PCBA Welding Quality Prediction Method and System Based on Time-Series Data

Abstract

The present application provides a method and system for predicting the welding quality of PCBA based on time series data, relating to the field of circuit board welding. The method includes: obtaining process time series data and detection data information, and performing missing value filling and data normalization processing; quantifying the correlation between each feature data and the thickness of the intermetallic compound; constructing an MPSO-LSTM prediction model and training the model to obtain an initial model; evaluating the accuracy of the initial model to adjust the hyperparameters of the initial model, obtaining a quality prediction model to analyze the real-time process data during the welding process, and realizing the prediction of the IMC thickness. The present application effectively improves the online quality prediction ability during the welding process by constructing an MPSO-LSTM prediction model. Through model training and hyperparameter adjustment, a quality prediction model is obtained, which can perform real-time analysis on the process data during the welding process, predict the thickness of the intermetallic compound, and assist technicians in quantitatively identifying the quality causes of solder joint defects.

Description

Technical Field

[0001] The present application relates to the technical field of circuit board soldering, and specifically relates to a method and system for predicting the soldering quality of PCBA based on time series data. Background Art

[0002] The soft soldering of electronic assembly is the core process to ensure the electrical performance reliability of Printed Circuit Board Assembly (PCBA). During the assembly soldering and debugging stages, obvious quality defects such as insufficient solder, bridging, and solder wicking can be identified by direct visual inspection. However, latent quality defects such as intermittent on-off and strong-weak fluctuations of circuit signals are mostly closely related to the thickness of the Intermetallic Compound (IMC) formed during the soldering process. During soldering, factors such as heating duration, wetting condition, and heating temperature are different, resulting in different IMC thicknesses. The IMC layer plays a crucial role in the mechanical, chemical, and electrical properties of the soldering.

[0003] In current soldering quality inspections, for the detection of IMC solder joint defects, mainly non-destructive and destructive detection methods are used. However, the quality inspection process is usually at the end of the production line, and the solder joint defects are often detected after soldering is completed, which leads to obvious lag and it is difficult to predict in real time and online the possible quality problems that may occur in the solder joints during the soldering process.

[0004] Traditional soldering processes have deficiencies in the utilization of real-time data. There is a lack of in-depth internal quantitative correlation analysis for IMC thickness data and real-time data of the process. This situation makes it difficult to conduct pre-quality prediction during the soldering process and is not conducive to designers and process personnel understanding the quality causes of solder joint defects from a quantitative perspective. Based on this, how to deeply explore the correlation between the real-time data of the electronic assembly soldering process and the IMC thickness, so as to realize the prediction of soldering reliability quality, has become a key problem to be solved urgently. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present application provides a method and system for predicting the soldering quality of PCBA based on time series data, and solves the problem that it is difficult to conduct real-time, online, accurate, and quantitative analysis and prediction of the soldering reliability of Printed Circuit Board Assembly (PCBA).

[0006] To achieve the above objectives, the present application is realized through the following technical solutions:

[0007] First aspect, an embodiment of the present application provides a method for predicting the welding quality of PCBA based on time series data. The welding quality prediction method includes: obtaining the process time series data during the printed circuit board assembly (PCBA) welding process and the detection data information characterizing the thickness of the intermetallic compound; respectively performing missing value filling and data normalization processing on the process time series data and the detection data information to obtain a first data set and a second data set; quantifying the correlation degree between each feature data in the first data set and the thickness of the intermetallic compound to determine more than two key features to screen the first data set and obtain a third data set; constructing an MPSO-LSTM prediction model and performing model training to obtain an initial model; obtaining the actual products of the same batch as the products corresponding to the detection data information, randomly extracting samples to detect the thickness data of the intermetallic compound, evaluating the accuracy of the initial model to adjust the hyperparameters of the initial model, and obtaining a quality prediction model; analyzing the real-time process data during the printed circuit board assembly (PCBA) welding process based on the quality prediction model to obtain welding quality prediction information.

[0008] According to the first aspect of the embodiment of the present application, the process time series data includes multiple feature data corresponding to multiple welding indicators; the multiple welding indicators include: working date, welding equipment operating status, welding product batch, welding time, welding temperature, wire length, wire feeding speed, welding speed, laser power, and spot diameter; the detection data information corresponds to either the average thickness of the intermetallic compound layer at the hole wall or the average thickness of the intermetallic compound layer at the lead.

[0009] According to the first aspect of the embodiment of the present application, during the process of missing value filling, the null value NAN is filled with 0, and the missing part of the non-null value is filled by mean imputation; during the training and testing process of the MPSO-LSTM prediction model, the mean absolute percentage error (MAPE), mean absolute error (MAE), and root mean square error (RMSE) are used for accuracy evaluation.

[0010] According to the first aspect of the embodiment of the present application, the foregoing quantification of the correlation degree between each feature data in the first data set and the thickness of the intermetallic compound to determine more than two key features to screen the first data set and obtain a third data set may specifically include the following steps: based on the first data set and the second data set, calculate the mutual information value between the feature data corresponding to each welding indicator in the welding process and the thickness of the intermetallic compound; sort the mutual information values from high to low, determine the welding indicators corresponding to the mutual information values greater than or equal to the preset threshold as key features, and determine the welding indicators corresponding to the mutual information values lower than the preset threshold as the target indicators to be deleted; delete the data corresponding to the target indicators from the first data set to obtain a third data set.

[0011] According to the first aspect of the embodiments of the present application, the steps of constructing the MPSO-LSTM prediction model and training the model to obtain the initial model may specifically include the following steps: dividing the second dataset and the third dataset into a training dataset and a test dataset according to a preset ratio; constructing an MPSO-LSTM prediction model based on the improved PSO particle swarm algorithm and the long short-term memory network LSTM, where the MPSO-LSTM prediction model includes an input layer, a hidden layer, and an output layer corresponding to the long short-term memory network LSTM; setting a dropout layer between the hidden layer and the output layer to introduce a dropout strategy in the long short-term memory network LSTM to suppress overfitting, and adding a ReLU activation layer after the output layer; performing model training and testing based on the training dataset and the test dataset, and iteratively optimizing the number of neurons and the dropout value of the long short-term memory network LSTM through the improved PSO particle swarm algorithm to determine the initial model.

[0012] According to the first aspect of the embodiments of the present application, the process of updating the velocity and position of the improved PSO particle swarm algorithm satisfies the expression:

[0013]

[0014]

[0015] where is the movement velocity of the i th particle in the (k + 1)th generation, is the movement velocity of the i th particle in the kth generation; is the inertia weight; is the local optimal value in the kth generation, is the global optimal value in the kth generation; 、 are learning factors; 、 are random numbers generated between 0 and 1; is the position of the i th particle in the kth generation, is the current position of the i th particle in the (k + 1)th generation.

[0016] According to the first aspect of the embodiments of the present application, the calculation of the inertia weight satisfies the expression:

[0017]

[0018] where represents the objective function, represents the objective function value of the i th particle;

[0019] represents the minimum value in the historical values of the objective function;

[0020] represents the average value of the historical values of the objective function;

[0021] represents the maximum weight; represents the minimum weight;

[0022] The objective function is a fitness function and satisfies the expression:

[0023]

[0024] wherein, is the number of iterations, is the prediction accuracy.

[0025] In a second aspect, an embodiment of the present application provides a PCBA welding quality prediction system based on time series data. The welding quality prediction system includes an acquisition module, a processing module, a quantization and screening module, a training module, an adjustment module, and an analysis and prediction module.

[0026] Specifically, the acquisition module is used to acquire the process time series data during the printed circuit board assembly (PCBA) welding process and the detection data information characterizing the thickness of the intermetallic compound; the processing module is used to respectively perform missing value filling and data normalization processing on the process time series data and the detection data information to obtain a first data set and a second data set; the quantization and screening module is used to quantify the correlation degree between each feature data in the first data set and the thickness of the intermetallic compound to determine two or more key features to screen the first data set to obtain a third data set; the training module is used to construct an MPSO-LSTM prediction model and perform model training to obtain an initial model; the adjustment module is used to obtain the actual products of the same batch as the products corresponding to the detection data information, and randomly select samples to detect the thickness data of the intermetallic compound, and evaluate the accuracy of the initial model to adjust the hyperparameters of the initial model to obtain a quality prediction model; the analysis and prediction module is used to analyze the real-time process data during the printed circuit board assembly (PCBA) welding process based on the quality prediction model to obtain welding quality prediction information.

[0027] In a third aspect, an embodiment of the present application provides an electronic device, which includes: a processor, a memory, and a program stored on the memory and executable on the processor. When the program is executed by the processor, it implements the PCBA welding quality prediction method based on time series data in the foregoing first aspect.

[0028] Fourthly, an embodiment of the present application provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the method for predicting the welding quality of PCBA based on time series data in the foregoing first aspect is implemented.

[0029] The present application provides a method and system for predicting the welding quality of PCBA based on time series data. Compared with the prior art, the following beneficial effects are achieved:

[0030] The present application obtains time series data on two levels, namely the process and the thickness of intermetallic compounds, during the welding process of printed circuit board assembly (PCBA). By performing missing value completion, the data set can be made more complete in structure, providing more comprehensive information for subsequent analysis and modeling, maintaining data integrity, and avoiding data deviation. By calculating the correlation between each welding index of the process and the thickness of intermetallic compounds through the mutual information method, a third data set that is closely correlated with the thickness of intermetallic compounds is screened out from the first data set, so as to accurately analyze key features after subsequent analysis, ensuring the accuracy of quality analysis and prediction while reducing the workload. The present application effectively improves the online quality prediction ability during the welding process by constructing an MPSO-LSTM prediction model. Through model training and hyperparameter adjustment, a quality prediction model is obtained, which can perform real-time analysis on the process data during the welding process, predict the thickness of intermetallic compounds, and assist technicians in quantitatively identifying the quality causes of solder joint defects. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 It is a schematic flowchart of a method for predicting the welding quality of PCBA based on time series data provided by an embodiment of the present application;

[0033] Figure 2 is Figure 1 An exemplary flowchart of S130 in

[0034] Figure 3 It is a schematic structural diagram of a system for predicting the welding quality of PCBA based on time series data provided by an embodiment of the present application;

[0035] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed Embodiments

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described. Apparently, the described embodiments are some, rather than all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] It should be noted that, in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including", or any other variation thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or device. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0038] The embodiments of the present application provide a method and system for predicting the welding quality of PCBA based on time series data, solving the problem that it is difficult to perform real-time online accurate quantitative analysis and prediction of the welding reliability of printed circuit board assembly (PCBA); it can dynamically analyze the influence of real-time data of the process during welding on the IMC thickness of solder joints.

[0039] The technical solutions in the embodiments of the present application for solving the above technical problems are generally as follows:

[0040] Electronic assembly soldering is the core process to ensure the electrical performance reliability of PCBA. During the assembly, soldering, and debugging stages, obvious quality defects such as insufficient solder, bridging, and solder wicking can be detected by direct visual inspection. However, latent quality defects such as intermittent on / off and strong / weak fluctuations of circuit signals are mostly closely related to the thickness of the intermetallic compound (IMC) formed during the soldering process. During soldering, differences in factors such as heating duration, wetting condition, and heating temperature result in different IMC thicknesses, and the IMC layer plays a crucial role in the mechanical, chemical, and electrical properties of the soldering.

[0041] In current welding quality inspection, for the inspection of IMC solder joint defects, non-destructive and destructive inspection methods are mainly used. However, the quality inspection process is usually at the end of the production line, and solder joint defects are often detected only after welding is completed, which leads to obvious lag and it is difficult to predict in real time and online the possible quality problems that may occur in the solder joints during the welding process. Traditional welding processes have deficiencies in the utilization of real-time data. For IMC thickness data and real-time data of the process, there is a lack of in-depth internal quantitative correlation analysis. This situation makes it difficult to conduct pre-quality prediction during the welding process and is also not conducive to designers and process personnel understanding the quality causes of solder joint defects from a quantitative perspective.

[0042] To better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0043] In this embodiment, a program is written using the python software tool, the internal TensorFlow deep learning framework and the Keras deep learning tool are called to build an LSTM model and the dropout strategy is introduced, and libraries such as Sklearn, pandas, numpy, and Matplotlib are imported for data processing, calculation, and data visualization.

[0044] First, a PCBA welding quality prediction method based on time series data provided by an embodiment of the present application will be introduced below.

[0045] A schematic flow chart of a PCBA welding quality prediction method based on time series data provided by an embodiment of the present application is as Figure 1 shown, and this welding quality prediction method may include the following steps S110-S160.

[0046] S110. Obtain the process time series data during the printed circuit board assembly (PCBA) welding process and the detection data information characterizing the thickness of the intermetallic compound.

[0047] S120. Respectively perform missing value filling and data normalization processing on the process time series data and the detection data information to obtain a first data set and a second data set.

[0048] S130. Quantify the correlation degree between each feature data in the first data set and the thickness of the intermetallic compound to determine two or more key features to screen the first data set and obtain a third data set.

[0049] S140. Build an MPSO-LSTM prediction model and perform model training to obtain an initial model.

[0050] S150. Obtain the actual products of the same batch as the products corresponding to the detection data information, randomly select samples to detect the thickness data of intermetallic compounds, evaluate the accuracy of the initial model to adjust the hyperparameters of the initial model, and obtain the quality prediction model.

[0051] S160. Analyze the real-time process data during the printed circuit board assembly (PCBA) welding process based on the quality prediction model to obtain the welding quality prediction information.

[0052] The above is the specific implementation manner of a PCBA welding quality prediction method based on time series data provided by this application. It can be understood that this application obtains the time series data at two levels, namely the process data and the thickness of intermetallic compounds, during the PCBA welding process. By filling in the missing values, the dataset can be made more complete in structure, providing more comprehensive information for subsequent analysis and modeling, maintaining data integrity, and avoiding data deviation. By calculating the correlation between each welding index of the process and the thickness of intermetallic compounds through the mutual information method, the third dataset that is closely related to the thickness of intermetallic compounds is screened out from the first dataset, so as to accurately analyze the key features after subsequent analysis, ensuring the accuracy of quality analysis and prediction while reducing the workload.

[0053] Furthermore, this application effectively improves the online quality prediction ability during the welding process by constructing the MPSO-LSTM prediction model. Through model training and hyperparameter adjustment, the quality prediction model is obtained, which can analyze the process data during the welding process in real time, predict the thickness of intermetallic compounds, and assist technicians in quantitatively identifying the quality causes of solder joint defects.

[0054] It should be noted that the utilization rate of traditional welding process data is low, and there is a lack of in-depth analysis of the internal relationship between IMC thickness data and process data, making it difficult to achieve pre-welding quality prediction and help designers and process engineers quantitatively understand the quality causes of solder joint defects. This application lays a foundation for the reliable prediction and analysis of subsequent welding quality by analyzing the internal coupling relationship between key welding process feature data and the thickness of intermetallic compounds.

[0055] It should be emphasized that traditional welding quality inspection (IMC solder joint defect inspection) generally uses non-destructive and destructive inspection methods, which takes a long time. The quality inspection process is generally located at the end of the production line and mostly conducts solder joint defect inspection after the fact, with a certain lag, making it difficult to predict online the quality problems that may occur in solder joints during the welding process. Based on the process data information during the welding process and the detection data information representing the thickness of intermetallic compounds, this application extracts the key features affecting the thickness of intermetallic compounds through mutual information, and constructs a quality prediction model based on MPSO-LSTM to achieve intelligent and reliable prediction of welding quality.

[0056] In some embodiments, the aforementioned data normalization process is carried out using the maximum - minimum method, mapping all data to between 0 and 1; this data normalization process satisfies the expression:

[0057]

[0058] wherein, is the data after normalization processing; is the original data to be processed; is the minimum value in the original data; is the maximum value in the original data, which is set manually.

[0059] In some embodiments, the process - time - series data includes multiple characteristic data corresponding to multiple welding indexes; the aforementioned multiple welding indexes include: working date, welding equipment operation status, welding product batch, welding time, welding temperature, wire length, wire feeding speed, welding speed, laser power, and spot diameter; the detection data information corresponds to either the average thickness of the intermetallic compound layer at the hole wall or the average thickness of the intermetallic compound layer at the lead.

[0060] It should be noted that the process - data information can be directly called by reading the data in the welding equipment database, and the detection data information characterizing the thickness of the intermetallic compound is obtained from the detection data packet.

[0061] In some embodiments, during the process of missing - value imputation, the null value NAN is imputed to 0, and the missing part of the non - null value is imputed using the mean value; this application can extract the time and numerical value of the characteristic data of the welding process of the same batch to impute the missing values of the data.

[0062] In the embodiments of this application, it can be understood that the existence of missing values will damage the integrity of the data. By imputing the missing values, the data set can be made more complete in structure, providing more comprehensive information for subsequent analysis and modeling. If the missing values are not processed, it may lead to deviations in the distribution of certain features in the data set. For example, when calculating statistical quantities such as the mean value and standard deviation, the missing values may cause large errors in the results, thereby affecting the grasp of the overall characteristics of the data. Imputing the missing values can reduce this deviation and make the statistical characteristics of the data more truly reflect the actual situation.

[0063] In some embodiments, during the training and testing process of the MPSO-LSTM prediction model, the mean absolute percentage error (MAPE), mean absolute error (MAE), and root mean square error (RMSE) are used for accuracy evaluation. In the case where the numerical values of the three indicators, namely MAPE, MAE, and RMSE, cannot simultaneously meet the model accuracy requirements, the number of neurons and the dropout value are further optimized, and the parameters of the LSTM are dynamically adjusted.

[0064] Exemplarily, the calculation method of the mean absolute percentage error (MAPE) satisfies the expression:

[0065]

[0066] The calculation method of the mean absolute error satisfies the expression:

[0067]

[0068] The calculation method of the root mean square error satisfies the expression:

[0069]

[0070] Wherein, represents the mean absolute percentage error value, represents the mean absolute error value, represents the root mean square error value, is the number of samples in the test set, ; is the actual measured value; is the model predicted value.

[0071] In some embodiments, please refer to Figure 2 , the foregoing quantifies the correlation between each feature data in the first dataset and the thickness of the intermetallic compound to determine two or more key features to screen the first dataset, and obtain the third dataset. That is, the foregoing S130 may specifically include the following steps:

[0072] S210. Based on the first dataset and the second dataset, calculate the mutual information value between the feature data corresponding to each welding index in the welding process and the thickness of the intermetallic compound.

[0073] S220. Sort the mutual information values from high to low, determine the welding indexes corresponding to the mutual information values greater than or equal to the preset threshold as key features, and determine the welding indexes corresponding to the mutual information values lower than the preset threshold as target indexes to be deleted.

[0074] S230. Delete the data corresponding to the target indexes from the first dataset to obtain the third dataset.

[0075] In the embodiments of the present application, it can be understood that since there are both linear correlations and non-linear correlations among multiple characteristic data in the welding process, it is difficult to solve using a single linear correlation or non-linear correlation analysis method. Therefore, in this embodiment, the mutual information method is used to extract the key characteristics affecting the thickness of the intermetallic compound.

[0076] In specific implementation, the present application calls the metric.mutual_info_score() function in the sklearn library inside python to calculate the mutual information values between the characteristic data corresponding to welding indexes such as welding date, welding equipment operation status, welding product batch, welding time, welding temperature, wire length, wire feeding speed, welding speed, etc. and the thickness of the intermetallic compound, and filters through a preset threshold to retain the key characteristics affecting the thickness of the intermetallic compound.

[0077] In some embodiments, constructing the MPSO-LSTM prediction model and performing model training to obtain the initial model, that is, the aforementioned S140 may specifically include the following steps:

[0078] S310: Divide the second data set and the third data set into a training data set and a test data set according to a preset ratio.

[0079] S320: Based on the improved PSO particle swarm algorithm and the long short-term memory network LSTM, construct the MPSO-LSTM prediction model, where the MPSO-LSTM prediction model includes an input layer, a hidden layer, and an output layer corresponding to the long short-term memory network LSTM.

[0080] S330: Set a dropout layer between the hidden layer and the output layer to introduce the dropout strategy in the long short-term memory network LSTM to suppress overfitting, and add a ReLU activation layer after the output layer.

[0081] S340: Based on the training data set and the test data set, perform model training and testing, and iteratively optimize the number of neurons and the dropout value of the long short-term memory network LSTM through the improved PSO particle swarm algorithm to determine the initial model.

[0082] In the embodiments of the present application, it can be understood that the present application can import the LSTM class and the Dropout class in the keras library of the TensorFlow deep learning framework to construct the LSTM model. To prevent overfitting, a dropout layer mechanism is added to the LSTM network to randomly discard a part of the neuron cells; the present application can define the LSTM network through Sequential, build a hidden layer and a fully connected layer, and use the Adam optimizer to train the model, and finally input the prediction result through the predict function.

[0083] It should be noted that since the number of neurons, learning rate, and dropout value of the LSTM affect the prediction accuracy of the LSTM, and the setting of the above parameters mostly depends on empirical settings, therefore, this application uses an improved PSO particle swarm algorithm to iteratively optimize the number of neurons and dropout value of the LSTM; when the fitness function meets the requirements, that is, the iteration terminates, the optimal parameters of the number of neurons and dropout value of the LSTM are output.

[0084] In one example, the process of updating the velocity and position of the improved PSO particle swarm algorithm satisfies the expression:

[0085]

[0086]

[0087] Among them, is the movement velocity of the i th particle in the (k + 1)th generation, is the movement velocity of the i th particle in the kth generation; is the inertia weight; is the local optimal value in the kth generation, is the global optimal value in the kth generation; 、 is the learning factor; 、 is a randomly generated number between 0 and 1; is the position of the i th particle in the kth generation, is the current position of the i th particle in the (k + 1)th generation.

[0088] In one example, the calculation of the inertia weight satisfies the expression:

[0089]

[0090] Among them, represents the objective function, represents the objective function value of the i th particle;

[0091] represents the minimum value in the historical values of the objective function;

[0092] represents the average value of the historical values of the objective function;

[0093] represents the maximum weight; represents the minimum weight;​

[0094] The objective function is a fitness function and satisfies the expression:

[0095]

[0096] wherein, is the number of iterations, is the prediction accuracy.

[0097] In some embodiments, the present application provides a PCBA welding quality prediction system 400 based on time series data, as Figure 3 shown, the system may include the following modules:

[0098] An acquisition module 410, configured to acquire process time series data during the printed circuit board assembly (PCBA) welding process, and detection data information characterizing the thickness of the intermetallic compound;

[0099] A processing module 420, configured to perform missing value filling and data normalization processing on the process time series data and the detection data information respectively, to obtain a first data set and a second data set;

[0100] A quantization and screening module 430, configured to quantify the correlation degree between each feature data in the first data set and the thickness of the intermetallic compound, to determine more than two key features to screen the first data set, and obtain a third data set;

[0101] A training module 440, configured to construct an MPSO-LSTM prediction model and perform model training to obtain an initial model;

[0102] An adjustment module 450, configured to obtain actual products of the same batch as the products corresponding to the detection data information, and randomly select samples to detect the thickness data of the intermetallic compound, evaluate the accuracy of the initial model to adjust the hyperparameters of the initial model, and obtain a quality prediction model;

[0103] An analysis and prediction module 460, configured to analyze the real-time process data during the printed circuit board assembly (PCBA) welding process based on the quality prediction model, and obtain welding quality prediction information.

[0104] According to the embodiments of the present application, any multiple of the acquisition module 410, the processing module 420, the quantization and screening module 430, the training module 440, the adjustment module 450, and the analysis and prediction module 460 may be combined and implemented in one module, or any one of them may be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module.

[0105] In some embodiments, the quantization and screening module 430 may specifically be configured to:

[0106] Based on the first data set and the second data set, calculate the mutual information value between the characteristic data corresponding to each welding index in the welding process and the thickness of the intermetallic compound;

[0107] Sort the mutual information values from high to low, determine the welding indexes corresponding to the mutual information values greater than or equal to the preset threshold as key features, and determine the welding indexes corresponding to the mutual information values lower than the preset threshold as target indexes to be deleted;

[0108] Delete the data corresponding to the target indexes from the first data set to obtain a third data set.

[0109] In some embodiments, the training module 440 may specifically be configured to:

[0110] Divide the second data set and the third data set into a training data set and a test data set according to a preset ratio;

[0111] Based on the improved PSO particle swarm optimization algorithm and the long short-term memory network LSTM, construct an MPSO-LSTM prediction model, where the MPSO-LSTM prediction model includes an input layer, a hidden layer, and an output layer corresponding to the long short-term memory network LSTM;

[0112] Set a dropout layer between the hidden layer and the output layer to introduce a dropout strategy in the long short-term memory network LSTM to suppress overfitting, and add a ReLU activation layer after the output layer;

[0113] Based on the training data set and the test data set, perform model training and testing, and iteratively optimize the number of neurons and the dropout value of the long short-term memory network LSTM through the improved PSO particle swarm optimization algorithm to determine the initial model.

[0114] Figure 3 Each module in the system shown has the function of implementing each step in the foregoing PCBA welding quality prediction method based on time series data and can achieve its corresponding technical effects. For the sake of brevity, it will not be described herein again.

[0115] In some embodiments, the present application provides an electronic device, and the schematic structural diagram of the electronic device is as Figure 4 shown.

[0116] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0117] Specifically, the above-mentioned processor 510 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits for implementing the embodiments of the present application.

[0118] The memory 520 may include a mass storage for data or instructions. By way of example and not limitation, the memory 520 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disc, a magneto-optical disc, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. In a suitable case, the memory 520 may include a removable or non-removable (or fixed) medium. In a suitable case, the memory 520 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 520 is a non-volatile solid state memory.

[0119] The memory 520 may include a read only memory (ROM), a random access memory (RAM), a magnetic disk storage media device, an optical storage media device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Thus, generally, the memory 520 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in any of the above-mentioned PCBA welding quality prediction methods based on time series data in the embodiments.

[0120] The processor 510 reads and executes the computer program instructions stored in the memory 520 to implement any of the above-mentioned PCBA welding quality prediction methods based on time series data in the embodiments.

[0121] In one example, the electronic device may further include a communication interface 530 and a bus 500. Among them, as Figure 4 shown, the processor 510, the memory 520, and the communication interface 530 are connected through the bus 500 and complete communication with each other.

[0122] The communication interface 530 is mainly used to implement communication between various modules, devices, units, and / or devices in the embodiments of the present application.

[0123] The bus 500 includes hardware, software, or both, and couples the components of the online data flow metering device to each other. By way of example and not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, the bus 500 may include one or more buses. Although the embodiments of the present application describe and illustrate specific buses, the present application contemplates any suitable bus or interconnect.

[0124] In addition, in combination with the PCBA welding quality prediction method based on time series data in the above embodiments, the embodiments of the present application can be implemented by providing a computer storage medium. Computer program instructions are stored on the computer storage medium; when the computer program instructions are executed by a processor, any one of the PCBA welding quality prediction methods based on time series data in the above embodiments is implemented.

[0125] It should be clear that the present application is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present application is not limited to the specific steps described and illustrated, and those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present application.

[0126] The functional blocks shown in the above structural block diagrams can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), appropriate firmware, a plug-in, a functional card, and so on. When implemented in software, the elements of the present application are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted via a data signal carried in a carrier wave on a transmission medium or a communication link. A "machine-readable medium" can include any medium that can store or transmit information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, Erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, Radio Frequency (RF) links, and so on. The code segment can be downloaded via a computer network such as the Internet, an intranet, and so on.

[0127] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps. That is, the steps can be executed in the order mentioned in the embodiments, or different from the order in the embodiments, or several steps can be executed simultaneously.

[0128] As described above with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block in the flowchart and / or block diagram, and the combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the functions / actions specified in one or more blocks of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It is also understood that each block in the block diagram and / or flowchart, and the combinations of blocks in the block diagram and / or flowchart, can also be implemented by dedicated hardware that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0129] In summary, compared with the prior art, this application has the following beneficial effects:

[0130] 1. This application obtains the timing data at two levels, namely the process and the thickness of the intermetallic compound, during the welding process of the printed circuit board assembly (PCBA); calculates the correlation between each welding index of the process and the thickness of the intermetallic compound through the mutual information method, and screens out the third data set that is closely related to the thickness of the intermetallic compound from the first data set, so as to accurately analyze the key features after subsequent analysis, ensuring the accuracy of quality analysis and prediction while reducing the workload; by analyzing the internal coupling relationship between the key welding process feature data and the thickness of the intermetallic compound, it can lay a foundation for the reliability prediction analysis of the subsequent welding quality.

[0131] 2. This application effectively improves the online quality prediction ability during the welding process by constructing an MPSO-LSTM prediction model and introducing a Dropout layer mechanism. Through model training and hyperparameter adjustment, a quality prediction model is obtained, which can perform real-time analysis on the process data during the welding process, predict the thickness of the intermetallic compound, and assist designers and process personnel in quantitatively identifying the causes of solder joint defect quality.

[0132] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A PCBA welding quality prediction method based on time series data, characterized in that: include: Obtaining the process time series data of the printed circuit board assembly (PCBA) welding process, as well as the detection data information characterizing the thickness of the intermetallic compounds; Performing missing value completion and data normalization processing on the process time series data and the detection data information respectively to obtain a first data set and a second data set; quantifying the correlation between each characteristic data in the first data set and the thickness of the intermetallic compound to determine two or more key characteristics to screen the first data set to obtain a third data set; Construct the MPSO-LSTM prediction model and perform model training to obtain the initial model; Acquire actual products from the same batch as the products corresponding to the detection data information, randomly select samples to detect intermetallic compound thickness data, evaluate the accuracy of the initial model to adjust the hyperparameters of the initial model, and obtain a quality prediction model; Based on the quality prediction model, real-time process data in the printed circuit board assembly (PCBA) welding process is analyzed to obtain welding quality prediction information.

2. The PCBA welding quality prediction method based on time series data according to claim 1, characterized in that: The process time series data includes multiple feature data corresponding to multiple welding indicators; the multiple welding indicators include: working date, welding equipment operating status, welding product batch, welding time, welding temperature, wire length, wire feeding speed, welding speed, laser power and spot diameter; the detection data information corresponds to one of the average thickness of the intermetallic compound layer at the hole wall and the average thickness of the intermetallic compound layer at the lead.

3. The PCBA welding quality prediction method based on time series data according to claim 1, characterized in that: In the process of missing value filling, the null value NAN is filled with 0, and the missing part of the non-null value is interpolated by the mean value; During the training and testing process of the MPSO-LSTM prediction model, the mean absolute percentage error (MAPE), the mean absolute error (MAE), and the normalized mean square error (RMSE) were used for accuracy evaluation.

4. The PCBA welding quality prediction method based on time series data according to any one of claims 1 to 3, characterized in that: The quantifying the correlation between each characteristic data in the first data set and the thickness of the intermetallic compound to determine two or more key features to screen the first data set, and obtaining a third data set includes: Based on the first data set and the second data set, calculating the mutual information value between the characteristic data corresponding to each welding index of the welding process and the thickness of the intermetallic compound; Sort the mutual information values ​​from high to low, determine the welding indicators corresponding to the mutual information values ​​greater than or equal to the preset threshold as key features, and determine the welding indicators corresponding to the mutual information values ​​lower than the preset threshold as target indicators to be deleted; The data corresponding to the target indicator is deleted from the first data set to obtain a third data set.

5. The PCBA welding quality prediction method based on time series data according to any one of claims 1 to 3, characterized in that: The construction of the MPSO-LSTM prediction model and the model training to obtain the initial model include: Dividing the second data set and the third data set into a training data set and a test data set according to a preset ratio; Based on the improved PSO particle swarm algorithm and the long short-term memory network LSTM, an MPSO-LSTM prediction model is constructed, wherein the MPSO-LSTM prediction model includes an input layer, a hidden layer and an output layer corresponding to the long short-term memory network LSTM; A dropout layer is set between the hidden layer and the output layer to introduce a dropout strategy in the long short-term memory network LSTM to suppress overfitting, and a ReLU activation layer is added after the output layer; Model training and testing are performed based on the training data set and the test data set, and the number of neurons and dropout value of the long short-term memory network LSTM are iteratively optimized through the improved PSO particle swarm algorithm to determine the initial model.

6. The PCBA welding quality prediction method based on time series data according to claim 5, characterized in that: The process of updating speed and position of the improved PSO particle swarm algorithm satisfies the expression: in, is the k+1th generation i The speed of a particle, is the kth generation i The speed of a particle; is the inertia weight; is the local optimal value of the kth generation, is the global optimal value of the kth generation; 、 is the learning factor; 、 A randomly generated number between 0 and 1; is the kth generation i The position of a particle, is the k+1th generation i The current position of a particle.

7. The PCBA welding quality prediction method based on time series data according to claim 6, characterized in that: The calculation of the inertia weight satisfies the expression: in, represents the objective function, Indicates i The objective function value of each particle; Represents the minimum value among the historical values ​​of the objective function; Represents the average value of the historical value of the objective function; represents the maximum weight; represents the minimum weight; The objective function is a fitness function and satisfies the expression: In the formula, is the number of iterations, For prediction accuracy.

8. A PCBA welding quality prediction system based on time series data, characterized in that: include: An acquisition module is used to acquire the process time series data during the soldering process of the printed circuit board assembly (PCBA) and the detection data information characterizing the thickness of the intermetallic compound; A processing module, used to perform missing value completion and data normalization processing on the process time series data and the detection data information respectively to obtain a first data set and a second data set; A quantitative screening module, used for quantifying the correlation between each characteristic data in the first data set and the thickness of the intermetallic compound, so as to determine two or more key characteristics to screen the first data set to obtain a third data set; The training module is used to build the MPSO-LSTM prediction model and perform model training to obtain the initial model; An adjustment module is used to obtain actual products from the same batch as the products corresponding to the detection data information, and randomly select samples to detect the intermetallic compound thickness data, evaluate the accuracy of the initial model to adjust the hyperparameters of the initial model, and obtain a quality prediction model; The analysis and prediction module is used to analyze the real-time process data in the printed circuit board assembly (PCBA) welding process based on the quality prediction model to obtain welding quality prediction information.

9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the method for predicting PCBA welding quality based on time series data as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the PCBA welding quality prediction method based on time series data as described in any one of claims 1 to 7 is implemented.

Images