Vehicle fault prediction method and system based on side cloud cooperation

Abstract

The invention provides a vehicle fault prediction method and system based on side cloud collaboration, and the method comprises the steps: obtaining sensor monitoring data through a sensor array disposed in each function system of a vehicle; carrying out vehicle state related feature extraction on the sensor monitoring data by utilizing an edge computing device arranged at a vehicle end, and sending obtained feature data to a cloud server; the cloud server inputs the feature data into a pre-constructed time sequence prediction model and a pre-constructed fault prediction model to obtain a first prediction result representing the health state of the vehicle and a second prediction result representing the potential fault of the vehicle; and performing dynamic weight fusion on the first prediction result and the second prediction result according to a preset fusion strategy to generate a comprehensive prediction result, and outputting a vehicle health status report. According to the invention, the problems of low prediction accuracy, high time delay and the like in the prior art are solved, and the accuracy, reliability and real-time performance of fault prediction are improved through edge cloud cooperative processing and hybrid model fusion analysis.

Description

Technical Field

[0001] This application relates to the field of vehicle health status management, and in particular to a vehicle fault prediction method and system based on edge-cloud collaboration. Background Technology

[0002] Vehicle fault prediction is a key technology for ensuring driving safety and optimizing maintenance strategies. Traditional vehicle fault detection mainly relies on offline diagnostics after the vehicle arrives at the shop or periodic preventative maintenance based on fixed mileage and time. This approach has a lag in response, cannot continuously monitor the real-time operating condition of the vehicle, and is even less capable of proactive fault prediction.

[0003] With the development of vehicle-to-everything (V2X) technology, it has become possible to collect multi-source sensor data via the vehicle bus and upload it to the cloud for analysis. Some existing solutions employ a vehicle-edge-cloud collaborative architecture, where preliminary data processing is performed on edge devices close to the data source to reduce the burden on the cloud and network transmission pressure. For example, data collection and online monitoring are performed at the edge, while a fault diagnosis model is trained and deployed in the cloud. When an anomaly is detected at the edge, the cloud model diagnoses and issues a warning. However, these solutions typically use only a single type of integrated model for analysis in the cloud, such as a single neural network model. Due to the complex and variable operating conditions of vehicles, influenced by various time-varying factors such as ambient temperature, road conditions, driving habits, and vehicle load, a single model struggles to simultaneously capture the dynamic trends of vehicle state changes over time and the nonlinear characteristics of the current operating condition. This results in insufficient generalization ability of the model across multiple vehicle types or complex, extreme conditions, and the prediction accuracy and reliability fail to meet the requirements of high-safety scenarios. Furthermore, the transmission of large amounts of raw or semi-processed data can still cause significant network latency, leading to delayed prediction results and making it difficult to handle high-priority fault scenarios requiring rapid response, such as braking system anomalies. Therefore, existing technologies still need improvement in terms of prediction accuracy, reliability, and real-time performance. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art, and proposes a vehicle fault prediction method and system based on edge-cloud collaboration.

[0005] In a first aspect, embodiments of the present invention provide a vehicle fault prediction method based on edge-cloud collaboration, comprising:

[0006] At least one type of sensor monitoring data is acquired by sensor arrays installed in various functional systems of the vehicle.

[0007] Using an edge computing device installed on the vehicle, vehicle state-related features are extracted from the sensor monitoring data to obtain feature data, and the feature data is sent to a cloud server.

[0008] The cloud server inputs the feature data into a pre-built time series prediction model and a fault prediction model to obtain a first prediction result characterizing the vehicle's health status and a second prediction result characterizing the vehicle's potential faults.

[0009] The first prediction result and the second prediction result are dynamically weighted and fused according to a preset fusion strategy to generate a comprehensive prediction result. Based on the comprehensive prediction result, a vehicle health status report is output to the user terminal.

[0010] Furthermore, the sensor array includes a temperature sensor, a pressure sensor, and a vibration sensor; the sensor monitoring data includes temperature data, pressure data, and vibration data.

[0011] The functional system includes at least one of the following: engine combustion chamber, transmission oil circuit, power battery thermal management system, and braking system.

[0012] Furthermore, before performing feature extraction on the sensor monitoring data, the method further includes:

[0013] The sensor monitoring data is preprocessed, specifically including:

[0014] Clock synchronization processing is performed on monitoring data from different types of sensors; and,

[0015] Outliers in the sensor monitoring data are removed based on preset criteria; and / or, missing values ​​in the sensor monitoring data are filled using interpolation; and / or, noise interference in the sensor data is suppressed through filtering or denoising processing; and / or, the sensor monitoring data is normalized.

[0016] Furthermore, the step of extracting vehicle state-related features from the sensor monitoring data to obtain feature data includes:

[0017] Statistical indicators are calculated for the time series of the sensor monitoring data to obtain time-domain features; the time-domain features include at least one of the following: mean, variance, standard deviation, root mean square, peak value, kurtosis, and skewness.

[0018] A Fourier transform is performed on the time series to extract frequency domain features; the frequency domain features include at least one of the center frequency, power spectral density, principal frequency, and frequency centroid.

[0019] The time-frequency features in the sensor monitoring data are extracted using short-time Fourier transform technology.

[0020] The time-domain features, the frequency-domain features, and the time-frequency features are combined to form a high-dimensional state feature vector, which is used as the feature data.

[0021] Furthermore, the time series prediction model is configured as a neural network structure based on a long short-term memory network, including an input gate, a forget gate, an output gate, and cell states;

[0022] The input gate is represented as:

[0023] in, Indicates the input gate; Represents the Sigmoid function; This represents the input gate weight matrix; This indicates the hidden state at the previous moment; This represents the input at the current moment; Indicates the input gate's corresponding bias;

[0024] The forget gate is represented as follows:

[0025] Among them, the Indicates the Gate of Oblivion; Represents the forget gate weight matrix; This indicates the bias corresponding to the forget gate;

[0026] The cell state is represented as follows:

[0027]

[0028] in, Indicates cell state; This indicates the cell state at the previous moment; Indicates the state of candidate cells; A weight matrix representing the state of candidate cells; This indicates the bias corresponding to the cell state;

[0029] The output gate is represented as:

[0030] Among them, the Indicates the output gate; This represents the output gate weight matrix; This indicates the offset corresponding to the output gate;

[0031] Hidden states are used to characterize the health status of a vehicle, and are represented as follows:

[0032]

[0033] in, Indicates a hidden state;

[0034] And / or,

[0035] The fault prediction model uses the random forest algorithm.

[0036] Furthermore, before inputting the feature data into the pre-built time series prediction model and fault prediction model respectively, the method further includes:

[0037] Training the time series prediction model using a historical feature dataset specifically includes:

[0038] Initialize the weight matrix and bias of the time series prediction model;

[0039] The historical feature dataset is used for forward and backward propagation to update the model parameters until the change in the loss function is less than a preset threshold, thus obtaining the trained time series prediction model.

[0040] And / or,

[0041] The fault prediction model is trained using a historical feature dataset, specifically including:

[0042] Step S1: Determine the first number of decision trees and the second number of features in each decision tree;

[0043] Step S2: Randomly select a specific number of samples with replacement from the historical feature dataset to form a new training subset;

[0044] Step S3: For each of the training subsets, randomly select a second number of features to construct a decision tree;

[0045] Step S4: Repeat steps S2 to S3 to construct the first number of decision trees and form a trained fault prediction model.

[0046] Further, the step of dynamically weighting and fusing the first prediction result and the second prediction result according to a preset fusion strategy to generate a comprehensive prediction result, and outputting a vehicle health status report based on the comprehensive prediction result, includes:

[0047] The comprehensive forecast result is calculated using the following formula:

[0048]

[0049] in, This indicates the overall forecast results; This indicates the first prediction result; This indicates the second prediction result; Indicates the weighting coefficient;

[0050] If the comprehensive prediction result is greater than the fault prediction threshold, a vehicle health status report indicating that the vehicle has a fault is output.

[0051] Furthermore, the method also includes:

[0052] When a vehicle malfunction is detected, a tiered early warning strategy is used to classify the comprehensive prediction results and execute response actions that match the judgment results, specifically including:

[0053] when When this occurs, an audible and visual alarm will be triggered immediately, and the vehicle will be forced into limp mode.

[0054] when At that time, preventative maintenance recommendations are pushed to the driver, and a fault tracing algorithm is activated to analyze the cause of the fault;

[0055] when When abnormal operating conditions occur, the abnormal data features corresponding to the abnormal operating conditions are included in the historical feature dataset used for model training.

[0056] in, This indicates the overall forecast results; , , This indicates the threshold for grading judgment. < < .

[0057] Secondly, embodiments of the present invention provide a vehicle fault prediction system based on edge-cloud collaboration, the system comprising a sensor array, an edge computing device, a cloud server, and a user terminal:

[0058] The sensor array is installed in various functional systems of the vehicle to acquire at least one type of sensor monitoring data.

[0059] The edge computing device is installed on the vehicle and is used to extract vehicle status-related features from the sensor monitoring data to obtain feature data, and then send the feature data to the cloud server.

[0060] The cloud server is used to input the feature data into a pre-built time series prediction model and a fault prediction model to obtain a first prediction result representing the vehicle's health status and a second prediction result representing the vehicle's potential faults; the first prediction result and the second prediction result are dynamically weighted and fused according to a preset fusion strategy to generate a comprehensive prediction result, and a vehicle health status report is output to the user terminal based on the comprehensive prediction result.

[0061] Furthermore, the edge computing device is also used to preprocess the sensor monitoring data before performing feature extraction on the sensor monitoring data;

[0062] The cloud server is also used to periodically acquire historical feature datasets from the edge computing device and use the historical feature datasets to train the time series prediction model and the fault prediction model;

[0063] The cloud server is also used to classify the comprehensive prediction results according to a graded early warning strategy when a vehicle is determined to have a fault, and to execute response operations that match the judgment results.

[0064] This invention provides a vehicle fault prediction method based on edge-cloud collaboration. It acquires at least one type of sensor monitoring data through sensor arrays installed in various functional systems of the vehicle. Using an edge computing device installed on the vehicle, it extracts vehicle state-related features from the sensor monitoring data to obtain feature data, which is then sent to a cloud server. The cloud server inputs the feature data into a pre-constructed time-series prediction model and a fault prediction model, respectively, to obtain a first prediction result representing the vehicle's health status and a second prediction result representing potential vehicle faults. The first and second prediction results are dynamically weighted and fused according to a preset fusion strategy to generate a comprehensive prediction result. Based on this comprehensive prediction result, a vehicle health status report is output to the user terminal. This method solves the problems of low prediction accuracy and high latency in existing technologies. Through edge-cloud collaborative processing and hybrid model fusion analysis, it improves the accuracy, reliability, and real-time performance of fault prediction. Attached Figure Description

[0065] Figure 1 This is a flowchart illustrating a vehicle fault prediction method based on edge-cloud collaboration provided in an embodiment of the present invention.

[0066] Figure 2 A flowchart illustrating a vehicle fault prediction method based on edge-cloud collaboration, provided as another embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of the training process for a fault prediction model provided in an embodiment of the present invention;

[0068] Figure 4 A schematic diagram illustrating the training process of a time series prediction model according to an embodiment of the present invention;

[0069] Figure 5 A schematic diagram of the structure of a vehicle fault prediction system based on edge-cloud collaboration provided in an embodiment of the present invention. Detailed Implementation

[0070] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0071] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0072] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0073] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0074] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0075] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information all comply with relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution follows relevant national laws and regulations (e.g., the "Information Security Technology - Personal Information Security Specification"). For example: appropriate measures are taken for personal information access control; restrictions are imposed on the display of personal information; the purpose of using personal information does not exceed the scope of direct or reasonable association; and explicit identity targeting is eliminated when using personal information to avoid precisely locating a specific individual.

[0076] In related technologies, vehicle health status assessment suffers from insufficient cross-domain generalization ability, high prediction latency, and extensive maintenance strategies. It is unable to continuously monitor the real-time operating conditions of vehicles, and it is even more difficult to achieve forward-looking fault prediction.

[0077] To address at least one of the technical problems existing in the aforementioned related technologies, this invention provides a vehicle fault prediction method based on edge-cloud collaboration. For example... Figure 1 As shown, the vehicle fault prediction method based on edge-cloud collaboration provided by this embodiment of the invention may include at least the following steps S101~S104:

[0078] Step S101: Acquire at least one type of sensor monitoring data through sensor arrays installed in various functional systems of the vehicle.

[0079] This invention constructs a heterogeneous sensor network covering temperature, pressure, vibration, etc., to analyze the coupled operating status of key components (such as engine block, transmission gearbox, power battery pack, and brake calipers). Specifically, it can monitor the operation of multiple functional systems, such as the engine combustion chamber, transmission oil circuit, power battery thermal management system, and braking system, by using monitoring devices such as temperature sensors, pressure sensors, and vibration sensors to monitor the temperature, pressure, and vibration data of each functional module during operation.

[0080] For example, a K-type thermocouple can be installed near the engine cylinder head to measure the combustion chamber wall temperature with an accuracy of ±0.5℃, accurately capturing changes in engine operating temperature; a piezoresistive pressure sensor can be integrated into the brake master cylinder with a range covering 0 to 10 MPa to monitor the dynamic load of the hydraulic system; and a triaxial microelectromechanical system vibration sensor can be installed on the gearbox input shaft to monitor gear meshing status with a sampling rate up to 10 kHz.

[0081] In practical applications, these sensors can operate continuously at a synchronous acquisition frequency of, for example, 200 Hz. The acquired multi-source heterogeneous monitoring data can be transmitted to edge computing devices in near real-time via in-vehicle Ethernet and using an open platform communication unified architecture protocol format.

[0082] Step S102: Using the edge computing device installed on the vehicle, vehicle status-related features are extracted from the sensor monitoring data to obtain feature data, and the feature data is sent to the cloud server.

[0083] It is understandable that edge computing devices are a key component in enabling localized data processing, reducing cloud load and communication latency. They are deployed inside vehicles (e.g., in-vehicle smart gateways or domain controllers) and are typically equipped with high-performance multi-core processors and sufficient memory to provide the necessary computing power.

[0084] like Figure 2 As shown, before feature extraction from the sensor monitoring data, this embodiment of the invention further includes step S105, which involves preprocessing the sensor monitoring data.

[0085] Specifically, after clock synchronization processing of different types of sensor monitoring data, outliers in the sensor monitoring data can be removed based on preset criteria (e.g., outlier identification based on the 3σ criterion); missing values ​​in the sensor monitoring data can be filled using interpolation; noise interference in the sensor data can be suppressed through filtering (moving average filter MAF) or denoising (wavelet denoising); and the sensor monitoring data can be normalized (mapped to the [0,1] interval). Through data preprocessing, the received raw and noisy sensor monitoring data is cleaned and normalized so that subsequent processes can extract key features that can effectively characterize the fault state from the clean data.

[0086] Furthermore, the feature engineering in this embodiment of the invention is used to extract key state-related features of the vehicle (such as temperature change rate, vibration frequency, pressure fluctuation, etc.) to assess the vehicle's health status through these key features. Specifically, statistical indicators can be calculated for the time series of sensor monitoring data to obtain time-domain features, including mean, variance, standard deviation, root mean square, peak value, kurtosis, and skewness. Fourier transform is performed on the time series to extract frequency-domain features, including center frequency, power spectral density, dominant frequency, and frequency centroid. Short-time Fourier transform technology is used to extract time-frequency features from the sensor monitoring data. The time-domain features, frequency-domain features, and time-frequency features are combined to form a high-dimensional state feature vector, which is then uploaded to a cloud server as feature data.

[0087] In step S103, the cloud server inputs the feature data into the pre-built time series prediction model and fault prediction model respectively to obtain the first prediction result representing the vehicle's health status and the second prediction result representing the vehicle's potential faults.

[0088] The cloud server periodically acquires historical feature datasets from edge computing devices. Based on these datasets, a time series prediction model is used to model the vehicle's health status, and a fault prediction model is used to classify and predict potential faults, resulting in a trained time series prediction model and a fault prediction model.

[0089] Preferably, the time series prediction model in this embodiment of the invention uses a bidirectional LSTM network (Bi-LSTM) to capture the forward and backward dependencies of the data sequence. The hidden layer has 128 memory units, the activation function is tanh, and the loss function is mean squared error (MSE). LSTM (Long Short-Term Memory) network, as a special type of recurrent neural network (RNN), is suitable for processing and predicting time series data. In vehicle health status monitoring, LSTM is used to model the sensor feedback data of key vehicle components, capture the long-term dependencies in the data, and predict the vehicle's health status. That is, the time series prediction model can be configured as a neural network structure based on a long short-term memory network, including an input gate, a forget gate, an output gate, and cell states.

[0090] The input gate is represented as:

[0091] in, Indicates the input gate; Represents the Sigmoid function; This represents the input gate weight matrix; This indicates the hidden state at the previous moment; This represents the input at the current moment; This indicates the bias corresponding to the input gate.

[0092] The forget gate is represented as:

[0093] in, Indicates the Gate of Oblivion; Represents the forget gate weight matrix; This indicates the offset corresponding to the forget gate.

[0094] Cell state is represented as:

[0095]

[0096] in, Indicates cell state; This indicates the cell state at the previous moment; Indicates the state of candidate cells; A weight matrix representing the state of candidate cells; This indicates the bias corresponding to the cell state.

[0097] The output gate is represented as:

[0098] in, Indicates the output gate; This represents the output gate weight matrix; This indicates the offset corresponding to the output gate.

[0099] Hidden states are used to characterize the health status of a vehicle, and are represented as follows:

[0100]

[0101] in, Indicates a hidden state.

[0102] Preferably, the fault prediction model uses a random forest algorithm to construct the classifier. Random forest, as an ensemble learning method, achieves classification and regression by constructing multiple decision trees and combining the prediction results. In this application scenario, random forest is used to classify and predict the vehicle's health status.

[0103] Furthermore, such as Figure 2 As shown, before inputting the feature data into the pre-built time series prediction model and fault prediction model respectively, this embodiment of the invention further includes: step S106, training the time series prediction model and fault prediction model using the historical feature dataset.

[0104] like Figure 3 As shown, training a time series prediction model may include the following steps S103-1-1 to S103-1-3:

[0105] Step S103-1-1: Initialize the weight matrix of the time series prediction model. and bias .

[0106] Step S103-1-2: Use the historical feature dataset to perform forward and backward propagation to update the model parameters.

[0107] Step S103-1-3: Determine that the change in the loss function (such as mean squared error) is less than a preset threshold, and obtain the trained time series prediction model.

[0108] This invention employs a specific learning mechanism to periodically acquire new samples from edge computing devices and uses online gradient descent (OGD) to update the LSTM model parameters, thereby obtaining a trained time series prediction model.

[0109] like Figure 4 As shown, training the fault prediction model may include the following steps S103-2-1 to S103-2-4:

[0110] Step S103-2-1: Determine the first number N of the decision tree and the second number m of the features in each decision tree.

[0111] Step S103-2-2: Randomly select a specific number of samples n with replacement from the historical feature dataset to form a new training subset.

[0112] Step S103-2-3: For each training subset, randomly select a second number m of features to construct a decision tree.

[0113] Step S103-2-4: Repeat steps S103-2-2 to S103-2-3 to construct the first number of decision trees N, forming a trained fault prediction model.

[0114] In practice, N can be set to 200, and a training subset can be generated using the bootstrap method. The random forest structure can then be optimized using the random subspace method.

[0115] For real-time sensor monitoring data, feature vectors are obtained after preprocessing and feature extraction. During the prediction process of the time series prediction model, the feature vector sequence... The hidden state is input into the LSTM network and then output. Used for subsequent fault classification and prediction. In the fault prediction model's prediction process, for a new data point, each decision tree gives a prediction result, and the random forest determines the final prediction result through majority voting (classification) or average (regression).

[0116] Step S104: Dynamically weight and fuse the first prediction result and the second prediction result according to the preset fusion strategy to generate a comprehensive prediction result, and output a vehicle health status report to the user terminal based on the comprehensive prediction result.

[0117] This invention utilizes a time series prediction model. For feature data Make a prediction and obtain the first prediction result. Using fault prediction models For feature data Prediction, to obtain the second prediction result Then, the comprehensive prediction result is calculated using the following formula:

[0118]

[0119] in, This indicates the overall forecast results; This indicates the first prediction result; This indicates the second prediction result; This represents the weighting coefficient.

[0120] In this embodiment of the invention, a dynamic weight adapter is used to automatically adjust the time series prediction model according to the complexity of the operating conditions. With fault prediction model The weighting coefficient α. Compare the overall prediction results. With fault prediction threshold If the overall prediction results Greater than the fault prediction threshold If the vehicle has a malfunction, a vehicle health status report will be output.

[0121] Furthermore, such as Figure 2 As shown, when a vehicle is determined to have a fault, the embodiment of the present invention further includes: step S107, using a graded early warning strategy to classify and judge the comprehensive prediction results, and executing a response operation that matches the judgment result.

[0122] The embodiments of the present invention are configured as follows: , , Level 3 judgment threshold < < This is used for tiered early warning. Specifically: when When a high-level warning is triggered, it indicates an extremely high risk of malfunction, immediately activating an audible and visual alarm and forcing the vehicle into limp mode. For example, a red warning light may appear on the vehicle's dashboard along with an audible alarm, while a serious malfunction warning may be pushed to the in-vehicle infotainment system. In some designs, engine power may even be actively limited to put the vehicle into limp mode, thereby ensuring safety. When a medium-level warning is issued, it indicates a potential risk. Preventative maintenance recommendations are sent to the driver, and a fault tracing algorithm is activated to analyze the cause of the fault. For example, a yellow warning message is sent to the user's device, suggesting that the user schedule a vehicle inspection or preventative maintenance as soon as convenient. (Low-level warning) indicates a situation that warrants attention but poses no immediate risk. The abnormal operating condition is recorded, and the corresponding abnormal data features are included in the historical feature dataset used for model training. For example, this event of concern is recorded only in the background for subsequent analysis and model optimization without disturbing the driver. This indicates the overall forecast results; , , This indicates the threshold for grading judgment. < < .

[0123] It is understandable that if the overall prediction result is less than the threshold... If the vehicle is considered to be in normal condition, the monitoring will continue, thus forming a continuously operating closed loop.

[0124] As an alternative implementation, time series prediction models can also use gated recurrent unit (GRU) modules to replace bidirectional long short-term memory neural network structures, or use the Transformer architecture to handle long sequence dependencies. While ensuring comparable prediction accuracy, these approaches can significantly reduce model training time and inference overhead, thereby reducing the consumption of computing resources on cloud servers and improving the overall operating efficiency of the system.

[0125] As another optional implementation, embodiments of the present invention can also introduce virtual sensor technology, which can build a model by accurately replicating physical laws, fill data gaps caused by sensor failures, environmental interference, etc., and optimize data reliability to avoid decision-making deviations or loss of control due to data problems.

[0126] This invention provides a vehicle fault prediction method based on edge-cloud collaboration, comprising: acquiring at least one type of sensor monitoring data through sensor arrays installed in various functional systems of the vehicle; extracting vehicle state-related features from the sensor monitoring data using an edge computing device installed on the vehicle to obtain feature data, and sending the feature data to a cloud server; the cloud server inputting the feature data into a pre-constructed time series prediction model and a fault prediction model to obtain a first prediction result representing the vehicle's health status and a second prediction result representing potential vehicle faults; dynamically weighting and fusing the first and second prediction results according to a preset fusion strategy to generate a comprehensive prediction result; and outputting a vehicle health status report to the user terminal based on the comprehensive prediction result. Compared with the prior art, this application has the following beneficial effects:

[0127] 1) Improve prediction accuracy and reliability. By using time series prediction models and classification models in parallel in the cloud and fusing their prediction results, it is possible to simultaneously take into account the dynamic trend of vehicle status evolution over time and the static nonlinear classification and discrimination of current operating conditions. This makes the comprehensive prediction results more robust and accurate than the prediction results of any single model or combination of serial models, effectively overcoming the problem of insufficient generalization ability of single models in existing technologies.

[0128] 2) Improved real-time prediction performance. By performing data preprocessing and feature extraction on the vehicle's edge computing device, only feature data containing key information is uploaded to the cloud, significantly reducing data transmission volume and network communication latency. This improves the response speed of fault warnings and solves the problem of high latency in existing centralized cloud architectures.

[0129] 3) Achieve refined management. Based on more accurate comprehensive forecasts, tiered early warnings can provide users with more reasonable maintenance suggestions and early warning responses, avoiding over-maintenance or delays in maintenance decisions caused by inaccurate forecasts, and helping to reduce the average maintenance cost and unplanned downtime throughout the vehicle's life cycle.

[0130] Based on the same inventive concept, as Figure 1 In a specific implementation, this invention also provides a vehicle fault prediction system based on edge-cloud collaboration, such as... Figure 5 As shown, the device may include: a sensor array 510, an edge computing device 520, a cloud server 530, and a user terminal 540.

[0131] The sensor array 510 is installed in various functional systems of the vehicle and can be used to acquire at least one type of sensor monitoring data.

[0132] The edge computing device 520 is installed on the vehicle and can be used to extract vehicle status-related features from sensor monitoring data, obtain feature data, and send the feature data to the cloud server 530.

[0133] The cloud server 530 can be used to input feature data into a pre-built time series prediction model and a fault prediction model to obtain a first prediction result representing the vehicle's health status and a second prediction result representing the vehicle's potential faults. The first prediction result and the second prediction result are dynamically weighted and fused according to a preset fusion strategy to generate a comprehensive prediction result. Based on the comprehensive prediction result, a vehicle health status report is output to the user terminal 540.

[0134] Optionally, the edge computing device 520 can also be used to preprocess the sensor monitoring data before feature extraction is performed on the sensor monitoring data;

[0135] The cloud server 530 can also be used to periodically acquire historical feature datasets from edge computing devices, and use these historical feature datasets to train time series prediction models and fault prediction models.

[0136] The cloud server 530 can also be used to classify the comprehensive prediction results according to the classification warning strategy when a vehicle is found to have a fault, and to execute response operations that match the classification results.

[0137] It should be noted that other corresponding descriptions of the functional modules involved in the vehicle fault prediction system based on edge-cloud collaboration provided in this embodiment of the invention can be found in [reference needed]. Figure 1 The corresponding description of the method shown will not be repeated here.

[0138] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0139] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable program instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CDROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0140] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0141] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0142] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0143] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0144] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0145] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0146] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0147] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A vehicle fault prediction method based on edge cloud cooperation, characterized in that, The method comprises the following steps: obtaining at least one type of sensor monitoring data through a sensor array arranged in each functional system of the vehicle; extracting vehicle state related features from the sensor monitoring data using an edge computing device arranged at the vehicle end to obtain feature data, and sending the feature data to a cloud server; inputting the feature data into a pre-constructed time series prediction model and a fault prediction model respectively to obtain a first prediction result representing the vehicle health state and a second prediction result representing the potential fault of the vehicle; dynamically weighting and fusing the first prediction result and the second prediction result according to a preset fusion strategy to generate a comprehensive prediction result, and outputting a vehicle health state report to a user end based on the comprehensive prediction result.

2. The method of claim 1, wherein, The sensor array comprises a temperature sensor, a pressure sensor and a vibration sensor; and the sensor monitoring data comprises temperature data, pressure data and vibration data. The functional system comprises at least one of an engine combustion chamber, a gearbox oil circuit, a power battery thermal management system and a braking system.

3. The method of claim 1, wherein, Before the feature extraction of the sensor monitoring data, the method further comprises: preprocessing the sensor monitoring data, specifically including: clock synchronization processing of different types of sensor monitoring data; and, eliminating outliers in the sensor monitoring data based on a preset criterion; and / or, filling in missing values in the sensor monitoring data by interpolation method; and / or, suppressing noise interference in the sensor data by filtering or denoising processing; and / or, normalizing the sensor monitoring data.

4. The method of claim 1, wherein, The feature extraction of the sensor monitoring data to obtain feature data comprises: calculating statistical indicators for the time series of the sensor monitoring data to obtain time domain features; the time domain features include at least one of mean, variance, standard deviation, root mean square, peak value, kurtosis and skewness; performing Fourier transform on the time series to extract frequency domain features; the frequency domain features include at least one of center frequency, power spectral density, main frequency and frequency gravity center; extracting time-frequency features in the sensor monitoring data by short-time Fourier transform technology; combining the time domain features, the frequency domain features and the time-frequency features into a high-dimensional state feature vector as the feature data.

5. The method of claim 1, wherein, The time series prediction model is configured as a neural network structure based on a long short-term memory network, including an input gate, a forget gate, an output gate and a cell state; The input gate is represented as: ; wherein, denotes an input gate; denotes a Sigmoid function; denotes an input gate weight matrix; denotes a hidden state at the previous time; denotes an input at the current time; denotes an input gate corresponding bias; The forget gate is represented as: ; Wherein the denotes a forget gate; denotes a forget gate weight matrix; denotes a forget gate corresponding bias; The cell state is represented as: ; ； wherein, represents a cell state; represents a cell state at a previous time; represents a candidate cell state; represents a weight matrix of a candidate cell state; represents a cell state corresponding bias; The output gate is represented as: ; wherein the denotes an output gate; denotes an output gate weight matrix; denotes an output gate corresponding bias; the hidden state is used to represent the vehicle health state, expressed as: ； wherein represents a hidden state; and / or, The fault prediction model adopts a random forest algorithm.

6. The method of claim 1, wherein, Before inputting the feature data into the pre-constructed time series prediction model and fault prediction model, the method further comprises: training the time series prediction model using a historical feature data set, specifically including: initializing the weight matrix and bias of the time series prediction model; performing forward propagation and back propagation using the historical feature data set to update the model parameters until the change of the loss function is less than a preset threshold, obtaining a trained time series prediction model; and / or, The fault prediction model is trained by using a historical feature data set, specifically comprising: Step S1: determining a first number of decision trees and a second number of features in each decision tree; Step S2: randomly sampling a certain number of samples from the historical feature data set with replacement to form a new training subset; Step S3: for each training subset, randomly selecting a second number of features to construct a decision tree; Step S4: repeating steps S2-S3 to construct a first number of decision trees to form a trained fault prediction model.

7. The method of claim 1, wherein, The first prediction result and the second prediction result are dynamically weighted and fused according to a preset fusion strategy to generate a comprehensive prediction result, and a vehicle health status report is output based on the comprehensive prediction result, comprising: The comprehensive prediction result is calculated using the following formula: ； wherein, represents the comprehensive prediction result; represents the first prediction result; represents the second prediction result; represents the weight coefficient; The comprehensive prediction result is compared with a fault prediction threshold, and if the comprehensive prediction result is greater than the fault prediction threshold, a vehicle health status report indicating that the vehicle has a fault is output.

8. The method according to any one of claims 1 to 7, characterized in that, The method further comprises: When it is determined that the vehicle has a fault, the comprehensive prediction result is judged by a hierarchical warning strategy, and a response operation matching the judgment result is performed, specifically comprising: When the sound and light alarm is triggered immediately, and the vehicle is forced into limp mode; When a preventive maintenance recommendation is pushed to the driver and a fault root cause algorithm is initiated to analyze the cause of the fault; When an abnormal working condition is recorded, and the abnormal data features corresponding to the abnormal working condition are included in the historical feature data set for model training. wherein represents the comprehensive prediction result; , , represents the hierarchical judgment threshold, < < . 9.A vehicle fault prediction system based on edge cloud cooperation, characterized in that, The system comprises a sensor array, an edge computing device, a cloud server and a user terminal: The sensor array is arranged in each functional system of the vehicle for acquiring at least one type of sensor monitoring data; The edge computing device is arranged at the vehicle end for extracting vehicle state related features from the sensor monitoring data to obtain feature data, and sending the feature data to the cloud server; The cloud server is used to input the feature data into a pre-constructed time series prediction model and a fault prediction model respectively to obtain a first prediction result representing the vehicle health status and a second prediction result representing the vehicle potential fault; the first prediction result and the second prediction result are dynamically weighted and fused according to a preset fusion strategy to generate a comprehensive prediction result, and a vehicle health status report is output to the user terminal based on the comprehensive prediction result.

10. The system of claim 9, wherein The edge computing device is further used to preprocess the sensor monitoring data before feature extraction; The cloud server is further used to periodically acquire a historical feature data set from the edge computing device, and train the time series prediction model and the fault prediction model by using the historical feature data set; The cloud server is further used to judge the comprehensive prediction result by a hierarchical warning strategy when it is determined that the vehicle has a fault, and perform a response operation matching the judgment result.

Images