Method, system and device for optimizing technological parameters of FCBGA chip carrier plate and medium

Abstract

The invention provides an FCBGA chip support plate process parameter optimization method, system and device and a medium, and belongs to the technical field of PCB production, the method comprises the following steps: collecting and preprocessing a historical data set in the FCBGA chip support plate production process, and screening data subsets with correlation greater than a preset threshold based on target features of a to-be-optimized production task; training a machine learning prediction model capable of mapping the process parameters to the quality indexes; constructing a multi-objective optimization function by taking a machine learning prediction model as a core; a multi-objective optimization algorithm is adopted to search in the process parameter constraint space, and a Pareto optimal solution set is solved; and selecting a final optimal process parameter according to a preset decision strategy, applying the final optimal process parameter to an actual production line of the FCBGA chip carrier plate, then collecting actual production result data, and feeding back the actual production result data to a historical data set. The method is based on historical data machine learning multi-objective optimization, incremental learning dynamically adapts to production line changes, and quality and efficiency are improved.

Description

Technical Field

[0001] This application belongs to the field of PCB manufacturing technology, specifically relating to a method, system, equipment, and medium for optimizing process parameters of an FCBGA chip carrier board. Background Technology

[0002] FCBGA, short for Flip Chip Ball Grid Array, is a high-density integrated circuit packaging technology widely used in high-performance chips. During the production of FCBGA chip substrates, the setting of process parameters (such as reflow soldering temperature, pressure, and time) directly affects packaging quality, yield, and reliability. Traditional process parameter optimization methods mainly rely on engineer experience and trial-and-error, which are inefficient and difficult to achieve global optimization. In existing technologies, Statistical Process Control (SPC) and Design of Experiments (DOE) are used for parameter optimization, but these methods require a large amount of experimental data, are costly, and cannot fully utilize historical production data for systematic comparative analysis.

[0003] With the development of industrial big data, some data-driven methods have been proposed, such as using regression analysis or machine learning models to predict quality indicators. However, these methods often lack effective comparison and similarity analysis of historical data, resulting in insufficient model accuracy and limited optimization effects. In addition, existing methods mostly focus on single-objective optimization, making it difficult to optimize multiple quality indicators (such as yield and defect rate) simultaneously, and they lack real-time feedback mechanisms, failing to adapt to dynamic changes in the production process.

[0004] Therefore, there is an urgent need for a method to optimize the process parameters of FCBGA chip carrier based on historical production data comparison, in order to overcome the shortcomings of existing technologies and improve production quality and efficiency. Summary of the Invention

[0005] In a first aspect, embodiments of this application provide a method for optimizing process parameters of an FCBGA chip carrier board, comprising the following steps: S1. Collect historical datasets from the FCBGA chip carrier board production process, including historical process parameters and corresponding historical quality indicators; S2. The historical dataset is preprocessed by cleaning and standardizing, and based on the target features of the current production task to be optimized, a subset of data with a correlation greater than a preset threshold is selected from the preprocessed historical dataset using a similarity measurement algorithm. S3. Train a machine learning prediction model using the selected subset of data, the machine learning prediction model being able to map process parameters to quality indicators; S4. Using a machine learning prediction model as the core, construct a multi-objective optimization function, which includes at least two quality indicators as optimization objectives; use a multi-objective optimization algorithm to search within the process parameter constraint space to find the Pareto optimal solution set; S5. Select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision-making strategy, and apply them to the actual production line of FCBGA chip carrier. S6. After applying the final optimal process parameters for production, collect actual production result data and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and realize dynamic updating of the model.

[0006] Furthermore, the specific steps of step S1 are as follows: S11. Extract historical process parameters from the manufacturing execution system, the historical process parameters including at least the set temperature of each temperature zone of the reflow oven, the conveyor belt speed, the pressure and duration of the lamination process, and the stencil thickness for solder paste printing; S12. Extract the corresponding historical quality indicators from the quality inspection system. The historical quality indicators include at least the solder ball void rate detected by X-ray, the bridging and cold solder defect rate detected by automatic optical inspection, and the final warpage of the product. S13. Associate and align the extracted process parameters and quality indicators by production batch to generate a structured historical dataset.

[0007] Furthermore, the specific steps of step S2 are as follows: S21. The 3σ principle is used to identify and remove abnormal temperature records in each temperature zone, and the K-nearest neighbor algorithm is used to interpolate the missing compression pressure data. S22. Based on the target features of the current production task to be optimized, perform similarity screening and calculate the cosine similarity between the target features and the features of each batch of products in the historical dataset; the target features include the chip size, number of packaging layers and substrate material of the currently produced chip; The cosine similarity formula is as follows:

[0008] in, The feature vector of the current task. The feature vectors of historical batches; S23. Filter out historical data with similarity greater than a preset threshold and generate a data subset for subsequent model training.

[0009] Furthermore, the specific steps of step S3 are as follows: S31. Construct a fully connected neural network as a machine learning prediction model. The input layer nodes of the fully connected neural network correspond to the screened process parameters, including peak temperature, time above the liquidus line, and pressing pressure. The output layer nodes of the fully connected neural network correspond to the quality indicators to be predicted, including solder ball void rate and warpage. S32. Train the machine learning prediction model using a subset of data, calculate the predicted values ​​using forward propagation, and perform backpropagation using the mean squared error as the loss function to adjust the network weights. The loss function formula is:

[0010] Where N is the number of samples, and These are the actual and predicted values ​​of the void ratio of the i-th sample solder ball, respectively. and These are the actual and predicted values ​​of the warp of the i-th sample, respectively; S33. After the loss function converges on the validation set, save the model weights to obtain the trained process parameter-quality index prediction model.

[0011] Furthermore, the specific steps of step S4 are as follows: S41. Using a machine learning prediction model as the evaluator, a bi-objective optimization problem is established that simultaneously minimizes the solder ball void rate and minimizes the warpage, mathematically expressed as:

[0012] in, For process parameter combinations, and These are the predicted values ​​of solder ball void rate and warpage obtained from the machine learning prediction model, respectively. S42. Set the constraint space for process parameters: The peak temperature of reflow soldering is limited to a preset temperature range, and the time above the liquidus is limited to a preset duration. S43. A non-dominated sorting genetic algorithm is used to search within the process parameter constraint space. Through selection, crossover, and mutation operations in the iteration, a set of non-dominated Pareto optimal solutions is generated. The Pareto optimal solution set represents the set of optimal process parameter combinations that balance the solder ball void ratio and warpage.

[0013] Furthermore, the specific steps of step S5 are as follows: S51. Assign weight coefficients to the two quality indicators, solder ball void rate and warpage, based on the priority of the current production task. Specifically, when the product is used in a high-performance computing scenario, assign a higher weight to the solder ball void rate; when the product is used in a large-size, low-cost scenario, assign a higher weight to the warpage. S52. Calculate the overall utility value of each Pareto solution in the Pareto optimal solution set using the weighted normalization method:

[0014] in, A larger summation utility value for a single Pareto solution indicates a better overall quality performance for the corresponding set of process parameters. and These are the weighting coefficients for the porosity and warpage of the solder balls, respectively. ; and These are the predicted values ​​of solder ball void ratio and warpage corresponding to the Pareto solution, calculated by the process parameter-quality index prediction model. and These are the maximum and minimum predicted values ​​of the solder ball void rate for all solutions in the current Pareto optimal solution set, respectively. and These are the maximum and minimum predicted warpage values ​​for all solutions in the current Pareto optimal solution set, respectively. S53. Select the solution with the best overall utility value as the final optimal combination of process parameters and send it to the production line control system for execution.

[0015] Furthermore, step S6 is detailed as follows: S61. After production using the new combination of process parameters, the actual solder paste volume and solder ball void rate of the corresponding new batch of products are collected by online SPI and X-ray inspection equipment as new actual production result data. S62. The validity of the collected actual production result data is determined. The current batch of data is considered valid if both of the following conditions are met simultaneously: For the current production batch, all types of process parameters and quality index data necessary for model training and validation have been successfully collected; The control chart for the current batch production process shows that all process parameters are under statistical control, and the first pass rate of the final product of the current batch is higher than the preset qualified threshold. S63. After accumulating K new production batches of data, the incremental learning process is automatically triggered. The machine learning prediction model is fine-tuned using a mixture of new and old historical datasets, and the network weights are updated so that the model can continuously adapt to changes in the production line status.

[0016] Secondly, embodiments of this application also provide an FCBGA chip carrier board process parameter optimization system, comprising: The data collection module is used to collect historical datasets during the production process of FCBGA chip carrier boards. The historical datasets include historical process parameters and corresponding historical quality indicators. The data preprocessing and filtering module is used to clean and standardize the historical dataset, and based on the target characteristics of the current production task to be optimized, it uses a similarity measurement algorithm to filter out a subset of data with a correlation greater than a preset threshold from the preprocessed historical dataset. The prediction model training module is used to train a machine learning prediction model using a selected subset of data. The machine learning prediction model can map process parameters to quality indicators. The multi-objective optimization module is used to construct a multi-objective optimization function based on a machine learning prediction model. The multi-objective optimization function includes at least two quality indicators as optimization objectives. The multi-objective optimization algorithm is used to search within the process parameter constraint space to find the Pareto optimal solution set. The decision and application module is used to select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision strategy and apply them to the actual production line of FCBGA chip carrier. The feedback update module is used to collect actual production result data after production is carried out using the final optimal process parameters, and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and realize dynamic updates of the model.

[0017] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the FCBGA chip carrier process parameter optimization method as described in the first aspect.

[0018] Fourthly, embodiments of this application also provide a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps of the FCBGA chip carrier board process parameter optimization method as described in the first aspect.

[0019] As can be seen from the above technical solutions, this application has the following advantages: The FCBGA chip carrier process parameter optimization method, system, equipment, and medium provided in this application systematically collect and process historical data and establish machine learning models, completely changing the traditional trial-and-error mode that relies on engineer experience, thus improving the efficiency and accuracy of process parameter optimization. It introduces product feature-based similarity screening to ensure that the data used to train the model is highly relevant to the current production task, improving the accuracy and generalization ability of the prediction model. It employs a multi-objective optimization algorithm to solve for the Pareto optimal solution set, finding the best balance between conflicting quality indicators and achieving optimal overall quality. By collecting production data in real time and triggering incremental model learning, it can continuously adapt to changes in the production process, achieving process quality control. It can quickly determine the optimal combination of process parameters, reducing product defect rates, increasing production yield, and simultaneously reducing the number of experimental debugging sessions, saving production costs. Attached Figure Description

[0020] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the method for optimizing the process parameters of the FCBGA chip carrier board according to the present invention.

[0022] Figure 2 This is a schematic diagram of the FCBGA chip carrier board process parameter optimization system of the present invention. Detailed Implementation

[0023] The various embodiments of this disclosure will be described more fully in the detailed steps of the FCBGA chip carrier process parameter optimization method described below. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0024] This embodiment provides a method for optimizing process parameters of FCBGA chip carrier boards. It uses historical data to train a machine learning model, solves the Pareto optimal solution through a multi-objective genetic algorithm, and combines scenario weight decision-making to achieve synergistic optimization of quality and efficiency, and incrementally learns to dynamically respond to production line drift.

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 The diagram shows a flowchart of a method for optimizing the process parameters of an FCBGA chip carrier board in a specific embodiment. The method includes the following steps: S1. Collect historical datasets from the FCBGA chip carrier board production process, including historical process parameters and corresponding historical quality indicators; It should be noted that by integrating process parameters from reflow soldering, lamination, solder paste printing, and other processes, as well as quality data from X-ray and optical inspection, and performing structured correlation by production batch, a complete historical dataset is formed, providing a data foundation for modeling and analysis and ensuring the integrity of model input. S2. The historical dataset is preprocessed by cleaning and standardizing, and based on the target features of the current production task to be optimized, a subset of data with a correlation greater than a preset threshold is selected from the preprocessed historical dataset using a similarity measurement algorithm. It should be noted that by ensuring data quality through data preprocessing and then focusing on relevant data through similarity filtering, the efficiency and accuracy of subsequent model training are improved. S3. Train a machine learning prediction model using the selected subset of data, the machine learning prediction model being able to map process parameters to quality indicators; It should be noted that a precise mapping relationship between process parameters and quality indicators was established through machine learning models, providing a reasonable prediction method for parameter optimization; S4. Using a machine learning prediction model as the core, construct a multi-objective optimization function, which includes at least two quality indicators as optimization objectives; use a multi-objective optimization algorithm to search within the process parameter constraint space to find the Pareto optimal solution set; It should be noted that the multi-objective optimization method resolves the contradictions between quality indicators, provides a variety of feasible optimization solutions, and offers ample choice for decision-making. S5. Select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision-making strategy, and apply them to the actual production line of FCBGA chip carrier. It should be noted that this step enables quantitative comparison of multi-objective decision-making. By transforming the Pareto front into a single optimal solution, it facilitates direct selection by engineers and enables the selected parameters to be sent to the production line control system for execution. S6. After applying the final optimal process parameters for production, collect actual production result data and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and realize dynamic updating of the model. It should be noted that this step enables the machine learning prediction model to gradually adapt to changes in the production line status, updating weights while retaining historical knowledge, and balancing the model's timeliness and stability.

[0027] This embodiment realizes intelligent management of the entire lifecycle from data collection to model application and self-updating.

[0028] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process in this embodiment, another method for optimizing FCBGA chip carrier board process parameters is provided. Taking the FCBGA chip carrier board production line of an electronics manufacturing company as an application scenario, which mainly produces two types of FCBGA products for high-performance computing (HPC) and consumer electronics as an example, the method includes the following steps: S1. Collect historical datasets from the FCBGA chip carrier board production process. These historical datasets include historical process parameters and corresponding historical quality indicators. The specific steps of step S1 are as follows: S11. Extract historical process parameters from the manufacturing execution system, the historical process parameters including at least the set temperature of each temperature zone of the reflow oven, the conveyor belt speed, the pressure and duration of the lamination process, and the stencil thickness for solder paste printing; For example, process parameters for 300 production batches from the past 12 months are extracted from the Manufacturing Execution System (MES). Example parameters are shown below: Reflow oven temperature settings: preheating zone 150℃, constant temperature zone 180℃, peak temperature zone 235℃, cooling zone 100℃; Conveyor belt speed: 50cm / min; Pressing process: pressure 18MPa, duration 90s; Solder paste printing stencil thickness: 0.12mm; S12. Extract the corresponding historical quality indicators from the quality inspection system. The historical quality indicators include at least the solder ball void rate detected by X-ray, the bridging and cold solder defect rate detected by automatic optical inspection, and the final warpage of the product. For example, quality data for the corresponding batch is extracted from the Quality Inspection System (QIS). Key indicators are shown in the following examples: Solder ball void rate (X-ray inspection): average 3.2%, range 1.5%-6.8%; Bridging and cold solder joint defect rate (AOI inspection): average 0.8%, range 0.2%-2.5%; Product warpage: average 0.15mm, range 0.08mm-0.32mm; S13. Associate and align the extracted process parameters and quality indicators by production batch to generate a structured historical dataset; For example, process parameters and quality indicators are linked and aligned by production batch, forming 300 structured records. Each record contains 12 process parameters and 5 quality indicators. An example of the structured records is shown in Table 1. Table 1

[0029] S2. The historical dataset undergoes cleaning and standardization preprocessing. Based on the target characteristics of the current production task to be optimized, a subset of data with a correlation greater than a preset threshold is selected from the preprocessed historical dataset using a similarity measurement algorithm. The specific steps of step S2 are as follows: S21. The 3σ principle is used to identify and remove abnormal temperature records in each temperature zone, and the K-nearest neighbor algorithm is used to interpolate the missing compression pressure data. For example, the 3σ principle is used to remove outlier data: the reflow soldering peak temperature of batch 20240315 was found to be 280℃ (exceeding the mean ± 3σ range), and it was removed. K-Nearest Neighbor Algorithm for Imputing Missing Values: The compression pressure data for batch 20240522 was missing, and it was imputed using the average pressure of the five similar batches before and after it (17.8 MPa). S22. Based on the target features of the current production task to be optimized, perform similarity screening and calculate the cosine similarity between the target features and the features of each batch of products in the historical dataset; the target features include the chip size, number of packaging layers and substrate material of the currently produced chip; The cosine similarity formula is as follows:

[0030] in, The feature vector of the current task. The feature vectors of historical batches; For example, the current production task to be optimized is an FCBGA chip carrier board for high-performance computing, with a target feature vector T = [chip size 15mm × 15mm, 8 packaging layers, substrate material BT resin]; calculate the cosine similarity between this feature vector and the feature vector H of historical batches, as shown in the example below: Historical batch 20240218 features H1=[14.8mm×14.8mm, 8 layers, BT resin], cosine similarity 0.97; Historical batch 20240403 features H2=[12mm×12mm, 6 layers, FR-4 resin], cosine similarity 0.72; S23. Filter out historical data with similarity greater than a preset threshold (e.g., 0.85) and generate a data subset for subsequent model training; For example, a similarity preset threshold of 0.85 is set to filter out 186 highly relevant batches of data to form a subset of model training data; S3. Using the selected subset of data, train a machine learning prediction model that can map process parameters to quality indicators; the specific steps of step S3 are as follows: S31. Construct a fully connected neural network as a machine learning prediction model. The input layer nodes of the fully connected neural network correspond to the screened process parameters, including peak temperature, time above the liquidus line, and pressing pressure. The output layer nodes of the fully connected neural network correspond to the quality indicators to be predicted, including solder ball void rate and warpage. S32. Train the machine learning prediction model using a subset of data, calculate the predicted values ​​using forward propagation, and perform backpropagation using the mean squared error as the loss function to adjust the network weights. The loss function formula is:

[0031] Where N is the number of samples, and These are the actual and predicted values ​​of the void ratio of the i-th sample solder ball, respectively. and These are the actual and predicted values ​​of the warp of the i-th sample, respectively; S33. After the loss function converges on the validation set, save the model weights to obtain the trained process parameter-quality index prediction model. S4. Construct a multi-objective optimization function based on a machine learning prediction model. This multi-objective optimization function includes at least two quality indicators as optimization objectives. Use a multi-objective optimization algorithm to search within the process parameter constraint space to find the Pareto optimal solution set. The specific steps of step S4 are as follows: S41. Using a machine learning prediction model as the evaluator, a bi-objective optimization problem is established that simultaneously minimizes the solder ball void rate and minimizes the warpage, mathematically expressed as:

[0032] in, For process parameter combinations, and These are the predicted values ​​of solder ball void rate and warpage obtained from the machine learning prediction model, respectively. For example, using the trained model as the evaluator, a bi-objective optimization function is established:

[0033] Where X = [peak temperature, pressing pressure, pressing time], the objective is to simultaneously minimize the solder ball void ratio and warpage; S42. Set the constraint space for process parameters: The peak temperature of reflow soldering is limited to a preset temperature range (e.g., 220°C to 250°C), and the time above the liquidus is limited to a preset duration range (e.g., 60s to 90s). For example, the constraint space is set as follows: Peak reflow soldering temperature: 220℃-250℃; Time above the liquidus line: 60s-90s; Pressing pressure: 15MPa-20MPa; S43. A non-dominated sorting genetic algorithm is used to search within the process parameter constraint space. Through selection, crossover and mutation operations in the iteration, a set of non-dominated Pareto optimal solutions is generated. The Pareto optimal solution set represents the set of optimal process parameter combinations that balance the solder ball void rate and warpage. For example, the non-dominated sorting genetic algorithm (NSGA-II) was used for the search. After 100 generations of iteration, 20 Pareto optimal solutions were generated. Some examples are shown in Table 2: Table 2

[0034] S5. Select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision-making strategy, and apply them to the actual production line of the FCBGA chip carrier; the specific steps of step S5 are as follows: S51. Assign weight coefficients to the two quality indicators, solder ball void rate and warpage, based on the priority of the current production task. Specifically, when the product is used in a high-performance computing scenario, assign a higher weight to the solder ball void rate; when the product is used in a large-size, low-cost scenario, assign a higher weight to the warpage. For example, the current production task is a high-performance computing scenario, with an emphasis on reliability, and the following weights are assigned: Weight of solder ball void ratio =0.7, warpage weight =0.3; S52. Calculate the overall utility value of each Pareto solution in the Pareto optimal solution set using the weighted normalization method:

[0035] in, A larger summation utility value for a single Pareto solution indicates a better overall quality performance for the corresponding set of process parameters. and These are the weighting coefficients for the porosity and warpage of the solder balls, respectively. ; and These are the predicted values ​​of solder ball void ratio and warpage corresponding to the Pareto solution, calculated by the process parameter-quality index prediction model. and These are the maximum and minimum predicted values ​​of the solder ball void rate for all solutions in the current Pareto optimal solution set, respectively. and These are the maximum and minimum predicted warpage values ​​for all solutions in the current Pareto optimal solution set, respectively. For example, using three sets of data from the Pareto optimal solution set, the overall utility value (U value) is calculated: Solution 1:

[0036] Solution 2:

[0037] Solution 3:

[0038] S53. Select the solution with the best overall utility value as the final optimal combination of process parameters (e.g., specific peak temperature and pressing pressure), and send it to the production line control system for execution; For example, solution 2 with the highest U value is selected as the final optimal process parameters: peak temperature 236℃, pressing pressure 17.8MPa, and pressing time 92s, and sent to the production line control system for execution; S6. After applying the final optimal process parameters for production, collect actual production result data and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and achieve dynamic updates of the model; the specific steps of step S6 are as follows: S61. After production using the new combination of process parameters, the actual solder paste volume and solder ball void rate of the corresponding new batch of products are collected by online SPI and X-ray inspection equipment as new actual production result data. For example, after producing 10 batches with optimal parameters, data was collected via SPI and X-ray detection: Actual solder ball void rate: average 2.2% (predicted value 2.3%); Actual warpage: average 0.11 mm (predicted value 0.10 mm); Solder paste volume: average 0.08 mm³; S62. The validity of the collected actual production result data is determined. The current batch of data is considered valid if both of the following conditions are met simultaneously: For the current production batch, all types of process parameters and quality index data necessary for model training and validation have been successfully collected; The control chart (such as the Xbar-R chart) of the current batch production process shows that all process parameters are under statistical control (i.e., there is no non-random variation in the pattern), and the first pass rate of the final product of the current batch is higher than the preset pass threshold (e.g., 98.5%). For example, perform a data validity check: All necessary process parameters and quality index data have been collected; The production process control chart (Xbar-R chart) shows no abnormal fluctuations, and the first-pass yield is 99.2% (higher than the preset threshold of 98.5%), indicating that the data is valid. S63. After accumulating K new production batches of data, the incremental learning process is automatically triggered. The machine learning prediction model is fine-tuned using a mixture of new and old historical datasets, and the network weights are updated so that the model can continuously adapt to changes in the state of the production line. For example, after accumulating 10 valid new batches of data (K=10), the incremental learning process is triggered, and a total of 316 batches of data are mixed to fine-tune the original model. After the update, the prediction error of the model for the solder ball void rate is reduced to ≤0.2%.

[0039] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0040] like Figure 2 As shown, the following are embodiments of the FCBGA chip carrier process parameter optimization system provided in this disclosure. This system and the FCBGA chip carrier process parameter optimization methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the FCBGA chip carrier process parameter optimization system, please refer to the embodiments of the above FCBGA chip carrier process parameter optimization methods.

[0041] The system includes: The data collection module is used to collect historical datasets during the production process of FCBGA chip carrier boards. The historical datasets include historical process parameters and corresponding historical quality indicators. The data preprocessing and filtering module is used to clean and standardize the historical dataset, and based on the target characteristics of the current production task to be optimized, it uses a similarity measurement algorithm to filter out a subset of data with a correlation greater than a preset threshold from the preprocessed historical dataset. The prediction model training module is used to train a machine learning prediction model using a selected subset of data. The machine learning prediction model can map process parameters to quality indicators. The multi-objective optimization module is used to construct a multi-objective optimization function based on a machine learning prediction model. The multi-objective optimization function includes at least two quality indicators as optimization objectives. The multi-objective optimization algorithm is used to search within the process parameter constraint space to find the Pareto optimal solution set. The decision and application module is used to select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision strategy and apply them to the actual production line of FCBGA chip carrier. The feedback update module is used to collect actual production result data after production is carried out using the final optimal process parameters, and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and realize dynamic updates of the model.

[0042] This embodiment achieves intelligent management of the entire lifecycle from data collection to model application and self-updating through the interactive collaboration of the data collection module, data preprocessing and filtering module, multi-objective optimization module, decision-making and application module, and feedback update module.

[0043] The FCBGA chip carrier process parameter optimization method provided in this application embodiment can be applied to electronic devices. Those skilled in the art will understand that the electronic device structure involved in the embodiments of this invention does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this invention, the electronic device includes, but is not limited to, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0044] Electronic devices may include processors, external memory interfaces, internal memory, universal serial bus (USB) interfaces, charging management modules, power management modules, batteries, wireless communication modules, audio modules, speakers, microphones, sensor modules, buttons, cameras, displays, and SIM card interfaces, etc.

[0045] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the electronic device. In other embodiments of this application, the electronic device may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0046] A processor may include one or more processing units, such as a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors.

[0047] The processor can serve as the nerve center and command center of an electronic device. The controller can generate operation control signals based on the instruction opcode and timing signals to control the fetching and execution of instructions.

[0048] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from this memory. This avoids repeated accesses, reduces processor latency, and thus improves system efficiency.

[0049] The aforementioned electronic device realizes the collection of historical datasets in the FCBGA chip carrier production process for the FCBGA chip carrier process parameter optimization method of this application. The historical datasets include historical process parameters and corresponding historical quality indicators. The historical dataset undergoes cleaning and standardization preprocessing. Based on the target characteristics of the current production task to be optimized, a subset of data with relevance greater than a preset threshold is selected from the preprocessed historical dataset using a similarity measurement algorithm. A machine learning prediction model is trained using this selected subset, mapping process parameters to quality indicators. A multi-objective optimization function is constructed around the machine learning prediction model, containing at least two quality indicators as optimization objectives. A multi-objective optimization algorithm searches within the process parameter constraint space to find the Pareto optimal solution set. A set of final optimal process parameters is selected from the Pareto optimal solution set according to a preset decision strategy and applied to the actual production line of the FCBGA chip carrier. After production using the final optimal process parameters, actual production result data is collected and fed back to the historical dataset to trigger incremental learning of the machine learning prediction model, achieving dynamic model updates. This approach achieves the beneficial effects of improving data relevance through similarity screening, accurate prediction of quality indicators by the machine learning model, finding the Pareto optimal front using a multi-objective genetic algorithm, personalized parameter recommendation through scenario-based weighted decision-making, and dynamic adaptation to production line changes through incremental learning.

[0050] The storage medium provided in this application stores a program product capable of implementing a method for optimizing the process parameters of an FCBGA chip carrier board.

[0051] The XFCBGA chip substrate process parameter optimization method includes: collecting historical datasets from the FCBGA chip substrate production process, the historical datasets including historical process parameters and corresponding historical quality indicators; cleaning and standardizing the historical datasets, and based on the target features of the current production task to be optimized, selecting a subset of data with a correlation greater than a preset threshold from the preprocessed historical datasets using a similarity measurement algorithm; training a machine learning prediction model using the selected subset of data, the machine learning prediction model being able to map process parameters to quality indicators; constructing a multi-objective optimization function with the machine learning prediction model as the core, the multi-objective optimization function including at least two quality indicators as optimization objectives; using a multi-objective optimization algorithm to search within the process parameter constraint space to find the Pareto optimal solution set; selecting a set of final optimal process parameters from the Pareto optimal solution set according to a preset decision strategy, and applying them to the actual production line of the FCBGA chip substrate; after production using the final optimal process parameters, collecting actual production result data and feeding it back to the historical dataset to trigger incremental learning of the machine learning prediction model, thereby achieving dynamic updates of the model.

[0052] In some possible implementations, the FCBGA chip carrier process parameter optimization method of this disclosure can be implemented as a program product, which includes program code. When the program product is run on a terminal device, the program code is used to cause the terminal device to perform the steps described in the "Exemplary Methods" section above according to various exemplary embodiments of this disclosure.

[0053] The storage medium disclosed herein may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0054] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for optimizing process parameters of an FCBGA chip carrier board, characterized in that, Includes the following steps: S1. Collect historical datasets from the FCBGA chip carrier board production process, including historical process parameters and corresponding historical quality indicators; S2. The historical dataset is preprocessed by cleaning and standardizing, and based on the target features of the current production task to be optimized, a subset of data with a correlation greater than a preset threshold is selected from the preprocessed historical dataset using a similarity measurement algorithm. S3. Train a machine learning prediction model using the selected subset of data, the machine learning prediction model being able to map process parameters to quality indicators; S4. Using a machine learning prediction model as the core, construct a multi-objective optimization function, which includes at least two quality indicators as optimization objectives; use a multi-objective optimization algorithm to search within the process parameter constraint space and solve for the Pareto optimal solution set; S5. Select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision-making strategy, and apply them to the actual production line of FCBGA chip carrier. S6. After applying the final optimal process parameters for production, collect actual production result data and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and realize dynamic updating of the model.

2. The method for optimizing FCBGA chip carrier board process parameters according to claim 1, characterized in that, The specific steps of step S1 are as follows: S11. Extract historical process parameters from the manufacturing execution system, the historical process parameters including at least the set temperature of each temperature zone of the reflow oven, the conveyor belt speed, the pressure and duration of the lamination process, and the stencil thickness for solder paste printing; S12. Extract the corresponding historical quality indicators from the quality inspection system. The historical quality indicators include at least the solder ball void rate detected by X-ray, the bridging and cold solder defect rate detected by automatic optical inspection, and the final warpage of the product. S13. Associate and align the extracted process parameters and quality indicators by production batch to generate a structured historical dataset.

3. The method for optimizing FCBGA chip carrier board process parameters according to claim 2, characterized in that, The specific steps of step S2 are as follows: S21. The 3σ principle is used to identify and remove abnormal temperature records in each temperature zone, and the K-nearest neighbor algorithm is used to interpolate the missing compression pressure data. S22. Based on the target features of the current production task to be optimized, perform similarity screening and calculate the cosine similarity between the target features and the features of each batch of products in the historical dataset; the target features include the chip size, number of packaging layers and substrate material of the current production; The cosine similarity formula is as follows: in, The feature vector of the current task. The feature vectors of historical batches; S23. Filter out historical data with similarity greater than a preset threshold and generate a data subset for subsequent model training.

4. The method for optimizing FCBGA chip carrier board process parameters according to claim 3, characterized in that, The specific steps of step S3 are as follows: S31. Construct a fully connected neural network as a machine learning prediction model. The input layer nodes of the fully connected neural network correspond to the screened process parameters, including peak temperature, time above the liquidus line, and pressing pressure. The output layer nodes of the fully connected neural network correspond to the quality indicators to be predicted, including solder ball void rate and warpage. S32. Train the machine learning prediction model using a subset of data, calculate the predicted values ​​using forward propagation, and perform backpropagation using the mean squared error as the loss function to adjust the network weights. The loss function formula is: Where N is the number of samples, and These are the actual and predicted values ​​of the void ratio of the i-th sample solder ball, respectively. and These are the actual and predicted values ​​of the warp of the i-th sample, respectively; S33. After the loss function converges on the validation set, save the model weights to obtain the trained process parameter-quality index prediction model.

5. The method for optimizing FCBGA chip carrier board process parameters according to claim 4, characterized in that, The specific steps of step S4 are as follows: S41. Using a machine learning prediction model as the evaluator, a bi-objective optimization problem is established that simultaneously minimizes the solder ball void rate and minimizes the warpage, mathematically expressed as: in, For process parameter combinations, and These are the predicted values ​​of solder ball void rate and warpage obtained from the machine learning prediction model, respectively. S42. Set the constraint space for process parameters: The peak temperature of reflow soldering is limited to a preset temperature range, and the time above the liquidus is limited to a preset duration. S43. A non-dominated sorting genetic algorithm is used to search within the process parameter constraint space. Through selection, crossover, and mutation operations in the iteration, a set of non-dominated Pareto optimal solutions is generated. The Pareto optimal solution set represents the set of optimal process parameter combinations that balance the solder ball void ratio and warpage.

6. The method for optimizing FCBGA chip carrier process parameters according to claim 5, characterized in that, The specific steps of step S5 are as follows: S51. Assign weight coefficients to the two quality indicators, solder ball void rate and warpage, based on the priority of the current production task. Specifically, when the product is used in a high-performance computing scenario, assign a higher weight to the solder ball void rate; when the product is used in a large-size, low-cost scenario, assign a higher weight to the warpage. S52. Calculate the overall utility value of each Pareto solution in the Pareto optimal solution set using the weighted normalization method: in, A larger summation utility value for a single Pareto solution indicates a better overall quality performance for the corresponding set of process parameters. and These are the weighting coefficients for the porosity and warpage of the solder balls, respectively. ; and These are the predicted values ​​of solder ball void ratio and warpage corresponding to the Pareto solution, calculated by the process parameter-quality index prediction model. and These are the maximum and minimum predicted values ​​of the solder ball void rate for all solutions in the current Pareto optimal solution set, respectively. and These are the maximum and minimum predicted warpage values ​​for all solutions in the current Pareto optimal solution set, respectively. S53. Select the solution with the best overall utility value as the final optimal combination of process parameters and send it to the production line control system for execution.

7. The method for optimizing FCBGA chip carrier process parameters according to claim 1, characterized in that, The specific steps of step S6 are as follows: S61. After production using the new combination of process parameters, the actual solder paste volume and solder ball void rate of the corresponding new batch of products are collected by online SPI and X-ray inspection equipment as new actual production result data. S62. The validity of the collected actual production result data is determined. The current batch of data is considered valid if both of the following conditions are met simultaneously: For the current production batch, all types of process parameters and quality index data necessary for model training and validation have been successfully collected; The control chart for the current batch production process shows that all process parameters are under statistical control, and the first pass rate of the final product of the current batch is higher than the preset qualified threshold. S63. After accumulating K new production batches of data, the incremental learning process is automatically triggered. The machine learning prediction model is fine-tuned using a mixture of new and old historical datasets, and the network weights are updated so that the model can continuously adapt to changes in the production line status.

8. A system for optimizing process parameters of an FCBGA chip carrier board, characterized in that, include: The data collection module is used to collect historical datasets during the production process of FCBGA chip carrier boards. The historical datasets include historical process parameters and corresponding historical quality indicators. The data preprocessing and filtering module is used to clean and standardize the historical dataset, and based on the target characteristics of the current production task to be optimized, it uses a similarity measurement algorithm to filter out a subset of data with a correlation greater than a preset threshold from the preprocessed historical dataset. The prediction model training module is used to train a machine learning prediction model using a selected subset of data. The machine learning prediction model can map process parameters to quality indicators. The multi-objective optimization module is used to construct a multi-objective optimization function based on a machine learning prediction model. The multi-objective optimization function includes at least two quality indicators as optimization objectives. The multi-objective optimization algorithm is used to search within the process parameter constraint space to find the Pareto optimal solution set. The decision and application module is used to select a set of final optimal process parameters from the Pareto optimal solution set according to the preset decision strategy and apply them to the actual production line of FCBGA chip carrier. The feedback update module is used to collect actual production result data after production is carried out using the final optimal process parameters, and feed it back to the historical dataset to trigger incremental learning of the machine learning prediction model and realize dynamic updates of the model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the FCBGA chip carrier process parameter optimization method as described in any one of claims 1 to 7.

10. A 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 FCBGA chip carrier process parameter optimization method as described in any one of claims 1 to 7.

Images