Training method, fault detection method, electronic equipment, vehicle, medium and product

Abstract

The invention discloses a fault detection model training method, a fault detection method, an electronic device, a vehicle, a computer readable storage medium and a computer program product, and a construction method comprises the steps: determining model parameters according to obtained first vehicle operation data and first fault state label data corresponding to the first vehicle operation data; and determining a fault detection model for detecting the fault state of the vehicle according to the model parameters. Thus, the model parameters of the fault detection model can be determined based on the first vehicle operation data and the corresponding first fault state label data, the situation that the model parameters need to be determined through a model training mode is avoided, the construction cost of the fault detection model is reduced, and the fault detection efficiency is improved. The relevance between the model parameters and the vehicle state data and the first fault state label data can be ensured, so that the detection precision of the fault detection model constructed based on the model parameters can be ensured.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and in particular to a fault detection model training method, a fault detection method, an electronic device, a vehicle, a computer-readable storage medium, and a computer program product. Background Art

[0002] In related technologies, vehicles can be equipped with machine learning-based classification models to identify the operating status of in-vehicle systems and devices, such as intelligent driving systems or engine failures. However, these machine learning-based classification models rely on manually pre-designed feature data, which can miss key features and require significant time to train. Consequently, the classification accuracy is low and the training and updating costs are high. Summary of the Invention

[0003] The present application provides a fault detection model training method, a fault detection method, an electronic device, a vehicle, a computer-readable storage medium, and a computer program product.

[0004] The present application provides a method for constructing a fault detection model, including:

[0005] determining model parameters based on the acquired first vehicle operation data and first fault state label data corresponding to the first vehicle operation data;

[0006] A fault detection model for detecting a vehicle fault state is determined according to the model parameters.

[0007] Thus, in an embodiment of the present application, the model parameters of a fault detection model for detecting a vehicle fault state can be determined by obtaining the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, so that the model parameters of the fault detection model can be determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, thereby avoiding the need to determine the model parameters through model training, and to a certain extent reducing the cost of constructing the fault detection model. Furthermore, since the model parameters are determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, the correlation between the model parameters and the vehicle state data and the first fault state label data can be guaranteed, thereby ensuring the detection accuracy of the fault detection model constructed based on the model parameters.

[0008] In certain embodiments of the present application, the first vehicle operation data includes vehicle engine status data, and the first fault status tag data is used to indicate a fault status of the vehicle engine.

[0009] Thus, in the embodiment of the present application, the model parameters can be correlated with the vehicle engine status data and the first fault state label data for the fault state of the vehicle engine, thereby enabling the fault detection model to accurately detect the fault state of the vehicle engine.

[0010] In certain embodiments of the present application, the model parameters include weight parameters.

[0011] In this way, in the embodiment of the present application, the fault detection model can be constructed based on the weight parameters determined by the first vehicle operation data and the first fault state label data, which reduces the difficulty of constructing the fault detection model to a certain extent, thereby further reducing the construction cost of the fault detection model.

[0012] In certain embodiments of the present application, determining the model parameters based on the acquired first vehicle operation data and the first fault status label data corresponding to the first vehicle operation data includes:

[0013] determining characteristic operating data based on the first vehicle operating data;

[0014] The model parameters are determined according to the characteristic operation data, the first vehicle operation data, and the first fault status label data.

[0015] In this way, in the embodiment of the present application, the characteristic operating data can be determined based on the first vehicle operating data, and the model parameters can be determined based on the characteristic operating data, the first vehicle operating data and the first fault status label data, thereby ensuring the reliability of the model parameters to a certain extent.

[0016] In certain embodiments of the present application, determining characteristic operating data based on the first vehicle operating data includes:

[0017] The first vehicle operation data is subjected to preset mapping and flattening processing to determine the characteristic operation data.

[0018] Thus, in the embodiment of the present application, the first vehicle operation data may be subjected to a preset mapping and flattening process to determine the characteristic operation data, thereby achieving determination of the characteristic operation data.

[0019] In certain embodiments of the present application, performing a preset mapping and flattening process on the first vehicle operation data to determine the characteristic operation data includes:

[0020] Performing mapping processing on the first vehicle operation data according to randomly determined parameters to determine vehicle state mapping data;

[0021] The vehicle state mapping data is flattened to determine the characteristic operation data.

[0022] In this way, in the embodiment of the present application, the first vehicle operation data can be mapped according to the randomly determined parameters to determine the vehicle state mapping data, and the vehicle state mapping data can be flattened to determine the characteristic operation data, thereby realizing the determination of the characteristic operation data.

[0023] In certain embodiments of the present application, determining the model parameters based on the characteristic operation data, the first vehicle operation data, and the first fault status label data includes:

[0024] determining an orthogonal matrix and an upper triangular matrix according to the characteristic operation data and the first vehicle operation data;

[0025] The model parameters are determined according to the orthogonal matrix, the upper triangular matrix and the first fault state label data.

[0026] Thus, in the embodiment of the present application, the model parameters can be determined based on the orthogonal matrix and the upper triangular matrix determined by the characteristic operation data and the first vehicle operation data, combined with the first fault state label data.

[0027] In certain embodiments of the present application, determining an orthogonal matrix and an upper triangular matrix based on the characteristic operation data and the first vehicle operation data includes:

[0028] determining an expansion matrix according to the characteristic operation data and the first vehicle operation data;

[0029] The orthogonal matrix and the upper triangular matrix are determined according to the extended matrix.

[0030] Thus, in the embodiment of the present application, the extended matrix can be determined according to the characteristic operation data and the first vehicle operation data, and the orthogonal matrix and the upper triangular matrix can be determined according to the extended matrix, thereby achieving the determination of the orthogonal matrix and the upper triangular matrix.

[0031] In certain embodiments of the present application, determining the orthogonal matrix and the upper triangular matrix according to the extended matrix includes:

[0032] Performing orthogonal decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix.

[0033] Thus, in the embodiment of the present application, the orthogonal matrix and the upper triangular matrix may be determined by orthogonal decomposition.

[0034] In certain embodiments of the present application, performing orthogonal decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix includes:

[0035] determining a Householder vector according to the expansion matrix;

[0036] Constructing a Householder matrix according to the Householder vector;

[0037] The orthogonal matrix and the upper triangular matrix are determined according to the expanded matrix and the Householder matrix.

[0038] In this way, in the implementation mode of the present application, the Householder vector can be determined based on the extended matrix, and the Householder matrix can be constructed based on the Householder vector, and the orthogonal matrix and the upper triangular matrix can be determined based on the extended matrix and the Householder matrix, thereby realizing the orthogonal decomposition based on the Householder transform.

[0039] The present application provides a fault detection method, including:

[0040] The vehicle fault state is determined according to the fault detection model and the acquired second vehicle operation data, wherein the fault detection model is constructed according to the above-mentioned fault detection model construction method.

[0041] Thus, in the embodiment of the present application, the vehicle fault state can be determined based on the fault detection model and the acquired second vehicle operation data, thereby realizing the detection of the vehicle fault state. Moreover, the model parameters of the fault detection model for detecting the vehicle fault state can be determined by the acquired first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, so that the model parameters of the fault detection model can be determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, thereby avoiding the need to determine the model parameters through model training, and reducing the construction cost of the fault detection model to a certain extent. In addition, since the model parameters are determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, the correlation between the model parameters and the vehicle state data and the first fault state label data can be guaranteed, thereby ensuring the detection accuracy of the fault detection model constructed based on the model parameters.

[0042] In certain embodiments of the present application, the method further comprises:

[0043] The fault detection model is updated according to the second vehicle operating data and the vehicle fault status.

[0044] In this way, in the embodiment of the present application, the fault detection model can be updated based on the second vehicle operation data obtained during the vehicle operation process, as well as the vehicle fault state determined by the vehicle fault state and the fault detection model. This can achieve real-time updating of the fault detection model to a certain extent, and then enable the fault detection model to be updated as the vehicle operation process progresses, thereby improving the detection accuracy of the fault detection model to a certain extent.

[0045] In certain embodiments of the present application, updating the fault detection model according to the second vehicle operating data and the vehicle fault status includes:

[0046] Determining second fault state label data according to the vehicle fault state;

[0047] The fault detection model is updated according to the second vehicle operation data and the second fault status label data.

[0048] Thus, in the embodiment of the present application, the second fault state label data can be determined according to the vehicle fault state, and the fault detection model can be updated according to the second vehicle operation data and the second fault state label data, thereby realizing the update of the fault detection model.

[0049] In certain embodiments of the present application, the vehicle fault state includes predicted probabilities of multiple fault types, the second fault state label data includes predicted label values of the multiple fault types, and determining the second fault state label data based on the vehicle fault state includes:

[0050] The predicted label value of a target fault among the multiple fault types is determined as a first value, and the predicted label values of the other faults are determined as a second value, wherein the target fault is the fault with the highest predicted probability among the multiple fault types.

[0051] Thus, in the embodiment of the present application, the predicted label value of the target fault among multiple fault types is determined as the first value, and the predicted label values of other faults are determined as the second value, thereby achieving the determination of the second fault state label data.

[0052] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the above-mentioned method for constructing a fault detection model or the above-mentioned fault detection method is implemented.

[0053] An embodiment of the present application provides a vehicle, comprising the above-mentioned electronic device.

[0054] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, it implements the above-mentioned method for constructing a fault detection model or the above-mentioned fault detection method.

[0055] An embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned method for constructing a fault detection model or the above-mentioned fault detection method.

[0056] The electronic device, vehicle, computer-readable storage medium, and computer program product provided by the embodiments of the present application can determine the model parameters of a fault detection model for detecting a vehicle fault state by obtaining the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, so that the model parameters of the fault detection model can be determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, thereby avoiding the need to determine the model parameters through model training, and reducing the construction cost of the fault detection model to a certain extent. In addition, since the model parameters are determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, the correlation between the model parameters and the vehicle state data and the first fault state label data can be guaranteed, thereby ensuring the detection accuracy of the fault detection model constructed based on the model parameters.

[0057] Additional aspects and advantages of the embodiments of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0059] Figure 1 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0060] Figure 2 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0061] Figure 3 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0062] Figure 4 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0063] Figure 5 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0064] Figure 6 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0065] Figure 7 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0066] Figure 8 A flowchart of a method for constructing a fault detection model in certain embodiments of the present application is provided;

[0067] Figure 9 This is a flowchart of a fault detection method in certain embodiments of the present application;

[0068] Figure 10 This is a flowchart of a fault detection method in certain embodiments of the present application;

[0069] Figure 11 This is a flowchart of a fault detection method in certain embodiments of the present application;

[0070] Figure 12 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0071] Figure 13 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0072] Figure 14 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0073] Figure 15 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0074] Figure 16 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0075] Figure 17 This is a schematic diagram of an application scenario in certain embodiments of the present application. DETAILED DESCRIPTION

[0076] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and should not be understood as limiting the embodiments of the present application.

[0077] With the continuous advancement of intelligent driving technology, intelligent driving systems are becoming increasingly complex, encompassing multiple links from sensor data acquisition to automatic vehicle control, integrating a large number of high-precision, highly automated devices. Understandably, intelligent driving systems are massive in scale, and the amount of data generated during driving is also extremely large and complex. However, this high complexity also brings a higher risk of failure. Once a failure occurs, it not only causes traffic disruption and economic losses, but can also trigger safety accidents, posing a serious threat to human life and property safety.

[0078] Therefore, real-time, accurate fault diagnosis is crucial for ensuring the stable operation of intelligent driving systems, improving traffic efficiency, and reducing maintenance costs. Online learning algorithms, with their powerful data processing and automated learning capabilities, offer unique advantages in real-time fault diagnosis. Online learning algorithms are essential for the following: 1) Real-time performance: Intelligent driving systems require extremely high real-time performance. Any minor fault, if not detected and addressed promptly, can rapidly escalate and lead to serious consequences. Online learning algorithms can process and analyze system data in real time, promptly identifying and reporting potential faults, significantly reducing the time required for fault detection and resolution. 2) Automation: By constructing neural network models, online learning algorithms can automatically extract features relevant to system status from massive amounts of data, enabling real-time monitoring and fault diagnosis. This automated process not only reduces manual intervention but also significantly improves diagnostic accuracy and efficiency. 3) Adaptability: The operating environment and conditions of intelligent driving systems are constantly changing. Online learning algorithms can continuously adapt to these new operating environments and conditions through continuous learning and optimization, improving the adaptability and robustness of fault diagnosis. 4) Intelligence: Advanced machine learning technologies such as deep learning empower online learning algorithms with powerful intelligent processing capabilities. By simulating the workings of neurons in the human brain, these algorithms can handle complex nonlinear problems and intelligently analyze and predict system states. This intelligent processing capability gives online learning algorithms enormous potential for real-time fault diagnosis in intelligent driving systems.

[0079] Therefore, it is clear that designing an algorithm with online learning capabilities and real-time fault diagnosis for intelligent driving systems is extremely necessary. This will not only improve the system's operational efficiency and stability, but also significantly reduce maintenance costs and safety risks, providing a strong guarantee for the development of modern transportation.

[0080] However, traditional fault diagnosis methods rely heavily on manual experience and regular maintenance. These methods have numerous limitations. For example, they are often only applicable to specific fault types or devices, limiting their ability to diagnose faults in new intelligent driving systems. Furthermore, they place high demands on data quality, often preventing accurate fault diagnosis when dealing with data with significant noise. Furthermore, manual inspections are inefficient and struggle to cope with complex and changing driving environments.

[0081] More specifically, in related technologies, most fault diagnosis solutions are based on traditional machine learning techniques, such as support vector machines (SVM), random forests (RF), and decision trees. These solutions predict or output fault detection results by extracting features and training classification models based on pre-collected fault data.

[0082] It's worth noting that these fault diagnosis solutions based on traditional machine learning rely on manual feature extraction. In other words, these solutions often require manually designing the feature extraction process, and the quality of the features directly impacts the effectiveness of the diagnosis. Understandably, this manual extraction process is not only time-consuming and labor-intensive, but also prone to missing key features, resulting in low diagnostic accuracy.

[0083] Furthermore, it's understandable that these solutions typically require a long time for model training and have limited real-time data processing capabilities, making them difficult to meet the real-time and efficiency requirements of intelligent driving systems. Furthermore, in the complex and ever-changing intelligent driving environment, these solutions have poor generalization capabilities for unseen fault modes, making them prone to misdiagnosis or missed diagnoses.

[0084] Therefore, to address these shortcomings, related technologies have proposed applying deep learning techniques, such as convolutional neural networks and long short-term memory networks, to fault detection, such as for fault diagnosis in intelligent vehicle driving systems. It is understood that in fault detection solutions based on deep learning, deep neural networks can automatically extract features and classify and predict fault modes.

[0085] However, it is understandable that deep learning models typically contain a large number of parameters, and the training and inference processes require high-performance computing resources, resulting in high computational complexity. This makes real-time diagnosis difficult in resource-constrained embedded devices. Furthermore, when data is insufficient or unevenly distributed, deep learning models are prone to overfitting, which reduces diagnostic reliability. Furthermore, deep learning models rely heavily on high-quality, well-labeled, large-scale datasets. However, fault data in actual intelligent driving systems is often scarce and difficult to label, limiting the method's application.

[0086] In addition, related technologies have also proposed fault detection solutions based on matrix decomposition techniques such as singular value decomposition and non-negative matrix factorization. It is understood that fault detection solutions implemented through matrix decomposition techniques can reduce the dimensionality and extract patterns from system operating data for fault diagnosis.

[0087] However, traditional matrix decomposition methods typically require global calculations across the entire dataset, making them unable to quickly respond to changes in newly added data and resulting in poor real-time performance. Furthermore, matrix decomposition methods are inherently linear and struggle to capture the complex nonlinear dynamic relationships found in intelligent driving systems. This lack of nonlinear modeling capabilities limits diagnostic accuracy.

[0088] Based on the above problems you may encounter, please refer to Figure 1 , an embodiment of the present application provides a method for constructing a fault detection model, comprising:

[0089] 01: Determine model parameters according to the acquired first vehicle operation data and first fault state label data corresponding to the first vehicle operation data;

[0090] 02: Determine the fault detection model used to detect vehicle fault status based on the model parameters.

[0091] The present application also provides an electronic device comprising a memory and a processor. The method for constructing a fault detection model according to the present application can be implemented by the electronic device according to the present application. Specifically, the memory stores a computer program, and the processor is configured to determine model parameters based on acquired first vehicle operation data and first fault status label data corresponding to the first vehicle operation data, and to determine a fault detection model for detecting a vehicle fault state based on the model parameters.

[0092] Specifically, in an embodiment of the present application, the electronic device can determine the model parameters based on the acquired first vehicle operation data and the first fault status label data corresponding to the first vehicle operation data, and then construct a fault detection model for detecting the vehicle fault status based on the determined model parameters.

[0093] Thus, in an embodiment of the present application, the model parameters of a fault detection model for detecting a vehicle fault state can be determined by obtaining the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, so that the model parameters of the fault detection model can be determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, thereby avoiding the need to determine the model parameters through model training, and to a certain extent reducing the cost of constructing the fault detection model. Furthermore, since the model parameters are determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, the correlation between the model parameters and the vehicle state data and the first fault state label data can be guaranteed, thereby ensuring the detection accuracy of the fault detection model constructed based on the model parameters.

[0094] In one example, the first vehicle operation data may be data collected based on sensors in the vehicle during the operation of the vehicle and used to characterize the vehicle operation status.

[0095] In one example, the first fault status label data corresponding to the first vehicle operating data can be understood as: when the current operating status of the vehicle can be represented by the first vehicle operating data, the fault condition of the vehicle at this time, such as engine abnormality, engine normal abnormality, engine crankshaft position offset, engine crankshaft position not offset, etc.

[0096] In one example, the first vehicle operation data may indicate relevant data of the vehicle intelligent driving system. For example, assuming that the intelligent driving system includes m sensors, the data collected by the m sensors are acquired n times, and the fault status of the vehicle is acquired while acquiring the data collected by the m sensors, then: the first vehicle operation data can be represented by X, X∈R n×m , the first fault state label data corresponding to the first vehicle operation data X can be represented as Y, Y∈R n×l , L is the number of types of fault conditions.

[0097] In one example, the fault detection model is implemented based on a random weight vector neural network, or the fault detection model is implemented based on a feedforward neural network, and the model parameters are weight parameters and / or biases of the fault detection model.

[0098] In one example, the fault detection model can be expressed as follows:

[0099]

[0100] Where, represents the vehicle fault state at time t, A(t) represents the matrix determined by the vehicle operation state data at time t, W b Represents model parameters.

[0101] In certain embodiments of the present application, the first vehicle operation data includes vehicle engine status data, and the first fault status tag data is used to indicate a fault status of the vehicle engine.

[0102] Specifically, in the embodiment of the present application, the first vehicle operation data may include vehicle engine status data, thereby indicating the status of the vehicle engine. Accordingly, in the embodiment of the present application, the first fault status tag data is used to indicate the fault status of the vehicle engine.

[0103] In one example, the vehicle engine status data includes detecting the vehicle engine through sensors such as Hall sensors, thereby determining the engine speed, engine crankshaft position, engine shaft position, etc.

[0104] In one example, the fault status of the vehicle engine includes fault states such as engine rotation fault, engine crankshaft position fault, engine shaft position fault, and no fault.

[0105] Thus, in the embodiment of the present application, the model parameters can be correlated with the vehicle engine status data and the first fault state label data for the fault state of the vehicle engine, thereby enabling the fault detection model to accurately detect the fault state of the vehicle engine.

[0106] In certain embodiments of the present application, the model parameters include weight parameters.

[0107] Specifically, in an embodiment of the present application, the electronic device can determine the weight parameters of the fault detection model based on the acquired first vehicle operation data and the first fault status label data corresponding to the first vehicle operation data, and construct the fault detection model based on the weight parameters.

[0108] In one example, the fault detection model can be expressed by formula (1), namely:

[0109]

[0110] Where, represents the vehicle fault state at time t, A(t) represents the matrix determined by the vehicle operation state data at time t, W b Represents the model parameters, which are also weight parameters.

[0111] In this way, in the embodiment of the present application, the fault detection model can be constructed based on the weight parameters determined by the first vehicle operation data and the first fault state label data, which reduces the difficulty of constructing the fault detection model to a certain extent, thereby further reducing the construction cost of the fault detection model.

[0112] See also Figure 2 In certain embodiments of the present application, step 01 includes:

[0113] 010: determining characteristic operating data according to the first vehicle operating data;

[0114] 011: Determine model parameters according to the characteristic operation data, the first vehicle operation data, and the first fault state label data.

[0115] The processor of the embodiment of the present application is also used to determine characteristic operation data based on the first vehicle operation data, and to determine model parameters based on the characteristic operation data, the first vehicle operation data and the first fault state label data.

[0116] Specifically, in an embodiment of the present application, characteristic operation data is determined based on the first vehicle operation data, and model parameters are determined based on the characteristic operation data, the first vehicle operation data and the first fault state label data.

[0117] Specifically, in order to ensure the validity of the model parameters, in the implementation mode of the present application, the electronic device can determine the characteristic operation data based on the first vehicle operation data, and then, the electronic device can determine the model parameters of the fault detection model based on the characteristic operation data, the first vehicle operation data and the first fault status label data.

[0118] In one example, the first vehicle operating data includes a plurality of sub-data, and the characteristic operating data is one or more of the plurality of sub-data.

[0119] In one example, the first vehicle operating data includes multiple sub-data. The electronic device may then calculate the correlation between the multiple sub-data and the first fault status tag data and use the sub-data with the highest correlation as the characteristic operating data, or use the top three sub-data with the highest correlation as the characteristic operating data. It should be understood that the number 3 is merely an example and may be adjusted based on actual circumstances.

[0120] In this way, in the embodiment of the present application, the characteristic operating data can be determined based on the first vehicle operating data, and the model parameters can be determined based on the characteristic operating data, the first vehicle operating data and the first fault status label data, thereby ensuring the reliability of the model parameters to a certain extent.

[0121] See also Figure 3 In certain embodiments of the present application, step 010 includes:

[0122] 0100: Perform preset mapping and flattening processing on the first vehicle operation data to determine characteristic operation data.

[0123] The processor of the embodiment of the present application is also used to perform preset mapping and flattening processing on the first vehicle operation data to determine characteristic operation data.

[0124] Specifically, in the embodiment of the present application, the first vehicle operation data may be subjected to preset mapping and flattening processing, so that the processed first vehicle operation data may be determined as characteristic operation data.

[0125] In one example, the mapping process can be understood as mapping the value of the first vehicle operating data to a preset interval, thereby obtaining a new first vehicle operating data.

[0126] In one example, the flattening process may be understood as performing an upward rounding or downward rounding operation on the first vehicle operation data.

[0127] Thus, in the embodiment of the present application, the first vehicle operation data may be subjected to a preset mapping and flattening process to determine the characteristic operation data, thereby achieving determination of the characteristic operation data.

[0128] See also Figure 4 In certain embodiments of the present application, step 0100 includes:

[0129] 01000: mapping the first vehicle operation data according to the randomly determined parameters to determine vehicle state mapping data;

[0130] 01001: Flatten the vehicle status mapping data to determine the characteristic operation data.

[0131] The processor of the embodiment of the present application is also used to map the first vehicle operation data according to randomly determined parameters to determine the vehicle state mapping data, and flatten the vehicle state mapping data to determine the characteristic operation data.

[0132] Specifically, in an embodiment of the present application, the electronic device may map the first vehicle operating data using randomly determined parameters to map the first vehicle operating data to a random range, thereby determining vehicle state mapping data. Subsequently, the vehicle state mapping data may be flattened to obtain characteristic operating parameters.

[0133] In one example, the process of acquiring characteristic operation data can be represented by the following formula:

[0134]

[0135]

[0136] Where X is the first vehicle operation data; η(·) represents the random mapping function; H iis the vehicle state mapping data obtained by random mapping for the i-th time; n e Indicates the number of random mapping operations; is to n e The matrix of the flattened combination of features obtained by random mapping, the matrix Each element in is the characteristic operation data; and are the random weight and constant of the i-th random mapping operation, that is, the randomly determined parameters.

[0137] In this way, in the embodiment of the present application, the first vehicle operation data can be mapped according to the randomly determined parameters to determine the vehicle state mapping data, and the vehicle state mapping data can be flattened to determine the characteristic operation data, thereby realizing the determination of the characteristic operation data.

[0138] See also Figure 5 In certain embodiments of the present application, step 011 includes:

[0139] 0110: Determine an orthogonal matrix and an upper triangular matrix according to the characteristic operation data and the first vehicle operation data;

[0140] 0111: Determine model parameters based on the orthogonal matrix, the upper triangular matrix and the first fault state label data.

[0141] The processor of the embodiment of the present application is also used to determine the orthogonal matrix and the upper triangular matrix based on the characteristic operation data and the first vehicle operation data, and to determine the model parameters based on the orthogonal matrix, the upper triangular matrix and the first fault state label data.

[0142] Specifically, in the implementation manner of the present application, an orthogonal matrix and an upper triangular matrix can be constructed based on the characteristic operation data and the first vehicle operation data, and then, the model parameters of the fault detection model can be determined based on the orthogonal matrix, the upper triangular matrix and the first fault state label data.

[0143] Thus, in the embodiment of the present application, the model parameters can be determined based on the orthogonal matrix and the upper triangular matrix determined by the characteristic operation data and the first vehicle operation data, combined with the first fault state label data.

[0144] See also Figure 6 In certain embodiments of the present application, step 0110 includes:

[0145] 01100: determining an expansion matrix according to the characteristic operation data and the first vehicle operation data;

[0146] 01101: Determine the orthogonal matrix and upper triangular matrix based on the expanded matrix.

[0147] The processor of the embodiment of the present application is further configured to determine an expansion matrix based on the characteristic operation data and the first vehicle operation data, and to determine an orthogonal matrix and an upper triangular matrix based on the expansion matrix.

[0148] Specifically, in an embodiment of the present application, the characteristic operation data and the first vehicle operation data can be used to determine an extended matrix, and then an orthogonal matrix and an upper triangular matrix are determined based on the extended matrix, such as by performing matrix decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix.

[0149] In one example, the process of obtaining the extended matrix can be represented by the above formula (3) and formula (4), as well as the following formula (4), namely:

[0150]

[0151] Where A is the expansion matrix.

[0152] Thus, in the embodiment of the present application, the extended matrix can be determined according to the characteristic operation data and the first vehicle operation data, and the orthogonal matrix and the upper triangular matrix can be determined according to the extended matrix, thereby achieving the determination of the orthogonal matrix and the upper triangular matrix.

[0153] See also Figure 7 In certain embodiments of the present application, step 01101 includes:

[0154] 011010: Perform orthogonal decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix.

[0155] The processor of the embodiment of the present application is further used to perform orthogonal decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix

[0156] Specifically, in the embodiments of the present application, an orthogonal matrix and an upper triangular matrix may be obtained by performing orthogonal decomposition on the extended matrix.

[0157] In one example, a fault detection model is implemented using a random weight vector neural network (RWVN), a feedforward neural network. The model parameters are typically calculated using a pseudo-inverse method. However, directly calculating the pseudo-inverse of a matrix can encounter numerical instability, especially when some singular values of the matrix are very close to zero. This can lead to inaccurate or unstable solutions.

[0158] Therefore, orthogonal decomposition, as a matrix decomposition method, is usually more stable numerically than directly calculating the pseudo-inverse. This is because orthogonal decomposition decomposes the matrix into the product of an upper triangular matrix and an orthogonal matrix through a series of orthogonal transformations. These transformations are more resistant to rounding errors in floating-point operations. Therefore, orthogonal decomposition can be used as a more reasonable alternative method to solve the model parameters of random weight vector neural networks.

[0159] Thus, in the embodiment of the present application, the orthogonal matrix and the upper triangular matrix may be determined by orthogonal decomposition.

[0160] See also Figure 8 In certain embodiments of the present application, step 011010 includes:

[0161] 0110100: Determine the Householder vector based on the expansion matrix;

[0162] 0110101: Construct Householder matrix based on Householder vector;

[0163] 0110102: Determine the orthogonal matrix and upper triangular matrix based on the expanded matrix and Householder matrix.

[0164] The processor of the embodiment of the present application is further used to determine the Householder vector according to the extended matrix, to construct the Householder matrix according to the Householder vector, and to determine the orthogonal matrix and the upper triangular matrix according to the extended matrix and the Householder matrix.

[0165] Specifically, in the embodiments of the present application, the orthogonal decomposition of the extended matrix can be achieved through Householder transformation.

[0166] In one example, the orthogonal matrix and the upper triangular matrix can be obtained according to the above formula (4) and the following steps, namely:

[0167] First, assume that the expansion matrix is

[0168] Next, let A 0 =A,a=A k (k:n A ,k), K=0,1,…,m A Among them, n A Indicates the number of rows of matrix A, m A Indicates the number of columns of matrix A.

[0169] Next, calculate the Householder vector v, which is:

[0170] v=a-sign(a1)‖a‖e1 (5)

[0171] in, a1 represents the first element of a.

[0172] Then, construct the Householder matrix H′ k ,Right now:

[0173]

[0174] u=v / ‖v‖ (7)

[0175] Then, the new Householder matrix H k It can be obtained by the following formula:

[0176]

[0177] Then, A k Perform Householder transformation, that is:

[0178] A k+1 =H k A k

[0179] Therefore, the orthogonal matrix Q and upper triangular matrix R of the extended matrix A can be obtained by the following formula:

[0180]

[0181] Finally, the model parameter W of the fault detection model is calculated according to the following formula:

[0182] W=R -1 Q T Y (11)

[0183] Wherein, W is a connection weight parameter of the fault detection model, and Y is the first fault state label data corresponding to the first vehicle operation data X.

[0184] Therefore, the fault detection model can be represented in the form of the above formula (1), that is:

[0185]

[0186] Where, W represents the vehicle fault state at time t, and A(t) represents the matrix determined by the vehicle operation state data at time t, or the expanded matrix determined by the vehicle operation state data at time t. b Represents model parameters.

[0187] In this way, in the implementation mode of the present application, the Householder vector can be determined based on the extended matrix, and the Householder matrix can be constructed based on the Householder vector, and the orthogonal matrix and the upper triangular matrix can be determined based on the extended matrix and the Householder matrix, thereby realizing the orthogonal decomposition based on the Householder transform.

[0188] In addition, to more clearly illustrate the process of constructing the above formula (1) based on the orthogonal matrix Q and the upper triangular matrix R of the extended matrix A, please refer to the following content, namely:

[0189] First, assume that the expansion matrix determined according to the first vehicle operating state data X is A b , and A b The orthogonal decomposition result of can be expressed as the following process, namely:

[0190] A b =Q b R b ,b=0,1,2,… (12)

[0191] Then, if the expansion matrix determined according to the vehicle operation status data obtained during the vehicle operation is A I , then the new expansion matrix can be expressed as follows:

[0192]

[0193] For A b+1 , A b+1 Based on Q b and R b The orthogonal decomposition result can be expressed as follows:

[0194]

[0195] Where h = A I (R b ) + , Q b+1 =[(Q b ) T h T ] T , R b+1 =R b .

[0196] Furthermore, based on the vehicle operating status data obtained during vehicle operation, the process of establishing the fault detection model can be expressed as follows:

[0197]

[0198] Among them, Y IIndicates the status label corresponding to the vehicle operation status data obtained during the vehicle operation process.

[0199] Assume that the following formula holds true:

[0200]

[0201] Based on Woodbury's theorem, P b+1 It is expressed as the following formula:

[0202] P b+1 =((Q b ) T Q b +h T h)=P b -P b h T (I+hP b h T )hP b (18)

[0203] Among them, P b =(Q b ) T Q b .

[0204] Furthermore, during vehicle operation, the model parameters can be expressed as:

[0205]

[0206] Among them, W b is a model parameter determined according to the first vehicle operating state data.

[0207] Therefore, for the vehicle operation status data x(t) collected at any time t, x(t) can be randomly mapped to obtain the characteristic operation data at time t:

[0208]

[0209] The vehicle operation status data x(t) and characteristic operation data collected at time t Flatten the combination as follows to get the real-time expansion matrix:

[0210]

[0211] Therefore, the fault detection model at time t can be expressed as:

[0212]

[0213] See also Figure 9Corresponding to the above-mentioned method for constructing a fault detection model, the present application also provides a fault detection method, which includes:

[0214] 03: Determine the vehicle fault state according to the fault detection model and the acquired second vehicle operation data, wherein the fault detection model is constructed according to the above-mentioned fault detection model construction method.

[0215] The present application also provides an electronic device comprising a memory and a processor. The fault detection method of the present application can be implemented by the electronic device of the present application. Specifically, the memory stores a computer program, and the processor is configured to determine a vehicle fault state based on a fault detection model constructed according to the aforementioned fault detection model construction method and acquired second vehicle operating data.

[0216] Specifically, in the implementation mode of the present application, the fault state of the vehicle at the current moment can be predicted based on the fault model constructed by the first vehicle operation data and the first fault state label data corresponding to the vehicle operation data, combined with the vehicle operation data obtained at the current moment, that is, the second vehicle operation data, to obtain the vehicle fault state at the current moment.

[0217] It is understandable that the fault detection model and the process of constructing the fault detection model can refer to the aforementioned content and formulas, and to avoid repetition, they will not be described here.

[0218] Thus, in the embodiment of the present application, the vehicle fault state can be determined based on the fault detection model and the acquired second vehicle operation data, thereby realizing the detection of the vehicle fault state. Moreover, the model parameters of the fault detection model for detecting the vehicle fault state can be determined by the acquired first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, so that the model parameters of the fault detection model can be determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, thereby avoiding the need to determine the model parameters through model training, and reducing the construction cost of the fault detection model to a certain extent. In addition, since the model parameters are determined based on the first vehicle operation data and the first fault state label data corresponding to the first vehicle operation data, the correlation between the model parameters and the vehicle state data and the first fault state label data can be guaranteed, thereby ensuring the detection accuracy of the fault detection model constructed based on the model parameters.

[0219] See also Figure 10 In certain embodiments of the present application, the fault detection method further includes:

[0220] 04: Update the fault detection model according to the second vehicle operating data and the vehicle fault status.

[0221] The processor of the embodiment of the present application is also used to update the fault detection model based on the second vehicle operating data and the vehicle fault status.

[0222] Specifically, during the actual operation of a vehicle, various external factors are complex and changeable. The operating conditions of the vehicle and its intelligent driving system will gradually change over time during long-term operation. Therefore, to ensure that the fault diagnosis model can adapt to real-time changing operating conditions, the model can continuously learn information from real-time operating process data during real-time fault detection and adjust model parameters to adapt to the changing operating conditions. Therefore, in the embodiments of the present application, the model can be updated based on the model output results, allowing the model to dynamically adapt to the changing operating conditions.

[0223] Specifically, in the embodiment of the present application, the electronic device can also update the fault detection model based on the status data of the vehicle at the current moment, that is, the second vehicle operation data, and the vehicle fault status determined by the second vehicle operation data and the fault detection model.

[0224] In this way, in the embodiment of the present application, the fault detection model can be updated based on the second vehicle operation data obtained during the vehicle operation process, as well as the vehicle fault state determined by the vehicle fault state and the fault detection model. This can achieve real-time updating of the fault detection model to a certain extent, and then enable the fault detection model to be updated as the vehicle operation process progresses, thereby improving the detection accuracy of the fault detection model to a certain extent.

[0225] See also Figure 11 In certain embodiments of the present application, step 04 further includes:

[0226] 040: Determine second fault state label data according to the vehicle fault state;

[0227] 041: Update the fault detection model according to the second vehicle operation data and the second fault status label data.

[0228] The processor of the embodiment of the present application is further configured to determine second fault state label data according to the vehicle fault state, and to update the fault detection model according to the second vehicle operation data and the second fault state label data.

[0229] Specifically, in an embodiment of the present application, in order to update the fault detection model, the second fault status label data corresponding to the second vehicle operation data can be determined according to the vehicle fault status, that is, the data corresponding to the second vehicle operation data and representing the actual fault status of the vehicle can be indicated.

[0230] It is understandable that in the embodiments of the present application, the vehicle fault state determined according to the fault detection model may be incorrect, or in other words, the vehicle fault state is the result of model reasoning, and the model may have reasoning errors.

[0231] It can also be understood that in order to ensure that the model is updated in the correct direction, the implementation method of the present application can determine the second fault state label data according to the vehicle fault state, so that the model can be updated based on the second fault state label data, thereby ensuring the correct update of the model.

[0232] Thus, in the embodiment of the present application, the second fault state label data can be determined according to the vehicle fault state, and the fault detection model can be updated according to the second vehicle operation data and the second fault state label data, thereby realizing the update of the fault detection model.

[0233] In certain embodiments of the present application, the vehicle fault state includes predicted probabilities of multiple fault types, and the second fault state label data includes predicted label values of multiple fault types. Furthermore, step 041 includes:

[0234] The predicted label value of a target fault among multiple fault types is determined as a first value, and the predicted label values of other faults are determined as a second value, wherein the target fault is the fault with the highest predicted probability among the multiple fault types.

[0235] The processor of the embodiment of the present application is also used to determine the predicted label value of the target fault among multiple fault types as a first value, and determine the predicted label values of other faults as a second value, wherein the target fault is the fault with the highest predicted probability among multiple fault types.

[0236] Specifically, in the embodiment of the present application, since it is difficult to obtain the second fault state label data, the embodiment of the present application can set the predicted label value of each type in the second fault state label data according to the vehicle fault state output by the model, such as setting it to 1 or 0.

[0237] In one example, the first value is 1 and the second value is 0.

[0238] In one example, the process of setting the predicted label value for each category in the second fault state label data can be expressed by the following formula:

[0239]

[0240] in, is the second fault state tag data; are the predicted label values for the first, second, …, and first fault types; is the vehicle fault status. l is the number of fault types.

[0241] In one example, the second fault status tag data is obtained by the above formulas (21) and (22): After updating the fault detection model, the updated fault model can be expressed as:

[0242]

[0243] Thus, in the embodiment of the present application, the predicted label value of the target fault among multiple fault types is determined as the first value, and the predicted label values of other faults are determined as the second value, thereby achieving the determination of the second fault state label data.

[0244] Furthermore, by using the above-mentioned method for determining the second fault state label data, the need for manual labeling of the second fault state label data can be reduced, thereby reducing the labeling cost.

[0245] To more clearly illustrate the acquisition and update process of the fault detection model in the embodiment of this application, please refer to Figure 12 , Figure 12 This is a schematic diagram of an application scenario in some embodiments of the present application, that is, Figure 12 As shown, the embodiment of the present application proposes a fault model construction method and a fault detection method based on matrix decomposition and random weight vector neural network, which specifically includes: in order to effectively extract feature data, a random mapping method is adopted; in order to enhance the parameter stability in the modeling and updating process, a model parameter solution method based on orthogonal decomposition is proposed; in order to be able to update model parameters according to real-time data, a strategy for updating the model according to the model output results is designed; in order to reduce the cost of model updating, the second fault state label data is determined based on the method shown in the above formulas (21) and (22).

[0246] To more clearly illustrate the validity of the vehicle fault status output by the fault detection model in the embodiment of the present application, please refer to Table 1, Table 2, Figure 13 、 Figure 14 、 Figure 15 、 Figure 16 and Figure 17 , Figure 13 、 Figure 14 、 Figure 15 、 Figure 16 and Figure 17 These are all schematic diagrams of application scenarios in certain embodiments of the present application.

[0247] Specifically, experiments were conducted based on a publicly available dataset of steel plate faults. That is, based on the publicly available dataset of steel plate faults and the method for constructing a fault model provided in the embodiments of the present application, a fault detection model for detecting steel plate faults was constructed, and the model was compared with the fault detection scheme in the related art. The fault types in the public dataset are shown in Table 1, and the process data collected in the public dataset are shown in Table 2. It can be understood that the content shown in Table 1 is equivalent to or similar to the first vehicle operation data in the embodiments of the present application, and the content shown in Table 2 is equivalent to or similar to the "first fault state label data corresponding to the first vehicle operation data" in the embodiments of the present application.

[0248] Table 1

[0249] serial number Fault type serial number Fault type 1 Pastry 5 Dirtiness 2 Z_Scratch 6 Bumps 3 K_Scatch 7 Other_Faults 4 Stains

[0250] Table 2

[0251]

[0252] Based on the steel plate failure dataset, parameter experiments and comparative experiments are carried out.

[0253] Among them, in the parameter experiment, the parameter n of the fault detection model is e and n f , select different experimental parameters to verify the performance of the fault detection model under different parameters, such as Figure 13 In the figure, the 3D grid search visualization results of the model performance of the fault detection model under different parameters are shown.

[0254] In the comparative experiment, in order to verify the method proposed in this application, it is compared with the adaptive random forest (ARF), broad learning system (BLS), broad learning system based on Sherman-Morrison formula (BLS-SMW), online sequential extreme learning machine (OSELM), and random vector function-link neural network (RVFLNN) in the related technologies.

[0255] Among them, the adaptive random forest is an integrated learning model for classifying evolving data streams. It adapts to concept drift in the data stream by using multiple decision trees with drift detection and model update mechanisms. The width learning system is a neural network framework that can operate without iterative depth adjustment. The width learning system based on the Sherman-Morrison formula is an improvement on the width learning system based on the Sherman-Morrison formula for processing online data stream tasks. The online sequential extreme learning machine is a variant of the feedforward neural network specifically designed to process streaming data so that regression and classification can be performed using a single hidden layer. The random weight vector neural network is a feedforward neural network with a single hidden layer, designed to effectively handle regression and classification.

[0256] Based on the above comparison method and the proposed method, experiments were conducted on the steel plate failure dataset. In order to avoid the uncertainty caused by the randomness of the model, 5 test experiments were conducted. In the experiment, 4 different experimental settings were selected: 1) the number of historical process data samples n = 50, the number of real-time process data n t =500, status tag category l = 3; 2) number of historical process data samples n = 50, number of real-time process data n t =1000, status tag category l = 6; 3) number of historical process data samples n = 100, number of real-time process data n t =500, status tag category l = 3; 4) number of historical process data samples n = 100, number of real-time process data n t =1000, state label category l = 6. Table 3 shows the average accuracy and standard deviation of each method involved in the experiment under the above four experimental settings. The learning curves of the experimental methods under different experimental settings are shown in Table 3. Figures 14 to 17 shown.

[0257] Table 3

[0258]

[0259] It is worth noting that in Table 3, Figures 14 to 17 In, n t is the test number of real-time process data, and “*” (or “RVFLNN-QR”) represents the “fault detection model for detecting steel plate faults” constructed based on the fault detection model construction method provided in the embodiment of the present application.

[0260] It is also worth noting that, according to Table 3, Figures 14 to 17 From the experimental results shown, it can be seen that the "fault detection model for detecting steel plate faults" constructed based on the fault detection model construction method provided in the embodiment of the present application has good performance.

[0261] An embodiment of the present application further provides a vehicle, which includes the above-mentioned electronic device.

[0262] The embodiments of the present application also provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, it implements the above-mentioned method for constructing a fault detection model or the above-mentioned fault detection method.

[0263] The embodiments of the present application also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned method for constructing a fault detection model, or implements the above-mentioned fault detection method.

[0264] In the description of this specification, the descriptions with reference to the terms "particularly", "further", "particularly", "understandably", etc. are intended to mean that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms are not intended to refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0265] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0266] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for constructing a fault detection model, characterized in that: include: determining model parameters based on the acquired first vehicle operation data and first fault state label data corresponding to the first vehicle operation data; A fault detection model for detecting a vehicle fault state is determined according to the model parameters.

2. The method according to claim 1, characterized in that The first vehicle operation data includes vehicle engine status data, and the first fault status tag data is used to indicate a fault status of the vehicle engine.

3. The method according to claim 1, characterized in that The model parameters include weight parameters.

4. The method according to claim 1, wherein The determining of the model parameters according to the acquired first vehicle operation data and the first fault status label data corresponding to the first vehicle operation data includes: determining characteristic operating data based on the first vehicle operating data; The model parameters are determined according to the characteristic operation data, the first vehicle operation data, and the first fault status label data.

5. The method according to claim 4, characterized in that The determining characteristic operation data according to the first vehicle operation data includes: The first vehicle operation data is subjected to preset mapping and flattening processing to determine the characteristic operation data.

6. The method according to claim 5, characterized in that The performing of preset mapping and flattening processing on the first vehicle operation data to determine the characteristic operation data includes: Performing mapping processing on the first vehicle operation data according to randomly determined parameters to determine vehicle state mapping data; The vehicle state mapping data is flattened to determine the characteristic operation data.

7. The method according to claim 4, characterized in that The determining the model parameters according to the characteristic operation data, the first vehicle operation data, and the first fault status label data includes: determining an orthogonal matrix and an upper triangular matrix according to the characteristic operation data and the first vehicle operation data; The model parameters are determined according to the orthogonal matrix, the upper triangular matrix and the first fault state label data.

8. The method according to claim 4, characterized in that Determining an orthogonal matrix and an upper triangular matrix according to the characteristic operation data and the first vehicle operation data includes: determining an expansion matrix according to the characteristic operation data and the first vehicle operation data; The orthogonal matrix and the upper triangular matrix are determined according to the extended matrix.

9. The method according to claim 8, characterized in that The determining the orthogonal matrix and the upper triangular matrix according to the extended matrix includes: Performing orthogonal decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix.

10. The method according to claim 9, characterized in that The performing orthogonal decomposition on the extended matrix to determine the orthogonal matrix and the upper triangular matrix includes: determining a Householder vector according to the expansion matrix; Constructing a Householder matrix according to the Householder vector; The orthogonal matrix and the upper triangular matrix are determined according to the expanded matrix and the Householder matrix.

11. A fault detection method, characterized in that: include: Determine a vehicle fault state according to a fault detection model and the acquired second vehicle operation data, wherein the fault detection model is constructed according to the method according to any one of claims 1-10.

12. The method according to claim 11, characterized in that The method further comprises: The fault detection model is updated according to the second vehicle operating data and the vehicle fault status.

13. The method according to claim 12, characterized in that The updating of the fault detection model according to the second vehicle operating data and the vehicle fault state includes: Determining second fault state label data according to the vehicle fault state; The fault detection model is updated according to the second vehicle operation data and the second fault status label data.

14. The method according to claim 13, characterized in that The vehicle fault state includes predicted probabilities of multiple fault types, the second fault state label data includes predicted label values of the multiple fault types, and determining the second fault state label data according to the vehicle fault state includes: The predicted label value of a target fault among the multiple fault types is determined as a first value, and the predicted label values of the other faults are determined as a second value, wherein the target fault is the fault with the highest predicted probability among the multiple fault types.

15. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method according to any one of claims 1 to 14 is implemented.

16. A vehicle, characterized in that: The vehicle includes the apparatus of claim 15 .

17. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by one or more processors, the method according to any one of claims 1 to 14 is implemented.

18. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 14 is implemented.

Images