A vehicle fault diagnosis method and system based on digital twins

Abstract

This disclosure provides a vehicle fault diagnosis method and system based on digital twins. The vehicle fault diagnosis method includes: (S1) collecting vehicle data; (S2) storing the collected vehicle data in association with corresponding generation timestamps and sensor identification codes; (S3) performing a preliminary diagnosis of whether the vehicle has a fault based on the collected vehicle sensor data; (S4) when the preliminary diagnosis indicates that the vehicle has a fault, performing a digital twin diagnosis of the vehicle fault; and (S5) generating a visual display of the fault. Compared with traditional diagnostic methods, the vehicle fault diagnosis method based on a real-time digital twin model of the vehicle disclosed in this disclosure can provide more efficient, accurate, and comprehensive fault detection.

Description

Technical Field

[0001] This invention relates to automated fault diagnosis of vehicles. More specifically, this invention relates to a vehicle fault diagnosis method and system based on digital twins. Background Technology

[0002] Digital twin technology, also known as digital twinning, is a technique that mirrors the state and behavior of an actual physical object by creating a virtual copy, enabling real-time monitoring and analysis. In the field of vehicle fault diagnosis, digital twin technology allows engineers and technicians to detect and predict faults without physically interacting with the vehicle by creating a virtual copy. This technology utilizes sensor data, operational history, and real-time performance information to build a dynamic vehicle model. This model can simulate real-world conditions in a virtual environment, thereby predicting and diagnosing potential problems. The application of digital twin technology in vehicle fault diagnosis not only improves the accuracy and efficiency of diagnosis but also contributes to the continuous operation and performance optimization of the vehicle.

[0003] Existing vehicle diagnostic technologies based on digital twins primarily rely on data collected by in-vehicle sensors, such as key indicators like engine temperature, oil pressure, and battery status. While this information is crucial for understanding the vehicle's internal workings, it doesn't fully reveal its real-world performance. To obtain more comprehensive data, it's necessary to combine driver input with external factors like weather conditions, road type, and traffic conditions—all significant influences on vehicle performance. Furthermore, another challenge with digital twin models is their iteration and updating. As vehicle information is continuously collected, the model needs constant updates to reflect the latest data. This process demands powerful computing capabilities to rapidly process large amounts of data and implement complex algorithms and simulations within the model. This not only increases the demand for hardware resources but also means that the simulation process may take a considerable amount of time before the model reaches sufficient accuracy. Summary of the Invention

[0004] The first aspect of this disclosure provides a vehicle fault diagnosis method based on digital twins. In some embodiments, the method may include: (S1) collecting vehicle data, including vehicle sensor data, user operation behavior data, and vehicle surrounding environment data, wherein the vehicle has an initial digital twin model M1; (S2) storing the collected vehicle data in a first database DB1 and a second database DB2 in association with corresponding generation timestamps and sensor identification codes; (S3) performing a preliminary diagnosis of whether the vehicle has a fault based on the collected vehicle sensor data to obtain a preliminary diagnosis result; (S4) when the preliminary diagnosis result indicates that the vehicle has a fault, performing a digital twin diagnosis of the vehicle fault; and (S5) generating a visual representation of the fault. In some instances, process (S4) may further include: (S41) determining a real-time digital twin model M2 of the vehicle corresponding to the time T in which the fault occurs, indicating the actual condition of the vehicle; and (S42) determining the location and fault level of the faulty component based on the initial digital twin model M1, the time T in which the fault occurs, and the real-time digital twin model M2.

[0005] In some embodiments, processing S2 may include: storing the collected vehicle data in the first database DB1 with the generation timestamp as the primary key and the sensor identification code as the secondary key; and storing the collected vehicle data in the second database DB2 with the sensor identification code as the primary key and the generation timestamp as the secondary key.

[0006] In some embodiments, process S41 may include: (S411) reading the collected vehicle data from a first database DB1 within a predetermined time period prior to time T; (S412) using the collected vehicle data read in (S411) as input data, inputting it into the initial digital twin model M1 of the vehicle in timestamp order for simulation to generate sensor simulation data; and (S413) comparing the sensor simulation data with the collected vehicle data to iteratively optimize the initial digital twin model M1 of the vehicle to obtain the real-time digital twin model M2 of the vehicle. The difference between the model parameter values ​​of the real-time digital twin model M2 of the vehicle and the initial digital twin model M1 of the vehicle indicates the location and fault level of the faulty component.

[0007] In some embodiments, processing S412 may include:

[0008] (S4121) Define the state transition equation as shown in Equation (1) to generate the current sensor estimation data based on the prior vehicle data:

[0009] x(t|t-1)=f(y(t-1),u(t-1),s(t-1)) (1)

[0010] Where x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, y(t-1) is the vehicle sensor data at time t-1, u(t-1) is the user operation behavior data at time t-1, s(t-1) is the vehicle's surrounding environment data at time t-1, and f(·) is the correlation feature equation; and

[0011] (S4122) Define the update equation as shown in formula (2):

[0012] x(t)=x(t|t-1)+x' (t-1)*Ts (2)

[0013] Where x(t) is the sensor simulation data at time t, x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, x'(t-1) is the derivative of the sensor simulation data at time t-1, and Ts is the time step. Ts is less than the acquisition interval of the vehicle data.

[0014] In some embodiments, processing S413 may include:

[0015] (S4131) Define the objective function for iterative optimization as shown in Equation (3):

[0016] min{|y(t)-x(t)|}(3)

[0017] Where y(t) is the vehicle data collected at time t; and

[0018] (S4132) Using the objective function of the iterative optimization, determine the real-time digital twin model M2 of the vehicle that corresponds to the time T in which the vehicle fault occurs and indicates the actual condition of the vehicle.

[0019] In some embodiments, process S42 may include: determining that one or more components have a general fault when the absolute value difference of parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle exceeds a first empirical threshold; determining that one or more components have a moderate fault when the absolute value difference of parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle exceeds a second empirical threshold; and determining that one or more components have a serious fault when the absolute value difference of parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle exceeds a third empirical threshold. The third empirical threshold may be greater than the second empirical threshold. The second empirical threshold may be greater than the first empirical threshold.

[0020] In some embodiments, processing S5 may include: reading the vehicle data at the time T of the fault occurrence from the second database DB2, displaying the real-time digital twin model M2 of the vehicle, and highlighting the model component in the real-time digital twin model M2 that corresponds to the faulty component.

[0021] In some embodiments, processing S5 may include: when an end user clicks or touches a component in the real-time digital twin model M2 of the vehicle, further displaying detailed model information of the component.

[0022] Another aspect of this disclosure provides a vehicle fault diagnosis system based on digital twins. In some embodiments, the vehicle fault diagnosis system may include: an initial digital twin model M1 of the vehicle; a vehicle data collection unit configured to collect vehicle data, including vehicle sensor data, user operation behavior data, and vehicle surrounding environment data; a first database DB1 and a second database DB2 configured to store the collected vehicle data in association with corresponding generation timestamps and sensor identification codes; a preliminary vehicle fault diagnosis unit configured to perform a preliminary diagnosis of whether the vehicle has a fault based on the collected vehicle sensor data to obtain a preliminary diagnosis result; a vehicle fault digital twin diagnosis unit configured to perform a digital twin diagnosis of the vehicle fault when the preliminary diagnosis result indicates that the vehicle has a fault; and a visual fault display unit configured to generate a visual display of the fault. In some instances, performing digital twin diagnostics of vehicle faults may include: determining a real-time digital twin model M2 of the vehicle that indicates the actual condition of the vehicle corresponding to the time T at which the fault occurred; and determining the location and fault level of the faulty component based on the initial digital twin model M1 of the vehicle, the time T at which the fault occurred, and the real-time digital twin model M2 of the vehicle.

[0023] In some embodiments, determining a real-time digital twin model M2 of the vehicle corresponding to the time T in which the fault occurs, indicating the actual condition of the vehicle, may include: (S411) reading the collected vehicle data from a first database DB1 within a predetermined time period before time T; (S412) using the collected vehicle data read in (S411) as input data, inputting it into the initial digital twin model M1 of the vehicle in timestamp order for simulation to generate sensor simulation data; and (S413) comparing the sensor simulation data with the collected vehicle data, thereby iteratively optimizing the initial digital twin model M1 of the vehicle to obtain the real-time digital twin model M2 of the vehicle. The difference between the model parameter values ​​of the real-time digital twin model M2 of the vehicle and the initial digital twin model M1 of the vehicle indicates the location and fault level of the faulty component.

[0024] Another aspect of this disclosure provides a system comprising one or more computer processors and a computer-readable storage device. The computer-readable storage device includes machine-executable code. When executed by the one or more computer processors, the machine-executable code implements a digital twin-based vehicle fault diagnosis method as described in the first aspect of this disclosure.

[0025] It should be understood that this disclosure is not intended to identify key or essential features of the embodiments thereof, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0026] The novel features of the present invention are specifically set forth in the appended claims. A better understanding of the features and advantages of the invention will be obtained by referring to the following detailed description and accompanying drawings, which illustrate illustrative embodiments utilizing the principles of the invention.

[0027] Figure 1 An example of the vehicle digital twin model disclosed herein is shown.

[0028] Figure 2 This is a schematic flowchart of a vehicle fault diagnosis method based on digital twins according to an embodiment of the present disclosure. Detailed Implementation

[0029] In existing technologies, vehicle fault diagnosis based on digital twins primarily relies on in-vehicle sensor data, which limits technicians' comprehensive understanding of actual vehicle usage. Real-world driving environments include variable weather conditions, complex traffic situations, and driver habits—factors difficult to capture solely from in-vehicle sensor data. Therefore, to make digital twin models more accurately reflect actual vehicle usage, external environmental data and driving behavior data need to be considered. For example, onboard cameras and external sensors can collect information on road conditions and traffic flow, while vehicle-to-everything (V2X) technology can provide driving data from surrounding vehicles. Integrating this external data will significantly enhance the predictive power and accuracy of digital twin models. Furthermore, another challenge of digital twin models is their high demand for computing resources. To achieve highly accurate simulations, the model must be constantly updated and iterated to reflect the latest collected data. This process requires powerful computing capabilities, especially when processing large amounts of data and performing complex calculations.

[0030] In view of the above-mentioned problems in the prior art, this disclosure proposes an innovative vehicle fault diagnosis method based on digital twins, which comprehensively considers vehicle sensor data and real driving environment data, and optimizes the update method of the vehicle's digital twin model, so that the digital twin model can run more efficiently and reduce the simulation time required.

[0031] Figure 1 An example of the vehicle digital twin model disclosed herein is shown. The initial digital twin model refers to a complete digital model constructed after the vehicle's production is completed, based on its design parameters and standard configuration. This model contains all important information about the vehicle, such as engine performance, body structure, electronic systems, etc., and their ideal state at the time of manufacture. The initial digital twin model provides a benchmark for the entire lifecycle of the vehicle. The real-time status of the vehicle, monitored during vehicle use, can be compared with the initial digital twin model. This comparison helps to promptly identify deviations and potential problems, thereby enabling preventative maintenance and fault diagnosis. For example, if the vehicle's real-time data shows that the engine temperature is consistently higher than the standard temperature in the initial model, the system can alert the owner or maintenance personnel to conduct inspection and maintenance.

[0032] As vehicles wear and age during use, their digital twin models are updated to reflect their current state. This means the digital twin model is dynamic. Synchronizing the initial digital twin model with actual vehicle data can be achieved through a series of automated processes. In one example, the vehicle's built-in sensors continuously collect various data about vehicle performance, including but not limited to engine temperature, fuel consumption, speed, and braking system status. This data is then transmitted in real-time to a cloud server via an in-vehicle communication system. In the cloud, a data processing system analyzes the collected vehicle information, detecting any discrepancies with the initial digital twin model. For example, it can identify normal wear and tear, potential fault signals, and even the impact of external environmental changes on vehicle performance. Once the system detects anomalies or deviations in the data, it can trigger an alarm and provide relevant feedback to the owner or maintenance team. The maintenance team can then analyze the data provided by the model to pinpoint the exact location and cause of the fault, thereby reducing downtime and improving the efficiency of repair work. This rapid response capability is particularly important for commercial fleet management, as it can significantly reduce operational disruptions caused by vehicle malfunctions.

[0033] Throughout the vehicle's lifespan, the digital twin model can be automatically updated. Based on real-time data, the digital twin model can be adjusted to reflect the vehicle's current state, which may include the degree of wear and tear on parts, minor changes in performance, or even updates to vehicle configuration. This dynamic updating ensures that the digital twin model always accurately represents the current condition of the physical vehicle.

[0034] In one embodiment, the digital twin model of the vehicle of this application may include models of the vehicle body structure, powertrain, braking system, suspension system, steering system, and auxiliary systems. For example, the digital twin model of the vehicle body structure includes not only the parameters of the frame and body but also simulates its response under different load and shock conditions. The powertrain model can simulate the performance of the engine and transmission system under various operating conditions, including acceleration, fuel efficiency, and emission levels. The digital twin model of the braking system can be used to analyze the effectiveness of the brake pedal and handbrake, as well as reaction time and distance in emergency situations. The suspension system can simulate the vehicle's handling under various road conditions while evaluating the durability of shock absorbers, wheels, and tires. The digital twin of the steering system can analyze the responsiveness and precision of the steering wheel, as well as the reliability of the entire system after long-term use. Digital twin models of auxiliary systems such as air conditioning and human-machine interaction systems can be used to evaluate the efficiency and user-friendliness of these systems. The initial digital twin model of the vehicle is a digital twin model of the vehicle in a standard state without any faults.

[0035] Figure 2This is a schematic flowchart of a vehicle fault diagnosis method based on digital twins according to an embodiment of the present disclosure. The vehicle may have an initial digital twin model M1. In one embodiment, the vehicle fault diagnosis method based on digital twins of the present disclosure may include: (S1) collecting vehicle data; (S2) associating the collected vehicle data with corresponding generation timestamps and sensor identification codes, and storing the associated vehicle data in a first database DB1 and a second database DB2; (S3) performing a preliminary diagnosis of whether the vehicle has a fault based on the collected vehicle sensor data to obtain a preliminary diagnosis result; (S4) when the preliminary diagnosis result indicates that the vehicle has a fault, performing a digital twin diagnosis of the vehicle fault; and (S5) generating a visual display of the fault.

[0036] In some embodiments, the vehicle data collected in step (S1) may include vehicle sensor data, user operation behavior data, and vehicle surrounding environment data. Vehicle sensor data provides the vehicle's real-time status, including but not limited to engine temperature, oil pressure, battery status, and tire pressure. These data points are fundamental to understanding the vehicle's current operating condition. User operation behavior data provides another dimension of information, reflecting how the driver interacts with the vehicle. For example, the driver's acceleration and braking habits, steering patterns, and frequency of using the vehicle control system can all affect the vehicle's performance and health. By analyzing this behavioral data, the context in which the fault occurred can be better understood. Vehicle surrounding environment data may include external conditions of vehicle operation, such as road conditions, weather conditions, traffic flow, and surrounding obstacles. These factors can all affect vehicle performance, so considering this external data is crucial in fault diagnosis. By comprehensively analyzing vehicle sensor data, user operation behavior data, and vehicle surrounding environment data, the digital twin-based vehicle fault diagnosis method of this disclosure can construct a comprehensive vehicle health profile.

[0037] In some embodiments, vehicle sensor data may include vehicle speed, acceleration, interior temperature, airflow, battery voltage, battery charge, coolant temperature, and vehicle level, collected from various in-vehicle sensors (e.g., vehicle speed sensor, acceleration sensor, interior temperature sensor, gas flow sensor, battery voltage sensor, battery charge sensor, coolant temperature sensor, vehicle level sensor, etc.). User operation behavior data may include: steering wheel angle, brake pedal displacement, accelerator pedal displacement, air conditioning on / off, air conditioning temperature adjustment amplitude, seat adjustment operations, etc. Vehicle surrounding environment data may include: distance to the vehicle in front, distance to the vehicle behind, traffic light data, lane information, road congestion conditions, weather conditions, traffic flow, etc.

[0038] In step (S2), the collected vehicle data is stored in the first database DB1 and the second database DB2. In the application of digital twin technology, data organization and management are crucial. To achieve efficient data processing, corresponding data formats need to be designed according to different data processing algorithms. This method not only improves the efficiency of data processing but also ensures the compatibility of the data format with the processing task. During digital twin simulation, vehicle data collected over a period of time is typically used as input. This data can be stored in different databases based on different combinations of generation timestamps and sensor identification codes. For example, the first database DB1 can use the generation timestamp as the primary key and the sensor identification code as the secondary key to store vehicle data. This data structure design facilitates the rapid retrieval of vehicle data at a specific point in time, thereby simulating the vehicle's operating status during that period. This helps in analyzing vehicle performance and behavior patterns, as well as predicting future operating trends. On the other hand, the second database DB2 can use the sensor identification code as the primary key and the generation timestamp as the secondary key to store vehicle data. This design enables the second database DB2 to efficiently handle fault analysis and localization tasks because it can quickly retrieve data based on the sensor identification code, and then analyze the operating status of components monitored by specific sensors at different points in time.

[0039] This data organization method also helps to reconstruct the spatial distribution characteristics of faults, which is invaluable for understanding the environment and conditions under which faults occur, optimizing maintenance strategies, and implementing preventative measures. In practice, when performing digital twin simulations, data can be extracted from the first database DB1 to simulate and analyze the vehicle's operating conditions over a certain period. Conversely, when performing fault analysis and localization, data can be extracted from the second database DB2, focusing on the operating status of the target components monitored by the sensors.

[0040] In step (S3), a preliminary diagnosis of whether the vehicle has a fault is performed based on the collected vehicle sensor data to obtain a preliminary diagnostic result. In one example, the preliminary diagnostic method of this disclosure may include comparing the vehicle sensor data at a certain time T with corresponding preset thresholds or threshold ranges. These thresholds or threshold ranges are preset based on a data model under normal vehicle operation conditions, and any data exceeding these thresholds or threshold ranges may indicate a potential fault in the vehicle. The preliminary diagnostic process is not just a simple data comparison. It also includes monitoring the rate of data change and analyzing user operating behavior. For example, if a sensor suddenly loses data, or its rate of data change differs significantly from the normal pattern, this may be a strong fault signal. Similarly, if the user's operating behavior shows abnormal changes, such as a sudden increase or decrease in the frequency of accelerator pedal use, this may also indicate that a certain system of the vehicle is malfunctioning.

[0041] In step (S4), when the preliminary diagnostic results indicate a vehicle malfunction, a digital twin diagnosis of the vehicle malfunction can be performed. In some embodiments, step (S4) may further include:

[0042] (S41) Determine a real-time digital twin model M2 of the vehicle that corresponds to the time T at which the fault occurred and indicates the actual condition of the vehicle; and

[0043] (S42) Determine the location and fault level of the faulty component based on the initial digital twin model M1 of the vehicle, the time T at which the fault occurred, and the real-time digital twin model M2 of the vehicle.

[0044] An initial digital twin model of the vehicle serves as a reference point, representing the vehicle's ideal state without any faults, and is typically built based on vehicle design and factory parameters. By comparing the initial digital twin model M1 with the real-time digital twin model M2 at the time of the fault (T), differences and inconsistencies between the two can be identified. These differences point to the specific location of the fault, such as an abnormal sensor reading or a performance degradation of a mechanical component. Furthermore, the severity of these differences can be assessed, thereby determining the fault level.

[0045] In some embodiments, step (S41) may include:

[0046] (S411) Read vehicle data collected within a predetermined time period before time T from the first database DB1;

[0047] (S412) The vehicle data collected in (S411) is used as input data and input into the initial digital twin model M1 of the vehicle in time-stamp order for simulation to generate sensor simulation data; and

[0048] (S413) Compare the sensor simulation data with the collected vehicle data to iteratively optimize the initial digital twin model M1 of the vehicle to obtain the real-time digital twin model M2 of the vehicle.

[0049] The difference in model parameter values ​​between the real-time digital twin model M2 and the initial digital twin model M1 of the vehicle can indicate the location and severity of the faulty component.

[0050] Step S411 extracts vehicle data within a specific time period from the first database DB1, providing foundational data for subsequent simulation. Step S412 uses this vehicle data as input, inputting it into the initial digital twin model M1 of the vehicle in timestamp order. The key to this step is to use the model for high-precision simulation to generate sensor simulation data. Step S413 compares the generated sensor simulation data with the actually collected vehicle data. This comparison process can be iterative, meaning that through continuous comparison and adjustment, the initial digital twin model M1 of the vehicle will be gradually optimized to more accurately reflect the actual operating conditions of the vehicle. In this way, the real-time digital twin model M2 of the vehicle is generated. This updated model includes various updates to the initial digital twin model M1 of the vehicle, enabling it to better simulate the real-time state of the vehicle.

[0051] In some embodiments, step S412 of generating sensor simulation data may include:

[0052] (S4121) Define the state transition equation as shown in Equation (1) to generate the current sensor estimation data based on the prior vehicle data:

[0053] x(t|t-1)=f(y(t-1),u(t-1),s(t-1)) (1)

[0054] Where x(t|t-1) is the sensor estimation data at time t generated based on vehicle data at time t-1, y(t-1) is the vehicle sensor data at time t-1, u(t-1) is the user operation behavior data at time t-1, s(t-1) is the vehicle's surrounding environment data at time t-1, and f(·) is the correlation characteristic equation; and

[0055] (S4122) Define the update equation as shown in formula (2):

[0056] x(t)=x(t|t-1)+x' (t-1)*Ts (2)

[0057] Where x(t) is the sensor simulation data at time t, x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, x'(t-1) is the derivative of the sensor simulation data at time t-1, and Ts is the time step.

[0058] Step S412 provides a method for generating current sensor estimation data based on historical vehicle data. This process involves two equations: a state transition equation (1) and an update equation (2). The state transition equation (1) uses vehicle sensor data from the previous moment, user operation behavior data, and vehicle surrounding environment data to predict the sensor data at the current moment. This equation reflects the dynamic nature of the system disclosed herein, enabling the generation of new estimates based on changing input data. The update equation (2) further refines the sensor estimates by incorporating the time step and the derivative of the sensor data, based on the state transition equation, thereby obtaining simulated sensor data. This equation considers the continuity of time and the changing trends of data, making the estimation more accurate. Different vehicles may have different sensor configurations and user operating habits, and this method can adapt to these differences by adjusting the correlation feature equation f(·). Simultaneously, the choice of the time step Ts provides flexibility. A suitable Ts value can be determined based on the sensor data acquisition frequency and the required prediction accuracy. In some instances, the time step Ts can be smaller than the vehicle data acquisition interval.

[0059] In some embodiments, step S413 of generating a real-time digital twin model M2 of the vehicle may include:

[0060] (S4131) Define the objective function for iterative optimization as shown in formula (3):

[0061] min{|y(t)-x(t)|}(3)

[0062] Where y(t) is the vehicle data collected at time t; and

[0063] (S4132) Using the objective function of iterative optimization, determine the real-time digital twin model M2 of the vehicle that corresponds to the time T when the vehicle fault occurs and indicates the actual condition of the vehicle.

[0064] The objective function defined in step S4131 is an iterative optimization problem that aims to find a solution that minimizes the absolute difference between y(t) (representing the actual vehicle data collected at time t) and x(t) (representing the model's prediction of the vehicle state). This difference can be understood as the error between the actual data and the model prediction, and the goal of the optimization process is to minimize this error. In practical applications, the optimization process may include gradient descent, genetic algorithms, or other iterative methods to gradually adjust the model parameters until the solution with the minimum error is found.

[0065] By iteratively optimizing the objective function, we can construct a highly accurate real-time digital twin model M2 of the vehicle. This real-time digital twin model M2 can essentially dynamically reproduce the current real-time fault conditions of the vehicle; that is, the real-time digital twin model M2 is a digital twin model with the current vehicle faults. This dynamic nature of the model means that it can adapt to constantly changing operating conditions and environmental factors, thus providing more accurate fault detection. For example, the real-time digital twin model M2 can simulate air filter blockage, cylinder-piston clearance exceeding a normal threshold, or coolant levels below a normal threshold by adjusting its model parameters. The real-time digital twin model M2 can also simulate more than one vehicle fault. Therefore, the real-time digital twin model M2 has a similar model structure to the initial digital twin model M1, but differs in its model parameter values. These different model parameter values ​​can be used to assist in determining the location of the faulty component and the fault level.

[0066] In some embodiments, step S42, determining the location and fault level of a faulty component, may include: determining that one or more components have a general fault when the absolute value of the parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from a first empirical threshold; determining that one or more components have a moderate fault when the absolute value of the parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from a second empirical threshold; and determining that one or more components have a severe fault when the absolute value of the parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from a third empirical threshold. In some instances, the third empirical threshold may be greater than the second empirical threshold, and the second empirical threshold may be greater than the first empirical threshold.

[0067] In some embodiments, step S5, which generates a visual representation of the fault, may include: reading vehicle data at time T of the fault occurrence from a second database DB2, displaying a real-time digital twin model M2 of the vehicle, and highlighting the model component in the real-time digital twin model M2 corresponding to the faulty component. By extracting vehicle data at a specific time T from the second database DB2, the fault diagnosis system of this disclosure can quickly locate the faulty component and visualize it in the digital twin model M2. This display not only includes the location of the faulty component but also shows other vehicle system information related to the fault, providing maintenance personnel with a comprehensive fault context.

[0068] Using three.js as the rendering tool, a digital twin model of the vehicle M2 can be displayed in 3D within a web browser, providing users with an intuitive and highly interactive fault diagnosis interface. three.js is a WebGL-based JavaScript library capable of rendering complex 3D graphics in most modern web browsers without requiring any special plugins. In the display of the digital twin model, the use of three.js ensures that the model not only visually appeals to users but also provides a smooth interactive experience.

[0069] Rendering 3D models is a computationally intensive task that can place a significant burden on computer resources. To address this, a multi-level display approach can be employed. In this method, the vehicle's digital twin model, M2, initially displays only the most critical features and components. This not only reduces the computational resources required for the initial rendering but also speeds up model loading, allowing users to interact with the model without delay. Only when the user requires more detail will the system load and display higher-resolution model components. This progressively detailed display method not only improves system responsiveness but also optimizes the user experience. Furthermore, this multi-level display method helps protect sensitive data. In the initial display phase, unnecessary details can be avoided, reducing the risk of data leakage. Only after explicit user actions, such as clicks or touches, will the system provide more details. This design ensures information security while meeting the user's information needs.

[0070] Another aspect of this disclosure provides a vehicle fault diagnosis system based on digital twins. In some embodiments, the vehicle fault diagnosis system of this disclosure may include: an initial digital twin model of the vehicle M1, a vehicle data collection unit, a first database DB1 and a second database DB2, a preliminary vehicle fault diagnosis unit, a vehicle fault digital twin diagnosis unit, and a visual fault display unit. In some embodiments, the vehicle data collection unit may be configured to collect vehicle data. The vehicle data may include vehicle sensor data, user operation behavior data, and vehicle surrounding environment data. The first database DB1 and the second database DB2 may be configured to store the collected vehicle data associated with corresponding generation timestamps and sensor identification codes. The preliminary vehicle fault diagnosis unit may be configured to perform a preliminary diagnosis of whether the vehicle has a fault based on the collected vehicle sensor data to obtain a preliminary diagnosis result. The vehicle fault digital twin diagnosis unit may be configured to perform a digital twin diagnosis of the vehicle fault when the preliminary diagnosis result indicates that the vehicle has a fault. Digital twin diagnosis of vehicle faults may include: (S41) determining a real-time digital twin model M2 of the vehicle corresponding to the time T of the fault occurrence, indicating the actual condition of the vehicle; and (S42) determining the location and fault level of the faulty component based on the initial digital twin model M1 of the vehicle, the time T of the fault occurrence, and the real-time digital twin model M2 of the vehicle. The visual fault display unit can be configured to generate a visual display of the fault.

[0071] In some embodiments, the first database DB1 can be configured to store the collected vehicle data, using the generation timestamp as the primary key and the sensor identification code as the secondary key. The second database DB2 can be configured to store the collected vehicle data, using the sensor identification code as the primary key and the generation timestamp as the secondary key.

[0072] In some embodiments, the operation S41 of the vehicle fault digital twin diagnostic unit may include:

[0073] (S411) Read vehicle data collected within a predetermined time period before time T from the first database DB1;

[0074] (S412) The vehicle data collected in (S411) is used as input data and input into the initial digital twin model M1 of the vehicle in time-stamp order for simulation to generate sensor simulation data; and

[0075] (S413) Compare the sensor simulation data with the collected vehicle data to iteratively optimize the initial digital twin model M1 of the vehicle to obtain the real-time digital twin model M2 of the vehicle.

[0076] The difference in model parameter values ​​between the real-time digital twin model M2 and the initial digital twin model M1 of the vehicle can indicate the location and severity of the faulty component.

[0077] In some embodiments, processing S412 may include:

[0078] (S4121) Define the state transition equation as shown in Equation (1) to generate the current sensor estimation data based on the prior vehicle data:

[0079] x(t|t-1)=f(y(t-1),u(t-1),s(t-1)) (1)

[0080] Where x(t|t-1) is the sensor estimation data at time t generated based on vehicle data at time t-1, y(t-1) is the vehicle sensor data at time t-1, u(t-1) is the user operation behavior data at time t-1, s(t-1) is the vehicle's surrounding environment data at time t-1, and f(·) is the correlation characteristic equation; and

[0081] (S4122) Define the update equation as shown in formula (2):

[0082] x(t)=x(t|t-1)+x' (t-1)*Ts (2)

[0083] Where x(t) is the sensor simulation data at time t, x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, x'(t-1) is the derivative of the sensor simulation data at time t-1, and Ts is the time step.

[0084] Ts can be less than the vehicle data collection interval.

[0085] In some embodiments, processing S413 may include:

[0086] (S4131) Define the objective function for iterative optimization as shown in formula (3):

[0087] min{|y(t)-x(t)|}(3)

[0088] Where y(t) is the vehicle data collected at time t; and

[0089] (S4132) Using the objective function of iterative optimization, determine the real-time digital twin model M2 of the vehicle that corresponds to the time T when the vehicle fault occurs and indicates the actual condition of the vehicle.

[0090] In some embodiments, the operation S42 of the vehicle fault digital twin diagnostic unit may include: determining that one or more components have a general fault when the absolute value difference of the parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle exceeds a first empirical threshold; determining that one or more components have a moderate fault when the absolute value difference of the parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle exceeds a second empirical threshold; and determining that one or more components have a serious fault when the absolute value difference of the parameters of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle exceeds a third empirical threshold. The third empirical threshold may be greater than the second empirical threshold. The second empirical threshold may be greater than the first empirical threshold.

[0091] In some embodiments, the visual fault display unit can be configured to: read vehicle data at the time T of the fault occurrence from the second database DB2, display the real-time digital twin model M2 of the vehicle, and highlight the model component in the real-time digital twin model M2 corresponding to the faulty component. In some embodiments, the visual fault display unit can be configured to: further display detailed model information of a component when an end user clicks or touches a component in the real-time digital twin model M2 of the vehicle.

[0092] This disclosure also provides a system including one or more computer processors and a computer-readable storage device. The computer-readable storage device may include machine-executable code that, when executed by the one or more computer processors, implements the digital twin-based vehicle fault diagnosis method of this disclosure.

[0093] The vehicle fault diagnosis method based on a real-time digital twin model disclosed herein provides more efficient, accurate, and comprehensive fault detection compared to traditional diagnostic methods. Traditional vehicle diagnostic methods typically rely on physical inspection and driver's intuitive judgment, which can be time-consuming and labor-intensive, and often proves inadequate in the face of complex or hidden faults. In contrast, the vehicle fault diagnosis method based on a real-time digital twin model disclosed herein utilizes real-time data and advanced algorithms to accurately reproduce and simulate the real-time state of the vehicle in a virtual environment, including various potential fault scenarios.

[0094] By continuously collecting vehicle operating data and comparing it with pre-set model parameters, the disclosed digital twin model M2 can monitor vehicle status in real time and promptly detect anomalies. This method not only improves the speed of fault detection but also reduces human error in the diagnostic process due to its high degree of automation and accuracy. Furthermore, the digital twin model M2 can simulate a variety of different fault scenarios, which is often difficult to achieve with traditional methods. For example, it can simultaneously simulate air filter blockage and insufficient coolant, which is invaluable for understanding the interactions between different faults and their impact on overall vehicle performance.

[0095] The digital twin model M2 also possesses the ability to predict potential future failures, a capability lacking in traditional diagnostic methods. By analyzing long-term data trends and patterns, the digital twin model can predict which components are likely to experience problems in the future, allowing maintenance teams to take proactive measures to prevent failures. This preventative maintenance strategy not only extends vehicle lifespan but also reduces unexpected downtime, improving vehicle reliability and safety. In the fields of autonomous driving and intelligent vehicles, the advantages of the digital twin model M2 are even more pronounced. It can interact with the vehicle's control system in real time, providing immediate feedback and adjustment suggestions to optimize vehicle performance and driving strategies. This real-time data analysis and decision support is unimaginable in traditional diagnostic methods. The application of the digital twin model M2 enables vehicles to maintain optimal performance under complex environments and conditions, ensuring driving safety.

[0096] Although preferred embodiments of the invention have been shown and described herein, it will be apparent to those skilled in the art that these embodiments are provided merely by way of example. Other variations, modifications, and substitutions may be conceived by those skilled in the art without departing from the invention.

Claims

1. A vehicle fault diagnosis method based on digital twin, the method comprising: (S1) Collect vehicle data, including vehicle sensor data, user operation behavior data and vehicle surrounding environment data, wherein the vehicle has an initial digital twin model M1. (S2) The collected vehicle data is associated with the corresponding generation timestamp and sensor identification code and stored in the first database DB1 and the second database DB2; (S3) Based on the collected vehicle sensor data, perform a preliminary diagnosis to determine whether the vehicle has a fault and obtain a preliminary diagnosis result. The preliminary diagnosis includes comparing the vehicle sensor data at a certain time T with a corresponding default threshold or threshold range. (S4) When the preliminary diagnostic results indicate that the vehicle has a fault, perform a digital twin diagnosis of the vehicle fault; as well as (S5) Generate a visual representation of the fault. The process (S4) further includes: (S41) Determine a real-time digital twin model M2 of the vehicle that corresponds to the time T at which the fault occurred and indicates the actual condition of the vehicle; Processing (S41) includes: Define the state transition equation as shown in Equation (1) to generate the current sensor estimation data based on the prior vehicle data: x(t|t-1)=f(y(t-1),u(t-1),s(t-1)) (1) Where x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, y(t-1) is the vehicle sensor data at time t-1, u(t-1) is the user operation behavior data at time t-1, s(t-1) is the vehicle surrounding environment data at time t-1, and f(·) is the correlation feature equation. and (S42) Based on the initial digital twin model M1 of the vehicle, the time T at which the fault occurred, and the real-time digital twin model M2 of the vehicle, determine the location and fault level of the faulty component. The process (S41) includes: (S411) Read the collected vehicle data from the first database DB1 within a predetermined time period before time T; (S412) The collected vehicle data read in (S411) is used as input data and input into the initial digital twin model M1 of the vehicle in timestamp order for simulation to generate sensor simulation data; and (S413) The sensor simulation data is compared with the collected vehicle data to iteratively optimize the initial digital twin model M1 of the vehicle, thereby obtaining the real-time digital twin model M2 of the vehicle. The difference in model parameter values ​​between the real-time digital twin model M2 and the initial digital twin model M1 indicates the location and fault level of the faulty component. The process (S42) includes: When the absolute value of the parameter of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from the first empirical threshold, it is determined that one or more components have a general fault. When the absolute value of the parameter of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from the second empirical threshold, it is determined that one or more components have a moderate fault; and When the absolute value of the parameter of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from the third empirical threshold, it is determined that one or more components have a serious fault. Wherein, the third experience threshold is greater than the second experience threshold, and the second experience threshold is greater than the first experience threshold.

2. The method according to claim 1, wherein the process (S2) comprises: The collected vehicle data is stored in the first database DB1, using the generated timestamp as the primary key and the sensor identification code as the secondary key. as well as The collected vehicle data is stored in the second database DB2, using the sensor identifier code as the primary key and the generated timestamp as the secondary key.

3. The method according to claim 1, The process (S412) includes: (S4122) Define the update equation as shown in formula (2): x(t)=x(t|t-1)+x' (t-1)*Ts (2) Where x(t) is the sensor simulation data at time t, x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, x'(t-1) is the derivative of the sensor simulation data at time t-1, and Ts is the time step, where Ts is less than the acquisition interval of the vehicle data. The process (S413) includes: (S4131) Define the objective function for iterative optimization as shown in Equation (3): min{|y(t)-x(t)|}(3) Where y(t) is the vehicle data collected at time t; and (S4132) Using the objective function of the iterative optimization, determine the real-time digital twin model M2 of the vehicle that corresponds to the time T in which the vehicle fault occurs and indicates the actual condition of the vehicle.

4. The method according to claim 1, wherein the process (S5) comprises: Read the vehicle data at the time T when the fault occurred from the second database DB2, display the real-time digital twin model M2 of the vehicle, and highlight the model component in the real-time digital twin model M2 that corresponds to the faulty component.

5. The method according to claim 1, wherein the process (S5) further comprises: When an end user clicks or touches a component in the real-time digital twin model M2 of the vehicle, detailed model information of the component is further displayed.

6. A vehicle fault diagnosis system based on digital twins, the vehicle fault diagnosis system comprising: Initial digital twin model of the vehicle, M1; A vehicle data collection unit is configured to collect vehicle data, including vehicle sensor data, user operation behavior data, and vehicle surrounding environment data. A first database DB1 and a second database DB1 are configured to store the collected vehicle data in association with corresponding generation timestamps and sensor identification codes; A vehicle fault preliminary diagnosis unit is configured to perform a preliminary diagnosis of whether a vehicle has a fault based on the collected vehicle sensor data to obtain a preliminary diagnosis result. The preliminary diagnosis includes comparing the vehicle sensor data at a certain time T with a corresponding default threshold or threshold range. A vehicle fault digital twin diagnostic unit is configured to perform digital twin diagnosis of the vehicle fault when the preliminary diagnostic result indicates that the vehicle has a fault. The digital twin diagnosis includes: determining a real-time digital twin model M2 of the vehicle corresponding to the time T at which the fault occurred, indicating the actual condition of the vehicle; and determining the location and fault level of the faulty component based on the initial digital twin model M1, the time T at which the fault occurred, and the real-time digital twin model M2. A visual fault display unit, configured to generate a visual display of the fault. The real-time digital twin model M2 of the vehicle, which corresponds to the time T when the fault occurs and indicates the actual condition of the vehicle, includes: (S41) Determine a real-time digital twin model M2 of the vehicle that corresponds to the time T at which the fault occurred and indicates the actual condition of the vehicle; Processing (S41) includes: Define the state transition equation as shown in Equation (1) to generate the current sensor estimation data based on the prior vehicle data: x(t|t-1)=f(y(t-1),u(t-1),s(t-1)) (1) Where x(t|t-1) is the sensor estimation data at time t generated based on the vehicle data at time t-1, y(t-1) is the vehicle sensor data at time t-1, u(t-1) is the user operation behavior data at time t-1, s(t-1) is the vehicle surrounding environment data at time t-1, and f(·) is the correlation feature equation. and (S42) Based on the initial digital twin model M1 of the vehicle, the time T at which the fault occurred, and the real-time digital twin model M2 of the vehicle, determine the location and fault level of the faulty component. The process (S41) includes: (S411) Read the collected vehicle data from the first database DB1 within a predetermined time period before time T. (S412) The collected vehicle data read in (S411) is used as input data and input into the initial digital twin model M1 of the vehicle in timestamp order for simulation to generate sensor simulation data; and (S413) The sensor simulation data is compared with the collected vehicle data to iteratively optimize the initial digital twin model M1 of the vehicle, thereby obtaining the real-time digital twin model M2 of the vehicle. The difference in model parameter values ​​between the real-time digital twin model M2 and the initial digital twin model M1 indicates the location and fault level of the faulty component. The process (S42) includes: When the absolute value of the parameter of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from the first empirical threshold, it is determined that one or more components have a general fault. When the absolute value of the parameter of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from the second empirical threshold, it is determined that one or more components have a moderate fault; and When the absolute value of the parameter of one or more components in the initial digital twin model M1 and the real-time digital twin model M2 of the vehicle differs from the third empirical threshold, it is determined that one or more components have a serious fault. Wherein, the third experience threshold is greater than the second experience threshold, and the second experience threshold is greater than the first experience threshold.

7. A system comprising one or more computer processors and a computer-readable storage device, the computer-readable storage device including machine-executable code that, when executed by the one or more computer processors, implements the vehicle fault diagnosis method based on digital twins as described in any one of claims 1-5.

Images