Modular defect detection method, system, and computer device for ic carriers

Abstract

The application relates to a modular defect detection method, system and computer equipment for an IC carrier plate, wherein the method comprises the following steps: acquiring original image data in an IC carrier plate production process, pre-processing the original image data to obtain IC carrier plate image data, and constructing a defect recognition model; introducing an attention mechanism into the defect recognition model to obtain an identification result of the IC carrier plate; acquiring process parameter data and equipment operation state data in the IC carrier plate production, performing correlation analysis on the historical defect data, mining potential relationships between the data, and determining key parameters that have a greater influence on defects; constructing a defect prediction model to determine whether the IC carrier plate has defects; and fusing the results of the defect recognition model and the defect prediction model, combining preset rules, and generating corresponding process parameter adjustment suggestions or equipment maintenance strategies, so that the purposes of accurately detecting IC carrier plate defects, predicting potential risks, optimizing processes and equipment maintenance, and improving product quality and production efficiency are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of IC carrier board detection, and particularly relates to a modular defect detection method and system for an IC carrier board and a computer device. BACKGROUND

[0002] In the field of IC carrier board production, the complexity of the process, the diversity of material properties, and the interference of environmental factors make it easy for various defects such as pinholes, scratches, and stains to appear on the surface of the carrier board. Once these defects are not detected in time and accurately, they will have a serious impact on the electrical performance and reliability of the carrier board, and may even cause the downstream chip to fail, thereby causing a large amount of economic loss. Currently, the traditional defect detection methods mainly include manual visual inspection and special optical detection equipment. However, manual visual inspection has obvious drawbacks, and its detection speed is slow and is easily disturbed by human subjective factors, making it difficult to guarantee the accuracy and stability of the detection results. Although the special optical detection equipment improves the detection efficiency to a certain extent, it lacks intelligent recognition and classification capabilities for defects of different shapes, and still requires a large amount of manual participation in judgment and screening in the actual operation process, which not only increases the labor cost but also reduces the overall detection efficiency.

[0003] In addition, the IC carrier board production process parameters are extremely numerous, and different batches of products have great differences in material properties, size accuracy, etc., which undoubtedly puts forward higher requirements for the cause judgment and prevention of defects. Relying only on post-detection of defects cannot timely obtain the dynamic change trend of the defects, and it is difficult to achieve effective prediction and early warning of the defects. The vast amount of test data generated in the production process cannot be fully and effectively utilized, and its data value needs to be deeply mined. SUMMARY

[0004] The main purpose of the present application is to provide a modular defect detection method, system and computer device for an IC carrier board to achieve the purpose of accurately detecting IC carrier board defects, predicting potential risks, optimizing process and equipment maintenance to improve product quality and production efficiency. To achieve the above purpose, the present application provides a modular defect detection method for an IC carrier board, comprising the following steps:

[0005] Obtain the original image data in the production process of the IC carrier board, pre-process the original image data to obtain the pre-processed IC carrier board image data;

[0006] According to the IC carrier board image data, a defect recognition model is constructed, and an attention mechanism is introduced into the defect recognition model to generate an attention weight map and obtain the recognition result of the IC carrier board;

[0007] Obtaining process parameter data and equipment running state data in the IC carrier plate production process, performing correlation analysis on the process parameter data, the equipment running state data and historical defect data, mining potential relationships between the data, and determining key parameters that have greater influence on defect formation;

[0008] According to the process parameter data, the equipment running state data and the key parameters, a defect prediction model is constructed, and it is judged whether the IC carrier plate has defects;

[0009] The results of the defect recognition model and the defect prediction model are fused, based on the fused results, combined with pre-set rules, corresponding process parameter adjustment suggestions or equipment maintenance strategies are generated, the process parameter adjustment suggestions include specific adjustment values of temperature and pressure, and the equipment maintenance strategies include equipment maintenance time and maintenance items.

[0010] Further, the step of obtaining original image data in the IC carrier plate production process and preprocessing the original image data to obtain preprocessed IC carrier plate image data includes:

[0011] Obtaining original image data in the IC carrier plate production process and preprocessing the original image data, including denoising, enhancement and standardization operations;

[0012] According to a pre-set IC carrier plate geometric feature parameter range, it is judged whether each connected region is a qualified IC carrier plate region;

[0013] Extracting the IC carrier plate region judged to be qualified to obtain the preprocessed IC carrier plate image data.

[0014] Further, the step of constructing a defect recognition model according to the IC carrier plate image data and introducing an attention mechanism in the defect recognition model to generate an attention weight map to obtain an identification result of the IC carrier plate includes:

[0015] According to the preprocessed IC carrier plate image data, a convolutional neural network is used to construct a defect recognition model, an attention mechanism is introduced in the convolutional neural network, the weights of different regions are automatically learned through the attention mechanism, and the defect region is highlighted;

[0016] In the training process of the defect recognition model, the attention weight is calculated according to the labeled defect region to obtain an attention weight map, wherein the region with high weight in the attention weight map indicates that there may be a defect;

[0017] The attention weight map and the original image data are fused to obtain fused image data, and the defect region is highlighted in the fused image data;

[0018] input the fused image data into the defect identification model for prediction;

[0019] output the prediction result of the defect identification model, including the position, type and confidence of the defect, and determine whether the IC carrier board has defects.

[0020] Further, the step of fusing the attention weight map with the original image data to obtain fused image data, in which the defect area is highlighted, further comprises:

[0021] Using data enhancement technology, the original image is rotated, translated and scaled to generate training samples and expand the training data set;

[0022] According to the fused image data and the expanded training data set, the model structure and hyperparameters in the convolutional neural network model are adjusted;

[0023] The convolutional neural network model is optimized by an adaptive learning rate adjustment algorithm.

[0024] Further, the step of obtaining process parameter data and equipment running state data in the IC carrier board production process, and correlating and analyzing the process parameter data, the equipment running state data and the historical defect data to mine the potential relationship between the data and determine the key parameters that have greater influence on defect formation, comprises:

[0025] Obtain process parameter data and equipment running state data in the IC carrier board production process, and preprocess the data, including data cleaning, data integration and data transformation operations

[0026] According to the characteristics of the IC carrier board production process, select the corresponding data mining algorithm, calculate the influence degree of each process parameter and equipment running state on the formation of IC carrier board defects, and determine the key parameters according to the influence degree, the data mining algorithm is used to analyze the correlation between process parameters, equipment running state and historical defect data.

[0027] Further, the step of constructing a defect prediction model according to the process parameter data, the equipment running state data and the key parameters, and determining whether the IC carrier board has defects, comprises:

[0028] According to the process parameter data, the equipment running state data and the key parameters, the data is preprocessed to obtain a preprocessed data set;

[0029] Using a feature selection algorithm, the preprocessed data set is selected to select a feature subset with high correlation with the carrier board defect;

[0030] A decision tree algorithm is used to construct a model, and the model is trained based on the feature subset to obtain a defect prediction model;

[0031] Process parameter data, equipment operation state data and key parameters of the IC carrier plate are acquired, and the acquired data is preprocessed to obtain preprocessed to-be-predicted data;

[0032] The preprocessed to-be-predicted data is input into the trained defect prediction model, and the operation of the model is used to determine whether the IC carrier plate has defects, and if the defect probability output by the model is greater than a preset threshold, it is determined that the IC carrier plate has defects.

[0033] Further, the results of the defect recognition model and the defect prediction model are fused, and based on the fused results, a corresponding process parameter adjustment suggestion or equipment maintenance strategy is generated in combination with a pre-set rule, comprising:

[0034] The output results of the defect recognition model and the defect prediction model are acquired, and the results of the two models are fused to obtain a fused defect analysis result;

[0035] According to a pre-set process parameter adjustment rule, it is determined whether the fused defect analysis result meets the condition of triggering process parameter adjustment, and if so, according to the defect type and severity, the specific adjustment value of temperature and pressure is determined to generate a process parameter adjustment suggestion;

[0036] According to a pre-set equipment maintenance strategy rule, it is determined whether the fused defect analysis result meets the condition of triggering equipment maintenance, and if so, according to the defect type and severity and the equipment historical maintenance record, the equipment repair time and repair item are determined to generate an equipment maintenance strategy.

[0037] The application also provides a modular defect detection system for an IC carrier plate, comprising:

[0038] An image preprocessing module is configured to acquire original image data in the production process of the IC carrier plate, and pre-process the original image data to obtain pre-processed IC carrier plate image data;

[0039] A recognition model construction module is configured to construct a defect recognition model according to the IC carrier plate image data, introduce an attention mechanism into the defect recognition model, generate an attention weight map, and obtain an identification result of the IC carrier plate;

[0040] a data analysis module, configured to acquire process parameter data and equipment operation state data in the IC carrier plate production process, perform correlation analysis on the process parameter data, the equipment operation state data and historical defect data, mine potential relationships between the data, and determine key parameters that have greater influence on defect formation;

[0041] a prediction model construction module, configured to construct a defect prediction model according to the process parameter data, the equipment operation state data and the key parameters, and determine whether the IC carrier plate has defects;

[0042] a strategy generation module, configured to fuse results of the defect recognition model and the defect prediction model, generate corresponding process parameter adjustment suggestions or equipment maintenance strategies based on the fused results and in combination with preset rules.

[0043] The application further provides a computer device including a memory and a processor, the memory has stored therein a computer program, and the processor implements the steps of the above-described modular defect detection method for an IC carrier plate when executing the computer program.

[0044] The application further provides a computer readable storage medium having stored therein a computer program, and the computer program implements the steps of the above-described modular defect detection method for an IC carrier plate when executed by a processor.

[0045] The modular defect detection method for an IC carrier plate, system and computer device provided by the application have the following beneficial effects: the application first utilizes a defect recognition model to quickly process image data, timely screens out surface defect products, avoids the surface defect products from entering subsequent processes, prevents resource waste, greatly improves production efficiency, and in the process, introduces an attention mechanism to accurately determine defect positions, types and confidence, provides reliable samples for defect recognition, and reduces overall detection misjudgment and omission rates. Based on the recognition results, key parameters are mined in combination with process and equipment data to construct a prediction model, potential defects are warned in advance, processes are timely adjusted, and defects are prevented from being generated again. Process adjustment suggestions and equipment maintenance strategies are generated in combination with preset rules based on the two model results, parameters are accurately adjusted, maintenance is reasonably arranged, product yield is improved, and production stability and product quality are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a flowchart of a modular defect detection method for an IC carrier plate in an embodiment of the application;

[0047] Figure 2 is a structural block diagram of a modular defect detection system for an IC carrier plate in an embodiment of the application;

[0048] Figure 3is a structural schematic block diagram of a computer device of an embodiment of the present application.

[0049] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0050] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application 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 application and should not be used to limit the present application.

[0051] Referring to Figure 1 A flowchart of a modular defect detection method for an IC carrier plate according to the present application is shown in FIG. 1, which includes the following steps:

[0052] S1, obtaining original image data in the production process of the IC carrier plate, pre-processing the original image data to obtain pre-processed IC carrier plate image data;

[0053] S2, constructing a defect recognition model according to the IC carrier plate image data, introducing an attention mechanism in the defect recognition model, generating an attention weight map, and obtaining a recognition result of the IC carrier plate;

[0054] S3, obtaining process parameter data and equipment running state data in the production process of the IC carrier plate, performing correlation analysis on the process parameter data, the equipment running state data and historical defect data, mining potential relationships between the data, and determining key parameters that have a greater impact on defect formation;

[0055] S4, constructing a defect prediction model according to the process parameter data, the equipment running state data and the key parameters, and determining whether the IC carrier plate has defects;

[0056] S5, fusing the results of the defect recognition model and the defect prediction model, generating corresponding process parameter adjustment suggestions or equipment maintenance strategies based on the fused results and in combination with pre-set rules, the process parameter adjustment suggestions including specific adjustment values of temperature and pressure, and the equipment maintenance strategies including equipment repair time and repair items.

[0057] As described in step S1, the original image data in the IC carrier plate production process is obtained, and the original image data is preprocessed to obtain preprocessed IC carrier plate image data. The image data of the IC carrier plate in the production process is collected, which is the original data without processing and may contain various noises, uneven illumination and other problems. Denoising, enhancement and standardization operations are performed on the original image data. Due to factors such as production environment and equipment, the original image may contain noise, and the denoising operation can reduce noise interference, improve image quality, and make subsequent feature extraction of the IC carrier plate more accurate. For example, Gaussian filtering algorithm is used to remove Gaussian noise in the image, enhance certain features of the image, and make the details of the IC carrier plate clearer for subsequent identification. For example, the histogram equalization is used to enhance the contrast of the image, so that different regions in the image are more easily distinguished. The image data is normalized according to a certain standard, so that the image data obtained under different batches and different conditions has a unified specification and range, which is convenient for subsequent model processing. According to the pre-set range of the geometric feature parameters of the IC carrier plate, it is judged whether each connected region in the image is a qualified IC carrier plate region. The qualified IC carrier plate region is extracted, and the image data composed of these regions is the preprocessed IC carrier plate image data, which can be used for subsequent operations such as constructing a defect recognition model.

[0058] As described in step S2 above, according to the IC carrier board image data, a defect recognition model is constructed, and an attention mechanism is introduced into the defect recognition model to generate an attention weight map, and an identification result of the IC carrier board is obtained. Based on the IC carrier board image data preprocessed in step S1, a convolutional neural network (CNN) is used to build a defect recognition model. An attention mechanism is introduced into the constructed convolutional neural network. The core role of the attention mechanism is to enable the model to automatically focus on the importance of different regions when processing images, i.e., by learning to assign different weights to different regions of the image. In IC carrier board defect recognition, the attention mechanism can highlight the defect regions in the image, so that the model focuses more on the parts that may have defects, and improves the sensitivity and recognition ability of the defects. In the training process of the defect recognition model, the attention weight is calculated according to the labeled defect region. The labeled defect region is the part of the image that actually has defects marked by human or other means in advance. According to these labeling information, the model calculates the importance of each region for recognizing defects through a specific algorithm, and then generates an attention weight map. In this weight map, the region with high weight indicates that this region is more likely to have defects. The generated attention weight map is fused with the original image data to obtain fused image data. In this fused image, since the attention weight map highlights the regions that may have defects, the defect regions will be more obviously displayed in the image. The fused image data is input into the constructed defect recognition model for prediction. The model judges whether the IC carrier board in the image has defects through the learned features and patterns, and outputs the prediction result. The prediction result includes the position, type and confidence of the defect. For example, the model may output that there is a scratch defect at a specific coordinate position of the IC carrier board, and the confidence of this judgment is 90%.

[0059] As described in step S3, the process parameter data and the equipment running state data in the IC carrier plate production process are obtained, the process parameter data and the equipment running state data are analyzed in association with historical defect data, potential relationships between the data are mined, and key parameters that have a greater impact on defect formation are determined. Process parameter data in the IC carrier plate production process, such as temperature, pressure, time, etc., is collected, and at the same time, equipment running state data, such as the speed of the equipment, the vibration condition, the working state of each component, etc., is obtained. The obtained process parameter data and equipment running state data are preprocessed, including data cleaning, data integration, and data transformation operations. According to the characteristics of the IC carrier plate production process, appropriate data mining algorithms are selected. These algorithms are used to analyze the association between the process parameters, the equipment running state, and the historical defect data. For example, decision tree algorithms, association rule mining algorithms, etc. are selected to analyze a large amount of historical data, and the contribution of each factor to defect formation is quantified. For example, it is found through analysis that the fluctuation of temperature within a certain range is positively correlated with the probability of pinhole defects on the IC carrier plate, and the specific impact degree value of temperature change on the formation of pinhole defects is calculated. According to the calculated impact degree, key parameters that have a greater impact on defect formation are determined. For example, if the analysis result shows that the pressure parameter and the vibration frequency of a certain component of the equipment have a much greater impact on the IC carrier plate scratch defect than other factors, then the pressure and the vibration frequency of the component are determined as key parameters, which need to be focused on and managed in subsequent production to reduce the possibility of scratch defects.

[0060] As described in step S4, a defect prediction model is constructed according to the process parameter data, the equipment running state data, and the key parameters, and it is determined whether the IC carrier plate has defects. According to the obtained process parameter data, equipment running state data, and key parameters determined in step S3, the data is preprocessed to obtain a data set suitable for model processing. A feature selection algorithm is used to analyze the preprocessed data set in depth, select a feature subset with a higher correlation with the carrier plate defect from numerous features (i.e., various process parameters, equipment running state parameters, etc.), thereby removing redundant or irrelevant features, reducing the data dimension, and reducing the complexity and computational amount of model training; and the model can focus more on key features, improving the prediction accuracy and efficiency of the model. A decision tree algorithm is used to construct a defect prediction model. Based on the selected feature subset, the decision tree model is trained. In the training process, the model learns the internal relationship between different feature combinations and IC carrier plate defects from the data, and adjusts its parameters and structure to achieve the best prediction performance.

[0061] The process parameter data, equipment running state data and key parameters of the current IC carrier to be detected are acquired, and the acquired data is also preprocessed to ensure that the data format and quality are consistent with those in the training model. The preprocessed data to be predicted is input into the trained defect prediction model. The model performs operation and analysis on the input data according to the rules and patterns learned in the training process. The model outputs a defect probability value, which is compared with a preset threshold. If the defect probability output by the model is greater than the preset threshold, it indicates that the IC carrier has a high possibility of defects according to the current process parameters, equipment running state and other data, and therefore it is determined that the IC carrier has defects. Conversely, if the defect probability is less than or equal to the preset threshold, it is determined that the IC carrier has no defects. In this way, the prediction and judgment of whether the IC carrier has defects are realized.

[0062] As described in step S5, the results of the defect identification model and the defect prediction model are fused, and based on the fused results, a corresponding process parameter adjustment suggestion or equipment maintenance strategy is generated according to the pre-set rules. The process parameter adjustment suggestion includes specific adjustment values of temperature and pressure, and the equipment maintenance strategy includes equipment maintenance time and maintenance items. The output results of the defect identification model and the defect prediction model are acquired. The defect identification model mainly outputs the information related to the defects that have occurred on the IC carrier, such as the position, type and confidence of the defects; the defect prediction model judges whether the IC carrier has defects and outputs the defect probability based on the process parameters and equipment running state and other data.

[0063] The results of the two models are fused. For example, the feature vectors output by the two models are spliced, or weighted summation is performed after assigning weights according to the importance of the results of different models, and the like. Through feature fusion, a fused defect analysis result that comprehensively reflects the defect situation of the IC carrier plate is obtained. This result integrates the judgment information of the IC carrier plate defects from different angles by the two models, and is more comprehensive and accurate than the single model result. According to the preset process parameter adjustment rules, it is judged whether the fused defect analysis result meets the condition of triggering process parameter adjustment. These preset rules are based on production experience, experimental data and understanding of process principles. For example, if the fusion result shows that a certain type of defect appears on the IC carrier plate, and the defect is closely related to abnormal changes in temperature and pressure in historical data, it meets the condition of triggering process parameter adjustment. When the condition is met, the specific adjustment value of temperature and pressure is determined according to the defect type and severity. Different types of defects may require different directions and magnitudes of process parameter adjustment. For example, if the defect is caused by excessive temperature leading to material deformation, a certain degree of temperature reduction may be required; the higher the defect severity, the greater the adjustment magnitude may be. According to the preset device maintenance strategy rules, it is judged whether the fused defect analysis result meets the condition of triggering device maintenance. These rules are also based on actual production conditions and device characteristics, for example, when the defect analysis result implies that a certain defect frequently occurs and is related to a specific operating state or component of the device, the triggering condition is met. If the condition is met, the device repair time and repair items are determined according to the defect type and severity and the device historical maintenance record. Different types of defects may indicate problems with different components of the device, for example, if the defect is related to unstable pressure of the device, and the historical maintenance record shows that the pressure control component has experienced similar failures, the repair items may include inspection and repair of the pressure control component; the higher the defect severity, the more timely the repair may be required, i.e. the repair time is earlier.

[0064] In one embodiment, high-definition camera is used to acquire image data during IC carrier board production process, image resolution is 1920x1080 pixels, color depth is 24 bits. Color image is converted to grayscale image using OpenCV image processing library, conversion formula is Gray=Rx299+Gx587+Bx114. Median filter algorithm with 5x5 is used to denoise grayscale image, remove salt and pepper noise and Gaussian noise. According to the grayscale characteristics of IC carrier board, set the binary threshold to 128, set the pixel points with grayscale value greater than or equal to 128 to 255, and set the pixel points with grayscale value less than 128 to 0, to obtain black and white binary image. First, use 3x3 rectangular structure element to perform erosion operation on binary image to eliminate small white noise points. Then use 5x5 rectangular structure element to perform inflation operation on the image to fill black breakpoints and holes. Connected region analysis algorithm is used to extract connected regions in binary image, and area, perimeter and circularity of each connected region are calculated. According to the design specifications of IC carrier board, the area range of qualified IC carrier board region is set to 100-400 square millimeters, the perimeter range is set to 40-80 millimeters, and the circularity range is set to 85-95. The connected regions meeting the parameter range are extracted to obtain preprocessed qualified IC carrier board image data.

[0065] Histogram equalization algorithm is used to enhance the acquired IC carrier board original image data, improve contrast and clarity. Median filter algorithm is used for denoising to preserve image edge and texture features. The image is normalized to scale the pixel value to 0 to 1.

[0066]

[0067] I norm represents the normalized image, I represents the original image, I min and I max represent the minimum and maximum pixel values of the image respectively. This formula is used to scale the pixel value to 0 to 1. Based on the preprocessed image data, ResNet-50 convolutional neural network is used to construct a defect recognition model. Channel attention mechanism and spatial attention mechanism are introduced after the last convolutional layer of the network to automatically learn the weights of different regions and highlight the defect regions. During model training, attention weights are calculated according to the labeled defect regions, the weight value of the defect region is set to 8, and the weight value of the non-defect region is set to 2, to obtain an attention weight map with the same size as the original image. The attention weight map and the original image are pixel-level weighted fusion, the formula is:

[0068] I fused =I·(1+αW)

[0069] I fusedThe fused image is represented by I, the original image is represented by I, the attention weight map is represented by W, and the fusion coefficient is represented by a. This formula is used to perform pixel-level weighted fusion of the attention weight and the original image. The pixel value of the defect area is amplified, and the pixel value of the non-defect area is reduced to obtain the fused image data. The fused image data is input into the defect recognition model for prediction, and the defect class probability is output through the softmax function. The confidence threshold is set to 8, and when the probability of a certain class is greater than 8, it is judged that the area exists this type of defect, and the defect position coordinates are output. Every week, 500 training data of different types of IC carrier boards are added. Data augmentation techniques are used to expand the data set, such as rotating the original image (within ±10 degrees), translating it (within ±10 pixels), scaling it (within 8 to 2 times), and other transformations. The grid search algorithm is used to adjust the model hyperparameters, such as learning rate, batch size, and regularization coefficient, and an adaptive learning rate adjustment algorithm is introduced to prevent model overfitting, with the formula:

[0070]

[0071] η t The learning rate of the t-th iteration is represented by η, the initial learning rate is represented by η0, the decay rate is represented by β, and the iteration number is represented by t. This formula is used to dynamically adjust the learning rate to prevent model overfitting and improve the defect recognition accuracy and generalization ability of the model.

[0072] During the production of IC carrier boards, process parameter data (such as temperature, pressure, speed, etc.) and equipment operating state data are collected in real time by online sensors, with a sampling frequency of 1000 Hz. The data is transmitted to a database server for storage via an industrial Ethernet network. The maximum entropy model is used to estimate missing data. The local anomaly factor algorithm is used to detect and remove abnormal data. Data standardization and normalization are performed to convert data from different sources and scales into a unified format and range. The Apriori association rule mining algorithm is selected to perform association analysis on process parameters, equipment states, and defect data, with a support threshold of 0.5, a confidence threshold of 0.8, and a minimum association rule length of 3. For example, it is found that in the electroplating process of IC carrier boards, when the current density exceeds 5 A / dm 2 and the plating solution temperature exceeds 60℃, the surface of the carrier board is prone to pinhole defects, with a defect occurrence rate exceeding 5%. Through experimental optimization analysis, it is determined that the current density should be controlled at 2 A / dm 2The plating solution temperature is controlled below 55°C, which can reduce the pinhole defect rate to within 1%. Current density and plating solution temperature are key parameters. Data cleaning is performed on process parameter data, equipment operation state data, and key parameters to remove missing values, abnormal values (using the 3σ criterion to remove data exceeding 3 times the standard deviation), and other noise data. The minimum-maximum normalization method is used to map the data to the [0, 1] interval, eliminating dimensional differences. The recursive feature elimination algorithm is used, with 10 iterations and 10% of the features eliminated each iteration. The top 20 features that have a greater impact on product defects are selected as the feature subset (based on the Pearson correlation score of the features and defects, features with a correlation of less than 1 are iteratively deleted).

[0073]

[0074] r represents the Pearson correlation coefficient, which is used to evaluate the correlation between the features and the defects. i and y i represent the feature values and defect values, respectively, and represent their average values, respectively.

[0075] Based on the feature subset, a CART decision tree algorithm is used to construct a defect prediction model, with a maximum decision tree depth of 5 and a minimum leaf node sample size of 10. Through 10-fold cross-validation, a model with an average prediction accuracy of 92% is obtained (or a support vector machine algorithm is used to construct a model, with 5-fold cross-validation and grid search optimization of hyperparameters such as penalty coefficient C=10 and Gaussian kernel, resulting in a prediction accuracy of more than 95%). After obtaining the relevant data of the product to be predicted and performing the same data preprocessing as the training data, the trained decision tree model is input for prediction, with the formula being:

[0076]

[0077] The formula represents a defect determination rule. When the defect probability P(defect) output by the model is greater than 6, it is determined that the product has a defect (Defect = 1), otherwise it is determined to be defect-free (Defect = 0). For products predicted to have defects, the key factors leading to defects are determined according to the importance scores (such as Gini importance scores) of each feature in the feature subset, such as temperature exceeding 450℃, pressure below 2MPa, etc., and defect localization analysis is performed. According to the analysis results, adjust the corresponding process parameters or equipment state, such as controlling the temperature below 420℃, increasing the pressure to 5MPa, predicting the defect probability again to make it drop below 2, and the product pass rate is increased to more than 98%. Obtain the output results of the defect identification model and the defect prediction model, assuming that the defect identification model outputs a crack defect type with a severity of 8, and the defect prediction model outputs a porosity defect type with a severity of 6. Feature fusion is performed using a weighted average method, with a crack weight of 7 and a porosity weight of 3, and the fused defect analysis results are obtained: the crack severity is 7.6 and the porosity severity is 1.8.

[0078] According to the preset process parameter adjustment rule, when the crack severity is greater than 7, the temperature needs to be reduced by 20℃ and the pressure needs to be increased by 5MPa; when the porosity severity is greater than 2, the temperature needs to be increased by 10℃ and the pressure needs to be reduced by 8MPa. Since the crack severity is 7.6, which meets the condition, the preliminary process parameter adjustment suggestion is to reduce the temperature by 20℃ and increase the pressure by 5MPa. Further combined with the real-time running parameters of the equipment and the past 24 hours historical trend data, the gradient descent algorithm is used to optimize the adjustment amplitude (the learning rate is set to 0.1 and the iteration is 100 times), and the final process parameter adjustment suggestion is to reduce the temperature by 15℃ and increase the pressure by 2.8MPa. According to the preset equipment maintenance strategy rule, when the crack severity is greater than 7.5 and the equipment running time exceeds 1000 hours, equipment maintenance needs to be performed within 3 days, including replacing parts and calibrating parameters. Query the historical maintenance record of the equipment, the equipment has been running for 1200 hours, and the preliminary equipment maintenance strategy is to perform equipment maintenance within 3 days, replace parts and calibrate parameters. Combined with the equipment health state evaluation result and the risk prediction result, the reinforcement learning algorithm (with the minimization of equipment failure risk as the reward function) is used to dynamically adjust the maintenance time and maintenance items, and after 500 iterations, the optimal equipment maintenance strategy is to perform equipment maintenance within 1 day, replace parts, calibrate parameters and add lubricating oil.

[0079] Reference Figure 2 The structural block diagram of the modular defect detection system for IC carrier boards in an embodiment of the present application comprises:

[0080] An image preprocessing module is configured to acquire original image data in the production process of the IC carrier board, and to preprocess the original image data to obtain preprocessed IC carrier board image data.

[0081] The recognition model construction module is configured to construct a defect recognition model according to the IC carrier board image data, introduce an attention mechanism into the defect recognition model, generate an attention weight map, and obtain a recognition result of the IC carrier board.

[0082] The data analysis module is configured to acquire process parameter data and equipment running state data in the production process of the IC carrier board, perform correlation analysis on the process parameter data, the equipment running state data and historical defect data, mine potential relationships between the data, and determine key parameters that have a greater impact on defect formation.

[0083] The prediction model construction module is configured to construct a defect prediction model according to the process parameter data, the equipment running state data and the key parameters, and determine whether the IC carrier board has defects.

[0084] The strategy generation module is configured to fuse results of the defect recognition model and the defect prediction model, generate corresponding process parameter adjustment suggestions or equipment maintenance strategies based on the fused results and in combination with preset rules.

[0085] The specific implementation of each module in the above device example can refer to the description in the above method embodiments, and will not be described here.

[0086] Referring to Figure 3 , the embodiment of the present application also provides a computer device, which can be a server, and the internal structure thereof can be as shown in Figure 3 . The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store corresponding data in the embodiment. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the above method.

[0087] Those skilled in the art can understand Figure 3 that the structure shown in the embodiment is only a block diagram of part of the structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied.

[0088] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the method. It can be understood that the computer readable storage medium in the embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0089] In summary, the original image data in the IC carrier plate production process is acquired, the original image data is preprocessed to obtain preprocessed IC carrier plate image data; a defect recognition model is constructed according to the IC carrier plate image data, and an attention mechanism is introduced into the defect recognition model to generate an attention weight map to obtain an IC carrier plate recognition result; process parameter data and equipment running state data in the IC carrier plate production process are acquired, the process parameter data and the equipment running state data are associated with historical defect data for correlation analysis to mine potential relationships between data and determine key parameters that have a greater impact on defect formation; a defect prediction model is constructed according to the process parameter data, the equipment running state data and the key parameters, and it is determined whether the IC carrier plate has defects; the results of the defect recognition model and the defect prediction model are fused, based on the fused results, a corresponding process parameter adjustment suggestion or equipment maintenance strategy is generated in combination with a pre-set rule, the process parameter adjustment suggestion includes specific adjustment values of temperature and pressure, and the equipment maintenance strategy includes equipment maintenance time and maintenance items, so as to realize the purpose of accurately detecting IC carrier plate defects, predicting potential risks, optimizing process and equipment maintenance to improve product quality and production efficiency.

[0090] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium provided by the present application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0091] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, device, article or method that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, device, article or method. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, device, article or method that includes the element.

[0092] The above description is only the preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, based on the content of the present application specification and drawings, are also included in the patent protection scope of the present application.

Claims

1. A modular defect detection method for IC substrates, characterized in that, Includes the following steps: The original image data of the IC carrier board production process is acquired, and the original image data is preprocessed to obtain preprocessed IC carrier board image data. Based on the preprocessed IC carrier image data, a defect recognition model is constructed using a ResNet-50 convolutional neural network. An attention mechanism is introduced into the ResNet-50 convolutional neural network to automatically learn the weights of different regions and highlight the defective regions. During the training process of the defect recognition model, attention weights are calculated based on the labeled defect regions to obtain an attention weight map. Regions with high weights in the attention weight map indicate potential defects. The attention weight map is then fused with the original image data to obtain fused image data, in which defect regions are highlighted. The fused image data is then input into the defect recognition model for prediction. The prediction results of the defect recognition model are output, including the location, type, and confidence level of the defect, to determine whether the IC carrier board has a defect. The process parameter data and equipment operating status data during the IC substrate production process are acquired and preprocessed, including data cleaning, data integration, and data transformation. Based on the characteristics of the IC substrate production process, the Apriori association rule mining algorithm is selected to calculate the degree of influence of each process parameter and equipment operating status on the formation of IC substrate defects. Based on the degree of influence, key parameters are determined. The data mining algorithm is used to analyze the correlation between process parameters, equipment operating status, and historical defect data. Based on the process parameter data, equipment operating status data, and key parameters, the data is preprocessed to obtain a preprocessed dataset. A feature selection algorithm is used to select features from the preprocessed dataset, identifying a subset of features highly correlated with carrier board defects. A decision tree algorithm is used to construct a model, which is then trained based on the feature subset to obtain a defect prediction model. The process parameter data, equipment operating status data, and key parameters of the IC carrier board are acquired, and the acquired data is preprocessed to obtain preprocessed data to be predicted. The preprocessed data to be predicted is input into the trained defect prediction model. Through the model's calculations, it is determined whether the IC carrier board has defects. If the defect probability output by the model is greater than a preset threshold, then the IC carrier board is determined to have defects. The results of the defect identification model and the defect prediction model are fused using a weighted average method. Based on the fused results and combined with pre-set rules, corresponding process parameter adjustment suggestions or equipment maintenance strategies are generated. The process parameter adjustment suggestions include specific adjustment values ​​for temperature and pressure, and the equipment maintenance strategies include equipment maintenance time and maintenance items.

2. The modular defect detection method for IC carrier boards according to claim 1, characterized in that, The step of acquiring raw image data during the IC carrier board manufacturing process and preprocessing the raw image data to obtain preprocessed IC carrier board image data includes: The raw image data of the IC substrate production process is acquired, and the raw image data is preprocessed, including noise reduction, enhancement and standardization operations. Based on the preset range of IC carrier geometric feature parameters, determine whether each connected region is a qualified IC carrier region; The qualified IC carrier board area is extracted to obtain the preprocessed IC carrier board image data.

3. The modular defect detection method for IC carrier boards according to claim 1, characterized in that, After fusing the attention weight map with the original image data to obtain fused image data, and highlighting defective areas in the fused image data, the method further includes: Data augmentation techniques are used to rotate, translate, and scale the original image to generate training samples and expand the training dataset. Based on the fused image data and the expanded training dataset, the model structure and hyperparameters in the convolutional neural network model are adjusted; Optimize convolutional neural network models using an adaptive learning rate adjustment algorithm.

4. The modular defect detection method for IC carrier boards according to claim 1, characterized in that, The step of fusing the results of the defect identification model and the defect prediction model, and generating corresponding process parameter adjustment suggestions or equipment maintenance strategies based on the fused results and pre-defined rules, includes: Obtain the output results of the defect identification model and the defect prediction model, perform feature fusion on the results of the two models, and obtain the fused defect analysis results; Based on the preset process parameter adjustment rules, determine whether the fused defect analysis results meet the conditions for triggering process parameter adjustment. If they do, determine the specific adjustment values ​​for temperature and pressure based on the defect type and severity, and generate process parameter adjustment suggestions. Based on the preset equipment maintenance strategy rules, it is determined whether the fused defect analysis results meet the conditions for triggering equipment maintenance. If they do, the equipment maintenance time and maintenance items are determined based on the defect type and severity, as well as the equipment's historical maintenance records, and an equipment maintenance strategy is generated.

5. A modular defect detection system for IC substrates, characterized in that, include: The image preprocessing module is used to acquire the original image data during the IC carrier board production process, preprocess the original image data, and obtain the preprocessed IC carrier board image data. The identification model building module is used to build a defect identification model using a ResNet-50 convolutional neural network based on the preprocessed IC carrier board image data. An attention mechanism is introduced into the ResNet-50 convolutional neural network to automatically learn the weights of different regions and highlight the defective regions. During the training process of the defect recognition model, attention weights are calculated based on the labeled defect regions to obtain an attention weight map. Regions with high weights in the attention weight map indicate potential defects. The attention weight map is then fused with the original image data to obtain fused image data, in which defect regions are highlighted. The fused image data is then input into the defect recognition model for prediction. The prediction results of the defect recognition model are output, including the location, type, and confidence level of the defect, to determine whether the IC carrier board has a defect. The data analysis module is used to acquire process parameter data and equipment operating status data during the IC substrate production process, and to preprocess the data, including data cleaning, data integration and data transformation operations; based on the characteristics of the IC substrate production process, the Apriori association rule mining algorithm is selected to calculate the degree of influence of each process parameter and equipment operating status on the formation of IC substrate defects, and to determine the key parameters based on the degree of influence. The prediction model building module is used to preprocess the data based on the process parameter data, the equipment operating status data, and the key parameters to obtain a preprocessed dataset; employ a feature selection algorithm to select features from the preprocessed dataset, choosing a subset of features highly correlated with carrier board defects; construct a model using a decision tree algorithm, and train the model based on the feature subset to obtain a defect prediction model; acquire the process parameter data, equipment operating status data, and key parameters of the IC carrier board, preprocess the acquired data to obtain preprocessed data to be predicted; input the preprocessed data to be predicted into the trained defect prediction model, and determine whether the IC carrier board has defects through the model's calculations; if the defect probability output by the model is greater than a preset threshold, then the IC carrier board is determined to have defects. The strategy generation module is used to fuse the results of the defect identification model and the defect prediction model using a weighted average method. Based on the fused results and combined with pre-set rules, it generates corresponding process parameter adjustment suggestions or equipment maintenance strategies.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the modular defect detection method for IC carrier boards according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the modular defect detection method for IC carrier boards as described in any one of claims 1 to 4.

Images