CN-122018489-A - Vehicle-mounted controller fault prediction method and device, electronic equipment and medium

CN122018489ACN 122018489 ACN122018489 ACN 122018489ACN-122018489-A

Abstract

The application provides a vehicle-mounted controller fault prediction method, a device, electronic equipment and a medium, wherein the method comprises the following steps: acquiring fault data of the vehicle-mounted controller, performing multi-category processing on the fault data to obtain processed fault data, training and verifying a preset basic vehicle-mounted controller fault prediction model based on the processed fault data to obtain a final vehicle-mounted controller fault prediction model, issuing the vehicle-mounted controller fault prediction model to the vehicle-mounted controller, deploying the vehicle-mounted controller based on a preset deployment mode, predicting real-time fault data of the vehicle-mounted controller through the vehicle-mounted controller fault prediction model, and processing faults based on a fault prediction result.

Inventors

JIANG JIAQIN
ZHANG HUIFENG
ZHANG DEDONG

Assignees

无锡车联天下智能科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260323

Claims (10)

1. A vehicle-mounted controller fault prediction method, characterized in that the method comprises: Acquiring fault data of a vehicle-mounted controller, and performing multi-class processing on the fault data to obtain processed fault data, wherein the multi-class processing at least comprises data cleaning, data labeling, data enhancement and data synthesis; training and verifying a preset basic vehicle-mounted controller fault prediction model based on the processed fault data to obtain a final vehicle-mounted controller fault prediction model; issuing the vehicle-mounted controller fault prediction model to the vehicle-mounted controller and deploying the vehicle-mounted controller based on a preset deployment mode; And predicting real-time fault data of the vehicle-mounted controller through the vehicle-mounted controller fault prediction model, and processing faults based on a fault prediction result.
2. The method of claim 1, wherein said multi-category processing of said fault data comprises: Searching lost data in the fault data, determining filling data matched with the lost data, and replacing the lost data with the filling data; And determining invalid data in the fault data based on a preset checking mechanism, and removing the invalid data to obtain cleaned fault data.
3. The method of claim 1, wherein said multi-category processing of said fault data comprises: The fault data are aggregated into a plurality of fault categories based on a preset classification model to obtain a corresponding target classification model, wherein the target classification model characterizes the classification model under the plurality of fault categories; And marking the fault data based on the target classification model so as to align the fault data with the corresponding fault category and obtain marked fault data.
4. The method of claim 1, wherein said multi-category processing of said fault data comprises: adding low-intensity Gaussian white noise into the marked fault data, and determining the fault data of a target type in the fault data; setting a randomly selected range, and scaling the fault data of the target type based on the randomly selected range to obtain enhanced fault data.
5. The method of claim 1, wherein said multi-category processing of said fault data comprises: Inputting the marked fault data and the enhanced fault data into a preset first cyclic neural network model to generate corresponding real hidden vectors; Inputting preset Gaussian white noise into a preset second cyclic neural network model to generate a corresponding false hidden vector; And inputting the true hidden vector and the false hidden vector into a preset third cyclic neural network model to distinguish true and false, and inputting the final true hidden vector and the false hidden vector into a preset fourth cyclic neural network model to obtain synthesized fault data.
6. The method of claim 1, wherein the training and validating a pre-set underlying on-board controller fault prediction model based on the processed fault data comprises: randomly splitting the synthesized fault data into first fault data of a first proportion and second fault data of a second proportion; fine-tuning the base on-board controller fault prediction model based on the first fault data; verifying model accuracy of the vehicle-mounted controller fault prediction model based on the second fault data; and if the model precision is greater than a preset precision threshold, inputting the second fault data into the vehicle-mounted controller fault prediction model for continuous training to obtain a final vehicle-mounted controller fault prediction model.
7. The method according to claim 1, wherein the method further comprises: Responding to the received updating instruction of a manufacturer, judging whether the updating information and the model precision of the current vehicle-mounted controller fault prediction model accord with the updating instruction or not, and feeding back to the manufacturer for secondary determination; Responding to the update confirmation instruction of a manufacturer, packaging the failure prediction model of the vehicle-mounted controller into a model installation package in a target format, and judging that the vehicle is in a dormant state within a preset update time; and in response to the vehicle being in an on-line state, downloading the model installation package through the vehicle and updating based on the model installation package.
8. An in-vehicle controller failure prediction apparatus, characterized by comprising: The processing module is used for acquiring fault data of the vehicle-mounted controller, and carrying out multi-class processing on the fault data to obtain processed fault data, wherein the multi-class processing at least comprises data cleaning, data marking, data enhancement and data synthesis; The acquisition module is used for training and verifying a preset basic vehicle-mounted controller fault prediction model based on the processed fault data to obtain a final vehicle-mounted controller fault prediction model; The deployment module is used for issuing the failure prediction model of the vehicle-mounted controller to the vehicle-mounted controller and deploying the failure prediction model of the vehicle-mounted controller based on a preset deployment mode; And the prediction module is used for predicting real-time fault data of the vehicle-mounted controller through the vehicle-mounted controller fault prediction model and processing faults based on a fault prediction result.
9. An electronic device comprising a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor in communication with the memory via the bus when the electronic device is in operation, the machine-readable instructions when executed by the processor performing the steps of the vehicle-mounted controller failure prediction method of any of claims 1-7.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the in-vehicle controller failure prediction method according to any one of claims 1 to 7.

Description

Vehicle-mounted controller fault prediction method and device, electronic equipment and medium Technical Field The application relates to the technical field of data processing, in particular to a vehicle-mounted controller fault prediction method, a device, electronic equipment and a medium. Background The background art of failure detection of on-board controllers has evolved mainly around the core need of how to find and locate anomalies in an automotive Electronic Control Unit (ECU) accurately in real time. Its evolution is closely related to the complexity of automotive electronics and electrical architecture. At present, a fault prediction model can be constructed through means such as big data, deep learning, data mining and the like to monitor faults of an automobile, so that accuracy and rapidness of vehicle fault diagnosis are realized, and the experience deficiency of after-sales maintenance personnel is solved. However, in the existing vehicle fault diagnosis scheme, the fault prediction model is difficult to adapt to scenes such as vehicle configuration upgrading, newly-added fault modes and the like, the dynamic adaptation capability is general, the occurrence of faults is difficult to predict in advance, and the prevention treatment on the controller is not available before the occurrence of the faults. Disclosure of Invention In view of the above, the present application aims to provide a vehicle-mounted controller fault prediction method, apparatus, electronic device and medium, which trains and verifies a basic vehicle-mounted controller fault prediction model through fault data processed in multiple categories to obtain a final vehicle-mounted controller fault prediction model, and issues the model to the vehicle-mounted controller for deployment, so that the vehicle-mounted controller is subjected to real-time fault prediction through the vehicle-mounted controller fault prediction model, and the fault is processed based on the fault prediction result, so that the vehicle-mounted controller fault prediction method, apparatus and medium can adapt to multiple scenarios, improve dynamic adaptation capability, and predict the occurrence of the fault in advance, and prevent and process the controller in advance before the occurrence of the fault, thereby avoiding the occurrence of the fault and protecting the vehicle-mounted controller. In a first aspect, an embodiment of the present application provides a method for predicting a failure of a vehicle-mounted controller, where the method includes: Acquiring fault data of a vehicle-mounted controller, and performing multi-class processing on the fault data to obtain processed fault data, wherein the multi-class processing at least comprises data cleaning, data labeling, data enhancement and data synthesis; training and verifying a preset basic vehicle-mounted controller fault prediction model based on the processed fault data to obtain a final vehicle-mounted controller fault prediction model; issuing the vehicle-mounted controller fault prediction model to the vehicle-mounted controller and deploying the vehicle-mounted controller based on a preset deployment mode; And predicting real-time fault data of the vehicle-mounted controller through the vehicle-mounted controller fault prediction model, and processing faults based on a fault prediction result. In a possible implementation manner, the performing multi-class processing on the fault data includes: Searching lost data in the fault data, determining filling data matched with the lost data, and replacing the lost data with the filling data; And determining invalid data in the fault data based on a preset checking mechanism, and removing the invalid data to obtain cleaned fault data. In a possible implementation manner, the performing multi-class processing on the fault data includes: The fault data are aggregated into a plurality of fault categories based on a preset classification model to obtain a corresponding target classification model, wherein the target classification model characterizes the classification model under the plurality of fault categories; And marking the fault data based on the target classification model so as to align the fault data with the corresponding fault category and obtain marked fault data. In a possible implementation manner, the performing multi-class processing on the fault data includes: adding low-intensity Gaussian white noise into the marked fault data, and determining the fault data of a target type in the fault data; setting a randomly selected range, and scaling the fault data of the target type based on the randomly selected range to obtain enhanced fault data. In a possible implementation manner, the performing multi-class processing on the fault data includes: Inputting the marked fault data and the enhanced fault data into a preset first cyclic neural network model to generate corresponding real hidden vectors; Inputting preset Gaussian white no