CN-121997167-A - Intelligent fault diagnosis method for automobile wire harness detection platform

CN121997167ACN 121997167 ACN121997167 ACN 121997167ACN-121997167-A

Abstract

The invention relates to the technical field of automobile wire harness detection and discloses an intelligent fault diagnosis method for an automobile wire harness detection platform, which comprises the following steps of utilizing the automobile wire harness detection platform to collect multi-source data and carrying out data enhancement pretreatment; the method comprises the steps of constructing a multi-source data fusion and parallel light one-dimensional convolutional neural network model based on a multi-branch attention mechanism, extracting multi-source data features by using a light one-dimensional convolutional neural network in the multi-source data fusion and parallel light one-dimensional convolutional neural network model based on the multi-branch attention mechanism, fusing the features by using the multi-branch attention mechanism, and outputting a final fault diagnosis result by using a Softmax classifier. According to the invention, voltage, resistance, temperature and air pressure signals are processed through a parallel network structure, the self-adaptive weighting fusion of characteristics is realized by utilizing a channel attention mechanism and a multi-branch attention mechanism, the complementarity of multi-source data is fully excavated, and the fault diagnosis accuracy of the automobile wire harness detection platform is improved.

Inventors

LI WEI
LI WEI
CHEN YAN
LI MENGJIANG
Cao Jinni
WANG CUI
DENG MIN

Assignees

长沙波特尼电气系统有限公司

Dates

Publication Date: 20260508
Application Date: 20251231

Claims (10)

1. An intelligent fault diagnosis method for an automobile wire harness detection platform is characterized by comprising the following steps of: Collecting multi-source data by using an automobile wire harness detection platform, and carrying out data enhancement pretreatment on the multi-source data; Constructing a multi-branch attention mechanism-based multi-source data fusion and parallel light one-dimensional convolutional neural network model, wherein the multi-branch attention mechanism-based multi-source data fusion and parallel light one-dimensional convolutional neural network model is marked as an MBAM-LM-1DCNN model; inputting the multi-source data subjected to data enhancement preprocessing into the MBAM-LM-1DCNN model; performing feature extraction by using a light one-dimensional convolutional neural network in the MBAM-LM-1DCNN model, wherein the light one-dimensional convolutional neural network is marked as LM-1DCNN; feature fusion is performed by utilizing a multi-branch attention mechanism in the MBAM-LM-1DCNN model, wherein the multi-branch attention mechanism is recorded as MBAM; And outputting the fault class probability of the automobile wire harness detection platform by using a Softmax classifier in the MBAM-LM-1DCNN model, and taking the class corresponding to the highest fault class probability as a final fault diagnosis result.
2. The intelligent fault diagnosis method for an automotive harness detection platform of claim 1, wherein the multi-source data comprises a voltage signal, a resistance signal, a temperature signal and a barometric pressure signal; The MBAM-LM-1DCNN model adopts a parallel network structure, and comprises a first data source processing branch and a second data source processing branch; The first data source processing branch is used for processing the voltage signal, and the second data source processing branch is used for processing the resistance signal; the LM-1DCNN is adopted by the first data source processing branch and the second data source processing branch; the LM-1DCNN consists of an input layer, a convolution layer, a batch normalization layer, an activation function layer, a pooling layer and a flattening layer.
3. The intelligent fault diagnosis method for the automobile harness detection platform according to claim 2, wherein the convolution layer performs sliding window calculation on an input signal by using a convolution kernel, and local features of the input signal are extracted through convolution operation; The batch normalization layer performs normalization processing on the output of the convolution layer; the activation function layer adopts a ReLU activation function; the pooling layer adopts the maximum pooling operation to perform downsampling on the feature map; The flattening layer is positioned at the tail end of the LM-1DCNN, expands the multidimensional feature map into a one-dimensional feature vector, and the one-dimensional feature vector is used as the input of the multi-branch attention mechanism.
4. The intelligent fault diagnosis method for an automobile wire harness detection platform according to claim 1, wherein the intelligent fault diagnosis method for an automobile wire harness detection platform introduces a channel attention mechanism in a feature extraction network; The channel attention mechanism is used for processing a feature map output by the one-dimensional convolutional neural network; the channel attention mechanism includes a compression operation, an excitation operation, and a feature recalibration operation.
5. The intelligent fault diagnosis method for an automotive harness detection platform according to claim 4, wherein the compressing operation compresses the feature map in a spatial dimension using global averaging pooling, compressing each two-dimensional feature channel by a real number; the excitation operation utilizes two full-connection layers to establish the correlation between channels, and weight of each characteristic channel is generated; the feature weight scaling operation applies the weights output by the excitation operation to the original feature map, weighting by multiplication channel by channel to the previous feature.
6. The intelligent fault diagnosis method for an automotive harness detection platform of claim 2, wherein the multi-branch attention mechanism comprises a first local branch, a second local branch, and a third global fusion branch; the first local branch is used for processing the characteristics of the first data source, the second local branch is used for processing the characteristics of the second data source, and the third global fusion branch is used for calculating the global weights of the characteristics of different data sources.
7. The intelligent fault diagnosis method for an automotive harness detection platform according to claim 6, wherein the first local branch receives a first flattened feature vector output by the lightweight one-dimensional convolutional neural network, and performs channel attention processing on the first flattened feature vector to obtain a first refined feature vector; And the second local branch receives a second flattened feature vector output by the lightweight one-dimensional convolutional neural network, and performs channel attention processing on the second flattened feature vector to obtain a second refined feature vector.
8. The intelligent fault diagnosis method for an automotive harness detection platform of claim 7, wherein the third global fusion branch receives the first refined feature vector and the second refined feature vector simultaneously; The third global fusion branch utilizes a full connection layer to perform global relevance analysis on the first refined feature vector and the second refined feature vector to generate attention scores; And the multi-branch attention mechanism normalizes the attention score by using a Softmax function to generate weight coefficients corresponding to different data sources.
9. The intelligent fault diagnosis method for an automotive harness detection platform according to claim 8, wherein the multi-branch attention mechanism performs weighted fusion on the refined feature vectors of the first and second local branches according to the generated weight coefficients to generate a final fused feature vector; the fusion feature vector is input to a full connection layer that connects the Softmax classifier.
10. The intelligent fault diagnosis method for an automotive harness detection platform according to claim 1, wherein the automotive harness detection platform collects the multi-source data based on a resistance detection function, a voltage detection function, a temperature test function and an air tightness test function; the resistance detection function is used for detecting the on-off state of the wire harness, the voltage detection function is used for detecting the voltage withstand value of the wire harness, the temperature test function is used for detecting the temperature resistance characteristic of the wire harness, and the air tightness test function is used for detecting the sealing performance of the wire harness.

Description

Intelligent fault diagnosis method for automobile wire harness detection platform Technical Field The invention relates to the technical field of automobile wire harness detection, in particular to an intelligent fault diagnosis method for an automobile wire harness detection platform. Background The automobile wire harness is a network main body of an automobile circuit, and the safety and reliability of the automobile wire harness directly determine the running stability of an entire automobile system. Along with the continuous improvement of the electronic degree of the automobile, the structure of the automobile wire harness is increasingly complex, and the types and the number of signals borne by the automobile wire harness are exponentially increased. The automobile wire harness detection platform is key equipment for guaranteeing the quality of automobile wire harnesses. The main task of the automobile wire harness detection platform is to timely and accurately identify various faults generated in the production or use process of the wire harness. Existing automotive harness detection techniques typically rely on a single type of signal parameter for judgment. The common detection mode mainly judges the on-off state of the wire harness through measuring the resistance value, or evaluates the voltage resistance of the wire harness through measuring the voltage value. It is difficult to comprehensively reflect the real health state of the automobile wire harness in a complex environment only depending on a single type of signal parameter. The automobile wire harness can be influenced by coupling of environmental factors such as temperature, air pressure and the like in actual work. The signal data with single dimension often contains one-sided information, and can not effectively capture early weak fault characteristics under the coupling of multiple physical fields, so that the condition of missed detection or misjudgment easily occurs when the automobile wire harness detection platform faces a complex fault mode. On the other hand, the conventional automobile harness fault diagnosis method often adopts a shallow machine learning model with a fixed threshold value or based on artificial feature extraction. The method of setting the fixed threshold value lacks adaptability to dynamic changes of signals, and intermittent faults with small signal fluctuation are difficult to detect. The method based on artificial feature extraction relies on expert experience, and the feature extraction process is tedious and has weak generalization capability. While some prior art techniques began to attempt to introduce deep convolutional neural networks to automatically extract features, standard convolutional neural network models typically have a significant amount of parameters and computation. The huge parameter and calculation amount make the requirement of the model on hardware computing resources extremely high, and efficient parallel processing and real-time diagnosis are difficult to realize on the embedded terminal of the automobile wire harness detection platform with limited resources. Furthermore, in scenarios involving multi-source data processing, the prior art lacks efficient data fusion mechanisms. The prior art typically employs simple data stitching or fixed weight averaging strategies to process multi-source data. Simple data stitching or fixed weight averaging strategies ignore differences in data heterogeneity between different data sources and their contribution to fault diagnosis. The importance of the voltage signal, the resistance signal, the temperature signal and the air pressure signal is dynamically variable under different fault categories. The lack of self-adaptive recalibration on the importance of the characteristic channel and dynamic analysis on the global relevance of the multi-source data make the prior art unable to fully utilize the complementary advantages of the multi-source data, and limit the further improvement of the fault diagnosis accuracy of the automobile wire harness detection platform. Therefore, the invention provides an intelligent fault diagnosis method for an automobile wire harness detection platform, which aims to solve the defects in the prior art. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an intelligent fault diagnosis method for an automobile wire harness detection platform, which solves the problems that weak faults under multi-physical field coupling are difficult to accurately identify by relying on single type signal parameters, the calculation complexity of a standard convolutional neural network model is high, the real-time detection requirement of an embedded terminal cannot be met, and the data complementarity cannot be fully utilized by multi-source data fusion strategy rigidification so as to limit the improvement of fault diagnosis accuracy. In order to achieve the above purpose, the invention is realiz