A detection system for high-power photovoltaic TMBS chip packaging

Abstract

The present application relates to photovoltaic chip packaging technical field, specifically to a kind of detection system for large power photovoltaic TMBS chip packaging, in the present application, the time length in solder joint group, power change and heat dissipation efficiency are analyzed using long short-term memory network, the correlation degree of parameters and defect occurrence is combined, the occurrence of defect is effectively predicted and potential high incidence path is identified, it has advantage in the processing of time series data, process complex nonlinear relationship, ensure the accurate identification of defect trend, through image recognition technology, the outline, deformation and pixel features of contact area of solder joint are analyzed, deformation deviation, outline closure degree and pixel loss rate are calculated, and defect area is screened out, improve the accuracy of defect screening, effectively distinguish different types of defects, use generative adversarial network to track crack point area, calculate the offset and diffusion trend of crack point path, effectively locate the occurrence and propagation of crack point, improve the packaging quality and long-term stability of photovoltaic chip.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic chip packaging, especially to a detection system for high-power photovoltaic TMBS chip packaging. BACKGROUND

[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, accurately detect possible defects during the packaging process, ensure that each chip packaging meets design and performance standards, and avoid affecting the overall performance and long-term stability of photovoltaic equipment due to packaging problems.

[0004] In the existing technology, there is a lack of accurate identification of thermal field changes and defects during the packaging process of high-power photovoltaic chips, and the dynamic changes of the thermal field during the welding process and its impact on the quality of the welding points cannot be fully considered. In addition, thermal imbalance cannot be accurately identified in time, resulting in excessive thermal stress or poor heat dissipation of the welding points, causing welding point cracks and falling off. Moreover, traditional detection methods have limitations in detecting dynamic thermal changes and small defects during long-term operation, lack in-depth analysis of welding point time series data, and cannot accurately identify potential defect occurrence paths and trends, resulting in many subtle defects that cannot be discovered and corrected in time, affecting the long-term stability of photovoltaic chips and the overall performance of photovoltaic equipment. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to provide a detection system for high-power photovoltaic TMBS chip packaging.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: a detection system for high-power photovoltaic TMBS chip packaging includes:

[0007] The thermal field variation monitoring module obtains local temperature change data through a sensor, extracts welding point thermal value change data, compares the temperature rise difference value and the thermal conduction rate, analyzes the transmission path through the temperature difference, selects the welding points with a temperature difference greater than a set value, calculates the thermal field non-uniformity, and generates a thermal field imbalance area distribution map;

[0008] The defect trend analysis module calculates the time length, power change and heat dissipation efficiency within the welding point group based on the thermal field imbalance area distribution map using a long short-term memory network, analyzes the correlation between parameters and defect occurrence, identifies abnormal patterns, and generates a defect high-risk path node sequence set;

[0009] Image recognition screening module: based on the defect high incidence path node sequence set, image data is extracted, pixel features of weld point contour, deformation and contact area are analyzed, deformation deviation, contour closure degree and pixel loss rate are calculated, defect area is screened, and a multi-level screening defect atlas set is generated;

[0010] Cleavage point area tracking module: based on the multi-level screening defect atlas set, a generative adversarial network is used to track the offset and diffusion trend of the cleavage point area, the closure of the cleavage point path is calculated, the area with closed path is selected, the cleavage point position is located, and a cleavage point tracking coordinate index set is generated;

[0011] Process parameter adjustment module: based on the cleavage point tracking coordinate index set, the process parameter configuration of the weld point is queried, the packaging process parameters are compared, the parameters associated with the cleavage point are selected, and a process change pointing parameter combination is generated.

[0012] As a further scheme of the present application, the thermal field variation monitoring module comprises:

[0013] Thermal field data acquisition submodule: based on the local temperature change data obtained by the sensor, different weld points in the chip packaging area are measured one by one, the temperature change curve of each weld point is recorded, the temperature rise rate of each weld point is calculated, the temperature change data of the weld point is matched with different areas of the packaging structure, and the weld point thermal value change data is obtained;

[0014] Temperature difference analysis submodule: based on the weld point thermal value change data, the temperature rise difference value of each weld point and the adjacent weld point is calculated, the weld point with a temperature difference value exceeding a set threshold is marked as a key weld point, the temperature conduction rate between different weld points is calculated and compared, the heat diffusion path is identified, the weld point area with a temperature difference greater than a set value is determined, and temperature difference conduction path data is generated;

[0015] Thermal field imbalance area identification submodule: based on the temperature difference conduction path data, the weld points with a temperature difference greater than a set value are analyzed, the actual situation of the temperature change rate and the conduction path is combined, the area with uneven heat distribution is identified, the thermal field uniformity of each area is calculated, and a thermal field imbalance area distribution map is generated.

[0016] As a further scheme of the present application, the defect trend judgment module comprises:

[0017] Parameter correlation analysis submodule: based on the thermal field imbalance area distribution map, the working time, power change and heat dissipation efficiency data in the weld point group are collected, a long short-term memory network is used for modeling, the change trend of each parameter in the time dimension is captured, and the relationship between the historical change of each parameter and the defect occurrence is analyzed, the correlation degree between each parameter and the defect occurrence probability is calculated, the key parameters are selected according to the correlation degree, and parameter defect correlation degree data is generated;

[0018] Anomaly pattern recognition submodule: based on the parameter defect correlation degree data, identify and extract the abnormal pattern of defect occurrence, identify the mode that may cause the defect by analyzing the relationship between each parameter and the defect, and generate an abnormal pattern recognition result;

[0019] Defect high incidence path recognition submodule: based on the abnormal pattern recognition result, select the path node with defect occurrence probability, analyze the influence that the node may be subjected to in the welding process, calculate the frequency and intensity of defect occurrence, and generate a defect high incidence path node sequence set.

[0020] As a further scheme of the present application, the long short-term memory network is according to the formula:

[0021]

[0022] Wherein: represents the correlation coefficient between the process parameters and the defect occurrence, 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 parameter, represents the mean value of the defect data, , , and represent the weight coefficient of each parameter, represents the total number of data points.

[0023] As a further scheme of the present application, the abnormal pattern identifies the characteristics and behavior patterns related to the defect by analyzing the relationship between the parameters and the defect occurrence, and the abnormal fluctuations or trends that appear in the time dimension and the space dimension, according to the parameters of the working time, power change, heat dissipation efficiency, heat value change, temperature rise difference, thermal conductivity rate, temperature difference, weld point contour deformation, contact area pixel feature, deformation deviation, contour closure degree and pixel loss rate.

[0024] As a further scheme of the present application, the image recognition screening module comprises:

[0025] The defect image extraction submodule extracts image data around the soldering point based on the defect high-occurrence path node sequence set, crops the image area, eliminates irrelevant background, enhances the contrast and definition of the area around the soldering point in the image, integrates the soldering point area image data, and generates an image data set;

[0026] The feature calculation submodule extracts the closure of the soldering point contour, analyzes the pixel point distribution of the deformation area, calculates the deformation deviation, compares the pixel loss rate of the soldering point and the contact area, records the numerical value of each feature, and generates a feature calculation result based on the image data set by detecting the edge of the soldering point area.

[0027] The defect area screening submodule compares the closure of the soldering point contour, the deformation deviation, and the pixel loss rate with the set threshold value based on the feature calculation result, screens out the area meeting the defect characteristics, marks the defect area, classifies the defect area according to the characteristics, and generates a multi-level screening defect atlas set.

[0028] As a further scheme of the present application, the crack point region tracking module comprises:

[0029] The crack point offset calculation submodule calculates the offset of the defect region at different time points by analyzing the position change of each defect region frame by frame based on the multi-level screening defect atlas set, records the offset direction and amplitude, obtains the displacement data of the crack point region, and generates crack point offset data.

[0030] The crack point path analysis submodule analyzes the diffusion trend of the crack point in the packaging area based on the crack point offset data, compares the generated virtual defect image with the historical data to determine whether the crack point expansion path is closed, evaluates the path closure condition by combining the defect images generated under different process conditions by the generative adversarial network, records the expansion trend of the crack point path, selects the region where the crack point path is closed, and generates a crack point path analysis result.

[0031] The crack point positioning submodule selects the crack point region in the path closed region based on the crack point path analysis result, accurately positions and calibrates the crack point position, obtains the spatial coordinates of the crack point, and generates a crack point tracking coordinate index set.

[0032] As a further scheme of the present application, the generative adversarial network is according to the formula:

[0033]

[0034] Wherein: represents the judgment value of the discriminator for the real data , represents the virtual defect image generated by the generator, a distribution representing real defect data, a distribution representing noise data of the generator input, a distribution representing generator weights, representing weight parameters in the generator, representing weight coefficients.

[0035] As a further scheme of the present application, the process parameter adjustment module comprises:

[0036] The parameter query sub-module queries the welding point process data corresponding to the crack point by locating the position of the crack point area based on the crack point tracking coordinate index set, retrieves the corresponding process configuration for each welding point one by one, obtains the specific process parameters of each welding point, and generates welding point process parameter data.

[0037] The correlation comparison sub-module compares the process parameters of different welding points based on the welding point process parameter data, analyzes the relationship between each welding parameter and the occurrence of the crack point, focuses on the influence of temperature, pressure, welding time and packaging speed on the occurrence of the crack point, calculates the strength of the association between each parameter and the crack point, selects the process parameters related to the crack point, and generates process parameter correlation data.

[0038] The process adjustment sub-module optimizes and adjusts the process parameters related to the occurrence of the crack point based on the process parameter correlation data, selects and adjusts the process parameters affecting the occurrence of the crack point, and generates process change pointing parameter combinations.

[0039] As a further scheme of the present application, the process parameters related to the crack point are selected by analyzing the process data of each welding point, identifying the process factors that have a significant association with the occurrence of the crack point, evaluating the relationship strength between each process parameter and the occurrence of the crack point by a quantitative method, selecting the key parameters that have a significant impact on the occurrence of the crack point, and determining the process parameters as temperature, pressure, welding time, packaging speed, welding power, heat dissipation efficiency, thermal conductivity rate, welding point deformation, contour closure degree, pixel loss rate and pixel features of the contact area.

[0040] Compared with the prior art, the present application has the advantages and positive effects that:

[0041] 1. In the present application, the long short-term memory network is used to analyze the time length, power change and heat dissipation efficiency in the welding point group, the correlation between the parameters and the occurrence of defects is combined, the occurrence of defects is effectively predicted and the potential high-risk path is identified, the processing of time series data has an advantage, the complex nonlinear relationship is processed, and the accurate identification of defect trend is ensured.

[0042] 2. In the present application, the contour, deformation and pixel characteristics of the contact area of the welding point are analyzed by image recognition technology, the deformation deviation, contour closure and pixel loss rate are calculated, the defect area is screened out, and the accuracy of defect screening is improved, and different types of defects are effectively distinguished;

[0043] 3. In the present application, the crack point area is tracked by using a generative adversarial network, the offset and diffusion trend of the crack point path are calculated, the occurrence and propagation of the crack point are effectively located, the crack point tracking and process parameters are combined, and the packaging process parameters are adjusted based on the defect position and diffusion trend, the quality of each welding point in the overall packaging process is optimized, the hidden defects in the welding process are avoided, and the packaging quality and long-term stability of the photovoltaic chip are improved. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 The system flowchart of the present application is shown in the figure;

[0045] Figure 2 The flowchart of the present application is shown in the figure;

[0046] Figure 3 The system framework diagram of the present application is shown in the figure.

[0047] DETAILED DESCRIPTION, which makes the purpose, technical scheme and advantages of the present application clearer and more apparent, is described below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0048] Please refer to Figure 1 and Figure 2 The present application provides a technical scheme: a detection system for high-power photovoltaic TMBS chip packaging includes:

[0049] The thermal field variation monitoring module obtains local temperature variation data through a sensor, extracts welding point thermal value variation data, compares temperature rise difference and thermal conductivity rate, analyzes transmission path through temperature difference, selects welding points with temperature difference greater than a set value, calculates thermal field non-uniformity, and generates a thermal field non-uniformity area distribution map;

[0050] The defect trend analysis module calculates the time length, power change and heat dissipation efficiency in the welding point group based on the thermal field non-uniformity area distribution map, analyzes the correlation between parameters and defect occurrence, identifies abnormal patterns, and generates a defect high-occurrence path node sequence set using a long short-term memory network;

[0051] The image recognition screening module extracts image data based on the defect high-occurrence path node sequence set, analyzes the pixel characteristics of the welding point contour, deformation and contact area, calculates the deformation deviation, contour closure and pixel loss rate, screens the defect area, and generates a multi-level screening defect atlas set;

[0052] Crack point region tracking module: based on the multi-level screening defect atlas set, using a generative adversarial network, tracking the offset and diffusion trend of the crack point region, calculating the closure of the crack point path, selecting the region with closed path, positioning the crack point position, and generating a crack point tracking coordinate index set;

[0053] Process parameter adjustment module: based on the crack point tracking coordinate index set, querying the process parameter configuration of the welding point, comparing the packaging process parameters, selecting the parameters associated with the crack point, and generating process change pointing parameter combination.

[0054] Please refer to Figure 3 , the thermal field variation monitoring module includes:

[0055] Thermal field data acquisition submodule: based on the local temperature change data obtained by the sensor, each welding point in the chip packaging area is measured one by one, the temperature change curve of each welding point is recorded, the temperature rise rate of each welding point is calculated, the temperature change data of the welding point is matched with different regions of the packaging structure, and the welding point thermal value change data is obtained;

[0056] Temperature difference analysis submodule: based on the welding point thermal value change data, the temperature rise difference value of each welding point and the adjacent welding point is calculated, the welding point with temperature difference value exceeding the set threshold is marked as the key welding point, the temperature conduction rate between different welding points is calculated and compared, the heat diffusion path is identified, the welding point region with temperature difference greater than the set value is determined, and the temperature difference conduction path data is generated;

[0057] Thermal field imbalance region identification submodule: based on the temperature difference conduction path data, the welding points with temperature difference greater than the set value are analyzed, the actual situation of temperature change rate and conduction path is combined, the region with uneven heat distribution is identified, the thermal field uniformity of each region is calculated, and the thermal field imbalance region distribution map is generated;

[0058] Thermal field data acquisition submodule: based on the local temperature change data obtained by the sensor, using temperature sensor acquisition module, specifically DHT22 temperature and humidity sensor, through its digital signal interface to obtain the real-time temperature data of each welding point, the temperature sampling frequency of each welding point is set to 1 time per second, the temperature range is -40℃ to 125℃, each welding point in the chip packaging area is measured one by one, the temperature change curve of each welding point is recorded, and the temperature rise rate calculation method is used to calculate the temperature rise rate of each welding point. Assuming that the temperature rise rate is greater than 0.5℃ / s, it is the normal threshold value, the temperature change data of the welding point is matched with different regions of the packaging structure, the welding point thermal value change data is obtained, and the temperature rise rate data of each welding point is compared with the temperature change of each region of the packaging, to generate the welding point thermal value change data;

[0059] The temperature difference analysis submodule: based on the heat value change data of the welding points, a difference analysis algorithm is used to calculate the temperature rise difference value of each welding point and the adjacent welding points, the welding points with temperature difference value exceeding the set threshold are marked as key welding points, the temperature conduction rate analysis method is used to calculate the temperature conduction rate between different welding points, the heat diffusion path is identified, the temperature conduction rate between different welding points is calculated and compared, the heat diffusion path is identified, the welding point area with temperature difference greater than the set value is determined, and the temperature difference conduction path data is generated;

[0060] The hot field imbalance area identification submodule: based on the temperature difference conduction path data, a region analysis algorithm is used to concentrate analysis on the welding points with temperature difference greater than the set value, combined with the actual situation of temperature change rate and conduction path, the hot field uniformity evaluation method is used to calculate the hot field uniformity of each region, if the standard deviation exceeds the set 1.2℃, it is marked as an imbalance area, and the hot field imbalance area distribution map is generated.

[0061] Please refer to Figure 3 , the defect trend judgment module includes:

[0062] The parameter correlation analysis submodule: based on the hot field imbalance area distribution map, the working time, power change and heat dissipation efficiency data in the welding point group are collected, a long short-term memory network is used for modeling, the change trend of each parameter in the time dimension is captured, and the relationship between the historical change of each parameter and the defect occurrence is analyzed, the correlation degree between each parameter and the defect occurrence probability is calculated, sorted by correlation degree, and the key parameters are selected to generate parameter defect correlation degree data;

[0063] The abnormal pattern recognition submodule: based on the parameter defect correlation degree data, the abnormal pattern of defect occurrence is identified and extracted, the relationship between each parameter and the defect is analyzed, the pattern that may cause defects is identified, and the abnormal pattern recognition result is generated;

[0064] The defect high incidence path identification submodule: based on the abnormal pattern recognition result, the path node with defect occurrence probability is selected, the influence of the node in the welding process is analyzed, the frequency and intensity of defect occurrence are calculated, and the defect high incidence path node sequence set is generated;

[0065] The parameter correlation analysis submodule collects data on working time, power variation, and heat dissipation efficiency within the solder joint group based on the distribution map of the uneven thermal field. It employs a Long Short-Term Memory (LSTM) network, with input parameters including historical working time, power variation, and heat dissipation efficiency for each solder joint within the group. A 300-neuron hidden layer is used, with Adam as the optimizer, a learning rate of 0.001, and 100 iterations. An LSTM network is trained to capture the changing trends of each parameter over time, analyzes the relationship between historical changes of each parameter and defect occurrence, calculates the correlation between each parameter and the probability of defect occurrence, and uses the Pearson correlation coefficient to calculate the correlation between each parameter and defect occurrence. The parameters are sorted from high to low correlation to identify key parameters and generate parameter-defect correlation data.

[0066] Anomaly pattern recognition submodule: Based on parameter defect correlation data, it uses an anomaly detection algorithm, specifically IsolationForest, with parameters set to 100 trees and a tree depth of 256. By analyzing the relationship between various parameters and defects, it identifies patterns that may lead to defects. It uses a tree-based isolated factor method for outlier identification, with an outlier threshold of 0.75. The anomaly pattern recognition results are output by the model to identify the anomaly patterns and generate anomaly pattern recognition results.

[0067] Defect-prone path identification submodule: Based on the abnormal pattern recognition results, select path nodes with a probability of defect occurrence, use the K-means clustering algorithm, set the number of clusters to 5, use random initialization for the initial cluster centers, set the number of iterations to 50, use Euclidean distance as the distance metric, and calculate the frequency and intensity of defects by analyzing the possible impacts on nodes during the welding process, and generate a sequence set of defect-prone path nodes.

[0068] Long Short-Term Memory (LSTM) networks, according to the formula:

[0069]

[0070] in: This represents the correlation coefficient between process parameters and the occurrence of defects. Indicates the first Process parameters at each time point Indicates the first Welding pressure data at various time points Indicates the first Defect occurrence data at each point in time Indicates the first Temperature data at each time point, This represents the average value of the process parameters. This represents the mean of the defect data. , 、 and represents the weight coefficient of each parameter, represents the total number of data points;

[0071] Execution process: first, collect the process parameters at different time points, input and process, help to capture the relationship between the trend of each process parameter in the time sequence and the occurrence of defects, then calculate the deviation between the process parameters and the corresponding mean value of the defect data at each time point, and give different importance to each process parameter through the weight coefficient , by calculating the covariance of the weighted process parameters and the defect occurrence data at each time point, and standardizing, the final correlation coefficient , the correlation between each process parameter and the occurrence of defects is evaluated at the same time, which provides the basis for subsequent defect prediction and optimization decision, and ensures timely adjustment of process parameters in the photovoltaic TMBS chip packaging process to reduce the defect rate.

[0072] Abnormal pattern, by analyzing the relationship between parameters and defects, identifying features and behavior patterns related to defects, abnormal fluctuations or trends in time and space dimensions, according to the parameters for the working time of the solder joint group, power change, heat dissipation efficiency, heat value change, temperature rise difference, thermal conductivity rate, temperature difference, solder joint profile deformation, pixel features of contact area, deformation deviation, profile closure degree and pixel loss rate.

[0073] Please refer to Figure 3 , the image recognition screening module includes:

[0074] Defect image extraction submodule: based on the defect high incidence path node sequence set, the image data around the solder joint is extracted by locating the image area of the solder joint, and the image area is cropped to eliminate irrelevant background and enhance the contrast and clarity of the area around the solder joint in the image. The solder joint area image data is integrated to generate an image data set;

[0075] Feature calculation submodule: based on the image data set, the edge of the solder joint area is detected, the closure degree of the solder joint profile is extracted, the pixel point distribution of the deformation area is analyzed, the deformation deviation is calculated, the pixel loss rate of the solder joint and the contact area is compared, the numerical value of each feature is recorded, and the feature calculation result is generated;

[0076] Defect area screening submodule: based on the feature calculation result, the closure degree of the solder joint profile, the deformation deviation and the pixel loss rate are compared with the set threshold value, the area meeting the defect characteristics is screened out, the defect area is marked, and the defect area is classified according to the characteristics of the defect area, and a multi-level screening defect atlas set is generated;

[0077] Defect image extraction submodule: based on the defect high incidence path node sequence set, the image positioning algorithm is used to locate the image area of the welding point, the findContours function in the OpenCV library is used to identify the contour of the welding point area, and the image data around the welding point is extracted. By setting the cropping frame, the irrelevant background is eliminated, the histogram equalization algorithm is used to enhance the contrast and clarity of the image around the welding point, the equalizeHist function of OpenCV is used to process the image, enhance the details, make the welding point area more clear, and integrate the welding point area image data to generate an image data set;

[0078] Feature calculation submodule: based on the image data set, the Canny edge detection algorithm is used to detect the edge of the welding point area by using the Canny function of OpenCV, the closure degree of the welding point contour is extracted, the contour closure degree calculation method is used to calculate the distance of the welding point contour points, analyze the pixel point distribution of the deformation area, calculate the deformation deviation, use the algorithm based on mean square error to compare the pixel loss rate of the welding point and the contact area, use the image difference method to match the welding point and the contact area by using the matchTemplate function of OpenCV, record the numerical value of each feature, and generate the feature calculation result;

[0079] Defect area screening submodule: based on the feature calculation result, the closure degree of the welding point contour, the deformation deviation and the pixel loss rate are compared with the set threshold value, the threshold value judgment algorithm is used, the closure degree threshold value is set to 0.8, the deformation deviation threshold value is set to 0.5, and the pixel loss rate threshold value is set to 0.2. The area meeting the defect characteristics is screened out, the defect area is marked, and the K-means clustering algorithm is used to classify the defect area with the number of clusters set to 3 and initialized randomly to generate a multi-level screening defect atlas set.

[0080] Please refer to Figure 3 , the crack point area tracking module comprises:

[0081] Crack point offset calculation submodule: based on the multi-level screening defect atlas set, the position change of each defect area is analyzed frame by frame, the offset of the defect area at different time points is calculated, the offset direction and amplitude are recorded, the displacement data of the crack point area is obtained, and the crack point offset data is generated;

[0082] Crack point path analysis submodule: based on the crack point offset data, the diffusion trend of the crack point in the packaging area is analyzed, the generated virtual defect image is compared with the historical data to judge whether the crack point expansion path is closed, the defect images under different process conditions generated by the generative adversarial network are combined to evaluate the path closure, the expansion trend of the crack point path is recorded, the region with closed crack point path is selected, and the crack point path analysis result is generated;

[0083] Crack point positioning submodule: based on the crack path analysis result, select the crack region in the path closed area, accurately position and calibrate the crack position, obtain the spatial coordinates of the crack point, and generate a crack tracking coordinate index set;

[0084] Crack point offset calculation submodule: based on the multi-stage screening defect map set, analyze the position change of each defect region frame by frame, use the findContours function of OpenCV to extract the defect region contour in each frame image and record its position, calculate the offset of the defect region centroid in adjacent two frames to obtain the displacement data of the crack region, 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, generate crack point offset data;

[0085] Crack path analysis submodule: based on the crack point offset data, analyze the diffusion trend of the crack point in the packaging area, use the virtual defect image generation method, generate defect images under different process conditions through the generative adversarial network, compare with historical data, use the peak signal-to-noise ratio algorithm to evaluate whether the crack expansion path is closed, set the closed condition as PSNR value greater than 30, record the expansion trend of the crack path, cluster the crack path through the K-means clustering algorithm, select the region where the crack path is closed, and generate the crack path analysis result;

[0086] Crack point positioning submodule: based on the crack path analysis result, select the crack region in the path closed area, use the image matching algorithm, use the matchTemplate function of OpenCV to accurately position the crack region, calculate the similarity between the template image and the current welding image, select the region with similarity greater than the set threshold 0.9, obtain the spatial coordinates of the crack point, and generate a crack tracking coordinate index set.

[0087] The generative adversarial network is generated according to the formula:

[0088]

[0089] Wherein: represents the judgment value of the discriminator to the real data , represents the virtual defect image generated by the generator, represents the distribution of real defect data, represents the distribution of noise data input by the generator, represents the distribution of generator weights, represents the weight parameter in the generator, represents the weight coefficient;

[0090] The execution process: first, receive the process parameters as input, generate possible defect images, the image represents the possible defects that may occur under different packaging conditions, the discriminator then evaluates the generated image, distinguishes the difference between the real image and the generated image, through training, the discriminator will learn the characteristics of the real defect image, and compare with the generated image, through the adversarial process, the generator is gradually optimized to make the generated image more close to the real defect image, the weight of the generator is determined The weight coefficient is used to balance the influence of the generator and the discriminator, so that the quality of the generated image is gradually improved, and finally a virtual image related to the packaging defect is generated, which provides accurate data support for crack path analysis and defect warning in high-power photovoltaic TMBS chip packaging.

[0091] Please refer to Figure 3 , the process parameter adjustment module includes:

[0092] Parameter query submodule: based on the crack tracking coordinate index set, the position of the crack point area is located, the corresponding solder joint process data is queried, the corresponding process configuration is retrieved for each solder joint, the specific process parameters of each solder joint are obtained, and the solder joint process parameter data is generated;

[0093] Correlation comparison submodule: based on the solder joint process parameter data, the process parameters of different solder joints are compared, the relationship between each welding parameter and the occurrence of crack points is analyzed, the influence of temperature, pressure, welding time and packaging speed on the occurrence of crack points is focused on, the strength of each parameter associated with crack points is calculated, the process parameters related to crack points are selected, and the process parameter correlation data is generated;

[0094] Process adjustment submodule: based on the process parameter correlation data, the process parameters related to the correlation of crack occurrence are optimized and adjusted, the process parameters affecting the occurrence of crack points are selected and adjusted, and the process change pointing parameter combination is generated;

[0095] Parameter query submodule: based on the crack tracking coordinate index set, the position of the crack point area is located by image positioning algorithm, the findContours function of OpenCV is used to extract the contour of the crack point area and calibrate the position, the corresponding solder joint process data is queried, the corresponding process configuration is retrieved for each solder joint, the SQL query statement is used to connect the solder joint ID in the process database and the corresponding process configuration table, the specific process parameters of each solder joint are obtained, and the solder joint process parameter data is generated;

[0096] The correlation comparison submodule: based on the welding point process parameter data, the process parameters of different welding points are compared, the Pearson correlation coefficient algorithm is used, the correlation between each pair of welding point parameters is calculated, the relationship between each welding parameter and the occurrence of crack points is analyzed, the influence of temperature, pressure, welding time and packaging speed on the occurrence of crack points is focused on, the strength of each parameter associated with crack points is calculated, a linear regression model is used to calculate the correlation and select the process parameters related to crack points, and process parameter correlation data is generated;

[0097] The process adjustment submodule: based on the process parameter correlation data, the particle swarm optimization algorithm is used, the particle number is set to 50, the iteration number is 1000, the maximum speed is 2, the minimum speed is-2, the particle position is updated, the process parameters related to the occurrence of crack points are optimized, the process parameters affecting the occurrence of crack points are selected and adjusted, the optimization method based on constraint conditions is used, the adjustment range of process parameters is set, and process change direction parameter combination is generated.

[0098] The process parameters related to crack points are screened out, the process data of each welding point is analyzed, the process factors significantly associated with the occurrence of crack points are identified, the relationship strength between each process parameter and the occurrence of crack points is evaluated by a quantitative method, the key parameters significantly affecting the occurrence of crack points are screened out, and the process parameters are temperature, pressure, welding time, packaging speed, welding power, heat dissipation efficiency, thermal conductivity rate, welding point deformation, contour closure, pixel loss rate and pixel characteristics of contact area.

[0099] The above is only the preferred embodiment of the present application, and does not limit the form of the present application, any skilled person in the art can use the disclosed technical content to make changes or modifications as equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application still belongs to the protection scope of the present application technical solution.

Claims

1. A testing system for high-power photovoltaic TMBS chip packaging, characterized in that, The system includes: Thermal field variation monitoring module: It obtains local temperature change data through sensors, extracts solder joint heat value change data, calculates the temperature difference between each solder joint and adjacent solder joints, marks solder joints with temperature differences exceeding a set threshold as critical solder joints, calculates and compares the temperature conduction rate between different solder joints, identifies the heat diffusion path, determines the solder joint area with a temperature difference greater than the set value, generates temperature difference conduction path data, performs centralized analysis on solder joints with a temperature difference greater than the set value based on the temperature difference conduction path data, and identifies areas with uneven heat distribution by combining the actual situation of temperature change rate and conduction path, calculates the thermal field non-uniformity of each area, and generates a thermal field non-uniformity area distribution map. Defect Trend Analysis Module: Based on the distribution map of the uneven thermal field, it collects data on working time, power changes, and heat dissipation efficiency within the weld joint group. It uses a Long Short-Term Memory (LSTM) network for modeling, capturing the changing trends of each parameter over time, and analyzing the relationship between historical changes of each parameter and defect occurrence. It calculates the correlation between each parameter and the probability of defect occurrence, sorts the parameters by correlation, filters out key parameters, and generates parameter-defect correlation data. Based on this correlation data, it identifies and extracts abnormal patterns of defect occurrence. By analyzing the relationship between various parameters and defects, it identifies the patterns leading to defects, generates abnormal pattern recognition results, and selects path nodes with a probability of defect occurrence based on these results. It analyzes the impact of these nodes during the welding process, calculates the frequency and intensity of defect occurrence, and generates a set of high-incidence path node sequences. Image recognition and screening module: Based on the high-incidence path node sequence set of defects, extract image data, analyze the pixel features of weld joint contour, deformation and contact area, calculate deformation deviation, contour closure and pixel loss rate, screen defect areas, and generate a multi-level screening defect map set; Crack point region tracking module: Based on the multi-level screened defect map set, it tracks the offset and diffusion trend of crack point regions, uses generative adversarial networks to generate defect images under different process conditions, calculates the closure of crack point paths, selects the areas where paths are closed, locates the crack point positions, and generates a crack point tracking coordinate index set. Process parameter adjustment module: Based on the crack point tracking coordinate index set, query the process parameter configuration of the weld point, generate weld point process parameter data, compare the process parameters of different weld points based on the weld point process parameter data, analyze the relationship between each welding parameter and the occurrence of crack points, select parameters with strong correlation to crack points, and generate process change pointing parameter combinations.

2. The testing system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The thermal field change monitoring module includes: Thermal field data acquisition submodule: Based on the local temperature change data obtained by the sensor, the module measures each solder joint in the chip packaging area, records the temperature change curve of each solder joint, calculates the temperature rise rate of each solder joint, and matches the temperature change data of the solder joint with different areas of the packaging structure to obtain the heat value change data of the solder joint. Temperature difference analysis submodule: Based on the heat value change data of the solder joints, calculate the temperature difference between each solder joint and its adjacent solder joints, mark the solder joints with temperature differences exceeding the set threshold as key solder joints, calculate and compare the temperature conduction rate between different solder joints, identify the path of heat diffusion, determine the area of ​​solder joints with temperature differences greater than the set value, and generate temperature difference conduction path data. The thermal field unevenness region identification submodule: Based on the temperature difference conduction path data, it performs centralized analysis on solder joints with temperature differences greater than the set value, and combines the actual situation of temperature change rate and conduction path to identify regions with uneven heat distribution, calculate the thermal field non-uniformity of each region, and generate a thermal field unevenness region distribution map.

3. The testing system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The defect trend analysis module includes: Parameter correlation analysis submodule: Based on the distribution map of the uneven thermal field, collect data on working time, power change and heat dissipation efficiency within the solder joint group, use a long short-term memory network for modeling, capture the changing 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 degree between each parameter and the probability of defect occurrence, sort by correlation degree, filter out key parameters, and generate parameter defect correlation degree data. Anomaly pattern recognition submodule: Based on the parameter defect correlation data, it identifies and extracts the abnormal patterns of defect occurrence, analyzes the relationship between various parameters and defects, identifies the patterns that lead to defects, and generates anomaly pattern recognition results; Defect-prone path identification submodule: Based on the abnormal pattern identification results, select path nodes with a probability of defect occurrence, analyze the impact of the nodes on the welding process, calculate the frequency and intensity of defect occurrence, and generate a sequence set of defect-prone path nodes.

4. The testing system for high-power photovoltaic TMBS chip packaging according to claim 3, characterized in that, The abnormal mode is identified by analyzing the relationship between parameters and defect occurrence, identifying defect-related features and behavioral patterns, abnormal fluctuations or trends in the time and space dimensions, based on parameters such as working time, power change, heat dissipation efficiency, heat value change, temperature rise difference, thermal conductivity, 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.

5. The testing 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 submodule: Based on the defect high-incidence path node sequence set, by locating the image region of the solder joint, extract the image data around the solder joint, crop the image region, eliminate irrelevant background, enhance the contrast and clarity of the region around the solder joint in the image, integrate the image data of the solder joint region, and generate an image dataset. Feature calculation submodule: Based on the image dataset, it detects the edges of the solder joint area, extracts the closure degree of the solder joint contour, analyzes the pixel distribution of the deformation area, calculates the deformation deviation, compares the pixel loss rate of the solder joint and the contact area, records the values ​​of each feature, and generates feature calculation results. Defect region screening submodule: Based on the feature calculation results, the closure degree, deformation deviation and pixel loss rate of the solder joint contour are compared with the set thresholds to screen out regions that meet the defect features, mark the defect regions, and classify them according to the features of the defect regions to generate a multi-level screening defect map set.

6. The testing system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The break point region tracking module includes: Crack point offset calculation submodule: Based on the multi-level screened defect map set, analyze the position change of each defect area frame by frame, calculate the offset of the defect area at different time points, record the offset direction and magnitude, obtain the displacement data of the crack point area, and generate crack point offset data. Crack point path analysis submodule: Based on the crack point offset data, analyze the spread trend of the crack point within the packaging area, compare the generated virtual defect image with historical data to determine whether the crack point expansion path is closed, combine defect images under different process conditions generated by the generative adversarial network to evaluate the path closure, 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. The crack point location submodule: Based on the crack point path analysis results, selects the crack point area within the closed path area, performs precise location and marks the crack point position, obtains the spatial coordinates of the crack point, and generates a crack point tracking coordinate index set.

7. The testing system for high-power photovoltaic TMBS chip packaging according to claim 1, characterized in that, The process parameter adjustment module includes: Parameter query submodule: Based on the crack point tracking coordinate index set, by locating the position of the crack point area, query the welding point process data corresponding to the crack point, retrieve the corresponding process configuration for each welding point one by one, obtain the specific process parameters of each welding point, and generate welding point process parameter data. Correlation Comparison Submodule: Based on the solder joint process parameter data, the process parameters of different solder joints are compared, the relationship between each welding parameter and the occurrence of cracks is analyzed, the welding parameters include the influence of temperature, pressure, welding time and packaging speed on the occurrence of cracks, the strength of the correlation between each parameter and cracks is calculated, the process parameters related to cracks are screened out, and process parameter correlation data is generated. Process adjustment submodule: Based on the process parameter correlation data, optimize and adjust the process parameters that are related to the occurrence of crack points, select the process parameters that affect the occurrence of crack points and adjust their settings, and generate process change pointing parameter combinations.

8. The testing system for high-power photovoltaic TMBS chip packaging according to claim 7, characterized in that, The process parameters related to cracking are screened out. By analyzing the process data of each solder joint, process factors that are significantly associated with cracking are identified. The relationship strength between each process parameter and cracking is evaluated by quantitative methods. Key parameters that have a significant impact on cracking are screened out. The process parameters are temperature, pressure, soldering time, packaging speed, soldering power, heat dissipation efficiency, thermal conductivity, solder joint deformation, contour closure, pixel loss rate, and pixel characteristics of the contact area.

Images