Fault detection method and system for automobile steering controller

Abstract

The invention discloses an automobile steering controller fault detection method and system, and relates to the technical field of automobile electronic control. The surface image acquisition module captures an image through a visual sensor, the quality is improved through the image preprocessing module, the convolutional neural network accurately recognizes appearance defects, and the three-dimensional contour detection and thermal imaging module is triggered to be linked to re-check an abnormal area; the three-dimensional contour module confirms patch offset or tombstone standing abnormity by using a laser scanning technology; the thermal imaging analysis module is matched with a machine learning technology to analyze welding spot temperature abnormity; the predictive fault diagnosis module predicts a defect risk by using deep learning and performs early warning in advance; and the collaborative decision optimization module integrates a multi-module data dynamic optimization detection strategy. Through integration of defect detection, re-checking and prediction, the quality detection precision, coverage and efficiency of the automobile steering controller are greatly improved, the high-quality standard and production stability of products are ensured, and the requirement of the automobile industry for efficient detection is met.

Description

Technical Field

[0001] The present invention relates to the field of automotive electronic control technology, and particularly to a method and system for detecting faults in an automotive steering controller. Background Art

[0002] With the continuous development of the automotive industry, as a key component, the surface quality of an automotive steering controller directly affects the safety and driving experience of the vehicle. Traditional surface defect detection methods mostly rely on manual visual inspection or simple automated equipment. These methods have problems such as low efficiency, poor accuracy, and high missed detection rate, and are difficult to meet the requirements of high-precision and high-efficiency quality control. Especially in the mass production process, manual inspection is not only time-consuming and laborious, but also prone to quality instability due to operator fatigue or technical differences.

[0003] In recent years, with the development of intelligent manufacturing technology, surface quality detection technologies based on machine vision and automated inspection have gradually been applied. Some current automated inspection systems mainly rely on two-dimensional image processing technology to detect the surface of the controller, and can find some obvious surface defects, but there are still limitations in being unable to comprehensively identify deep-seated defects, especially in effectively detecting welding quality, assembly accuracy, and defects in complex shapes.

[0004] In addition, most traditional defect detection systems work in isolation, lacking an effective cooperation mechanism among modules, and it is difficult to dynamically adjust the detection strategy to adapt to different production processes and environmental conditions, resulting in unsatisfactory detection effects. On the other hand, traditional fault diagnosis methods usually rely on historical experience and expert systems, lacking data-driven intelligent analysis means, and unable to give early warnings of potential faults, thus affecting the stability and safety of the production process.

[0005] Therefore, there is an urgent need for a more efficient, intelligent, and comprehensive automotive steering controller fault detection system that can detect and identify various surface defects in real time and accurately, while providing intelligent predictive analysis and decision support to meet the production requirements of high quality, low latency, and high reliability in the modern automotive industry. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for detecting faults in an automotive steering controller, aiming to solve the problems of low efficiency, poor accuracy, high missed detection rate, and insufficient cooperation among modules in the surface quality detection of automotive steering controllers in the prior art. By introducing a variety of intelligent detection technologies, combining image processing, three-dimensional contour scanning, thermal imaging analysis, and deep learning algorithms, the present invention can detect various defects on the surface of the controller in all directions and with high precision, while providing timely fault prediction and dynamic optimization control, providing an efficient and reliable closed-loop solution for the quality control of automotive electronic components.

[0007] To achieve the above object, the present invention provides a method and system for detecting faults in an automotive steering controller, including the following main modules: a surface image acquisition module, an image preprocessing module, a defect intelligent recognition module, a three-dimensional contour detection module, a thermal imaging analysis module, a predictive fault diagnosis module, and a collaborative decision-making optimization module.

[0008] The surface image acquisition module collects high-definition images of the controller surface in real time and generates surface raw image data; The image preprocessing module performs image denoising, enhancement, and normalization processing on the basis of the raw image data to generate preprocessed image data; The defect intelligent recognition module uses a pre-trained convolutional neural network to identify appearance defects based on the preprocessed image data and generates defect types and defect locations; The three-dimensional contour detection module collects three-dimensional contour data by using laser scanning technology based on the defect location information, detects abnormal patch heights, and generates three-dimensional contour analysis data; The thermal imaging analysis module monitors the solder joint temperature distribution by using infrared thermal imaging technology based on the defect location information, and cooperates with a machine learning algorithm to identify virtual soldering or cold soldering defects, generating welding thermal imaging analysis data; The predictive fault diagnosis module uses a deep learning algorithm to perform fault diagnosis prediction and generates a defect prediction report; The collaborative decision-making optimization module dynamically adjusts the detection strategies of each module based on the data of the defect intelligent recognition module, the three-dimensional contour detection module, the thermal imaging analysis module, and the predictive fault diagnosis module to achieve system performance optimization.

[0009] Furthermore, the surface image acquisition module includes an image capture sub-module and an image quality detection sub-module; The image capture sub-module collects images of the automotive steering controller surface by means of real-time shooting. The visual sensor performs real-time focus and exposure calibration on the controller surface through automatic focus and automatic exposure algorithms to ensure stable image brightness and clarity; The image quality detection sub-module receives the raw image data output by the image capture sub-module in real time, detects whether the image clarity meets a predetermined standard, and feeds back to the image capture sub-module in real time when the standard is not met; The detection of image clarity uses an image clarity detection algorithm based on Fourier transform to quickly evaluate the raw image, calculate the image clarity value, and determine whether there are problems such as blurring, jitter, and defocusing in the image.

[0010] Furthermore, the noise reduction process in the image preprocessing module specifically adopts a Gaussian filtering algorithm, the enhancement process specifically adopts a histogram equalization algorithm, and the standardization process specifically adopts an image standardization algorithm.

[0011] Further, the defect intelligent identification module includes a defect classification submodule and a defect location submodule; The defect classification submodule identifies the pre-processed image data through a defect recognition model and outputs the defect type; The defect location submodule adopts a multi-scale image segmentation algorithm to achieve accurate location of the defect area through the following steps: Defect boundary detection function definition: Constructing multi-scale gradient response functions , used to enhance the edge features of defects of different sizes: in, is the Gaussian function at the pixel point The output result, its standard deviation Generated by geometric sequence ( , ) to accommodate the small size differences of the surface components of the automotive steering controller; is the weight coefficient, which is optimized through historical preprocessed image data training to meet , prioritizes amplifying edge information at high-resolution scale; is the gradient operator, represents the convolution operation, is the number of preset convolution kernel scales.

[0012] Defect area extraction and positioning: Image segmentation: Based on The value of the adaptive threshold segmentation algorithm is used, and the threshold , for Mean, for Standard deviation, Points greater than the threshold are regarded as defect areas; Further, the three-dimensional contour detection module includes a laser scanning submodule, a contour analysis submodule and a data linkage submodule; The laser scanning submodule uses a high-precision laser profile scanner to scan the surface of the controller element in real time to obtain accurate three-dimensional profile height point cloud data; The contour analysis sub-module analyzes and processes the three-dimensional height information of the component surface based on the three-dimensional height point cloud data, using point cloud processing algorithms and surface fitting techniques. By calculating the difference between the actual height of the component and the standard design height, it automatically identifies tombstoning anomalies, offset anomalies, and other height anomalies of the chip, and generates a detailed anomaly analysis report. The data linkage sub-module automatically focuses on and locates the abnormal defect area according to the defect position information output by the defect intelligent recognition module, and triggers the laser scanning sub-module to perform targeted secondary fine scanning verification to improve the defect detection accuracy and review efficiency.

[0013] Furthermore, the thermal imaging analysis module includes a temperature monitoring sub-module, a defect analysis sub-module, and a dynamic review sub-module. The temperature capture sub-module uses an infrared thermal imager to capture the temperature change of the solder joint in real time and generates solder joint temperature data. The defect analysis sub-module identifies solder joint voids through an improved solder joint temperature anomaly determination model, and its temperature difference threshold is dynamically calculated as: where, is the historical qualified solder joint temperature data, is the total number of historical qualified solder joint temperature data, is the temperature standard deviation, and are preset parameters. When the solder joint temperature data is greater than the threshold , it is judged as an abnormal temperature area. The dynamic review sub-module judges the abnormal temperature area according to the solder joint temperature data and the three-dimensional contour height point cloud data, and generates a final welding defect report.

[0014] Furthermore, the predictive fault diagnosis module includes a historical data collection sub-module, a prediction modeling sub-module, and a warning output sub-module. The historical data collection sub-module integrates the defect historical data and the real-time detection data in real time to form a fault data set. The prediction modeling sub-module uses a deep neural network to train the fault data set to generate a fault prediction model. The warning output sub-module based on the fault prediction model gives an early warning of potential defects and generates a fault risk report.

[0015] Furthermore, the collaborative decision-making optimization module includes a data integration sub-module, a decision analysis sub-module, and a strategy adjustment sub-module. The data integration sub-module integrates the data from the defect intelligent recognition module, the three-dimensional contour detection module, the thermal imaging analysis module, and the predictive fault diagnosis module. The decision analysis sub-module comprehensively evaluates the detection data by using decision tree algorithms and reinforcement learning algorithms; The policy adjustment sub-module dynamically adjusts the detection parameters based on the analysis results, including detection sensitivity, scanning frequency, and defect threshold, to achieve closed-loop feedback control among modules.

[0016] Furthermore, an intelligent detection and fault prediction method for the surface performance of an automotive steering controller includes the following steps: Step S1: Collect high-definition images of the controller surface and perform image preprocessing; Step S2: Use a convolutional neural network for defect type identification and location; Step S3: Link laser three-dimensional profile detection and thermal imaging analysis technologies to recheck abnormal defect areas; Step S4: Perform predictive fault diagnosis based on historical defect data and real-time detection data to early warn of potential defects; Step S5: Integrate the data of each detection module and dynamically adjust the detection strategy through a collaborative decision-making mechanism.

[0017] Compared with the prior art, the advantages of the present invention are as follows: (1) Omnidirectional detection ability: The present invention realizes comprehensive automated detection of surface defects, welding quality, assembly accuracy, and functional testing of automotive steering controllers by combining machine vision, three-dimensional laser scanning, infrared thermal imaging technology, and deep learning algorithms. Compared with traditional technologies, the present invention effectively makes up for the deficiencies of existing detection methods in multi-dimensional detection capabilities, and significantly improves the accuracy and coverage of detection.

[0018] (2) Intelligent prediction and real-time adjustment: The present invention introduces deep learning and adaptive algorithms. Based on real-time detection data and historical defect data, it conducts fault prediction and early warning, avoiding the problem of the lack of intelligent analysis and warning mechanisms in traditional systems. In addition, the system can dynamically adjust the detection strategy according to the detection results, further improving the detection accuracy and efficiency.

[0019] (3) Multi-module collaborative optimization: The present invention enables each module to automatically adjust parameters according to the previous detection results (such as high-precision re-inspection of surface defect areas and secondary confirmation of temperature anomaly areas) through data sharing and collaborative control mechanisms among modules, solving the problems of isolated data and poor collaboration among modules in traditional detection systems, and effectively improving resource utilization and overall detection efficiency.

[0020] (4)Scalability and Adaptability: The present invention supports the detection requirements of automotive steering controllers of different types and models, including differential tests for those with and without motors. Through a flexible modular design, it can expand functions and configure according to different application scenarios, significantly enhancing the versatility and adaptability of the system and ensuring its wide application in different industrial environments. Brief Description of the Drawings

[0021] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 It is the system module diagram of the present invention; Figure 2 It is the structural diagram of the image preprocessing module of the present invention; Figure 3 It is the structural diagram of the predictive fault diagnosis module of the present invention. Detailed Embodiments

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0024] Embodiment 1: Please refer to Figure 1 As shown, in this embodiment, a fault detection system for an automotive steering controller includes a surface image acquisition module, an image preprocessing module, a defect intelligent recognition module, a three-dimensional contour detection module, a thermal imaging analysis module, a predictive fault diagnosis module, and a collaborative decision-making optimization module.

[0025] The surface image acquisition module real-time acquires high-definition images of the controller surface and generates surface original image data; The surface image acquisition module includes an image capture sub-module and an image quality detection sub-module; The image capture sub-module acquires images of the automotive steering controller surface through real-time shooting. The visual sensor calibrates the automatic focus and automatic exposure of the controller surface to ensure the stability of image brightness and clarity; The image quality detection sub-module receives the original image data output by the image capture sub-module in real time, detects whether the image clarity meets the predetermined standard, and feeds back to the image capture sub-module in real time when the standard is not met. The detection of image clarity uses an image clarity detection algorithm based on Fourier transform to quickly evaluate the original image, calculate the image clarity value, and determine whether there are problems such as blur, jitter, and defocus in the image.

[0026] The autofocus algorithm uses hybrid focusing to quickly locate using phase detection, and further optimizes the accuracy through contrast detection, which is suitable for various lighting and motion scenarios.

[0027] The contrast detection depends on evaluating the clarity in the image and adjusts the lens position to maximize the contrast; the phase detection quickly calculates the focus position by detecting the phase difference in the image.

[0028] The automatic exposure algorithm automatically adjusts the camera's exposure settings by analyzing the brightness, contrast, and regional distribution of the image.

[0029] The image preprocessing module, as shown in Figure 2 Based on the original image data, performs image denoising processing, enhancement processing, and normalization processing to generate preprocessed image data. The denoising processing in the image preprocessing module specifically uses the Gaussian filtering algorithm, and the Gaussian function is expressed as follows: Among them, are the coordinates of the pixel point, is the standard deviation of the Gaussian function, which controls the smoothness; the Gaussian filtering is achieved by convolving the Gaussian kernel with the image, expressed as follows: Among them, are the coordinates of the pixel point; is the radius of the Gaussian kernel; is the offset relative to the central pixel ; is the output result of the Gaussian function at the pixel point ; represents the image at the position ; is the image after Gaussian filtering; The enhancement processing specifically uses the histogram equalization algorithm, which redistributes the pixel intensity values of the image to make the histogram of the image more evenly distributed, expressed as follows: Among them, is the enhanced image; are the coordinates of the pixel point; is the image after the Gaussian filtering process; is to the value obtained by inputting into the cumulative distribution function and normalizing; The normalization process specifically uses an image normalization algorithm, which is expressed as follows: where, is the image after normalization; is the enhanced image; is the minimum pixel value of the enhanced image, is the maximum pixel value of the enhanced image; The defect intelligent recognition module, based on the preprocessed image data, uses a pre-trained convolutional neural network to recognize appearance defects and generate defect types and defect locations; The defect intelligent recognition module includes a defect classification sub-module and a defect localization sub-module; The defect classification sub-module uses a defect recognition model to recognize the preprocessed image data and output the defect type; The defect localization sub-module uses a multi-scale image segmentation algorithm to accurately locate the defect area through the following steps: Definition of the defect boundary detection function: Construct a multi-scale gradient response function , which is used to enhance the edge features of defects of different sizes: where, is the output result of the Gaussian function at the pixel point , and its standard deviation is generated in a geometric sequence ( , ) to adapt to the small size differences of the surface components of the automotive steering controller; is the weight coefficient, which is trained and optimized through historical preprocessed image data and satisfies , giving priority to amplifying the edge information at high-resolution scales; is the gradient operator, represents the convolution operation, is the preset number of convolution kernel scales.

[0030] Defect area extraction and localization: Image segmentation: Based on the value of , an adaptive threshold segmentation algorithm is used, and the threshold , is​ Mean value, is standard deviation, Points greater than the threshold are regarded as defective areas; Based on the defective position information, the three-dimensional contour detection module uses laser scanning technology to collect three-dimensional contour data, detect abnormal patch heights, and generate three-dimensional contour analysis data; The three-dimensional contour detection module includes a laser scanning sub-module, a contour analysis sub-module, and a data linkage sub-module; The laser scanning sub-module uses a high-precision laser contour scanner to scan the surface of the controller component in real time to obtain accurate three-dimensional contour height point cloud data; Based on the three-dimensional height point cloud data, the contour analysis sub-module uses point cloud processing algorithms and surface fitting techniques to analyze and process the three-dimensional height information of the component surface. By calculating the difference between the actual height of the component and the standard design height, it automatically identifies tombstoning abnormalities, offset abnormalities, and other height abnormalities of the patch, and generates a detailed abnormality analysis report; According to the defective position information output by the defective intelligent recognition module, the data linkage sub-module automatically focuses on and locates the abnormal defective area, triggering the laser scanning sub-module to perform targeted secondary fine scanning verification to improve the accuracy of defective detection and the review efficiency.

[0031] Based on the defective position information, the thermal imaging analysis module uses infrared thermal imaging technology to monitor the temperature distribution of solder joints, and cooperates with machine learning algorithms to identify defective soldering or cold soldering defects, and generates welding thermal imaging analysis data; The thermal imaging analysis module includes a temperature monitoring sub-module, a defective analysis sub-module, and a dynamic review sub-module; The temperature capture sub-module uses an infrared thermal imager to capture the temperature change of the solder joint in real time and generates solder joint temperature data; The defective analysis sub-module identifies defective soldering through an improved solder joint temperature abnormality determination model, and its temperature difference threshold is dynamically calculated as: where is the solder joint temperature data of historical qualified ones, is the total number of solder joint temperature data of historical qualified ones, is the temperature standard deviation, and are preset parameters. When the solder joint temperature data is greater than the threshold , it is judged as an abnormal temperature area; The dynamic review sub-module judges the abnormal temperature area based on the solder joint temperature data and the three-dimensional contour height point cloud data, and generates a final welding defect report.

[0032] The predictive fault diagnosis module uses deep learning algorithms to conduct fault diagnosis and prediction, and generates a defect prediction report; The predictive fault diagnosis module includes a historical data collection sub-module, a prediction modeling sub-module, and a warning output sub-module, as shown in Figure 3 the figure; The historical data collection sub-module integrates defect historical data and real-time detection data in real time to form a fault data set; The prediction modeling sub-module uses a deep neural network to train the fault data set and generates a fault prediction model; The warning output sub-module, based on the fault prediction model, gives early warnings of potential defects and generates a fault risk report.

[0033] The collaborative decision-making optimization module dynamically adjusts the detection strategies of each module based on the data of the defect intelligent recognition module, the three-dimensional contour detection module, the thermal imaging analysis module, and the predictive fault diagnosis module to achieve system performance optimization.

[0034] The collaborative decision-making optimization module includes a data integration sub-module, a decision analysis sub-module, and a strategy adjustment sub-module; The data integration sub-module integrates data from the defect intelligent recognition module, the three-dimensional contour detection module, the thermal imaging analysis module, and the predictive fault diagnosis module; The decision analysis sub-module comprehensively evaluates the detection data using decision tree algorithms and reinforcement learning algorithms; The strategy adjustment sub-module dynamically adjusts the detection parameters based on the analysis results, including detection sensitivity, scanning frequency, and defect threshold, to achieve closed-loop feedback control between each module; The working process of the decision tree algorithm: The input data includes defect type, three-dimensional contour outliers, thermal imaging temperature distribution, and predictive risk score; Decision rules: First-layer judgment: If the defect type is an appearance defect and the confidence level is greater than 90%, it is directly marked as "confirmed defect", triggering a local fine scan of the three-dimensional contour module; Second-layer judgment: If the defect type is solder joint temperature anomaly and the predictive risk score is greater than or equal to the high-risk threshold, immediately increase the scanning frequency of the thermal imaging module to the highest gear; Third-layer judgment: If the deviation of the patch height detected by the three-dimensional contour is less than the standard tolerance, but the thermal imaging shows local temperature anomaly, reduce the defect determination threshold and require re-inspection; The output actions include adjusting detection sensitivity, controlling module collaboration, and optimizing resource allocation.

[0035] The working process of the reinforcement learning algorithm: The reinforcement learning algorithm realizes dynamic optimization through a closed-loop mechanism of environmental perception, action decision-making, reward feedback, and policy iteration: Environmental perception: Real-time collection of detection data as state input; Action decision-making: Generation of adjustment instructions based on the current state; Reward feedback: Quantifying the reward value based on the action execution effect and evaluating the quality of the policy; Policy iteration: Through maximizing the long-term cumulative reward, optimizing the parameter configuration and module cooperation logic, and finally achieving the adaptive balance of detection accuracy and efficiency.

[0036] The rule triggering result of the decision tree serves as the initial policy reference for reinforcement learning, and the long-term learning result of reinforcement learning feeds back to the decision tree to update its rule base. An intelligent detection and fault prediction method for the surface performance of an automotive steering controller, comprising the following steps: Step S1: Collect high-definition images of the controller surface and perform image preprocessing; Step S2: Use a convolutional neural network for defect type recognition and location; Step S3: Link the laser three-dimensional contour detection and thermal imaging analysis technologies to review the abnormal defect area; Step S4: Perform predictive fault diagnosis based on historical defect data and real-time detection data to early warn of potential defects; Step S5: Synthesize the data of each detection module and dynamically adjust the detection strategy through a collaborative decision-making mechanism.

[0037] Example 2: After assembly, the patch capacitor of the steering controller has an offset of 0.2 mm (the standard tolerance is ±0.1 mm): Detection process: First, the image preprocessing module preprocesses the collected original surface image data to generate preprocessed image data; Second, the defect localization sub-module uses a multi-scale image segmentation algorithm to achieve defect localization: Set parameters , weights , calculate the gradient response function , calculate the threshold and determine the abnormal area; Finally, based on the three-dimensional height point cloud data, use a point cloud processing algorithm and surface fitting technology to analyze and process the three-dimensional height information of the component surface, and automatically identify the abnormal patch offset by calculating the difference between the actual height and the standard design height of the component; Dynamic policy adjustment: Automatically increase the detection sensitivity according to the overlimit ratio of the offset amount to expand the defect capture range; Notify the production line to pause the batch of controllers through the system and mark the defect coordinates.

[0038] Example 3: The solder joints of the steering controller have a risk of virtual soldering due to uneven reflow soldering temperature: Detection process: The temperature capture sub-module uses an infrared thermal imager to capture the temperature change of the solder joint in real time and generates solder joint temperature data; the defect analysis sub-module identifies a cold solder joint through an improved solder joint temperature anomaly determination model. First, a threshold value is calculated , when the solder joint temperature data is greater than the threshold value , then the dynamic review sub-module generates a local temperature difference based on the solder joint temperature data and a solder joint height difference based on the three-dimensional contour height point cloud data. When both the local temperature difference and the solder joint height difference are not within the preset safety threshold range, the abnormal temperature area is determined to be a cold solder joint.

[0039] Dynamic response: The scanning frequency of the thermal imaging module is increased to focus on the abnormal area; Early warning mechanism: Trigger an audible and visual alarm, and at the same time push a maintenance work order to the designated workstation through the system.

[0040] In the description of this specification, the description with reference to terms such as "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0041] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to the specific embodiments shown. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and utilize the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A vehicle steering controller fault detection system, characterized in that: The system comprises: The surface image acquisition module acquires high-definition images of the controller surface in real time and generates original surface image data; The image preprocessing module performs image noise reduction, enhancement and standardization processing based on the original image data to generate preprocessed image data; The defect intelligent recognition module uses a pre-trained convolutional neural network to recognize appearance defects based on the pre-processed image data and generates defect types and defect locations; The 3D contour detection module uses laser scanning technology to collect 3D contour data based on the defect position information, detects abnormal patch height, and generates 3D contour analysis data; The thermal imaging analysis module uses infrared thermal imaging technology to monitor the temperature distribution of solder joints based on the defect location information, and cooperates with machine learning algorithms to identify false solder joints or cold solder joints to generate welding thermal imaging analysis data; The predictive fault diagnosis module uses deep learning algorithms to predict fault diagnosis and generate defect prediction reports; The collaborative decision-making optimization module dynamically adjusts the detection strategies of each module based on the data of the defect intelligent recognition module, three-dimensional contour detection module, thermal imaging analysis module and predictive fault diagnosis module to achieve system performance optimization.

2. The system according to claim 1, characterized in that The surface image acquisition module includes an image capture submodule and an image quality detection submodule; The image capture submodule collects images of the surface of the vehicle steering controller by real-time shooting, and the visual sensor performs real-time focus and exposure calibration on the controller surface by automatic focus and automatic exposure algorithms to ensure stable image brightness and clarity; The image quality detection submodule receives the original image data output by the image capture submodule in real time, detects whether the image clarity meets the predetermined standard, and provides feedback to the image capture submodule in real time if the standard is not met; The image clarity detection adopts the image clarity detection algorithm based on Fourier transform to quickly evaluate the original image, calculate the image clarity value, and determine whether the image has the problems of blur, shaking, out of focus, etc.

3. The system according to claim 1, characterized in that The noise reduction process in the image preprocessing module specifically adopts a Gaussian filtering algorithm, the enhancement process specifically adopts a histogram equalization algorithm, and the standardization process specifically adopts an image standardization algorithm.

4. The system according to claim 1, characterized in that The defect intelligent identification module includes a defect classification submodule and a defect location submodule; The defect classification submodule identifies the pre-processed image data through a defect recognition model and outputs the defect type; The defect location submodule adopts a multi-scale image segmentation algorithm to achieve accurate location of the defect area through the following steps: Defect boundary detection function definition: Constructing multi-scale gradient response functions , used to enhance the edge features of defects of different sizes: in, is the Gaussian function at the pixel point The output result, its standard deviation Generated by geometric sequence ( , ) to accommodate the small size differences of the surface components of the automotive steering controller; is the weight coefficient, which is optimized through historical preprocessed image data training to meet , prioritizes amplifying edge information at high-resolution scale; is the gradient operator, represents the convolution operation, The number of preset convolution kernel scales; Defect area extraction and positioning: Image segmentation: Based on The value of the adaptive threshold segmentation algorithm is used, and the threshold , for Mean, for Standard deviation, Points larger than the threshold are regarded as defect areas.

5. The system according to claim 1, characterized in that The three-dimensional contour detection module includes a laser scanning submodule, a contour analysis submodule and a data linkage submodule; The laser scanning submodule uses a high-precision laser profile scanner to scan the surface of the controller element in real time to obtain accurate three-dimensional profile height point cloud data; The contour analysis submodule analyzes and processes the three-dimensional height information of the component surface based on the three-dimensional height point cloud data by using the point cloud processing algorithm and surface fitting technology, and automatically identifies the abnormality of the patch's tombstone, abnormality of offset and other abnormal height conditions by calculating the difference between the actual height of the component and the standard design height, and generates a detailed abnormality analysis report; The data linkage submodule automatically focuses on and locates the abnormal defect area according to the defect position information output by the defect intelligent recognition module, and triggers the laser scanning submodule to perform targeted secondary fine scanning verification to improve the defect detection accuracy and review efficiency.

6. The system according to claim 1, characterized in that The thermal imaging analysis module includes a temperature monitoring submodule, a defect analysis submodule, and a dynamic review submodule; The temperature capture submodule uses an infrared thermal imager to capture solder point temperature changes in real time and generate solder point temperature data; The defect analysis submodule identifies cold solder joints through an improved solder joint temperature anomaly determination model, and its temperature difference threshold The dynamic calculation is: in, For historical qualified solder joint temperature data, is the total number of historical qualified solder joint temperature data, is the temperature standard deviation, and For preset parameters, when the solder point temperature data is greater than the threshold , it is judged as an abnormal temperature area; The dynamic review submodule determines the abnormal temperature area according to the welding point temperature data and the three-dimensional profile height point cloud data, and generates a final welding defect report.

7. The system according to claim 1, characterized in that The predictive fault diagnosis module includes a historical data collection submodule, a predictive modeling submodule, and an early warning output submodule; The historical data collection submodule integrates defect historical data and real-time detection data in real time to form a fault data set; The prediction modeling submodule uses a deep neural network to train the fault data set to generate a fault prediction model; The warning output submodule provides early warning of potential defects and generates a fault risk report based on the fault prediction model.

8. The system according to claim 1, characterized in that The collaborative decision-making optimization module includes a data integration submodule, a decision analysis submodule, and a strategy adjustment submodule; The data integration submodule integrates data from the defect intelligent identification module, the three-dimensional contour detection module, the thermal imaging analysis module and the predictive fault diagnosis module; The decision analysis submodule uses a decision tree algorithm and a reinforcement learning algorithm to comprehensively evaluate the detection data; The strategy adjustment submodule dynamically adjusts the detection parameters based on the analysis results, including detection sensitivity, scanning frequency and defect threshold, to achieve closed-loop feedback control between modules.

9. A method for detecting a fault in an automobile steering controller, characterized in that: The following steps are involved: Step S1: collecting high-definition images of the controller surface and performing image preprocessing; Step S2: using convolutional neural network to identify defect types and locate positions; Step S3: Linking laser 3D profile detection and thermal imaging analysis technology to review abnormal defect areas; Step S4: Perform predictive fault diagnosis based on historical defect data and real-time detection data to provide early warning of potential defects; Step S5: Integrate the data of each detection module and dynamically adjust the detection strategy through a collaborative decision-making mechanism.

Images