Search

CN-122020389-A - Welding quality detection method for flexible circuit board of audio equipment

CN122020389ACN 122020389 ACN122020389 ACN 122020389ACN-122020389-A

Abstract

The invention discloses a welding quality detection method for a flexible circuit board of audio equipment, which relates to the technical field of electronic manufacturing quality detection and comprises the steps of synchronously collecting multispectral images of a welding area and infrared thermal imaging time sequence data, and forming a standardized welding spot image sequence and a temperature change curve through registration and enhancement processing. And extracting geometric texture features of the image and heat conduction and cooling features of a temperature curve, fusing and constructing a multidimensional feature descriptor, and inputting the multidimensional feature descriptor into a deep learning network to obtain the quality grade and defect probability of the welding spot. Combining the probability of defects with the temperature profile, locating defects and analyzing the process stages they produce, and finally generating an assessment report containing repair suggestions. The method realizes multi-mode dynamic detection and effectively improves the identification accuracy of the hidden defects and the traceability of the process problems.

Inventors

  • WANG JIE
  • CAO JUAN

Assignees

  • 衡阳市硕丰电子有限公司

Dates

Publication Date
20260512
Application Date
20260205

Claims (10)

  1. 1. The method for detecting the welding quality of the flexible circuit board of the audio equipment is characterized by comprising the following steps of: acquiring multispectral scanning image data and infrared thermal imaging time sequence data of the surface of a flexible circuit board of the audio equipment to be detected, and forming an original welding area multimodal data set; Performing image enhancement and data registration processing on the original welding area multi-mode data set to generate a standardized welding spot characteristic image sequence and a corresponding temperature distribution time sequence curve; Extracting geometric feature vectors and texture distribution feature vectors of welding spots from the standardized welding spot feature image sequence, and extracting heat conduction feature vectors and cooling feature vectors of the welding spots from the temperature distribution time sequence curve; Performing feature fusion on the geometric feature vector, the texture distribution feature vector, the heat conduction feature vector and the cooling feature vector to construct a multidimensional welding spot quality feature descriptor; inputting the multidimensional welding spot quality characteristic descriptors into a preset deep learning classification network, and outputting quality grade labels and defect type probability distribution of each welding spot by the deep learning classification network; Positioning the specific position of the welding defect and analyzing the time phase of defect generation based on the probability distribution of the defect type and the temperature distribution time sequence curve; And generating a welding quality assessment report containing repair priority and repair suggestions according to the quality grade label and defect position information of each welding spot.
  2. 2. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 1, wherein the performing image enhancement and data registration processing on the original multi-mode data set of the welding area to generate a standardized welding spot characteristic image sequence and a corresponding temperature distribution time sequence curve includes: denoising the multispectral scanning image data, eliminating image noise by adopting a self-adaptive filtering algorithm, and enhancing the contrast ratio of welding spots and the background; Performing temperature calibration on the infrared thermal imaging time sequence data, converting an original thermal imaging gray value into an accurate temperature value, and establishing a time-temperature corresponding relation; recognizing the outline position of a welding spot in the multispectral image, and establishing a local coordinate system by taking the center of the welding spot as a reference; Mapping the infrared thermal imaging data into the local coordinate system to ensure that the position of each welding spot in the thermal imaging data and the multispectral image corresponds accurately; performing size normalization processing on the registered multispectral image sequence, and adjusting all welding spot images to uniform resolution to form the standardized welding spot characteristic image sequence; and extracting temperature change data of each welding spot in the registered thermal imaging data, and arranging the temperature change data in time sequence to form the corresponding temperature distribution time sequence curve.
  3. 3. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 2, wherein the extracting the geometric feature vector and the texture distribution feature vector of the welding spot from the standardized welding spot feature image sequence comprises: performing edge detection on each frame of image in the standardized welding spot characteristic image sequence, and extracting the complete boundary of the welding spot outline; Calculating the area, perimeter, circularity, length-width ratio and eccentricity of the outline of the welding spot, and combining the geometric parameters to form the geometric form feature vector; dividing a welding spot image into a plurality of sub-areas, and calculating a gray level co-occurrence matrix in each sub-area; extracting four texture characteristic indexes of contrast, correlation, energy and homogeneity from the gray level co-occurrence matrix; And carrying out statistical analysis on texture feature indexes of all the subareas, and calculating a mean value, a variance and an extremum to form the texture distribution feature vector.
  4. 4. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 2, wherein the extracting the heat conduction characteristic vector and the cooling characteristic vector of the welding spot from the temperature distribution time sequence curve comprises: Identifying a heating stage, a peak platform stage and a cooling stage in a temperature distribution time sequence curve; In the heating stage, calculating an initial slope, a maximum slope and an average heating rate of temperature rise to form a heating characteristic group; in the peak value platform stage, calculating the stability index of peak value temperature and the duration of the platform to form a peak value characteristic group; In the cooling stage, calculating an initial cooling rate, an average cooling rate and a final stable temperature of temperature reduction to form a cooling characteristic group; combining the heating characteristic group, the peak characteristic group and the cooling characteristic group in time sequence to form the heat conduction characteristic vector; and (3) analyzing a temperature drop curve in a cooling stage, fitting an exponential decay model, and extracting a decay time constant and a steady-state temperature deviation to form the cooling characteristic vector.
  5. 5. The method for detecting the welding quality of the flexible circuit board of the audio equipment according to claim 1, wherein the feature fusion of the geometric feature vector, the texture distribution feature vector, the heat conduction feature vector and the cooling feature vector is performed, and the multi-dimensional welding spot quality feature descriptor is constructed, and the method comprises the following steps: carrying out normalization processing on each geometric parameter in the geometric feature vector to eliminate dimension influence; Performing principal component analysis on texture indexes in the texture distribution feature vector, and extracting principal texture feature components; performing time sequence coding on each stage of characteristics in the heat conduction characteristic vector, and reserving time sequence correlation; Carrying out logarithmic transformation on attenuation parameters in the cooling characteristic vector, and enhancing characteristic distinction degree; Splicing the processed geometric form features, main texture feature components, time sequence coding features and logarithmically transformed cooling features according to a preset sequence by adopting a feature level fusion mode; And performing dimension reduction processing on the spliced feature vectors, and reserving the feature component with the largest information quantity to form the multidimensional welding spot quality feature descriptor with fixed dimension.
  6. 6. The method for detecting the welding quality of the flexible circuit board of the audio equipment according to claim 1, wherein the step of inputting the multidimensional welding spot quality feature descriptor into a preset deep learning classification network and outputting the quality grade label and the defect type probability distribution of each welding spot comprises the following steps: the deep learning classification network comprises a feature extraction layer, an attention mechanism layer and a full-connection classification layer; the feature extraction layer receives the multidimensional welding spot quality feature descriptors and extracts deep feature representations through a plurality of convolution layers and a pooling layer; The attention mechanism layer analyzes importance weights of different feature dimensions in the deep feature representation, enhances key features and suppresses secondary features; The fully-connected classification layer maps the weighted characteristic representation to a predefined classification space and outputs probability values of each welding spot belonging to each quality grade; taking the quality grade with the maximum probability value as the quality grade label; Meanwhile, the full-connection classifying layer also outputs probability distribution of welding spots belonging to various defect types, and the probability distribution of the defect types is formed, wherein the defect types comprise cold welding, bridging, insufficient welding flux and excessive welding flux.
  7. 7. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 6, wherein the steps of locating a specific position of the welding defect and analyzing a time period of the defect generation based on the probability distribution of the defect type and the time sequence curve of the temperature distribution include: for defect types with probability exceeding a threshold value in the probability distribution of the defect types, backtracking to an original multi-mode data set of the welding area; positioning the accurate position of a welding spot with defects in a standardized welding spot characteristic image sequence, and recording row coordinates of the welding spot; analyzing abnormal temperature points of the welding spots in the welding process in the corresponding temperature distribution time sequence curve; Comparing the abnormal temperature points with a standard temperature curve of a welding process, and identifying a time period when the temperature deviates from a normal range; judging specific process stages possibly generating defects according to the type and the degree of temperature deviation, wherein the process stages comprise a preheating stage, a welding stage and a cooling stage; and combining the welding spot position information and the defect generation stage information to form a complete defect positioning and stage analysis result.
  8. 8. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 1, wherein the generating a welding quality evaluation report including repair priority and repair advice according to the quality class label and the defect position information of each welding spot comprises: Counting the quality grade distribution of all welding spots, and calculating the proportion of qualified welding spots and the quantity of various defect welding spots; according to the defect type and the defect severity, assigning a repair priority score to each defective welding spot, wherein a higher priority score indicates that early repair is required; For each defect type, retrieving corresponding repair process parameters and operation key points from a repair knowledge base; Combining the position information of specific welding spots to generate personalized repair path planning, so as to avoid influencing other welding spots in the repair process; and integrating the quality grade distribution statistical result, the repair priority ordering of the defective welding spots, the personalized repair suggestion and the repair path planning to form a structured welding quality evaluation report.
  9. 9. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 7, wherein comparing the abnormal temperature point with a standard temperature curve of the welding process, the identifying the period of time when the temperature deviates from the normal range, comprises: Retrieving a standard welding temperature curve corresponding to the current product model from a welding process parameter database, wherein the standard welding temperature curve defines an ideal temperature range and duration of each process stage; aligning a temperature distribution time sequence curve of a welding spot to be detected with a standard welding temperature curve on a time axis; Calculating the temperature difference value of the two curves at each time point to form a temperature difference curve; Setting a temperature difference threshold value, and marking the time point as a temperature abnormal point when the temperature difference exceeds the threshold value; analyzing the distribution condition of continuous temperature anomaly points, and identifying a temperature anomaly duration period; and determining the specific stage of defect generation according to the process stage of the temperature anomaly duration period.
  10. 10. The method for detecting the welding quality of the flexible circuit board of the audio device according to claim 8, wherein the retrieving the corresponding repair process parameters and operation points from the repair knowledge base for each defect type comprises: Establishing a mapping relation table of defect types and repair processes, wherein the mapping relation table records repair methods, required tools, temperature parameters and time parameters corresponding to different defect types; Searching a corresponding repairing method in a mapping relation table according to the defect type identified in the defect type probability distribution; acquiring detailed operation steps, safety precautions and common problem solutions of the repair method from a repair knowledge base; local parameters in the repairing process parameters are adjusted by combining the position characteristics of the specific welding spots and the layout of surrounding elements; a personalized repair operation guidance is generated that includes a sequence of repair steps, parameter settings, tool selections, and risk cues.

Description

Welding quality detection method for flexible circuit board of audio equipment Technical Field The invention belongs to the technical field of electronic manufacturing quality detection, and particularly relates to a welding quality detection method for a flexible circuit board of audio equipment. Background The welding quality detection of the flexible circuit board of the current audio equipment mainly depends on automatic optical detection or manual visual detection. Based on a two-dimensional image under visible light, the welding quality is judged by comparing preset static characteristics such as the outline, the position and the glossiness of the welding spot. Part of the high-end detection can introduce X-ray imaging to observe structural defects such as holes, cracks and the like in the welding spot. However, these techniques perform static analysis on the final state after welding, and the data source and the information dimension are single. The prior art solutions have drawbacks. Static optical or X-ray detection cannot capture dynamic physical change information during the welding process. This results in a detection that remains at the appearance level, a low recognition rate for hidden defects closely related to the thermal process, such as cold welding, and the like, and complete inability to determine whether the defect is generated during the heating, melting, or cooling stage. When facing tiny, dense and material-specific welding spots on a flexible circuit board, the limited information of a single image mode is difficult to support accurate classification and root tracing of complex defects. A method of detecting the multi-dimensional static characteristics of a weld joint by fusing dynamic information of the welding process is needed. The technical problem of evolution from single static detection to multi-mode dynamic analysis is required to be solved, so that deeper judgment of welding quality and fault diagnosis with better guidance on production links are realized. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art; therefore, the invention provides a welding quality detection method for a flexible circuit board of audio equipment, which comprises the following steps: acquiring multispectral scanning image data and infrared thermal imaging time sequence data of the surface of a flexible circuit board of the audio equipment to be detected, and forming an original welding area multimodal data set; Performing image enhancement and data registration processing on the original welding area multi-mode data set to generate a standardized welding spot characteristic image sequence and a corresponding temperature distribution time sequence curve; Extracting geometric feature vectors and texture distribution feature vectors of welding spots from the standardized welding spot feature image sequence, and extracting heat conduction feature vectors and cooling feature vectors of the welding spots from the temperature distribution time sequence curve; Performing feature fusion on the geometric feature vector, the texture distribution feature vector, the heat conduction feature vector and the cooling feature vector to construct a multidimensional welding spot quality feature descriptor; inputting the multidimensional welding spot quality characteristic descriptors into a preset deep learning classification network, and outputting quality grade labels and defect type probability distribution of each welding spot by the deep learning classification network; Positioning the specific position of the welding defect and analyzing the time phase of defect generation based on the probability distribution of the defect type and the temperature distribution time sequence curve; And generating a welding quality assessment report containing repair priority and repair suggestions according to the quality grade label and defect position information of each welding spot. Further, the image enhancement and data registration processing are performed on the original multi-mode data set of the welding area, and a standardized welding spot characteristic image sequence and a corresponding temperature distribution time sequence curve are generated, which comprises the following steps: denoising the multispectral scanning image data, eliminating image noise by adopting a self-adaptive filtering algorithm, and enhancing the contrast ratio of welding spots and the background; Performing temperature calibration on the infrared thermal imaging time sequence data, converting an original thermal imaging gray value into an accurate temperature value, and establishing a time-temperature corresponding relation; recognizing the outline position of a welding spot in the multispectral image, and establishing a local coordinate system by taking the center of the welding spot as a reference; Mapping the infrared thermal imaging data into the local coordinate s