Detection system for packaging high-power photovoltaic TMBS chip

Abstract

The invention relates to the technical field of photovoltaic chip packaging, in particular to a detection system for packaging a high-power photovoltaic TMBS chip, which adopts a long-short-term memory network to analyze the duration, power change and heat dissipation efficiency in a welding spot group, combines parameters with the correlation degree of defect occurrence, effectively predicts the occurrence of defects and identifies potential high-incidence paths, and improves the detection accuracy of the high-power photovoltaic TMBS chip packaging. The method has advantages in time sequence data processing, processes complex nonlinear relations, ensures accurate identification of defect trends, analyzes contour and deformation of welding spots and pixel features of contact areas through an image identification technology, calculates deformation deviation, contour closure and pixel loss rate, screens out defect areas, improves defect screening accuracy, and improves detection accuracy. According to the method, defects of different types are effectively distinguished, a generative adversarial network is used for tracking a crack point area, the offset and diffusion trend of a crack point path are calculated, occurrence and propagation of crack points are effectively positioned, and the packaging quality and long-term stability of a photovoltaic chip are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic chip packaging, and in particular to a detection system for high-power photovoltaic TMBS chip packaging. Background Art

[0002] The technical field of photovoltaic chip packaging aims to improve the performance, reliability, and long-term stability of photovoltaic chips through innovative packaging methods, protect photovoltaic chips from external environmental factors, and ensure that the chips can operate efficiently under various environmental conditions.

[0003] The detection system for high-power photovoltaic TMBS chip packaging aims to ensure the quality and reliability of high-power photovoltaic TMBS chips during the packaging process. By accurately detecting possible defects during the packaging process, it ensures that each chip packaging meets the design and performance standards, and avoids affecting the overall performance and long-term stability of photovoltaic devices due to packaging problems.

[0004] In the existing technology during the packaging process of high-power photovoltaic chips, there is a lack of the ability to accurately identify thermal field changes and defects. It cannot fully consider the dynamic changes in the thermal field during the welding process and its impact on the quality of solder joints. Moreover, thermal imbalance cannot be identified in a timely and accurate manner, resulting in excessive thermal stress or poor heat dissipation in the solder joints, leading to cracks and detachment of the solder joints. In addition, traditional detection methods have limitations in detecting dynamic thermal changes and minute defects during long-term operation. There is a lack of in-depth analysis of solder joint timing data, and it is unable to accurately identify potential defect occurrence paths and their trends, resulting in many subtle defects not being discovered and corrected in a timely manner, affecting the long-term stability of photovoltaic chips and the overall performance of photovoltaic devices. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose a detection system for high-power photovoltaic TMBS chip packaging.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A detection system for high-power photovoltaic TMBS chip packaging includes: Thermal field variation monitoring module: Obtain local temperature change data through sensors, extract solder joint heat value change data, compare the temperature rise difference and heat conduction rate, analyze the transfer path through temperature difference, select solder joints with a temperature difference greater than the set value, calculate the thermal field non-uniformity, and generate a thermal field non-uniform region distribution map; Defect trend judgment module: Based on the thermal field non-uniform region distribution map, use a long short-term memory network to calculate the duration, power change, and heat dissipation efficiency within the solder joint group, analyze the correlation between parameters and defect occurrence, identify abnormal patterns, and generate a defect high-occurrence path node sequence set; Image recognition and screening module: Based on the set of defect-prone path node sequences, extract image data, analyze the pixel features of solder joint contours, deformations, and contact areas, calculate deformation deviations, contour closure degrees, and pixel loss rates, screen defect areas, and generate a multi-level screened defect atlas set; Break point area tracking module: Based on the multi-level screened defect atlas set, use a generative adversarial network to track the offset and diffusion trend of the break point area, calculate the closure of the break point path, select the area with a closed path, locate the break point position, and generate a break point tracking coordinate index set; Process parameter adjustment module: Based on the break point tracking coordinate index set, query the process parameter configuration of the solder joints, compare the packaging process parameters, select the parameters strongly associated with the break points, and generate a process change pointing parameter combination.

[0007] As a further aspect of the present invention, the thermal field change monitoring module includes: Thermal field data acquisition sub-module: Based on the local temperature change data obtained by the sensor, measure each solder joint in the chip packaging area one by one, record the temperature change curve of each solder joint, calculate the temperature rise rate of each solder joint, match the temperature change data of the solder joints with different areas of the packaging structure, and obtain the solder joint heat value change data; Temperature difference analysis sub-module: Based on the solder joint heat value change data, calculate the temperature rise difference between each solder joint and its adjacent solder joints, mark the solder joints with a temperature difference exceeding the set threshold as key solder joints, calculate and compare the temperature conduction rates between different solder joints, identify the heat diffusion path, determine the solder joint area with a temperature difference greater than the set value, and generate temperature difference conduction path data; Thermal field non-uniform area recognition sub-module: Based on the temperature difference conduction path data, conduct a centralized analysis of the solder joints with a temperature difference greater than the set value, combine the actual situation of the temperature change rate and the conduction path, identify the areas with uneven heat distribution, calculate the thermal field non-uniformity of each area, and generate a thermal field non-uniform area distribution map.

[0008] As a further aspect of the present invention, the defect trend judgment module includes: Parameter correlation analysis sub-module: Based on the thermal field non-uniform area distribution map, collect data on working hours, power changes, and heat dissipation efficiency within the solder joint group, use a long short-term memory network for modeling, capture the change trends of each parameter in the time dimension, analyze the relationship between the historical changes of each parameter and the occurrence of defects, calculate the correlation degree between each parameter and the probability of defect occurrence, sort by the correlation degree, screen out the key parameters, and generate parameter defect correlation data; Abnormal mode recognition sub-module: Based on the parameter defect correlation data, identify and extract the abnormal modes of defect occurrence, and by analyzing the relationship between each parameter and the defect, identify the modes that may cause defects, and generate the abnormal mode recognition results; Defect high - incidence path recognition sub - module: Based on the abnormal pattern recognition results, select path nodes with a probability of defect occurrence, analyze the possible impacts on the nodes during the welding process, calculate the frequency and intensity of defect occurrence, and generate a set of defect high - incidence path node sequences.

[0009] As a further solution of the present invention, the long - short - term memory network follows the formula:

[0010] Where: represents the correlation coefficient between process parameters and defect occurrence, represents the process parameters at the th time point, represents the welding pressure data at the th time point, represents the defect occurrence data at the th time point, represents the temperature data at the th time point, represents the mean value of process parameters, represents the mean value of defect data, , , and represent the weight coefficients of each parameter, represents the total number of data points.

[0011] As a further solution of the present invention, the abnormal pattern identifies the features and behavior patterns related to defects by analyzing the relationship between parameters and defect occurrence, the abnormal fluctuations or trends that appear in the time dimension and space dimension, and the parameters are the working duration, power change, heat dissipation efficiency, calorific value change, temperature rise difference, heat conduction rate, temperature difference, solder joint contour deformation, pixel features of the contact area, deformation deviation, contour closure degree, and pixel loss rate within the solder joint group.

[0012] As a further solution of the present invention, the image recognition and screening module includes: Defect image extraction sub - module: Based on the set of defect high - incidence path node sequences, by locating the image area of the solder joint, extract the image data around the solder joint, crop the image area, eliminate the irrelevant background, enhance the contrast and clarity of the area around the solder joint in the image, and integrate the image data of the solder joint area to generate an image data set; Feature calculation sub - module: Based on the image data set, by detecting the edge of the solder joint area, extract the closure degree of the solder joint contour, analyze the pixel point distribution in the deformation area, calculate the deformation deviation, compare the pixel loss rate between the solder joint and the contact area, record the values of each feature, and generate a feature calculation result; Defect area screening sub-module: Based on the feature calculation results, compare the closeness, deformation deviation, and pixel loss rate of the solder joint contour with the set thresholds, screen out the areas that meet the defect characteristics, mark the defect areas, and classify them according to the characteristics of the defect areas to generate a multi-level screened defect atlas set.

[0013] As a further solution of the present invention, the crack point area tracking module includes: Crack point offset calculation sub-module: Based on the multi-level screened defect atlas set, analyze the position changes of each defect area frame by frame, calculate the offsets of the defect areas at different time points, record the offset directions and amplitudes, obtain the displacement data of the crack point areas, and generate crack point offset data; Crack point path analysis sub-module: Based on the crack point offset data, analyze the diffusion trend of the crack points in the packaging area, compare the generated virtual defect images with historical data to determine whether the crack point expansion path is closed, combine the defect images under different process conditions generated by the generative adversarial network, evaluate the path closure situation, record the expansion trend of the crack point path, select the areas where the crack point path is closed, and generate the crack point path analysis result; Crack point positioning sub-module: Based on the crack point path analysis result, select the crack point areas within the path closed areas, perform precise positioning and calibrate the crack point positions to obtain the spatial coordinates of the crack points, and generate a crack point tracking coordinate index set.

[0014] As a further solution of the present invention, the generative adversarial network is in accordance with the formula:

[0015] Where: Represents the judgment value of the discriminator for the real data ; Represents the virtual defect image generated by the generator, Represents the distribution of the real defect data, Represents the distribution of the noise data input to the generator, Represents the distribution of the generator weights, Represents the weight parameters in the generator, Represents the weight coefficient.

[0016] As a further solution of the present invention, the process parameter adjustment module includes: Parameter query sub-module: Based on the crack point tracking coordinate index set, by locating the positions of the crack point areas, query the solder joint process data corresponding to the crack points, retrieve the corresponding process configurations one by one for each solder joint, obtain the specific process parameters of each solder joint, and generate solder joint process parameter data; Relevance comparison sub-module: Based on the solder joint process parameter data, compare the process parameters of different solder joints, analyze the relationship between each welding parameter and the occurrence of cracking points, focus on the effects of temperature, pressure, welding duration, and encapsulation speed on the occurrence of cracking points, calculate the strength of the association between each parameter and the cracking points, screen out the process parameters related to the cracking points, and generate process parameter correlation data; Process adjustment sub-module: Based on the process parameter correlation data, optimize and adjust the process parameters related to the occurrence of cracking points, select the process parameters that affect the occurrence of cracking points and adjust the settings to generate a process change target parameter combination.

[0017] As a further solution of the present invention, to screen out the process parameters related to the cracking points, by analyzing the process data of each solder joint, identify the process factors that have a significant association with the occurrence of cracking points, evaluate the relationship strength between each process parameter and the occurrence of cracking points through a quantification method, screen out the key parameters that have a significant impact on the occurrence of cracking points, and the process parameters are temperature, pressure, welding duration, encapsulation speed, welding power, heat dissipation efficiency, heat conduction rate, solder joint deformation, contour closure degree, pixel loss rate, and pixel characteristics of the contact area.

[0018] Compared with the prior art, the advantages and positive effects of the present invention are as follows: 1. In the present invention, a long short-term memory network is used to analyze the duration, power change, and heat dissipation efficiency within the solder joint group, combine the correlation between the parameters and the occurrence of defects, effectively predict the occurrence of defects and identify potential high-incidence paths, have advantages in the processing of time-series data, handle complex non-linear relationships, and ensure the accurate identification of defect trends; 2. In the present invention, through image recognition technology, analyze the contour, deformation, and pixel characteristics of the contact area of the solder joint, calculate the deformation deviation, contour closure degree, and pixel loss rate, screen out the defect area, improve the accuracy of defect screening, and effectively distinguish different types of defects; 3. In the present invention, use a generative adversarial network to track the cracking point area, calculate the offset and diffusion trend of the cracking point path, effectively locate the occurrence and propagation of the cracking point, combine the cracking point tracking with the comparison of process parameters, and adjust the encapsulation process parameters based on the defect position and diffusion trend to optimize the quality of each solder joint in the overall encapsulation process, avoid hidden defects in the welding process, and improve the encapsulation quality and long-term stability of the photovoltaic chip. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is the system flow chart of the present invention; Figure 2 is the flow chart of the present invention; Figure 3 is the schematic diagram of the system framework of the present invention.

[0020] DETAILED IMPLEMENTATION MODE In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0021] Please refer to Figure 1 and Figure 2 , the present invention provides a technical solution: A detection system for high-power photovoltaic TMBS chip packaging includes: Thermal field variation monitoring module: Obtain local temperature change data through sensors, extract solder joint heat value change data, compare the temperature rise difference and heat conduction rate, analyze the transfer path through temperature difference analysis, select solder joints with a temperature difference greater than the set value, calculate the thermal field non-uniformity, and generate a thermal field non-uniform region distribution map; Defect trend judgment module: Based on the thermal field non-uniform region distribution map, use long short-term memory networks to calculate the duration, power change and heat dissipation efficiency within the solder joint group, analyze the correlation between parameters and defect occurrence, identify abnormal patterns, and generate a defect high-occurrence path node sequence set; Image recognition screening module: Based on the defect high-occurrence path node sequence set, extract image data, analyze the pixel features of the solder joint contour, deformation and contact area, calculate the deformation deviation, contour closure degree and pixel loss rate, screen the defect area, and generate a multi-level screening defect map set; Break point area tracking module: Based on the multi-level screening defect map set, use a generative adversarial network to track the offset and diffusion trend of the break point area, calculate the closure of the break point path, select the area with a closed path, locate the break point position, and generate a break point tracking coordinate index set; Process parameter adjustment module: Based on the break point tracking coordinate index set, query the process parameter configuration of the solder joint, compare the packaging process parameters, select the parameters strongly associated with the break point, and generate a process change pointing parameter combination.

[0022] Please refer to Figure 3 , the thermal field variation monitoring module includes: Thermal field data acquisition sub-module: Based on the local temperature change data obtained by the sensor, measure each solder joint in the chip packaging area one by one, record the temperature change curve of each solder joint, calculate the temperature rise rate of each solder joint, and match the temperature change data of the solder joint with different areas of the packaging structure to obtain the solder joint heat value change data; Temperature difference analysis sub-module: Based on the solder joint heat value change data, calculate the temperature rise difference between each solder joint and its adjacent solder joints, mark the solder joints with a temperature difference exceeding the set threshold as key solder joints, calculate and compare the temperature conduction rates between different solder joints, identify the heat diffusion path, determine the solder joint area with a temperature difference greater than the set value, and generate temperature difference conduction path data; Thermal field imbalance area identification sub-module: Based on the temperature difference conduction path data, conduct a centralized analysis of solder joints with a temperature difference greater than the set value. Combine the actual situation of the temperature change rate and the conduction path to identify the areas with uneven heat distribution, calculate the thermal field non-uniformity of each area, and generate a thermal field imbalance area distribution map; Thermal field data acquisition sub-module: Based on the local temperature change data obtained by the sensor, use a temperature sensor acquisition module, specifically the DHT22 temperature and humidity sensor, to obtain the real-time temperature data of each solder joint through its digital signal interface. The temperature sampling frequency of each solder joint is set to 1 time per second, and the temperature range is from -40°C to 125°C. Measure each solder joint in the chip packaging area one by one, record the temperature change curve of each solder joint, and use the temperature rise rate calculation method to calculate the temperature rise rate of each solder joint. Assume that when the temperature rise rate is greater than 0.5°C / s, it is the normal threshold. Match the temperature change data of the solder joint with different areas of the packaging structure to obtain the solder joint heat value change data, and compare the temperature rise rate data of each solder joint with the temperature change of each area of the package to generate the solder joint heat value change data; Temperature difference analysis sub-module: Based on the solder joint heat value change data, use the difference analysis algorithm to calculate the temperature rise difference between each solder joint and its adjacent solder joints. Mark the solder joints with a temperature difference exceeding the set threshold as key solder joints. Use the temperature conduction rate analysis method to calculate the temperature conduction rate between different solder joints, identify the heat diffusion path, calculate and compare the temperature conduction rate between different solder joints, identify the heat diffusion path, determine the solder joint area with a temperature difference greater than the set value, and generate the temperature difference conduction path data; Thermal field imbalance area identification sub-module: Based on the temperature difference conduction path data, use the area analysis algorithm to conduct a centralized analysis of solder joints with a temperature difference greater than the set value. Combine the actual situation of the temperature change rate and the conduction path, and use the thermal field uniformity evaluation method to calculate the thermal field non-uniformity of each area. If the standard deviation exceeds the set 1.2°C, it is marked as an imbalance area, and a thermal field imbalance area distribution map is generated.

[0023] Please refer to Figure 3 , the defect trend judgment module includes: Parameter correlation analysis sub-module: Based on the thermal field imbalance area distribution map, collect the working duration, power change, and heat dissipation efficiency data within the solder joint group, use a long short-term memory network for modeling, capture the change trends of each parameter in the time dimension, and analyze the relationship between the historical changes of each parameter and the occurrence of defects. Calculate the correlation degree between each parameter and the defect occurrence probability, sort by the correlation degree, screen out the key parameters, and generate the parameter defect correlation degree data; Abnormal pattern recognition sub-module: Based on the parameter-defect correlation data, identify and extract the abnormal patterns of defect occurrence. By analyzing the relationships between various parameters and defects, identify the patterns that may lead to defects, and generate the abnormal pattern recognition results; High-defect path recognition sub-module: Based on the abnormal pattern recognition results, select the path nodes with the probability of defect occurrence, analyze the possible impacts on the nodes during the welding process, calculate the frequency and intensity of defect occurrence, and generate the high-defect path node sequence set; Parameter correlation analysis sub-module: Based on the distribution map of the uneven thermal field region, collect the data of working duration, power change, and heat dissipation efficiency within the solder joint group. Use the long short-term memory network. The input parameters include the time series data such as the historical working duration, power change, and heat dissipation efficiency of each solder joint within the solder joint group. Use a hidden layer with 300 neurons, the optimizer is Adam, the learning rate is 0.001, and the number of iterations is set to 100 times. Through the LSTM network training, capture the change trends of each parameter in the time dimension, and analyze the relationship between the historical changes of each parameter and the occurrence of defects. Calculate the correlation degree between each parameter and the probability of defect occurrence. Use the Pearson correlation coefficient to calculate the correlation between each parameter and the occurrence of defects, sort them from high to low according to the correlation degree, screen out the key parameters, and generate the parameter-defect correlation data; Abnormal pattern recognition sub-module: Based on the parameter-defect correlation data, use the anomaly detection algorithm, specifically IsolationForest, with the parameters set as the number of trees being 100 and the depth of the tree being 256. By analyzing the relationships between various parameters and defects, identify the patterns that may lead to defects. Use the tree-structured isolation factor method for outlier recognition, with the outlier determination threshold being 0.75. The abnormal pattern recognition results are output through the model to identify the abnormal patterns and generate the abnormal pattern recognition results; High-defect path recognition sub-module: Based on the abnormal pattern recognition results, select the path nodes with the probability of defect occurrence. Use the K-means clustering algorithm, with the number of clusters set to 5, the initial cluster centers randomly initialized, the number of iterations set to 50 times, and the distance metric using the Euclidean distance. By analyzing the possible impacts on the nodes during the welding process, calculate the frequency and intensity of defect occurrence, and generate the high-defect path node sequence set.

[0024] Long short-term memory network, according to the formula:

[0025] Where: Represents the correlation coefficient between the process parameter and the occurrence of defects, Represents the th time point of the process parameter, Represents the Welding pressure data at each time point, indicating the defect occurrence data at the th time point, indicating the temperature data at the th time point, indicating the mean value of the process parameters, , , and indicating the weight coefficient of each parameter, indicating the total number of data points; Execution process: First, collect the process parameters at different time points, input and process them to help capture the relationship between the change trend of each process parameter in the time series and the defect occurrence. Then, calculate the deviation between the process parameter and defect data at each time point and the corresponding mean value, and assign different importance to each process parameter through the weight coefficient . By calculating the covariance between the weighted process parameter and defect occurrence data at each time point and normalizing it, the final correlation coefficient is obtained. At the same time, evaluate the correlation degree between each process parameter and defect occurrence, provide a basis for subsequent defect prediction and optimization decision-making, and ensure timely adjustment of process parameters during the photovoltaic TMBS chip packaging process to reduce the defect occurrence rate.

[0026] Abnormal patterns. By analyzing the relationship between parameters and defect occurrence, identify the features and behavior patterns related to defects, the abnormal fluctuations or trends that appear in the time dimension and space dimension. The parameters are the working duration, power change, heat dissipation efficiency, heat value change, temperature rise difference, heat conduction rate, temperature difference, solder joint contour deformation, pixel features of the contact area, deformation deviation, contour closure degree, and pixel loss rate within the solder joint group.

[0027] Please refer to Figure 3 . The image recognition and screening module includes: Defect image extraction sub-module: Based on the defect high-occurrence path node sequence set, by locating the image area of the solder joint, extract the image data around the solder joint, and crop the image area to eliminate the irrelevant background, enhance the contrast and clarity of the area around the solder joint in the image, and integrate the image data of the solder joint area to generate an image data set; Feature calculation sub-module: Based on the image data set, by detecting the edge of the solder joint area, extract the closure degree of the solder joint contour, analyze the pixel point distribution in the deformation area, calculate the deformation deviation, compare the pixel loss rate between the solder joint and the contact area, record the numerical values of each feature, and generate the feature calculation result; Defect area screening sub-module: Based on the feature calculation results, compare the closure degree, deformation deviation, and pixel loss rate of the solder joint contour with the set thresholds, screen out the areas that meet the defect characteristics, mark the defect areas, and classify them according to the characteristics of the defect areas to generate a multi-level screening defect atlas set; Defect image extraction sub-module: Based on the defect high-occurrence path node sequence set, use the image localization algorithm to locate the image area of the solder joint, use the findContours function in the OpenCV library to identify the contour of the solder joint area, extract the image data around the solder joint, use the image cropping method, set the cropping frame to eliminate the irrelevant background, use the histogram equalization algorithm to enhance the contrast and clarity of the area around the solder joint in the image, use the equalizeHist function of OpenCV to process the image, enhance the details, make the solder joint area clearer, and integrate the image data of the solder joint area to generate an image data set; Feature calculation sub-module: Based on the image data set, use the Canny edge detection algorithm and the Canny function of OpenCV to perform edge detection on the solder joint area, detect the edge of the solder joint area, extract the closure degree of the solder joint contour, use the contour closure degree calculation method, analyze the pixel point distribution of the deformation area by calculating the distance of the solder joint contour points, calculate the deformation deviation, use the mean square error-based algorithm to compare the pixel loss rate between the solder joint and the contact area, use the image difference method, perform the matching of the solder joint and the contact area through the matchTemplate function of OpenCV, record the values of each feature, and generate the feature calculation results; Defect area screening sub-module: Based on the feature calculation results, compare the closure degree, deformation deviation, and pixel loss rate of the solder joint contour with the set thresholds, use the threshold determination algorithm, set the closure degree threshold to 0.8, the deformation deviation threshold to 0.5, and the pixel loss rate threshold to 0.2, screen out the areas that meet the defect characteristics, mark the defect areas, and use the K-means clustering algorithm with the number of clusters set to 3 and initialized randomly to classify the defect areas and generate a multi-level screening defect atlas set.

[0028] Please refer to Figure 3 , the crack point area tracking module includes: Crack point offset calculation sub-module: Based on the multi-level screening defect atlas set, analyze the position changes of each defect area frame by frame, calculate the offset of the defect area at different time points, record the offset direction and amplitude, obtain the displacement data of the crack point area, and generate the crack point offset data; Breakpoint Path Analysis Sub-module: Based on the breakpoint offset data, analyze the diffusion trend of breakpoints within the encapsulation area. Compare the generated virtual defect images with historical data to determine whether the breakpoint expansion path is closed. Combine the defect images under different process conditions generated by the generative adversarial network to evaluate the path closure situation, record the expansion trend of the breakpoint path, select the area where the breakpoint path is closed, and generate the breakpoint path analysis result; Breakpoint Location Sub-module: Based on the breakpoint path analysis result, select the breakpoint area within the path-closed area, perform precise positioning and calibration of the breakpoint position to obtain the spatial coordinates of the breakpoint, and generate the breakpoint tracking coordinate index set; Breakpoint Offset Calculation Sub-module: Based on the multi-level screened defect atlas set, analyze the position change of each defect area frame by frame. Use the findContours function in OpenCV to extract the contour of the defect area in each frame image and record its position. By calculating the offset of the centroid of the defect area between two adjacent frames, obtain the displacement data of the breakpoint area. Use the Euclidean distance formula to calculate the offset, record the offset direction and amplitude. Assume that the offset direction is obtained by calculating the change angle of the centroid position of each frame, and generate the breakpoint offset data; Breakpoint Path Analysis Sub-module: Based on the breakpoint offset data, analyze the diffusion trend of breakpoints within the encapsulation area. Adopt the virtual defect image generation method, and through the generative adversarial network, generate defect images under different process conditions. Combine the comparison with historical data, and use the peak signal-to-noise ratio algorithm to evaluate whether the breakpoint expansion path is closed. Set the closing condition as the PSNR value being greater than 30. Record the expansion trend of the breakpoint path, perform clustering on the breakpoint path through the K-means clustering algorithm, select the area where the breakpoint path is closed, and generate the breakpoint path analysis result; Breakpoint Location Sub-module: Based on the breakpoint path analysis result, select the breakpoint area within the path-closed area, adopt the image matching algorithm, use the matchTemplate function in OpenCV to precisely locate the breakpoint area, calculate the similarity between the template image and the current welding image, select the area where the similarity is greater than the set threshold of 0.9, obtain the spatial coordinates of the breakpoint, and generate the breakpoint tracking coordinate index set.

[0029] Generative Adversarial Network, according to the formula:

[0030] Where: Represents the judgment value of the discriminator for the real data ; Represents the virtual defect image generated by the generator, Represents the distribution of real defect data, Represents the distribution of the noise data input to the generator, Represents the distribution of the generator weights, Represents the weight parameters in the generator, Represents the weight coefficient; Execution process: First, receive process parameters as input, generate possible defect images. The images represent the defects that may occur under different packaging conditions. The discriminator then evaluates the generated images to distinguish the differences between real images and generated images. Through training, the discriminator will learn the characteristics of real defect images and compare them with the generated images. Through the adversarial process, the generator is gradually optimized to make the generated images closer to real defect images. The weights of the generator Determine the characteristics of the defect images. The generator optimizes the output by adjusting the weights. The weight coefficient Is used to balance the influence of the generator and the discriminator, gradually improving the quality of the generated images, and finally generating virtual images related to packaging defects, providing accurate data support for the crack point path analysis and defect warning in the packaging of high-power photovoltaic TMBS chips.

[0031] Please refer to Figure 3 , the process parameter adjustment module includes: Parameter query sub-module: Based on the crack point tracking coordinate index set, by locating the position of the crack point area, query the solder joint process data corresponding to the crack point, retrieve the corresponding process configuration for each solder joint one by one, obtain the specific process parameters of each solder joint, and generate solder joint process parameter data; Relevance comparison sub-module: Based on the solder joint process parameter data, compare the process parameters of different solder joints, analyze the relationship between each welding parameter and the occurrence of crack points, focus on the influence of temperature, pressure, welding duration, and packaging speed on the occurrence of crack points, calculate the strength of the association between each parameter and the crack point, screen out the process parameters related to the crack point, and generate process parameter correlation degree data; Process adjustment sub-module: Based on the process parameter correlation degree data, optimize and adjust the process parameters related to the occurrence of crack points, select the process parameters that affect the occurrence of crack points and adjust the settings, and generate a process change pointing parameter combination; Parameter query sub-module: Based on the crack point tracking coordinate index set, through the image localization algorithm, locate the position of the crack point area, use the findContours function of OpenCV to extract the contour of the crack point area and calibrate the position, query the solder joint process data corresponding to the crack point, retrieve the corresponding process configuration for each solder joint one by one, use the SQL query statement, and obtain the specific process parameters of each solder joint by connecting the solder joint ID in the process database and the corresponding process configuration table, and generate solder joint process parameter data; Relevance comparison sub-module: Based on the solder joint process parameter data, compare the process parameters of different solder joints. Adopt the Pearson correlation coefficient algorithm. By calculating the correlation between each pair of solder joint parameters, analyze the relationship between each welding parameter and the occurrence of cracking points, with a focus on the influence of temperature, pressure, welding duration, and encapsulation speed on the occurrence of cracking points. Calculate the strength of the association between each parameter and the cracking point. Use a linear regression model to calculate the correlation and screen out the process parameters related to the cracking point, generating process parameter association degree data; Process adjustment sub-module: Based on the process parameter association degree data, adopt the particle swarm optimization algorithm. By setting the number of particles to 50, the number of iterations to 1000, the maximum speed to 2, and the minimum speed to -2, update the particle positions, optimize the process parameters related to the occurrence of cracking points, select the process parameters that affect the occurrence of cracking points and adjust the settings. Adopt an optimization method based on constraint conditions to set the adjustment range of the process parameters, generating a process change direction parameter combination.

[0032] Screen out the process parameters related to the cracking point. By analyzing the process data of each solder joint, identify the process factors that have a significant association with the occurrence of cracking points. Evaluate the strength of the relationship between each process parameter and the occurrence of cracking points through a quantification method, and screen out the key parameters that have a significant impact on the occurrence of cracking points. The process parameters are temperature, pressure, welding duration, encapsulation speed, welding power, heat dissipation efficiency, heat conduction rate, solder joint deformation, contour closure degree, pixel loss rate, and pixel characteristics of the contact area.

[0033] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A detection system for high-power photovoltaic TMBS chip packaging, characterized in that The system includes: Thermal field variation monitoring module: Obtain local temperature change data through sensors, extract solder joint heat value change data, compare the temperature rise difference and heat conduction rate, analyze the transfer path through temperature difference analysis, select solder joints with a temperature difference greater than the set value, calculate the thermal field non-uniformity, and generate a thermal field imbalance area distribution map; Defect trend judgment module: Based on the thermal field imbalance area distribution map, use a long short-term memory network to calculate the duration, power change, and heat dissipation efficiency within the solder joint group, analyze the correlation between parameters and defect occurrence, identify abnormal patterns, and generate a set of defect high-occurrence path node sequences; Image recognition and screening module: Based on the set of defect high-occurrence path node sequences, extract image data, analyze the pixel features of the solder joint contour, deformation, and contact area, calculate the deformation deviation, contour closure degree, and pixel loss rate, screen the defect area, and generate a multi-level screening defect atlas set; Breakpoint area tracking module: Based on the multi-level screening defect atlas set, use a generative adversarial network to track the offset and diffusion trend of the breakpoint area, calculate the closure of the breakpoint path, select the area with a closed path, locate the breakpoint position, and generate a set of breakpoint tracking coordinate indexes; Process parameter adjustment module: Based on the set of breakpoint tracking coordinate indexes, query the process parameter configuration of the solder joint, compare the packaging process parameters, select the parameters strongly associated with the breakpoint, and generate a process change pointing parameter combination.

2. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The thermal field variation monitoring module includes: Thermal field data acquisition sub-module: Based on the local temperature change data obtained by the sensor, measure each solder joint in the chip packaging area one by one, record the temperature change curve of each solder joint, calculate the temperature rise rate of each solder joint, and match the temperature change data of the solder joint with different areas of the packaging structure to obtain the solder joint heat value change data; Temperature difference analysis sub-module: Based on the solder joint heat value change data, calculate the temperature rise difference between each solder joint and its adjacent solder joints, mark the solder joints with a temperature difference exceeding the set threshold as key solder joints, calculate and compare the temperature conduction rates between different solder joints, identify the heat diffusion path, determine the solder joint area with a temperature difference greater than the set value, and generate temperature difference conduction path data; Thermal field imbalance area identification sub-module: Based on the temperature difference conduction path data, conduct a centralized analysis of the solder joints with a temperature difference greater than the set value, combine the actual situation of the temperature change rate and the conduction path, identify the area with uneven heat distribution, calculate the thermal field non-uniformity of each area, and generate a thermal field imbalance area distribution map.

3. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The defect trend judgment module includes: Parameter correlation analysis sub-module: Based on the thermal field imbalance area distribution map, collect data on the working duration, power change, and heat dissipation efficiency within the solder joint group, use a long short-term memory network for modeling, capture the change trend of each parameter in the time dimension, analyze the relationship between the historical changes of each parameter and the occurrence of defects, calculate the correlation between each parameter and the defect occurrence probability, sort by the correlation, screen out the key parameters, and generate parameter defect correlation data; Abnormal pattern recognition sub-module: Based on the parameter defect correlation data, identify and extract the abnormal patterns of defect occurrence. By analyzing the relationships between various parameters and defects, identify the patterns that may cause defects, and generate the abnormal pattern recognition results; High-defect path recognition sub-module: Based on the abnormal pattern recognition results, select the path nodes with the probability of defect occurrence, analyze the possible impacts on the nodes during the welding process, calculate the frequency and intensity of defect occurrence, and generate the high-defect path node sequence set.

4. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The long short-term memory network, according to the formula: Wherein: represents the correlation coefficient between the process parameters and the occurrence of defects, represents the process parameter at the -th time point, represents the welding pressure data at the -th time point, represents the defect occurrence data at the -th time point, represents the temperature data at the -th time point, represents the mean value of the process parameters, represents the mean value of the defect data, , , and represent the weight coefficients of each parameter, represents the total number of data points.

5. The detection system for high-power photovoltaic TMBS chip packaging according to claim 3, wherein The abnormal pattern, by analyzing the relationship between parameters and defect occurrence, identifies the features and behavior patterns related to defects, the abnormal fluctuations or trends that appear in the time dimension and space dimension. The parameters are the working duration, power change, heat dissipation efficiency, heat value change, temperature rise difference, heat conduction rate, temperature difference, solder joint contour deformation, pixel features of the contact area, deformation deviation, contour closure degree, and pixel loss rate within the solder joint group.

6. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The image recognition and screening module includes: Defect image extraction sub-module: Based on the high-defect path node sequence set, by locating the image area of the solder joint, extract the image data around the solder joint, and crop the image area to eliminate the irrelevant background, enhance the contrast and clarity of the area around the solder joint in the image, and integrate the image data of the solder joint area to generate the image data set; Feature calculation sub-module: Based on the image data set, by detecting the edge of the solder joint area, extract the contour closure degree of the solder joint, analyze the pixel point distribution in the deformation area, calculate the deformation deviation, compare the pixel loss rate between the solder joint and the contact area, record the numerical values of each feature, and generate the feature calculation results; Defect area screening sub-module: Based on the feature calculation results, compare the contour closure degree, deformation deviation, and pixel loss rate of the solder joint with the set thresholds, screen out the areas that meet the defect characteristics, mark the defect areas, and classify them according to the characteristics of the defect areas to generate the multi-level screening defect atlas set.

7. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The crack point area tracking module includes: Crack point offset calculation sub-module: Based on the multi-level screening defect atlas set, analyze the position changes of each defect area frame by frame, calculate the offset of the defect area at different time points, record the offset direction and amplitude, obtain the displacement data of the crack point area, and generate the crack point offset data; Crack point path analysis sub-module: Based on the crack point offset data, analyze the diffusion trend of the crack point in the packaging area. By comparing the generated virtual defect images with historical data, judge whether the crack point expansion path is closed. Combine the defect images under different process conditions generated by the generative adversarial network to evaluate the path closure situation, record the expansion trend of the crack point path, select the area where the crack point path is closed, and generate the crack point path analysis results; Crack point positioning sub-module: Based on the crack point path analysis results, select the crack point area within the path closed area, perform precise positioning and calibrate the crack point position to obtain the spatial coordinates of the crack point, and generate the crack point tracking coordinate index set.

8. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, wherein The generative adversarial network, according to the formula: Wherein: represents the discriminator's judgment value for real data ; represents the virtual defect image generated by the generator represents the distribution of real defect data represents the distribution of the noise data input to the generator represents the distribution of the generator's weights represents the weight parameter in the generator represents the weight coefficient.

9. The detection system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that The process parameter adjustment module includes: Parameter query sub-module: Based on the break point tracking coordinate index set, by locating the position of the break point area, query the solder joint process data corresponding to the break point, retrieve the corresponding process configuration for each solder joint one by one, obtain the specific process parameters of each solder joint, and generate solder joint process parameter data; Relevance comparison sub-module: Based on the solder joint process parameter data, compare the process parameters of different solder joints, analyze the relationship between each welding parameter and the occurrence of break points, focus on the influence of temperature, pressure, welding duration, and encapsulation speed on the occurrence of break points, calculate the strength of the association between each parameter and the break point, screen out the process parameters related to the break point, and generate process parameter correlation degree data; Process adjustment sub-module: Based on the process parameter correlation degree data, optimize and adjust the process parameters related to the occurrence of break points, select the process parameters that affect the occurrence of break points and adjust the settings, and generate a process change pointing parameter combination.

10. The detection system for high-power photovoltaic TMBS chip packaging according to claim 9, wherein The process parameters related to the break point are screened out by analyzing the process data of each solder joint, identifying the process factors that have a significant association with the occurrence of break points, evaluating the relationship strength between each process parameter and the occurrence of break points through a quantification method, screening out the key parameters that have a significant impact on the occurrence of break points, and the process parameters are temperature, pressure, welding duration, encapsulation speed, welding power, heat dissipation efficiency, heat conduction rate, solder joint deformation, contour closure degree, pixel loss rate, and pixel characteristics of the contact area.

Images