aspects. Fig. Figure 10 is a block diagram of an electronic vehicle operating unit and sensors connected to it via a bus to provide and store information regarding a fuel cell of a fuel cell vehicle, according to some aspects. Detailed description

[0008] Before any arrangements, examples, aspects, or features are explained in detail, it should be understood that these arrangements, examples, aspects, and features are not limited in their application to the details of the construction and arrangement of components set forth in the following description or illustrated in the following drawings. Other arrangements, examples, aspects, and features are possible and can be implemented or executed in various ways.

[0009] Furthermore, it is understood that the language and terminology used herein are for descriptive purposes only and should not be considered restrictive. The terms "attached," "connected," and "coupled" are used broadly and include both direct and indirect attachment, connection, and coupling. Moreover, "connected" and "coupled" are not limited to physical or mechanical connections and couplings and may include electrical connections or couplings, whether direct or indirect. Furthermore, electronic communications and notifications may be made using any known means, including wired connections, wireless connections, etc.

[0010] It should also be noted that a plurality of hardware- and software-based devices, as well as a plurality of other structural components, can be used in various implementations. Aspects, features, and instances may include hardware, software, and electronic components or modules, which, for the purpose of discussion, may be illustrated and described as if the majority of the components were implemented in hardware. However, a person skilled in the art would recognize, based on reading this detailed description, that in at least one instance the electronics-based aspects of the invention may be implemented in software (for example, stored on a non-volatile, computer-readable medium) that is executable by one or more electronic processors.Consequently, it should be noted that a multitude of hardware- and software-based devices, as well as a multitude of other structural components, can be used to implement the invention. For example, the "operating units" and "controllers" described in the patent specification may include one or more electronic processors, one or more memory modules including a non-volatile, computer-readable medium, one or more input / output interfaces, and various connections (for example, a system bus) linking the components. It should be understood that, although certain drawings illustrate hardware and software located within specific devices, these illustrations are for illustrative purposes only. In some cases, the illustrated components may be combined or divided into separate software, firmware, and / or hardware.Instead of residing within and being performed by a single electronic processor, logic and processing can, for example, be distributed among multiple electronic processors. Regardless of how they are combined or shared, hardware and software components can reside on the same computing device or be distributed among different computing devices connected by one or more networks or other suitable communication links.

[0011] For ease of description, some or all of the exemplary systems presented herein are illustrated with a single instance of each of its component parts. Some examples may not describe or illustrate all components of the systems. Other cases may include more or less of each of the illustrated components, may combine some components, or may include additional or alternative components.

[0012] Fig. Figure 1 illustrates a block diagram of a trained artificial intelligence (AI) model 20 that receives multiple vehicle inputs x1, x2, x3, x4, x... over time (t). The trained AI model 20 provides a predicted vehicle signal Y*(t), a tolerance band (σ), and a domain validity signal according to conditions under which the trained AI model is designed to operate.

[0013] The in Fig. The artificial intelligence model 20 shown is trained using training data 24 and a model training algorithm. The model training proceeds as follows: All available data are identified, and an assessment of the data's relevance and quality is performed. The data are separated into training data 24 and test datasets. Iterative model training takes place, during which the model structure and hyperparameters are selected. The quality of the trained model is evaluated and validated based on relevant metrics. The iterative process generates many candidate models. Only one model is selected for use. Model selection is based on a trade-off between quality and complexity. An analysis of the model's performance with test data is performed. The selected model is integrated into the artificial intelligence model 20 and formatted as a digital file.

[0014] In operation, the trained AI model 20, which in some configurations is a trained algorithm, receives the various inputs x1(t)-x4(t) to determine a value for the predicted vehicle signal, along with a tolerance band value σ and a validity signal indicating that the output of the trained AI model 20 is useful. The domain validity signal corresponds to conditions under which the trained AI model is designed to operate.

[0015] Fig. Figure 2 shows an example of a graph 30 of a measured signal that is to be stored and monitored over time. An anomaly is identified where the measured signal lies outside the tolerance band or range represented by the respective dashed lines. A normal predicted signal value is shown by the line that remains within the tolerance band or range defined by the dashed lines.

[0016] In actual operation, the trained artificial intelligence model 20 is designed to: receive a plurality of input signals x1-x4 correlated with the vehicle signal, and analyze the plurality of input signals. The analysis by the trained AI model 20 determines a predicted vehicle signal Y*(t) from the plurality of input signals; determines a tolerance band value σ from the plurality of input signals; and determines whether the model is valid, generating a model validity signal. The predicted vehicle signal Y*(t) is shown to change over time in response to the inputs x1(t)-x4(t), as shown in Fig. Figure 2 illustrates this. The predicted vehicle signal and the tolerance band also change over time, as shown in Fig. 2 shown. An indication statement (true / false) about when the model is valid is provided by the trained AI model 20, which generates the model validity signal.

[0017] Fig. Figure 2A shows an electronic operating device 40 for the trained AI model 20 of Fig. 1, comprising an electronic processor 44 (for example, a microprocessor, an application-specific integrated circuit, etc.), a memory 48, and an input interface 52, together with an output interface 56 and a visual display 58. The memory 48, coupled to the electronic processor 44, stores the trained artificial intelligence model. The memory can consist of one or more non-volatile, computer-readable media and includes at least a program memory area and a data memory area. The program memory area and the data memory area can include combinations of different types of memory, such as read-only memory (“ROM”), random-access memory (“RAM”), electrically erasable programmable read-only memory (“EEPROM”), flash memory, or other suitable storage devices.The electronic processor 44 is coupled to the memory 48, the input interface 52, and the output interface 56. The electronic processor 44 is designed to send and receive information (for example, from the memory 48 and / or the input interface 52) and to process the information by executing the trained AI model 20 via one or more software instructions or one or more software modules. The electronic processor 44 is also capable of storing processed information in the memory 48 or in another non-volatile, computer-readable medium. The output interface 56 provides information or results for further evaluation by another device.

[0018] The in Fig. Figure 2A shows a portable electronic operating device 40 in an arrangement. The portable electronic operating device 40 includes a connector for connecting to a vehicle connector to read stored vehicle data from a separate vehicle memory or to read stored, time-sequenced vehicle data from the vehicle memory.

[0019] Fig. Figure 3 shows a system 75 including the trained AI model 20 of Fig. 1, which is connected to a comparator or comparison device 80 to evaluate an anomaly for a vehicle signal. In Fig. 3. The comparator 80 is designed to receive the predicted vehicle signal, the tolerance band, and the model validity signal generated by the trained AI model 20. Furthermore, the comparator 80 receives the measured vehicle signal stored in the vehicle's memory and the vehicle's operating conditions. After receiving the signals and conditions, the comparator 80 determines whether the measured vehicle signal is an anomaly, based on the measured vehicle signal, the predicted vehicle signal Y*(t), the tolerance band value σ, and whether the model is valid. In an arrangement, the anomalies are aggregated by an aggregator 84, which counts and stores the times and occurrences or the number of anomalies.In this arrangement, the aggregated anomalies are further processed by aggregator 84 to determine a failure / fault of a vehicle component, system, or sensor for display. Vehicle component failures include sensor failures / faults or system failures / faults.

[0020] Fig. Figure 4 shows a specific example of a vehicle system 90 for storing and providing information, including a measured vehicle signal. The vehicle system 90 includes an electronic vehicle operating unit 100, which comprises an electronic processor 104 (for example, a microprocessor, an application-specific integrated circuit, etc.), a memory 108, and an input / output interface 112. The memory 108 can consist of one or more non-volatile, computer-readable media and includes at least a program memory area and a data memory area. The program memory area and the data memory area can include combinations of different types of memory, such as read-only memory (“ROM”), random-access memory (“RAM”), electrically erasable programmable read-only memory (“EEPROM”), flash memory, or other suitable storage devices.The electronic processor 104 is coupled to the memory 108 and the input / output interface 112. The electronic processor 104 is designed to send and receive information (for example, from the memory 108 and / or the input / output interface 112). The electronic processor 104 is also capable of storing processed information in the memory 108 or on another non-volatile, computer-readable medium. The software can include firmware, one or more applications, program data, filters, rules, one or more program modules, and other executable instructions.

[0021] The electronic processor 44 is designed, among other things, to retrieve and execute software from memory 48, including a trained AI algorithm of the trained AI model 20, for the execution of the procedures described herein. The input / output interface 52 retrieves information originally from the vehicle system 90 externally to the electronic operating unit 40. It should be understood that the electronic vehicle operating unit 100 may include additional components beyond those described in Fig. 4 and are illustrated in various configurations. For example, in some examples, the electronic vehicle operating unit 100 includes multiple electronic processors 104, multiple memories 108, multiple input / output interfaces 112, or a combination thereof.

[0022] In some cases, the in Fig. The communication bus 116 shown in Figure 4 is a Controller Area Network (CAN) bus. In some cases, the communication bus 116 is automotive Ethernet, a FlexRay™ communication bus, or another suitable bus. Alternatively, some or all of the vehicle's components may be communicatively coupled using suitable wireless modalities (for example, Bluetooth™ or near-field communication links).

[0023] Fig. Figure 4 shows sensors that measure one or more attributes of a vehicle / vehicle engine and the surrounding environment and communicate information regarding these attributes to other vehicle components, for example, using messages transmitted on communication bus 116. The in Fig. The four sensors shown include, for example, an ambient or outside temperature sensor 120, an ambient air pressure sensor 122, an engine speed sensor 124, a vehicle speed sensor 126, and an exhaust gas recirculation (EGR) flow sensor 128. In some cases, the listed sensors are similar to or identical with sensor sets used in an electronic stability control (ESC) system and / or a vehicle engine control system. In other configurations, the vehicle includes a diesel engine for a diesel-powered vehicle.

[0024] In the arrangement of Fig. 4. During operation, the signals from an ambient or outside temperature sensor 120, an ambient air pressure sensor 122, an engine speed sensor 124, a vehicle speed sensor 126, and an exhaust gas recirculation current sensor 128 are retrieved by the electronic processor 44 of the electronic operating device 40 from a storage device, for example, a memory outside the vehicle. A wireless or wired connector provides the various signals to this device. The trained Kl model 20, which is stored in memory 48 or elsewhere, is executed by the electronic processor 44. The electronic processor 44 uses the ambient temperature signal, the ambient pressure signal, the engine speed signal, and the vehicle speed signal, which are received by the vehicle system 90, to determine a predicted exhaust gas recirculation (EGR) current signal, a tolerance band σ, and a model validity signal.If the model signal is valid, the value of the predicted exhaust gas recirculation current signal is compared with the value of the measured and stored exhaust gas recirculation current signal from the EGR current sensor 128, as determined by the comparator 80 in . Fig. Figure 3 shows that in another arrangement, the electronic processor 44 performs a comparison of received input signals. If the measured EGR current signal is outside the tolerance band σ, the electronic processor 44 determines an anomaly. When the electronic processor 44 determines that an anomaly exists, it is operated to store the anomaly in memory 48. In another arrangement, the comparator 80 provides anomaly information to the aggregator 84. The aggregator 84 is operated to count and store occurrences of anomalies. When a predetermined number of anomalies occur within a predetermined time, a warning is provided to a technician via a visual display and / or an audio speaker (not shown).

[0025] In another arrangement, the comparator 80 and the aggregator 84 are integrated into the electronic processor 44 and the memory 48, which are located in Fig. Figure 2A shows an integrated anomaly detection software algorithm and a vehicle component failure indication arrangement. Memory 48 stores the anomalies and times of their occurrence received from the vehicle. Electronic processor 44 is operated to indicate a vehicle component failure based on a number of anomalies over a period of time. The vehicle component and the type of failure can be indicated on visual display 58. In one arrangement, the vehicle component is an EGR system.

[0026] Fig. Figure 5 shows another specific example of a vehicle system 180 for storing and providing vehicle information, including vehicle signals. The vehicle system 180 includes an electronic vehicle operating unit 200, which comprises an electronic processor 204 (for example, a microprocessor, an application-specific integrated circuit, etc.), a memory 208, and an input / output interface 212. The memory 208 can consist of one or more non-volatile, computer-readable media and includes at least a program memory area and a data memory area. The program memory area and the data memory area can include combinations of different types of memory, such as read-only memory (“ROM”), random-access memory (“RAM”), electrically erasable programmable read-only memory (“EEPROM”), flash memory, or other suitable storage devices.The electronic processor 204 is coupled to the memory 208 and the input / output interface 212. The electronic processor 204 sends and receives information (for example, from the memory 208 and / or the input / output interface 212). The electronic processor 204 is also capable of storing and retrieving one or more software instructions or one or more software modules in and from the memory 208 or another non-volatile, computer-readable medium. The software can include firmware, one or more applications, program data, filters, rules, one or more program modules, and other executable instructions.

[0027] The in Fig. The electronic processor 44 shown in Figure 2A is designed to retrieve and execute, among other things, software including a trained AI algorithm, from memory 48 for the execution of the procedures described herein. The input / output interface 52 retrieves information originally from the electronic vehicle operating unit 200 externally to the electronic operating unit 40. It should be understood that the electronic operating unit 200 may contain additional components beyond those shown in Figure 2A. Fig. 5 and are illustrated in various configurations. For example, in some examples, the electronic operating unit 200 includes multiple electronic processors 204, multiple memories 208, multiple input / output interfaces 212, or a combination thereof.

[0028] In some cases, the in Fig. The communication bus 216 shown in Figure 4 is a Controller Area Network (CAN) bus. In some cases, the communication bus 216 is automotive Ethernet, a FlexRay™ communication bus, or another suitable bus. Alternatively, some or all of the vehicle's components may be communicatively coupled using suitable wireless modalities (for example, Bluetooth™ or near-field communication links).

[0029] Fig. Figure 5 shows sensors that measure one or more attributes of a vehicle / vehicle combustion engine and the surrounding environment and communicate information regarding these attributes to other vehicle components, for example, using messages transmitted on communication bus 216. The in Fig. The five sensors shown include, for example, an ambient temperature sensor 220, an ambient air pressure sensor 222, an engine speed sensor 224, a TrbCH actuator sensor 226, and a boost pressure sensor 228. In some cases, the listed sensors are similar to or identical with sensor sets used in an electronic stability control (ESC) system and / or a vehicle engine control system. In some arrangements, the internal combustion engine is intended for a gasoline-powered or flex-fuel vehicle.

[0030] In the arrangement of Fig. 5. During operation, signals from the ambient temperature sensor 220, the ambient pressure sensor 222, the engine speed sensor 224, and the TrbCH actuator sensor 226 are retrieved by the electronic processor 44 of the electronic operating device 40 from a storage device, for example, a memory location outside the vehicle. A wireless or wired connector provides the signals there. The trained Kl model 20, which is stored in memory 48 or elsewhere, is executed by the electronic processor 44. The electronic processor 44 uses the ambient temperature signal, the ambient pressure signal, the engine speed signal, and the value of the actuator sensor signal TrbCH to determine a predicted boost pressure signal value, a tolerance band σ, and a model validity signal.If the model signal is valid, the predicted boost pressure signal value is compared to the actual value of the measured boost pressure signal from boost pressure sensor 228. If the signal value is outside the tolerance band σ, an anomaly is detected and stored in memory 48. In this arrangement, the electronic processor 44 is operated to count anomalies, and if a predetermined number of anomalies occur within a predetermined time, a boost pressure signal failure warning is stored for reading by a technician examining the vehicle data.

[0031] In another arrangement, the comparator 80 and the aggregator 84 are integrated into the electronic processor 44 and the memory 48 and are provided as an anomaly detection software algorithm that is executed by the electronic processor 44.

[0032] Fig. Figure 6 illustrates a graph 230, which is based on properties of the in Fig. The boost pressure signal arrangement shown in Figure 5 is aligned. Graph 230 includes characteristics of the outputs from the sensors and shows the tolerance band σ in a shaded area 234. The lower side of Fig. Figure 6 shows four different time periods 235A, 235B, 235C, 235D, during which the boost pressure value is not within the tolerance band σ. The amount of time for which the measured boost pressure signal 237 is not within the tolerance band is stored in memory 208. In some arrangements, if the measured boost pressure signal 237, or its value, is not within the tolerance band for more than a predetermined period, a warning or boost pressure failure is stored. At the bottom of Fig. 6 shows the model validity signal 236 as valid for the entire period shown. A boost pressure setpoint 239 is also shown in Fig. 6 shown. The top side of Fig. Figure 6 shows the motor speed 238, which is represented by revolutions per minute (rpm) in a range of 2700 rpm to 3500 rpm. The top of Fig. Figure 6 also shows an accelerator pedal position 240, which is provided in a range of zero to 100% of the accelerator pedal movement. Fig. Figure 6 also shows the turbocharger actuator signal 242, which corresponds to a scale of 0 to 100. Another throttle valve actuator signal 244 also corresponds to a scale of 0 to 100. An engine load signal 246 is shown in Fig. 6 shown. Fig. Figure 6 also shows the ambient temperature signal of 248 degrees Celsius, displayed as 0 to 10 degrees on the scale. Finally, it shows Fig. 6 the ambient pressure signal 250 over time according to a scale of 900 to 1000 hPa.

[0033] Fig. Figure 7 illustrates a block diagram of a trained artificial intelligence (AI) model 260 for exhaust gas recirculation (EGR) ratios, which receives multiple vehicle inputs. The multiple vehicle inputs shown include: 1) representing a modeled pressure at an inlet of a turbocharged turbine (MPT) corresponding to a pressure upstream of an exhaust gas recirculation (EGR) valve, 2) a measured pressure in an air intake manifold (MPAIM) of the vehicle located downstream of the exhaust gas recirculation (EGR) valve, 3) a model fresh air mass flow (MAMF) signal corresponding to a model air mass flow, and 4) a dynamically adjusted requested EGR ratio (AGRR) corresponding to an exhaust gas recirculation (EGR) rate setpoint. The EGR Ratio Artificial Intelligence Model 260 is operated to provide an estimated or predicted exhaust gas recirculation (EVGR) rate.

[0034] In the training, the exhaust gas recirculation (EGR) ratio artificial intelligence model 260 uses model training as described above with reference to the model of Fig. 7. Model training is carried out as follows. All available data are identified, and an assessment of the data's relevance and quality is performed. The data are separated into training data (264) and test datasets. Iterative model training takes place, during which the model structure and hyperparameters are selected. The quality of the trained model is evaluated and validated based on relevant metrics. The iterative process results in many candidate models. Only one model is selected for use. Model selection is based on a trade-off between quality and complexity. An analysis of the model's performance using test data is performed. The selected model is integrated into the trained AGR ratio artificial intelligence model (260) and formatted as a digital file.The trained AGR ratio artificial intelligence model 260 is provided with or for use by an electronic operating device 40, as in . Fig. 2A shown in an arrangement.

[0035] In operation, the trained EGR ratio artificial intelligence model 260 is designed to provide an output VAGR that corresponds to the estimated or predicted EGR rate for an EGR system. The trained EGR ratio artificial intelligence model 260 also determines a tolerance band value σ from the plurality of received input signals and determines when the model is valid in order to generate a model validity signal. The predetermined or estimated EGR rate VAGR and the tolerance band also change over time, potentially along with the model validity signal.

[0036] Fig. Figure 8 shows a specific example of a vehicle system 280 for storing and providing vehicle information, including vehicle signals and modulated input signals. The vehicle system 280 includes an electronic operating unit 300, which comprises an electronic processor 304 (for example, a microprocessor, an application-specific integrated circuit, etc.), a memory 308, and an input / output interface 312. The memory 308 can consist of one or more non-volatile, computer-readable media and includes at least a program memory area and a data memory area. The program memory area and the data memory area can include combinations of different types of memory, such as read-only memory (“ROM”), random-access memory (“RAM”), electrically erasable programmable read-only memory (“EEPROM”), flash memory, or other suitable storage devices.The electronic processor 304 is coupled to the memory 308 and the input / output interface 312. The electronic processor 304 sends and receives information (for example, from the memory 308 and / or the input / output interface 312). The electronic processor 304 is also capable of storing and retrieving one or more software instructions or one or more software modules in and from the memory 44 or another non-volatile, computer-readable medium.

[0037] The input / output interface 312 transmits and receives information from the electronic vehicle operating unit 300 (for example, vehicle components via a communication bus 316). It should be understood that the electronic vehicle operating unit 300 may include additional components beyond those specified in Fig. 8 and are illustrated in various configurations. For example, in some examples, the electronic vehicle operating unit 300 includes multiple electronic processors 304, multiple memories 308, multiple input / output interfaces 312, or a combination thereof.

[0038] In some cases, the in Fig. The communication bus 316 shown in Figure 8 is a Controller Area Network (CAN) bus. In some cases, the communication bus 316 is automotive Ethernet, a FlexRay™ communication bus, or another suitable bus. Alternatively, some or all of the vehicle's components may be communicatively coupled using suitable wireless modalities (for example, Bluetooth™ or near-field communication links).

[0039] Fig. Figure 8 shows arrangements and sensors that measure or determine one or more attributes useful for an exhaust gas recirculation (EGR) system of a vehicle / vehicle engine, which includes an EGR circuit controlling an EGR valve and an EGR bypass valve to provide EGR mass flow and fresh air mass flow, which are mixed and discharged at an EGR rate. Sensors also detect the environment surrounding the vehicle and communicate information regarding these attributes to other vehicle components, for example, using messages transmitted on the 316 communication bus. Fig. The arrangement shown in Figure 8 includes, for example, a model pressure arrangement 320 for determining a model pressure signal MPAIM upstream of an exhaust gas recirculation (EGR) valve, a pressure sensor 322 for detecting a pressure signal downstream of the EGR valve, a fresh air mass flow arrangement 324 for determining a model air mass flow signal, and an exhaust gas recirculation rate arrangement 328 for obtaining a requested exhaust gas recirculation rate AGRR. In some cases, the model signals provided by the arrangements 320, 324, and 328 are obtained by a separate trained Kl model, distinct from the trained Kl model 260, and executed by an electronic processor or other device. In some cases, the vehicle exhaust gas recirculation (EGR) system is intended for a diesel-powered vehicle. In other cases, the vehicle EGR system is considered for a gasoline engine.An EGR rate sensor 330 measures the exhaust gas recirculation rate.

[0040] In the arrangement of Fig. 8. Signals stored during operation from the model pressure arrangement 320 upstream of the EGR valve, the downstream pressure sensor 322, the model fresh air mass flow arrangement 324, and the requested EGR rate arrangement 328 are retrieved by the processor 44 of the electronic operating device 40 from a storage device, for example, a memory outside the vehicle, in an example. The trained EGR ratio model 260, which is stored in memory 48 or elsewhere, is executed by the electronic processor 44 of the electronic operating device 40.The electronic processor 44 uses the pressure sensor 320, which detects a pressure upstream of an EGR valve, the pressure sensor 322, which detects a pressure downstream of the EGR valve, and the fresh air mass flow sensor 324, which detects a fresh air mass flow signal, and a dynamically adapted exhaust gas recirculation rate setpoint calculated from detected values.

[0041] The model pressure arrangement 320 receives a model pressure signal MPT for a turbine of a vehicle engine. The downstream pressure sensor 322 provides a pressure signal corresponding to an intake manifold pressure MPAIM. The model fresh air mass flow arrangement 324 provides a model fresh air mass flow to a mixer of the EGR system to provide a model fresh air mass flow signal MAMF.

[0042] The retrieved signals corresponding to the model pressure array 320, the pressure sensor 322, and the model fresh air mass flow signal from the model fresh air mass flow array 324, together with the requested EGR rate setpoint AGRR, are used by the electronic processor 44, which executes the trained EGR ratio Kl model 260 to determine an estimated or predicted exhaust gas recirculation rate VAGR, a tolerance band σ, and a model validity signal for the EGR system. In some aspects, other acquired values, such as ambient pressure, ambient temperature, and other additional vehicle parameters, are used by the electronic processor 44, which executes the trained EGR ratio Kl model 260 by running an algorithm to obtain the predicted EGR rate VAGR.

[0043] Fig. Figure 9 shows a graph 400 of an example displaying the estimated or predicted EGR rate VAGR 410, a requested EGR setpoint AGRR 420, and a measured EGR rate 430 retrieved from memory. In operation, the trained EGR ratio Kl model 260 is not valid in certain cases, including when the EGR valve is closed, when the EGR valve ramps up from an open loop, and when EGR system limits are reached. Fig. However, figure 9 shows the model validity signal 434 as valid for the entire period shown. An anomaly 438 occurs only for a small period. Fig. Figure 9 also shows a fault model 440 for the arrangement. The top side of Fig. Figure 9 shows an EGR valve position signal 442 corresponding to a percentage that the valve is open. Fig. Figure 9 also shows a measured pressure signal 444 in an air intake manifold, where the scale of 0-100 represents a range from 1000 hPa to 1500 hPa. The model pressure signal 446 at the inlet of a turbocharger corresponds to a range from 1000 hPa to 2000 hPa. Fig. Figure 9 also shows the model air mass flow signal 448, which corresponds to a scale from 100 mg / stroke to 500 mg / stroke.

[0044] During operation, when the EGR ratio Kl model 260 is valid, the predicted EGR rate is compared with the measured and stored EGR rate from the EGR rate sensor 330. If the measured EGR rate is outside the tolerance band σ (in Fig. (9 not shown), an anomaly is determined by the electronic processor 44. When the electronic processor 44 determines that an anomaly exists, it is operated to store the anomaly in memory 48. In this arrangement, the electronic processor 44 is operated to count and store anomalies, and when a predetermined number of anomalies occur within a predetermined time, a warning is provided to a technician analyzing the vehicle data.

[0045] In an arrangement, a comparison device is included, which is in Fig. The comparison device 80 shown in section 3 resembles the electronic processor 44 and the memory 48, which are located in Fig. The system shown in Figure 2A is integrated and provided as an anomaly detection software algorithm. In another aspect, a physical comparison device performs the comparison instead of the electronic processor 44.

[0046] Another vehicle system, 480, which is in Fig. Figure 10 shows a system for receiving and storing vehicle information, including vehicle signals for a fuel cell thermal system in a fuel cell vehicle. The vehicle system 480 includes an electronic operating unit 500, which comprises an electronic processor 504 (for example, a microprocessor, an application-specific integrated circuit, etc.), a memory 508, and an input / output interface 512. The electronic processor 504 is coupled to the memory 508 and the input / output interface 512.The electronic processor 504 sends and receives information (for example, from the memory 508 and / or the input / output interface 512) to the electronic operating device 40 for the trained AI model 20 via one or more software instructions or one or more software modules that can be stored in the memory 508 or another non-volatile, computer-readable medium.

[0047] Fig. Figure 10 shows sensors that measure one or more attributes useful for the thermal system of a fuel cell in a fuel cell vehicle. The in Fig. The 10 sensors shown include, for example, a power sensor 520 for measuring the power supplied by the fuel cell, an actuator position sensor 522 for measuring an actuator position signal for the fuel cell of the fuel cell vehicle, an ambient temperature sensor 524 for measuring an ambient temperature signal, and a fuel cell temperature sensor 528 for measuring a fuel cell temperature signal. The signals are stored in the memory 508 or transmitted to another memory that is not provided in the fuel cell vehicle.

[0048] During operation, signals from the power sensor 520, the actuator position sensor 522, and the ambient temperature sensor 524 are retrieved over time by the processor 44 of the electronic operating device 40 from a storage device, for example, an external memory location. The trained Kl model 20, stored in memory 48 or elsewhere, is executed by the electronic processor 44 to determine a predicted fuel cell heat / temperature value for the fuel cell's thermal system, a tolerance band σ, and a model validity signal based on the power signal from the fuel cell, the actuator position signal, and the ambient temperature signal. If the model signal is valid, the predicted fuel cell heat / temperature signal is compared with the measured and stored fuel cell temperature signal from the fuel cell temperature sensor 528.If the compared value lies outside the tolerance band σ, an anomaly is determined and stored in memory 48.

[0049] In a fuel cell vehicle arrangement, if the measured fuel cell temperature signal is not within the tolerance band for the predicted fuel cell temperature signal, as indicated by a number of detected anomalies, a fuel cell temperature sensor warning is provided on a visual display 58 of the electronic operating device 40. In addition to the above, the electronic operating device 40 is also capable of determining sensor failure and other anomalies, such as engine performance anomalies, and providing a visual display 58 of other warning information.

[0050] In another arrangement, the fuel cell temperature sensor 528 is a first fuel cell temperature sensor, and the vehicle includes a second fuel cell temperature sensor (not shown). If the vehicle component of the first fuel cell temperature sensor is reported as failing, depending on a number of anomalies over a period of time, the electronic operating device 40 for the trained AI model 20 will not continue to use the first fuel cell temperature sensor. In one arrangement, the second fuel cell temperature sensor for a fuel cell vehicle is activated when the first fuel cell temperature sensor 528 exhibits anomalies.

[0051] Although Fig. 4, Fig. 5, Fig. 8 and Fig. As certain types of sensors are indicated, additional sensors are being considered to assist in determining a predicted vehicle signal. Such sensors include ascent sensors, descent sensors, trailer load sensors, and other types of sensors. Other sensors determine whether a vehicle engine or fuel cell is cold or hot, plus environmental conditions such as altitude and ambient temperature. In one configuration, the fuel cell vehicle is a hydrogen fuel cell vehicle. Other configurations are being considered.

[0052] The in Fig. The artificial intelligence (AI) model 20 shown is capable of being trained in various computing environments using different microprocessors and computer configurations. The training data and datasets, such as vehicle sensor data, are stored in any type of memory or in cloud-based storage. After training is complete, the trained AI model can be loaded onto a plurality of electronic operating devices 40 for use by a technician.

[0053] In addition to determining the presence of anomalies, the electronic operating device 40 is designed to indicate defective components in a vehicle under investigation. For example, in an arrangement, if the measured and stored exhaust gas recirculation signal is not within the tolerance band for the predicted EGR signal, as shown by a number of detected anomalies, an indication of an EGR system failure is provided as an indication on a visual display of the electronic operating device 40.

[0054] In another arrangement, if the measured boost pressure signal is not within the tolerance band for the predicted boost pressure signal, as indicated by detected anomalies, a boost pressure signal failure is provided by an indication on a visual display 58 of the electronic operating device 40. In another arrangement, the boost pressure sensor 228 is a first boost pressure sensor, and the vehicle includes a second boost pressure sensor (not shown). If the vehicle component of the first boost pressure sensor 228 is indicated as a failure, depending on a number of anomalies over a period of time, the electronic operating device 40 does not continue to use the first boost pressure sensor 228 for the trained Kl model 20, which sends a signal to the vehicle. In one arrangement, a second boost pressure sensor is activated for use by an internal combustion engine vehicle.

[0055] Throughout the text above, sensor or model signals include a stored signal value and a stored time of occurrence of the measured / determined and stored signal value. Measured signals also include a stored signal value and a time of occurrence. Thus, both signal values ​​and a timestamp or time of occurrence are generally provided for all of the input vehicle signals and measured vehicle signals outlined above in some arrangements.

[0056] The above arrangements are used to provide various methods for anomaly detection. The correlated input signals depend on the type of vehicle signal of interest. Although a plurality of sensors are listed in the patent specification and claims, the arrangements in Fig. 4, Fig. 5, Fig. 8 and Fig.10 inclusive, and may also include additional sensors which are set forth herein and not specifically disclosed in the arrangements.

[0057] Unless the context of their use clearly indicates otherwise, the articles "ein" and "eine" should not be interpreted as meaning "one" or "only one". Instead, these articles should be interpreted as meaning "at least one" or "one or more".

[0058] Although certain arrangements, examples, features and aspects have been described and illustrated, variations and modifications exist within the scope of protection and meaning of the subject explained and shown.

Claims

[1] A system for indicating a failure of a vehicle component by determining anomalies for a measured vehicle signal of a vehicle, the system comprising: a memory that stores a trained artificial intelligence model; an electronic processor coupled to the memory and executing the trained artificial intelligence model, the electronic processor being configured to: Retrieving a plurality of input signals correlated with the measured vehicle signal; Determining a predicted vehicle signal from the plurality of input signals; Determining a tolerance band from the plurality of input signals; Determining whether the model is valid, and if the model is valid, generating a model validity signal, a comparison device designed to: Retrieving the predicted vehicle signal, the tolerance band and the model validity signal, Retrieving the measured vehicle signal, and Determining whether the measured vehicle signal is an anomaly based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal, and an aggregator coupled to the comparison device, the aggregator being configured to: Counting and storing times and occurrences of anomalies; and Specify a vehicle component failure depending on a number of anomalies over a period of time. [2] The system of claim 1, wherein the predicted vehicle signal is a predicted exhaust gas recirculation flow signal and the measured vehicle signal is an exhaust gas recirculation flow signal. [3] The system of claim 2, wherein the plurality of input signals includes an ambient temperature signal, an ambient pressure signal, an engine speed signal, and a vehicle speed signal, and wherein the vehicle component is an exhaust gas recirculation flow sensor. [4] The system of claim 1, wherein the predicted vehicle signal is a boost pressure signal for an internal combustion engine and the measured vehicle signal is a measured boost pressure signal, and wherein the comparison device is configured to receive operating conditions of the vehicle. [5] The system of claim 4, wherein the plurality of input signals includes an ambient temperature signal, an ambient pressure signal, an engine speed signal, and an actuator sensor signal, and wherein the vehicle component is a boost pressure sensor. [6] The system of claim 1, wherein the predicted vehicle signal is a predicted fuel cell temperature signal and the measured vehicle signal is a measured fuel cell temperature signal, wherein the plurality of input signals include a power supplied by the fuel cell, an actuator position signal, an ambient temperature signal, and a measured fuel cell temperature signal, wherein the vehicle component is a fuel cell temperature sensor that provides the measured fuel cell temperature signal, and wherein the vehicle is a fuel cell vehicle. [7] The system of claim 6, wherein the fuel cell temperature sensor is a first fuel cell temperature sensor and the fuel cell vehicle includes a second fuel cell temperature sensor, and wherein, when the vehicle component of the first fuel cell temperature sensor is indicated as the failure, depending on the number of anomalies over the period, the use of the first fuel cell temperature sensor is discontinued and the second fuel cell temperature sensor is activated. [8] The system of claim 1, wherein the predicted vehicle signal is a predicted exhaust gas recirculation rate, and the plurality of input signals includes a model pressure signal determined for upstream of an exhaust gas recirculation valve, a pressure signal downstream of the exhaust gas recirculation valve corresponding to an air intake manifold pressure, a model fresh air mass flow signal, and a dynamically adjusted exhaust gas recirculation rate setpoint. [9] A method for indicating a failure of a vehicle component by determining anomalies for a measured vehicle signal of a vehicle, the method comprising: Executing a trained artificial intelligence model with an electronic processor by: Retrieving a plurality of input signals correlated with the measured vehicle signal; Determining a predicted vehicle signal from the plurality of input signals; Determining a tolerance band from the plurality of input signals; Determining whether the model is valid, If the model is valid, generating a model validity signal; retrieving the measured vehicle signal; Determining whether the measured vehicle signal is an anomaly based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal; Counting a number of and storing times and occurrences of the anomalies; and Specify the failure of the vehicle component as a function of the number of anomalies over a period of time. [10] The method of claim 9, wherein the predicted vehicle signal is a predicted exhaust gas recirculation flow signal and the measured vehicle signal is an exhaust gas recirculation flow signal, and wherein the vehicle component is an exhaust gas recirculation flow sensor. [11] The method of claim 10, wherein the plurality of input signals includes an ambient temperature signal, an ambient pressure signal, an engine speed signal, and a vehicle speed signal. [12] The method of claim 9, wherein the predicted vehicle signal is a boost pressure signal for an internal combustion engine, and wherein the plurality of input signals includes an ambient temperature signal, an ambient pressure signal, an engine speed signal, and an actuator sensor signal, and wherein the vehicle component is a boost pressure sensor. [13] The method of claim 9, wherein the predicted vehicle signal is a predicted exhaust gas recirculation rate, and the plurality of input signals includes a model pressure signal determined for upstream of an exhaust gas recirculation valve, a pressure signal downstream of the exhaust gas recirculation valve corresponding to an air intake manifold pressure, a model fresh air mass flow signal, and a dynamically adjusted exhaust gas recirculation rate setpoint. [14] The method of claim 9, including a memory coupled to the electronic processor, the memory storing the trained artificial intelligence model. [15] The method of claim 9, wherein the predicted vehicle signal is a predicted fuel cell temperature signal and the measured vehicle signal is a measured fuel cell temperature signal, wherein the plurality of input signals include a power delivered by the fuel cell, an actuator position signal, and an ambient temperature signal, wherein the vehicle component is a fuel cell temperature sensor, and wherein the vehicle is a fuel cell vehicle. [16] The method of claim 15, wherein the fuel cell temperature sensor is a first fuel cell temperature sensor and the vehicle includes a second fuel cell temperature sensor, and wherein, when the vehicle component of the first fuel cell temperature sensor is indicated as the failure, depending on the number of anomalies over the period, a signal is provided to a vehicle system to discontinue use of the first fuel cell temperature and to enable use of the second fuel cell temperature sensor for the fuel cell vehicle. [17] A method for indicating a failure of a vehicle component of a vehicle, the method comprising: retrieving a plurality of input signals correlated with a vehicle signal; Determining a predicted vehicle signal from the plurality of input signals using a trained artificial intelligence model; Determining a tolerance band from the plurality of input signals; Determining whether the model is valid, If the model is valid, generating a model validity signal; retrieving a measured vehicle signal; Determining whether the measured vehicle signal is an anomaly based on the measured vehicle signal, the predicted vehicle signal, the tolerance band, and the model validity signal; Counting a number of and storing times of occurrence of the anomalies, and Specify the failure of a vehicle component depending on the number of anomalies over a period of time. [18] The method of claim 17, wherein the predicted vehicle signal is a predicted exhaust gas recirculation flow signal and the measured vehicle signal is an exhaust gas recirculation flow signal, and wherein the plurality of input signals includes an ambient temperature signal, an ambient pressure signal, an engine speed signal, and a vehicle speed signal. [19] The method of claim 17, wherein the predicted vehicle signal is a boost pressure signal for an internal combustion engine, and wherein the plurality of input signals includes an ambient temperature signal, an ambient pressure signal, an engine speed signal, and an actuator sensor signal. [20] The method of claim 17, wherein the predicted vehicle signal is a predicted fuel cell temperature signal and the measured vehicle signal is a measured fuel cell temperature signal, wherein the plurality of input signals includes a power supplied by a fuel cell, an actuator position signal, and an ambient temperature signal.

Images