Vehicle fault prediction method, device and equipment and storage medium

Abstract

The invention discloses a vehicle fault prediction method, device and equipment and a storage medium, and relates to the technical field of vehicle engineering, and the method comprises the steps: carrying out the preprocessing of vehicle data reflecting the vehicle operation state, carrying out the working condition division of the obtained processed data through a target working condition recognition model, and obtaining a target working condition recognition model; determining corresponding time sequence features based on the processed data corresponding to the obtained working condition segments, performing weighted fusion on the time sequence features by using historical maintenance records of the vehicle, and screening obtained dynamic feature vectors based on a preset feature selection algorithm, training the initial vehicle fault prediction model by using the obtained target feature vector to obtain a target vehicle fault prediction model; and performing statistical analysis on the historical fault data of the vehicle, determining a fault threshold value based on an analysis result and the real-time operation state data of the vehicle, and determining a fault prediction result based on the to-be-detected vehicle data by using the target vehicle fault prediction model and the fault threshold value. And the accuracy of vehicle fault prediction is improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle engineering technology, and in particular to a vehicle fault prediction method, device, equipment and storage medium. Background Art

[0002] Existing vehicle fault prediction solutions mainly rely on rule engines and simple statistical analysis, lacking the ability to deeply mine and model complex data. Not only can they not accurately predict potential faults, resulting in unscientific maintenance plans and possible over-maintenance or under-maintenance, but they also cannot achieve real-time fault prediction, resulting in the inability to take timely maintenance measures when a fault occurs, increasing vehicle safety risks.

[0003] As can be seen from the above, how to improve the accuracy of vehicle fault prediction is an urgent problem to be solved. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a vehicle fault prediction method, device, equipment and storage medium that can improve the accuracy of vehicle fault prediction. The specific solution is as follows:

[0005] In a first aspect, the present application provides a vehicle fault prediction method, comprising:

[0006] collecting vehicle data reflecting the operating status of the vehicle, preprocessing the vehicle data to obtain processed data, and then dividing the processed data into operating conditions using a target operating condition identification model to obtain operating condition segments; the target operating condition identification model is an operating condition identification model determined based on a long short-term memory network or a convolutional neural network;

[0007] Determining corresponding time series features based on the processed data corresponding to each of the operating conditions, weighting and fusing the time series features using the vehicle's historical maintenance records to obtain a dynamic feature vector, then screening the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and using the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm;

[0008] A statistical analysis is performed on the historical fault data of the vehicle, and a fault threshold is determined based on the analysis result and the real-time operating status data of the vehicle, and a fault prediction result is determined based on the data of the vehicle to be detected and using the target vehicle fault prediction model and the fault threshold.

[0009] Optionally, the target operating condition identification model is used to divide the processed data into operating condition segments to obtain operating condition segments, including:

[0010] Constructing an initial operating condition recognition model based on a long short-term memory network, and training the initial operating condition recognition model using the processed data to obtain a target operating condition recognition model;

[0011] Using the target operating condition recognition model to extract features from the time series data corresponding to the processed data, and converting the obtained data features to obtain an operating condition classification result;

[0012] Based on the working condition classification result and using a preset activation function, the probability distribution of each working condition category is determined, and the working condition categories corresponding to the processed data are determined using the probability distribution and a preset decision rule to obtain each working condition segment.

[0013] Optionally, the target operating condition identification model is used to divide the processed data into operating condition segments to obtain operating condition segments, including:

[0014] Constructing an initial operating condition recognition model based on a convolutional neural network, and training the initial operating condition recognition model using the processed data to obtain a target operating condition recognition model;

[0015] performing a convolution operation on the processed data using the target operating condition recognition model to obtain local features of different scales, and performing a pooling operation on the local features to obtain pooled features;

[0016] The processed data is classified into working condition categories based on the pooled features to obtain various working condition segments.

[0017] Optionally, determining corresponding time series features based on the processed data corresponding to each of the operating conditions, and weightedly fusing the time series features using historical maintenance records of the vehicle to obtain a dynamic feature vector, includes:

[0018] Determine corresponding time series features based on the change pattern and feature information of the processed data corresponding to each of the operating sections in the time dimension;

[0019] The vehicle's historical maintenance records are used to assign weights to the time series features, and weighted fusion is performed based on the time series features and the corresponding weights to obtain a dynamic feature vector.

[0020] Optionally, the screening of the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and using the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model, includes:

[0021] Screening the dynamic feature vector using a chi-square test and recursive feature elimination to obtain a target feature vector;

[0022] Dividing the target feature vector based on a preset division ratio to obtain corresponding training set, validation set, and test set;

[0023] Constructing an initial vehicle fault prediction model based on a machine learning algorithm, and training the initial vehicle fault prediction model using the training set to obtain a trained vehicle fault prediction model;

[0024] The trained vehicle fault prediction model is verified and tested using the verification set and the test set to obtain a target vehicle fault prediction model.

[0025] Optionally, performing statistical analysis on historical fault data of the vehicle and determining a fault threshold based on the analysis result and real-time operating status data of the vehicle includes:

[0026] Determining vehicle fault data based on historical fault data and fault-related data of the vehicle, and classifying the vehicle fault data into fault types to obtain each fault type corresponding to the vehicle fault data;

[0027] Analyze the characteristic parameters corresponding to each fault type, and determine the distribution range of the characteristic parameters under different fault types based on the analysis results and a preset statistical method;

[0028] Acquiring real-time operating status data of the vehicle, and adjusting weights corresponding to the characteristic parameters based on the real-time operating status data to obtain adjusted parameter weights;

[0029] Performing weighted fusion on the historical fault data of the characteristic parameters and the real-time operating status data based on an exponential sliding average algorithm to obtain an exponential sliding average;

[0030] The fault threshold is determined by utilizing the exponential sliding average value in combination with the corresponding distribution range and the adjusted parameter weight, and based on a preset interval time, the step of determining the vehicle fault data based on the historical fault data and fault-related data of the vehicle is jumped to update the fault threshold.

[0031] Optionally, determining a fault prediction result based on the to-be-detected vehicle data and using the target vehicle fault prediction model and the fault threshold includes:

[0032] Acquiring data of a vehicle to be detected, inputting the data of the vehicle to be detected into the target vehicle fault prediction model to obtain a fault type and a corresponding fault probability, and determining whether the fault probability exceeds the fault threshold;

[0033] If the fault occurrence probability exceeds the fault threshold, a corresponding target fault type is determined based on the fault occurrence probability, and a corresponding maintenance suggestion is generated based on the target fault type, triggering a fault warning for the vehicle to be detected.

[0034] In a second aspect, the present application provides a vehicle fault prediction device, comprising:

[0035] a working condition classification module, configured to collect vehicle data reflecting the vehicle operating status, pre-process the vehicle data to obtain processed data, and then classify the processed data into working conditions using a target working condition identification model to obtain various working condition segments; the target working condition identification model is a working condition identification model determined based on a long short-term memory network or a convolutional neural network;

[0036] a fault prediction model training module, configured to determine corresponding time series features based on the processed data corresponding to each of the operating conditions, weightedly fuse the time series features using the vehicle's historical maintenance records to obtain a dynamic feature vector, then filter the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and use the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm;

[0037] A fault prediction result determination module is used to perform statistical analysis on the historical fault data of the vehicle, and determine a fault threshold based on the analysis results and the real-time operating status data of the vehicle, and determine a fault prediction result based on the vehicle data to be detected and using the target vehicle fault prediction model and the fault threshold.

[0038] In a third aspect, the present application provides an electronic device, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor is used to execute the computer program to implement the aforementioned vehicle fault prediction method.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program implements the aforementioned vehicle fault prediction method when executed by a processor.

[0042] The present application collects vehicle data used to reflect the operating status of a vehicle, and pre-processes the vehicle data to obtain processed data, and then uses a target operating condition identification model to divide the processed data into operating conditions to obtain various operating condition segments; the target operating condition identification model is an operating condition identification model determined based on a long short-term memory network or a convolutional neural network; the corresponding time series features are determined based on the processed data corresponding to each of the operating condition segments, and the time series features are weighted and fused using the vehicle's historical maintenance records to obtain a dynamic feature vector, and then the dynamic feature vector is screened based on a preset feature selection algorithm to obtain a target feature vector, and the target feature vector is used to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm; the historical fault data of the vehicle is statistically analyzed, and a fault threshold is determined based on the analysis results and the real-time operating status data of the vehicle, and a fault prediction result is determined based on the vehicle data to be detected and using the target vehicle fault prediction model and the fault threshold.

[0043] As can be seen from the above, the present application utilizes the constructed target operating condition classification model to divide the processed vehicle data into operating conditions, so as to extract the time series features of the processed data corresponding to each operating condition segment obtained by the division, and utilizes the historical maintenance records of the vehicle to weightedly fuse the time series features. The obtained dynamic feature vector is then screened using a preset feature selection algorithm to obtain a target feature vector, and the target feature vector is used to train the initial vehicle fault prediction model to improve the training efficiency and prediction accuracy of the target vehicle fault prediction model. The fault threshold is then determined based on the vehicle's historical fault data and real-time operating status data. In this way, the use of the target vehicle fault prediction model and the fault threshold can achieve accurate prediction and timely warning of vehicle faults, thereby improving the operational safety and reliability of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0045] Figure 1 This is a flow chart of a vehicle fault prediction method disclosed in this application;

[0046] Figure 2 This is a schematic structural diagram of a vehicle fault prediction device disclosed in this application;

[0047] Figure 3 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] Currently, existing vehicle fault prediction solutions rely heavily on rule engines and simple statistical analysis, lacking the ability to deeply mine and model complex data. This makes it difficult to accurately predict potential faults, leading to unscientific maintenance plans and inadequate maintenance. Furthermore, the inability to predict in real time makes it difficult to initiate timely maintenance when a fault occurs, further increasing vehicle safety risks. To this end, this application provides a vehicle fault prediction method that utilizes a target vehicle fault prediction model and fault thresholds to accurately predict and promptly warn of vehicle faults, thereby improving the operational safety and reliability of the vehicle.

[0050] See also Figure 1 As shown, an embodiment of the present invention discloses a vehicle fault prediction method, comprising:

[0051] Step S11: Collect vehicle data reflecting the vehicle operating status, pre-process the vehicle data to obtain processed data, and then use the target operating condition identification model to divide the processed data into operating conditions to obtain various operating condition segments; the target operating condition identification model is a operating condition identification model determined based on a long short-term memory network or a convolutional neural network.

[0052] In this embodiment, sensors are deployed in key vehicle components (such as the engine, transmission, brake system, and battery) to collect data reflecting the vehicle's operating status, such as temperature, pressure, vibration, and current, to obtain vehicle data. A distributed storage system is used to store the raw data corresponding to this vehicle data, and a time series database is used to store the time-related sensor data within this vehicle data. Interpolation and filtering techniques are then used to remove noise, missing values, and outliers from this vehicle data to obtain processed data. An initial operating condition recognition model is then constructed based on a long-short-term memory network or a convolutional neural network. This initial operating condition recognition model is trained using this processed data to obtain a target operating condition recognition model. This target operating condition recognition model is then used to perform operating condition recognition on the processed data to obtain various operating condition segments.

[0053] In one specific embodiment, an initial operating condition identification model is constructed based on a long short-term memory network, and the initial operating condition identification model is trained using the processed data to obtain a target operating condition identification model. The processed data is then used to obtain corresponding time series data, and feature extraction is performed on the time series data based on the target operating condition identification model to obtain data features, such as mean, variance, and peak value. The data features are then transformed to obtain an operating condition classification result. Based on the operating condition classification result and using a softmax activation function, a probability distribution for each operating condition category is determined. The operating condition categories corresponding to the processed data are determined using the probability distribution and a preset decision rule to obtain each operating condition segment. Specifically, using the target operating condition identification model to classify the processed data into operating conditions to obtain various operating condition segments includes: constructing an initial operating condition identification model based on a long short-term memory network, and training the initial operating condition identification model using the processed data to obtain a target operating condition identification model; using the target operating condition identification model to extract features from time series data corresponding to the processed data, and converting the obtained data features to obtain an operating condition classification result; based on the operating condition classification result and using a preset activation function, determining the probability distribution of each operating condition category, and using the probability distribution and preset decision rules to determine the operating condition category corresponding to the processed data to obtain various operating condition segments. If the probability distribution of the acceleration operating condition is the largest, the sample corresponding to the processed data is classified as an acceleration operating condition segment.

[0054] In another specific embodiment, an initial working condition recognition model is constructed based on a convolutional neural network, and the processed data is used to train the initial working condition recognition model to obtain a target working condition recognition model; the processed data is convolved using the convolution kernel in the target working condition recognition model to obtain local features of different scales, and the local features are pooled to reduce the dimension of the data while retaining the main features to obtain pooled features, and then the processed data is classified according to working conditions based on the pooled features and using the fully connected layer of the target working condition recognition model to obtain various working condition segments. Specifically, the use of the target working condition identification model to divide the processed data into working conditions to obtain various working condition segments includes: constructing an initial working condition identification model based on a convolutional neural network, and training the initial working condition identification model using the processed data to obtain a target working condition identification model; using the target working condition identification model to perform a convolution operation on the processed data to obtain local features of different scales, and performing a pooling operation on the local features to obtain pooled features; and classifying the processed data into working conditions based on the pooled features to obtain various working condition segments.

[0055] Step S12: determining corresponding time series features based on the processed data corresponding to each of the operating conditions, performing weighted fusion on the time series features using the historical maintenance records of the vehicle to obtain a dynamic feature vector, then screening the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and using the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm.

[0056] In this embodiment, after obtaining each operating condition segment, the corresponding time series features, such as short-term root mean square, peak value, and spectrum, are determined based on the variation patterns and characteristic information of the processed data corresponding to each operating condition segment in the time dimension. A domain knowledge-driven feature weight allocation mechanism is then designed in combination with the vehicle's historical maintenance records to assign weights to each of the time series features based on the feature weight allocation mechanism, and weighted fusion is performed based on each of the time series features and the corresponding weights to obtain a dynamic feature vector. Specifically, determining the corresponding time series features based on the processed data corresponding to each of the operating conditions segments, and weighted fusion of the time series features using the vehicle's historical maintenance records to obtain a dynamic feature vector, includes: determining the corresponding time series features based on the variation patterns and characteristic information of the processed data corresponding to each of the operating conditions segments in the time dimension; assigning weights to each of the time series features using the vehicle's historical maintenance records, and weighted fusion of each of the time series features and the corresponding weights to obtain a dynamic feature vector.

[0057] It is understood that after obtaining the dynamic feature vector, the dynamic feature vector is screened based on a preset feature selection algorithm to obtain a target feature vector; the preset feature selection algorithm includes a chi-square test and recursive feature elimination. The target feature vector is then partitioned according to a preset partitioning ratio to obtain corresponding training sets, validation sets, and test sets. An initial vehicle fault prediction model is then constructed based on a machine learning algorithm. The initial vehicle fault prediction model is trained using the training set to obtain a trained vehicle fault prediction model. The trained vehicle fault prediction model is then validated and tested using the validation set and the test set to improve the model's performance and obtain a target vehicle fault prediction model.

[0058] Specifically, the dynamic feature vector is screened based on a preset feature selection algorithm to obtain a target feature vector, and the target feature vector is used to train the initial vehicle fault prediction model to obtain a target vehicle fault prediction model, including: using a chi-square test and recursive feature elimination to screen the dynamic feature vector to obtain a target feature vector; dividing the target feature vector based on a preset division ratio to obtain corresponding training sets, validation sets, and test sets; constructing an initial vehicle fault prediction model based on a machine learning algorithm, and using the training set to train the initial vehicle fault prediction model to obtain a trained vehicle fault prediction model; using the validation set and the test set to verify and test the trained vehicle fault prediction model to obtain a target vehicle fault prediction model.

[0059] Step S13: Statistically analyze the historical fault data of the vehicle, determine a fault threshold based on the analysis results and the real-time operating status data of the vehicle, and determine a fault prediction result based on the data of the vehicle to be detected and using the target vehicle fault prediction model and the fault threshold.

[0060] In this embodiment, after obtaining the target vehicle fault prediction model, vehicle fault data is determined based on the vehicle's historical fault data and fault-related data. The vehicle fault data is then classified into fault types to obtain the fault types corresponding to the vehicle fault data. The characteristic parameters corresponding to each fault type are then analyzed to obtain corresponding analysis results. For example, for an engine misfire, relevant characteristics may include the engine speed range, vehicle speed range, and fuel injection volume at the time of the fault. Based on the analysis results and a preset statistical method, the distribution range of the characteristic parameters for different fault types is determined; the preset statistical method includes mean, standard deviation, minimum, and maximum values. The vehicle's real-time operating status data is then obtained, and the weights corresponding to the characteristic parameters are adjusted based on the real-time operating status data to obtain adjusted parameter weights. The historical fault data and the real-time operating status data of the characteristic parameters are weighted and fused using an exponential moving average algorithm to obtain an exponential moving average. The exponential moving average is used in combination with the corresponding distribution range and the adjusted parameter weights to determine a fault threshold. When a preset interval is reached, the process jumps to the step of determining vehicle fault data based on the vehicle's historical fault data and fault-related data to update the fault threshold. In one specific embodiment, if the vehicle frequently travels long distances, the fault threshold can be recalculated and updated every 1,000 kilometers. If the vehicle is a city commuter with high usage but relatively low mileage, the fault threshold can be recalculated and updated every month. It is worth noting that the preset interval can be adjusted based on the vehicle and actual conditions and is not specifically limited here.

[0061] Specifically, the statistical analysis of the historical fault data of the vehicle and the determination of the fault threshold based on the analysis results and the real-time operating status data of the vehicle include: determining the vehicle fault data based on the historical fault data and fault-related data of the vehicle, and classifying the fault types of the vehicle fault data to obtain each fault type corresponding to the vehicle fault data; analyzing the characteristic parameters corresponding to each fault type, and determining the distribution range of the characteristic parameters under different fault types based on the analysis results and a preset statistical method; obtaining the real-time operating status data of the vehicle, and adjusting the weights corresponding to the characteristic parameters based on the real-time operating status data to obtain adjusted parameter weights; weighted fusion of the historical fault data and the real-time operating status data of the characteristic parameters based on an exponential sliding average algorithm to obtain an exponential sliding average; determining the fault threshold using the exponential sliding average in combination with the corresponding distribution range and the adjusted parameter weight, and jumping to the step of determining the vehicle fault data based on the historical fault data and fault-related data of the vehicle based on a preset interval time to update the fault threshold.

[0062] It is understood that after obtaining the fault threshold, the target vehicle data is input into the target vehicle fault prediction model to obtain the fault type and corresponding fault probability, and to determine whether the fault probability exceeds the fault threshold. If the fault probability exceeds the fault threshold, the corresponding target fault type is determined based on the fault probability, and corresponding maintenance recommendations are generated based on the target fault type, vehicle maintenance manuals, and expert experience, triggering a fault warning for the target vehicle. For example, if the prediction result is low engine oil pressure, the maintenance recommendation may include checking the oil level, oil pump operating status, and oil filter blockage.

[0063] Specifically, the method of determining a fault prediction result based on the data of the vehicle to be detected and using the target vehicle fault prediction model and the fault threshold includes: obtaining the data of the vehicle to be detected, inputting the data of the vehicle to be detected into the target vehicle fault prediction model to obtain the fault type and the corresponding fault probability, and judging whether the fault probability exceeds the fault threshold; if the fault probability exceeds the fault threshold, determining the corresponding target fault type based on the fault probability, and generating corresponding maintenance suggestions based on the target fault type to trigger a fault warning for the vehicle to be detected.

[0064] As can be seen from the above, the present application utilizes the constructed target operating condition classification model to divide the processed vehicle data into operating conditions, so as to extract the time series features of the processed data corresponding to each operating condition segment obtained by the division, and utilizes the historical maintenance records of the vehicle to weightedly fuse the time series features. The obtained dynamic feature vector is then screened using a preset feature selection algorithm to obtain a target feature vector, and the target feature vector is used to train the initial vehicle fault prediction model to improve the training efficiency and prediction accuracy of the target vehicle fault prediction model. The fault threshold is then determined based on the vehicle's historical fault data and real-time operating status data. In this way, the use of the target vehicle fault prediction model and the fault threshold can achieve accurate prediction and timely warning of vehicle faults, thereby improving the operational safety and reliability of the vehicle.

[0065] Accordingly, see Figure 2 As shown, the present application also provides a vehicle fault prediction device, comprising:

[0066] The operating condition classification module 11 is used to collect vehicle data reflecting the vehicle operating status, pre-process the vehicle data to obtain processed data, and then use a target operating condition identification model to classify the processed data into operating conditions to obtain various operating condition segments; the target operating condition identification model is a operating condition identification model determined based on a long short-term memory network or a convolutional neural network;

[0067] a fault prediction model training module 12 for determining corresponding time series features based on the processed data corresponding to each of the operating conditions, performing weighted fusion on the time series features using the vehicle's historical maintenance records to obtain a dynamic feature vector, then screening the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and using the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm;

[0068] The fault prediction result determination module 13 is used to perform statistical analysis on the historical fault data of the vehicle, and determine the fault threshold based on the analysis results and the real-time operating status data of the vehicle, and determine the fault prediction result based on the vehicle data to be detected and using the target vehicle fault prediction model and the fault threshold.

[0069] As can be seen from the above, the present application utilizes the constructed target operating condition classification model to divide the processed vehicle data into operating conditions, so as to extract the time series features of the processed data corresponding to each operating condition segment obtained by the division, and utilizes the historical maintenance records of the vehicle to weightedly fuse the time series features. The obtained dynamic feature vector is then screened using a preset feature selection algorithm to obtain a target feature vector, and the target feature vector is used to train the initial vehicle fault prediction model to improve the training efficiency and prediction accuracy of the target vehicle fault prediction model. The fault threshold is then determined based on the vehicle's historical fault data and real-time operating status data. In this way, the use of the target vehicle fault prediction model and the fault threshold can achieve accurate prediction and timely warning of vehicle faults, thereby improving the operational safety and reliability of the vehicle.

[0070] In some specific implementations, the working condition division module 11 may specifically include:

[0071] a first recognition model training unit, configured to construct an initial operating condition recognition model based on a long short-term memory network, and train the initial operating condition recognition model using the processed data to obtain a target operating condition recognition model;

[0072] a feature extraction unit, configured to extract features from the time series data corresponding to the processed data using the target operating condition recognition model, and convert the obtained data features to obtain an operating condition classification result;

[0073] A probability distribution determination unit is used to determine the probability distribution of each operating condition category based on the operating condition classification result and using a preset activation function, so as to determine the operating condition category corresponding to the processed data using the probability distribution and preset decision rules to obtain each operating condition segment.

[0074] In some specific implementations, the working condition division module 11 may specifically include:

[0075] a first recognition model training unit, configured to construct an initial operating condition recognition model based on a convolutional neural network, and train the initial operating condition recognition model using the processed data to obtain a target operating condition recognition model;

[0076] a local feature determination unit, configured to perform a convolution operation on the processed data using the target operating condition recognition model to obtain local features of different scales, and perform a pooling operation on the local features to obtain pooled features;

[0077] The operating condition classification unit is used to classify the processed data into operating conditions based on the pooled features to obtain various operating condition segments.

[0078] In some specific implementations, the fault prediction model training module 12 may specifically include:

[0079] a time series feature determination unit, configured to determine corresponding time series features based on the variation pattern and feature information of the processed data corresponding to each of the operating sections in the time dimension;

[0080] The weight allocation unit is used to allocate weights to the time series features using the historical maintenance records of the vehicle, and perform weighted fusion based on the time series features and the corresponding weights to obtain a dynamic feature vector.

[0081] In some specific implementations, the fault prediction model training module 12 may specifically include:

[0082] a feature vector screening unit, configured to screen the dynamic feature vector using a chi-square test and recursive feature elimination to obtain a target feature vector;

[0083] A feature vector division unit, configured to divide the target feature vector based on a preset division ratio to obtain corresponding training sets, validation sets, and test sets;

[0084] A prediction model training unit, configured to construct an initial vehicle fault prediction model based on a machine learning algorithm, and train the initial vehicle fault prediction model using the training set to obtain a trained vehicle fault prediction model;

[0085] The prediction model testing unit is used to verify and test the trained vehicle fault prediction model using the verification set and the test set to obtain a target vehicle fault prediction model.

[0086] In some specific implementations, the fault prediction result determination module 13 may specifically include:

[0087] a fault data determining unit, configured to determine vehicle fault data based on the historical fault data and fault-related data of the vehicle, and classify the vehicle fault data into fault types to obtain each fault type corresponding to the vehicle fault data;

[0088] A characteristic parameter analysis unit, configured to analyze the characteristic parameters corresponding to each of the fault types, and determine the distribution range of the characteristic parameters under different fault types based on the analysis results and a preset statistical method;

[0089] a weight adjustment unit, configured to obtain real-time operating status data of the vehicle and adjust the weights corresponding to the characteristic parameters based on the real-time operating status data to obtain adjusted parameter weights;

[0090] a weighted fusion unit, configured to perform weighted fusion on the historical fault data and the real-time operating status data of the characteristic parameters based on an exponential sliding average algorithm to obtain an exponential sliding average value;

[0091] A fault threshold determination unit is used to determine a fault threshold using the exponential sliding average in combination with the corresponding distribution range and the adjusted parameter weight, and jump to the step of determining vehicle fault data based on the historical fault data and fault-related data of the vehicle based on a preset interval time to update the fault threshold.

[0092] In some specific implementations, the fault prediction result determination module 13 may specifically include:

[0093] a fault occurrence probability determination unit, configured to obtain data of a vehicle to be detected, input the data of the vehicle to be detected into the target vehicle fault prediction model to obtain a fault type and a corresponding fault occurrence probability, and determine whether the fault occurrence probability exceeds the fault threshold;

[0094] A target fault type determination unit is configured to determine a corresponding target fault type based on the fault probability if the fault probability exceeds the fault threshold, and to generate corresponding maintenance recommendations based on the target fault type to trigger a fault warning for the vehicle to be detected.

[0095] Furthermore, the embodiment of the present application also discloses an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of use of this application. The electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps of the vehicle fault prediction method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0096] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device. The communication protocol it follows is any communication protocol that can be applied to the technical solution of this application and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world. Its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0097] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or CD, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0098] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, and can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of implementing the vehicle fault prediction method executed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 may further include a computer program capable of implementing other specific tasks.

[0099] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when executed by a processor, the computer program implements the vehicle fault prediction method disclosed above. The specific steps of this method can be referred to the corresponding contents disclosed in the aforementioned embodiments and will not be repeated here.

[0100] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0101] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0102] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0103] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0104] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A vehicle fault prediction method, characterized in that: include: collecting vehicle data reflecting the operating status of the vehicle, preprocessing the vehicle data to obtain processed data, and then dividing the processed data into operating conditions using a target operating condition identification model to obtain operating condition segments; the target operating condition identification model is an operating condition identification model determined based on a long short-term memory network or a convolutional neural network; Determining corresponding time series features based on the processed data corresponding to each of the operating conditions, weighting and fusing the time series features using the vehicle's historical maintenance records to obtain a dynamic feature vector, then screening the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and using the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm; A statistical analysis is performed on the historical fault data of the vehicle, and a fault threshold is determined based on the analysis result and the real-time operating status data of the vehicle, and a fault prediction result is determined based on the data of the vehicle to be detected and using the target vehicle fault prediction model and the fault threshold.

2. The vehicle fault prediction method according to claim 1, characterized in that: The target operating condition identification model is used to divide the processed data into operating condition segments to obtain various operating condition segments, including: Constructing an initial operating condition recognition model based on a long short-term memory network, and training the initial operating condition recognition model using the processed data to obtain a target operating condition recognition model; Using the target operating condition recognition model to extract features from the time series data corresponding to the processed data, and converting the obtained data features to obtain an operating condition classification result; Based on the working condition classification result and using a preset activation function, the probability distribution of each working condition category is determined, and the working condition categories corresponding to the processed data are determined using the probability distribution and a preset decision rule to obtain each working condition segment.

3. The vehicle fault prediction method according to claim 1, characterized in that: The target operating condition identification model is used to divide the processed data into operating condition segments to obtain various operating condition segments, including: Constructing an initial operating condition recognition model based on a convolutional neural network, and training the initial operating condition recognition model using the processed data to obtain a target operating condition recognition model; performing a convolution operation on the processed data using the target operating condition recognition model to obtain local features of different scales, and performing a pooling operation on the local features to obtain pooled features; The processed data is classified into working condition categories based on the pooled features to obtain various working condition segments.

4. The vehicle fault prediction method according to claim 1, characterized in that: Determining corresponding time series features based on the processed data corresponding to each of the operating conditions, and weighting and fusing the time series features using historical maintenance records of the vehicle to obtain a dynamic feature vector, includes: Determine corresponding time series features based on the change pattern and feature information of the processed data corresponding to each of the operating sections in the time dimension; The vehicle's historical maintenance records are used to assign weights to the time series features, and weighted fusion is performed based on the time series features and the corresponding weights to obtain a dynamic feature vector.

5. The vehicle fault prediction method according to claim 1, characterized in that: The method of screening the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and using the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model, includes: Screening the dynamic feature vector using a chi-square test and recursive feature elimination to obtain a target feature vector; Dividing the target feature vector based on a preset division ratio to obtain corresponding training set, validation set, and test set; Constructing an initial vehicle fault prediction model based on a machine learning algorithm, and training the initial vehicle fault prediction model using the training set to obtain a trained vehicle fault prediction model; The trained vehicle fault prediction model is verified and tested using the verification set and the test set to obtain a target vehicle fault prediction model.

6. The vehicle fault prediction method according to claim 1, characterized in that: The statistical analysis of the historical fault data of the vehicle and determining the fault threshold based on the analysis result and the real-time operating status data of the vehicle include: Determining vehicle fault data based on historical fault data and fault-related data of the vehicle, and classifying the vehicle fault data into fault types to obtain each fault type corresponding to the vehicle fault data; Analyze the characteristic parameters corresponding to each fault type, and determine the distribution range of the characteristic parameters under different fault types based on the analysis results and a preset statistical method; Acquiring real-time operating status data of the vehicle, and adjusting weights corresponding to the characteristic parameters based on the real-time operating status data to obtain adjusted parameter weights; Performing weighted fusion on the historical fault data of the characteristic parameters and the real-time operating status data based on an exponential sliding average algorithm to obtain an exponential sliding average; The fault threshold is determined by utilizing the exponential sliding average value in combination with the corresponding distribution range and the adjusted parameter weight, and based on a preset interval time, the step of determining the vehicle fault data based on the historical fault data and fault-related data of the vehicle is jumped to update the fault threshold.

7. The vehicle fault prediction method according to any one of claims 1 to 6, characterized in that: The determining of the fault prediction result based on the to-be-detected vehicle data and using the target vehicle fault prediction model and the fault threshold comprises: Acquiring data of a vehicle to be detected, inputting the data of the vehicle to be detected into the target vehicle fault prediction model to obtain a fault type and a corresponding fault probability, and determining whether the fault probability exceeds the fault threshold; If the fault occurrence probability exceeds the fault threshold, a corresponding target fault type is determined based on the fault occurrence probability, and a corresponding maintenance suggestion is generated based on the target fault type, triggering a fault warning for the vehicle to be detected.

8. A vehicle fault prediction device, characterized in that: include: a working condition classification module, configured to collect vehicle data reflecting the vehicle operating status, pre-process the vehicle data to obtain processed data, and then classify the processed data into working conditions using a target working condition identification model to obtain various working condition segments; the target working condition identification model is a working condition identification model determined based on a long short-term memory network or a convolutional neural network; a fault prediction model training module, configured to determine corresponding time series features based on the processed data corresponding to each of the operating conditions, weightedly fuse the time series features using the vehicle's historical maintenance records to obtain a dynamic feature vector, then filter the dynamic feature vector based on a preset feature selection algorithm to obtain a target feature vector, and use the target feature vector to train an initial vehicle fault prediction model to obtain a target vehicle fault prediction model; the initial vehicle fault prediction model is a prediction model determined based on a machine learning algorithm; A fault prediction result determination module is used to perform statistical analysis on the historical fault data of the vehicle, determine a fault threshold based on the analysis results and the real-time operating status data of the vehicle, and determine a fault prediction result based on the data of the vehicle to be detected and using the target vehicle fault prediction model and the fault threshold.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the vehicle fault prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Used to store a computer program, wherein when the computer program is executed by a processor, the vehicle fault prediction method according to any one of claims 1 to 7 is implemented.

Images