An experimental data statistics and visualization method and system for IC carrier plate material research and development

Abstract

This invention discloses a method and system for statistical analysis and visualization of experimental data in IC substrate material research and development, aiming to address issues such as insufficient adaptability to existing technologies, limited statistical dimensions, weak visualization, and lack of intelligent optimization and iteration mechanisms. The method includes: collecting and preprocessing data throughout the entire research and development process; assigning traceability codes to construct a hierarchical database; establishing a multi-dimensional statistical indicator system; employing multi-factor interactive analysis to mine parameter correlations and filter key parameters; constructing a standardized visualization model and configuring interactive functions; establishing an anomaly warning mechanism and generating optimization suggestions; and storing data with tags and iteratively optimizing the model. The system includes five collaborative modules. This invention achieves standardized management and control, in-depth analysis, intuitive presentation, and closed-loop optimization of research and development data, improving data reliability and decision-making accuracy, reducing the interpretation threshold and research and development costs, shortening the cycle, achieving full-scenario coverage, and continuous upgrading of research and development capabilities. It is applicable to various IC substrate material research and development scenarios.

Description

Technical Field

[0001] This invention belongs to the field of IC substrate material research and development and data processing technology, specifically involving an experimental data statistics and visualization method and system for IC substrate material research and development, applicable to the research and development experiments of core IC substrate materials such as ABF, BT resin, PI film, and copper foil. Background Technology

[0002] In existing technologies, the development of IC substrate materials requires extensive orthogonal experiments, performance testing, and stability verification. This involves highly coupled data across multiple dimensions, including material formulation, preparation process, environmental parameters, and performance indicators. Effective management and in-depth analysis of this data are crucial for supporting R&D decisions. However, current IC substrate-related information processing technologies suffer from significant shortcomings in terms of accuracy and depth in adapting to R&D scenarios.

[0003] In practical applications, existing technologies focus on information processing during the IC substrate production stage. Their core objective is to achieve redundant storage, loss prevention, and traceability of production data. They only involve simple preprocessing and basic comparative analysis of production information, without designing technical solutions for the core needs of the R&D stage. Limited by its production-oriented positioning, this technology cannot solve the problem of standardized integration of multi-dimensional data during R&D, struggles to uncover the nonlinear relationships and multi-factor interactions between formulations, processes, and performance, and lacks the ability to dynamically visualize and generate intelligent optimization suggestions to support R&D decisions. This results in R&D personnel spending a significant amount of time interpreting data, poor reusability of historical experimental data, long R&D cycles, and high trial-and-error costs.

[0004] This demonstrates that existing technologies often suffer from shortcomings such as insufficient compatibility with IC substrate material development scenarios, limited data statistics dimensions, weak visualization capabilities, and a lack of R&D-oriented intelligent optimization and iteration mechanisms. These are the deficiencies of existing technologies.

[0005] In view of this, it is very necessary to provide an experimental data statistics and visualization method and system for IC substrate material research and development to solve the above-mentioned defects in the prior art. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies, such as insufficient adaptability to IC substrate material R&D scenarios, limited data statistics dimensions, weak visualization, and lack of R&D-oriented intelligent optimization and iteration mechanisms. This invention provides a method for experimental data statistics and visualization in IC substrate material R&D to solve the aforementioned technical problems.

[0007] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for statistical analysis and visualization of experimental data for IC substrate material research and development, specifically including the following steps: Step S1: Deploy multi-source data acquisition terminals to collect experimental data related to the core aspects of IC substrate material research and development. After preprocessing the collected data, assign a unique R&D traceability code and associate it with metadata to build a hierarchical database. Step S2: Establish a multi-dimensional statistical indicator system, dynamically adjust the indicator weights for different IC substrate material types; use a multi-factor interactive analysis method to mine parameter correlations and screen key parameters, establish an iterative statistical mechanism based on a hierarchical database; output the statistical analysis results. Step S3: Construct a standardized visualization model to present the data; configure multi-dimensional interactive functions; Step S4: Establish an anomaly warning mechanism based on statistical analysis results and visualization models, set parameter statistical thresholds, and automatically trigger warnings and push relevant information when data exceeds the thresholds or parameter correlations fluctuate abnormally; combine multi-factor statistical analysis results and historical data iteration patterns to generate optimization suggestions. Step S5: Store the relevant data by labeling according to preset dimensions and supplement it to the hierarchical database to form a standardized R&D data asset library; iteratively optimize the statistical analysis algorithm and visualization model based on the historical data of the data asset library.

[0008] Furthermore, in step S1, the hierarchical database is a three-tier storage architecture of "basic database - experimental database - result database"; wherein the basic database stores static data; the experimental database stores preprocessed real-time collected data, metadata and experimental process records; the result database stores statistically analyzed data, performance evaluation reports and optimization suggestions; and a data linkage mechanism between the three databases is established through the development of traceability codes.

[0009] The preprocessing includes data format unification, outlier removal, and missing value completion. The data format standardization operation includes converting the formula ratio into a mass fraction format, standardizing the temperature unit to degrees Celsius, and converting all other types of data according to the corresponding unified standard format. The outlier removal operation is based on The criteria and experimental logic are double-checked to eliminate abnormal data caused by equipment failure or operational errors; The missing value completion operation uses interpolation or the mean of similar experiments to complete the missing data, and marks the completed data with clear completion labels.

[0010] Furthermore, in step S2, the multi-dimensional statistical indicator system includes basic statistical indicators, correlation indicators, process window indicators, and performance optimization indicators; wherein, the basic statistical indicators include mean, variance, standard deviation, and range; the correlation indicators include Pearson correlation coefficient and Spearman rank correlation coefficient; the process window indicators include CPK value and parameter fluctuation range; and the performance optimization indicators include signal-to-noise ratio, orthogonal experimental range, and orthogonal experimental variance.

[0011] The multi-factor interaction analysis method includes orthogonal experimental analysis, response surface methodology, and analysis of variance. The response surface methodology is used to construct a three-dimensional statistical model of multiple parameters and performance indicators, calculate the optimal combination of parameters and the process window range, and the analysis of variance is used to set the significance level. Based on the significance level, key parameters that have a significant impact on material performance are screened and invalid parameters are removed.

[0012] Furthermore, in step S3, the standardized visualization model includes a parameter trend model, a correlation analysis model, a process window model, and a traceability visualization model; Among them, the parameter trend model uses dynamic line charts or bar charts; the correlation analysis model uses heat maps or scatter plot matrices; the process window model uses contour plots or 3D surface plots; and the traceability visualization model uses time axis plots.

[0013] Furthermore, in step S5, the iterative optimization statistical analysis algorithm and visualization model include: dynamically adjusting the weights of statistical indicators, optimizing the process window threshold, updating the visualization presentation logic, and feeding back the implementation effect of the optimization suggestions to the standardized R&D data asset library to form a closed-loop R&D system.

[0014] The standardized R&D data asset library is constructed by tagging data assets according to preset classification dimensions and then integrating the tagged data assets into a hierarchical database. A regular cleaning and updating mechanism is also established to check the data asset library and remove invalid data according to a preset cycle.

[0015] Secondly, the present invention provides an experimental data statistics and visualization system for IC substrate material research and development, comprising: The data acquisition and storage module is used to deploy multi-source data acquisition terminals to collect experimental data related to the entire process of IC substrate material research and development. After preprocessing the collected data, a unique R&D traceability code is assigned to each data point and associated with metadata. A hierarchical database is built to realize data linkage and full life cycle traceability. The multi-dimensional statistical analysis module is used to establish a multi-dimensional statistical indicator system, dynamically adjust the indicator weights for different IC substrate material types, use multi-factor interactive analysis methods to explore the correlation between formula parameters, process parameters and performance indicators and screen key parameters, establish an iterative statistical mechanism based on historical data in a hierarchical database, and output statistical analysis results. The dynamic visualization module is used to build standardized visualization models to present data in layers, and configure multi-dimensional interactive functions to support data adjustment, filtering, drill-down and result export; The anomaly warning and optimization suggestion module is used to establish an anomaly warning mechanism based on statistical analysis results and visualization models. It sets parameter statistical thresholds, and automatically triggers warnings and pushes relevant information when data exceeds the thresholds or when parameter correlation fluctuates abnormally. It also generates optimization suggestions by combining multi-factor statistical analysis results and historical data iteration patterns. The data accumulation and iteration module is used to store statistical analysis results, visualization charts, anomaly handling records, and optimization suggestions in a tagged manner according to preset dimensions, supplementing them to a hierarchical database to form a standardized R&D data asset library. Based on the historical data in the data asset library, the statistical analysis algorithm and visualization model are iteratively optimized.

[0016] Furthermore, the standardized visualization model constructed by the dynamic visualization presentation module includes a parameter trend model, a correlation analysis model, a process window model, and a traceability visualization model; The parameter trend model uses dynamic line charts or bar charts, the correlation analysis model uses heat maps or scatter plot matrices, the process window model uses contour plots or 3D surface plots, and the traceability visualization model uses time axis plots.

[0017] The beneficial effects of this invention are as follows: By standardizing data collection, preprocessing, and hierarchical storage throughout the entire process, it solves the problems of fragmented, chaotic, and difficult-to-trace data in R&D scenarios, thereby improving data reliability and traceability; by constructing a multi-dimensional statistical system, dynamically adapting material weights, and conducting multi-factor interactive analysis, it solves the problems of single statistical dimensions and inability to uncover parameter correlations, thereby improving the accuracy of R&D decisions; by using standardized visualization models and interactive functions, it solves the problems of cumbersome data interpretation and inefficient application, lowering the interpretation threshold and improving application efficiency; by providing intelligent anomaly warnings and precise optimization suggestions, it solves the problems of difficulty in detecting anomalies and high levels of blind trial and error, reducing ineffective trial and error and lowering R&D costs; and by using data tagging and accumulation, closed-loop iteration, and multi-scenario adaptation, it solves the problems of low historical data reuse rate, lack of continuous optimization, and poor scenario adaptability, thereby improving data reuse rate, achieving the goal of covering all R&D scenarios and continuously upgrading R&D capabilities, shortening the R&D cycle, and improving R&D efficiency.

[0018] Furthermore, the design principle of this invention is reliable, the structure is simple, and it has a very wide range of application prospects.

[0019] Therefore, it is evident that the present invention has substantial features and progress compared with the prior art, and the beneficial effects of its implementation are also obvious. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This embodiment provides a flowchart of an experimental data statistics and visualization method for IC substrate material research and development.

[0022] Figure 2 This embodiment provides a principle block diagram of an experimental data statistics and visualization system for IC substrate material research and development.

[0023] The module consists of 1-data acquisition and storage module, 2-multi-dimensional statistical analysis module, 3-dynamic visualization module, 4-anomaly warning and optimization suggestion module, and 5-data accumulation and iteration module. Detailed Implementation

[0024] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following implementation methods.

[0025] Example 1: This embodiment provides a method for statistical analysis and visualization of experimental data for IC substrate material research and development, such as... Figure 1 As shown, the specific steps include: Step S1: Deploy multi-source data acquisition terminals to collect experimental data, preprocess the collected data, assign unique R&D traceability codes and associate metadata, and build a hierarchical database; Step S2: Establish a multi-dimensional statistical indicator system, dynamically adjust the indicator weights for different IC substrate material types; use a multi-factor interactive analysis method to mine parameter correlations and screen key parameters, establish an iterative statistical mechanism based on a hierarchical database; output the statistical analysis results. Step S3: Construct a standardized visualization model to present the data; configure multi-dimensional interactive functions; Step S4: Establish an anomaly warning mechanism based on statistical analysis results and visualization models, set parameter statistical thresholds, and automatically trigger warnings and push relevant information when data exceeds the thresholds or parameter correlations fluctuate abnormally; combine multi-factor statistical analysis results and historical data iteration patterns to generate optimization suggestions. Step S5: Store the relevant data by labeling according to preset dimensions and supplement it to the hierarchical database to form a standardized R&D data asset library; iteratively optimize the statistical analysis algorithm and visualization model based on the historical data of the data asset library.

[0026] Furthermore, step S1 specifically includes the following steps: Step S11: Review the core experimental steps in the entire process of IC substrate material R&D; the core experimental steps specifically include the formulation preparation step, the preparation process step, the performance testing step, and the environmental control step; the preparation process step includes hot pressing, curing, and cutting sub-steps; the performance testing step includes thermal conductivity testing, peel strength testing, insulation resistance testing, and damp heat resistance testing sub-steps. Deploy a multi-source data acquisition terminal; the multi-source data acquisition terminal is compatible with the communication interfaces of experimental equipment such as electronic balances, hot presses, universal testing machines, and impedance analyzers, and supports data acquisition using multiple protocols such as RS485, USB, and Ethernet. The multi-source data acquisition terminal enables real-time automatic acquisition of experimental data such as formula ratio, molding temperature, molding pressure, molding time, performance index data, and environmental temperature and humidity data. The acquisition frequency is set to 1Hz-10Hz according to actual R&D needs to completely avoid data errors caused by manual input.

[0027] Step S12: Establish standardized specifications for IC substrate material R&D data, and preprocess the collected multi-source experimental data according to the standardized specifications for IC substrate material R&D data; the preprocessing includes data format unification, outlier removal, and missing value completion. The data format standardization operation includes converting the formula ratio into a mass fraction format, standardizing the temperature unit to degrees Celsius (°C), and converting all other types of data according to the corresponding unified standard format. The outlier removal operation is based on The criteria and experimental logic are double-checked to accurately identify and eliminate abnormal data caused by non-experimental design factors such as equipment failure and operational errors. The missing value completion operation uses interpolation or the mean of similar experiments to complete the missing data, and marks the completed data with clear completion labels; Each piece of preprocessed experimental data is assigned a unique R&D traceability code. The R&D traceability code is uniquely associated with metadata such as experimental batch, material type, R&D stage, operator, and experimental equipment, so as to achieve precise binding of experimental data with specific experimental scenarios.

[0028] Step S13: Construct a hierarchical database with a three-tier architecture of "basic database - experimental database - results database". The basic database is specifically used to store static data such as material standard parameters, experimental specification documents, and equipment calibration information; the experimental database is specifically used to store real-time acquired data after preprocessing, corresponding metadata, and experimental process record data; the results database is specifically used to store data after statistical analysis, performance evaluation reports, and optimization suggestion documents. By developing traceability codes, a data linkage mechanism is established between three layers of databases to ensure that experimental data can be accurately traced throughout its entire lifecycle, from collection and processing to analysis and application.

[0029] Furthermore, step S2 specifically includes the following steps: Step S21: Based on the core requirements of IC substrate material research and development, establish a multi-dimensional statistical indicator system; the statistical indicator system specifically includes four categories: basic statistical indicators, correlation indicators, process window indicators, and performance optimization indicators. Among them, basic statistical indicators include, but are not limited to, mean, variance, standard deviation, and range; correlation indicators include, but are not limited to, Pearson correlation coefficient and Spearman rank correlation coefficient; process window indicators include, but are not limited to, CPK value and parameter fluctuation range; performance optimization indicators include, but are not limited to, signal-to-noise ratio, orthogonal experiment range and orthogonal experiment variance.

[0030] For different IC substrate core material types, the weight allocation of various statistical indicators is dynamically adjusted: for resin materials, the statistical weight of indicators related to curing degree and viscosity is emphasized; for copper foil materials, the statistical weight of indicators related to tensile strength and surface roughness is emphasized; for thin film materials, the corresponding indicator weights are configured according to their core performance requirements.

[0031] Step S22: Orthogonal experimental design and response surface methodology (RSM) are used to conduct in-depth statistical analysis of the multi-dimensional experimental data to uncover the interaction effects and nonlinear correlations between formulation parameters, process parameters, and performance indicators. Specifically, this includes: A three-dimensional statistical model of multiple parameters and performance indicators is constructed using the response surface methodology. Based on this model, the optimal combination of parameters and the corresponding process window range are accurately calculated. The significance of each parameter is tested by analysis of variance (ANOVA), with a significance level of P < 0.05. Key parameters that have a significant impact on material performance are screened out, while invalid parameters with no substantial impact are eliminated. This simplifies the direction of R&D optimization and improves the targeting of R&D.

[0032] Step S23: Based on the historical experimental data stored in the hierarchical database, establish a standardized iterative statistical mechanism; perform precise comparative statistics between historical experimental data and current experimental data according to material type and experimental purpose, analyze the quantitative influence trend between parameter adjustment behavior and material performance changes, calculate key evaluation indicators such as performance improvement rate and parameter optimization range, and provide direct data support for the design of subsequent experimental schemes. Meanwhile, clustering algorithms are used to cluster and statistically analyze similar experimental data, integrating them into a standardized experimental dataset. This reduces repetitive trial and error operations in similar scenarios and improves the efficiency of experimental data reuse.

[0033] Through the implementation of steps S21 to S23, step 2 finally outputs multi-dimensional experimental data statistical analysis results. These results specifically include: weighted basic statistical index quantification data, correlation index quantification data, process window index quantification data, and performance optimization index quantification data. The basic statistical index quantification data includes mean, variance, standard deviation, and range. The correlation index quantification data includes Pearson correlation coefficient and Spearman rank correlation coefficient. The process window index quantification data includes CPK value and parameter fluctuation range. The performance optimization index quantification data includes signal-to-noise ratio, orthogonal experimental range, and orthogonal experimental variance. The three-dimensional statistical model of multi-parameter performance indicators, the optimal combination scheme of parameters, process window range data, and key parameter screening results; the quantitative impact trend conclusions of parameter adjustment and material performance changes, quantitative data of performance improvement rate and parameter optimization magnitude, and standardized experimental datasets; quantitative data of material performance differences and parameter-performance correlation differences under cross-material research and development scenarios. The above statistical analysis results provide core data support for subsequent dynamic visualization, data anomaly warning and optimization suggestion generation.

[0034] Furthermore, step S3 specifically includes the following steps: Step S31: Construct a visualization model intelligently adapted and presented to achieve hierarchical and accurate visualization of multi-dimensional data. The specific execution process is as follows: The system receives the multi-dimensional statistical analysis results from step S2, the experimental metadata from step S1, the differentiated requirements parameters for IC substrate material R&D, the original experimental data, and related records. The experimental metadata includes material type, experimental batch, R&D stage, and equipment information. The differentiated requirements parameters include key performance indicators and data presentation accuracy requirements. The original experimental data includes formula parameters, process parameters, performance test data, and environmental parameters.

[0035] The input data is classified, organized, and feature-mapped. The statistical analysis results are split according to material type and performance index dimension. The time dimension characteristics, numerical distribution characteristics, and correlation characteristics of the data are extracted. The original experimental data and metadata are linked and bound through the R&D traceability code to form a standardized and visualized data source, ensuring the compatibility of data and model. Intelligent configuration is performed for the four types of standardized visualization models: Parametric trend model: Based on time dimension characteristics and numerical change amplitude, a dynamic threshold algorithm is used to determine the data sampling interval and local magnification trigger conditions. The dynamic threshold calculation formula is as follows:

[0036] in, The data sampling interval, The maximum value of the parameter. The minimum value of the parameter is n, where n is the preset number of data points for the visualization chart. The variation magnitude is used as a weighting coefficient; at the same time, outliers are automatically identified, and corresponding experimental node information is linked and labeled through the R&D traceability code. Association Analysis Model: Based on the correlation index data from step S2, an adaptive threshold algorithm is used to set the correlation strength labeling standard. The adaptive threshold calculation formula is as follows:

[0037] in, The threshold for determining strong correlation. Here, r is the threshold for determining a weak correlation, and r is the mean of all correlation coefficients. This represents the standard deviation of the correlation coefficient; based on this threshold, strong correlations are marked in red, weak correlations in blue, and moderate correlations are presented according to the default visual rules. Process window model: Based on the process window range data and parameter-performance correlation characteristics obtained in step S2, a mesh generation algorithm is used to determine the region division accuracy. The mesh density calculation formula is as follows:

[0038] Where d is the mesh cell size. This is the upper limit of the process window. The lower limit of the process window is k, which is the preset number of grids; according to the division results, qualified process windows are marked in green and unqualified areas are marked in red. Traceability visualization model: Based on the temporal characteristics of the experimental process and the associated data of the R&D traceability code, a time axis calibration algorithm is used to determine the time scale density and node labeling position. The time axis calibration formula is:

[0039] in, This represents the time axis scale interval. This is the end time of the experiment. The experiment start time is m, and the time axis is the preset number of scales; the entire experimental data trajectory is fully restored, and abnormal nodes, causes of abnormalities, and handling measures are associated and marked. The presentation of the four types of models was uniformly optimized by adjusting color contrast, line thickness, and legend position to ensure the intuitiveness and readability of the data presentation. By integrating the associated data from various models and establishing data linkage between models through the development of traceability codes, the experimental data can be traced and reviewed throughout the entire process.

[0040] The output parameter trend model, correlation analysis model, process window model, and traceability visualization model have complete configuration files and presentation results. The configuration files include intelligent configuration parameters such as sampling interval, threshold standard, grid density, and time scale. The presentation results are hierarchical visualization charts that meet differentiated needs.

[0041] Step S32: Construct a multi-dimensional intelligent interaction model. This model is used to perform calculations to achieve highly flexible data interpretation and application functions. The specific execution process is as follows: It receives interactive operation instructions from R&D personnel, visualization model data output in step S31, and raw experimental data and metadata in the hierarchical database. The interactive operation instructions include dimension adjustment instructions, model switching instructions, data drill-down instructions, and export instructions. The visualization model data includes data sources and configuration parameters for various charts.

[0042] The instruction recognition algorithm is used to analyze the interaction intent of the R&D personnel and convert the operation instructions into standardized control parameters. For example, the instruction "screen thermal conductivity data by ABF material" is converted into screening parameters with material type = ABF and performance index = thermal conductivity. Based on preset dimensions such as material type, experimental batch, and performance indicators, a multi-dimensional filtering index is constructed, allowing R&D personnel to independently adjust filtering conditions and quickly locate target data through the index; the formula for constructing the filtering index is:

[0043] in, Filter the data using index values, where p is the number of filter dimensions. Let be the weight coefficient of the i-th dimension. Let i be the feature value of the i-th dimension; Establish a switching mapping relationship for four types of visualization models to support rapid switching; construct a data drill-down correlation index so that when developers click on a data point, they can retrieve the corresponding original data and metadata through an index formula. The correlation index formula is:

[0044] Where Link is the associated index code, R is the R&D traceability code, and q is the number of data levels. Hierarchical encoding for the j-th layer of data; Based on the export command, the export format is determined, including PNG, PDF, and Excel. The core information of the visualization chart is automatically extracted, including data range, key patterns, abnormal nodes, and core conclusions, and standardized data interpretation notes are generated. By developing traceability codes, we can achieve real-time linkage between interactive operations and hierarchical databases, ensuring that the filtered and drill-down data is the latest and most valid data. We can also generate downloadable file packages according to the configured export format and notes, adapting to different scenarios such as R&D reports and data archiving.

[0045] The model output includes an interactive control module, filtered data results, drill-down retrieved raw data and metadata, standardized export files, and data interpretation notes; providing highly flexible support for R&D personnel's data interpretation and application.

[0046] Furthermore, step S4 specifically includes the following steps: Step S41: Based on the multi-dimensional experimental data statistical analysis results of step S2 and the dynamic visualization model of step S3, establish a standardized data anomaly early warning mechanism; the standardized data anomaly early warning mechanism specifically includes: A threshold intelligent generation model is constructed, and the statistical thresholds for each experimental parameter are obtained through calculation using this model. The specific execution process is as follows: It receives historical qualified data, IC substrate material industry standard documents, and R&D design schemes from the hierarchical database. The historical qualified data includes historical records of formula parameters, process parameters, performance indicators, and environmental parameters, while the R&D design schemes include R&D target parameters and R&D phase requirements.

[0047] The threshold intelligent generation model performs data cleaning on the input historical qualified data, removing invalid data. After cleaning, it performs standardization on the historical qualified data using Z-score transformation to ensure data quality meets the requirements of subsequent calculations. The Z-score transformation formula is:

[0048] in, The value is the standardized value of the data. This is the original value from the historical qualified data. This is the average of the historical qualified data for this parameter. This represents the standard deviation of the historical qualified data for this parameter.

[0049] Feature extraction operations are performed on the preprocessed historical qualified data, industry standard documents, and R&D design schemes respectively. Distribution features are extracted from the historical qualified data, including mean, standard deviation, and range. Constraint features are extracted from the industry standard documents, including upper limit, lower limit, and allowable fluctuation range. Target features are extracted from the R&D design schemes, including optimal value and acceptable range. The extracted distribution features, constraint features, and target features are processed using a weighted fusion algorithm. Weights are assigned to the distribution features, constraint features, and target features. The weighted fusion algorithm then calculates the upper and lower statistical thresholds for each experimental parameter. The weighted fusion calculation formula is as follows:

[0050] in, As the initial statistical threshold for the parameters, These are the distribution characteristic weighting coefficients. The parameter threshold is derived from the distribution characteristics. To constrain the feature weight coefficients, The parameter thresholds derived for constraining features, The target feature weight coefficients, The parameter threshold is derived for the target feature.

[0051] To address the key performance indicators in IC substrate material development, a safety factor is introduced to correct the statistical thresholds, thereby improving the rigor of the threshold warnings. The formula for correcting the key performance indicator thresholds is as follows:

[0052] in, This is the upper limit of the corrected statistical threshold. The initial statistical threshold is the upper limit, and k is the safety factor. This is the lower limit of the corrected statistical threshold. This is the lower limit of the initial statistical threshold.

[0053] A two-way mapping relationship is established between the calculated and corrected statistical thresholds and the IC substrate material type and R&D stage. A threshold standard library is formed based on this two-way mapping relationship. The threshold standard library is dynamically iterated as historical data is updated, and each statistical threshold is precisely matched with the corresponding material type and R&D stage. The final threshold intelligent generation model outputs a unique statistical threshold, threshold applicability range, and threshold confidence level for each experimental parameter. The statistical threshold includes an upper threshold and a lower threshold, the threshold applicability range includes the matching material type and R&D stage, and the threshold confidence level is used to characterize the reliability of the statistical threshold.

[0054] An anomaly intelligent judgment model is constructed, which receives the current experimental data after preprocessing in step S1, the parameter correlation coefficients output in step S2, the threshold standard library output by the threshold intelligent generation model, and the correlation stability interval obtained by training based on historical qualified data. The parameter correlation coefficients include Pearson correlation coefficient and Spearman rank correlation coefficient, and the current experimental data includes real-time collected formula parameters, process parameters, performance indicators, and environmental parameters.

[0055] The system uniformly receives and formats all types of input data to ensure the consistency of subsequent calculations. It then compares the formatted current experimental data with the statistical thresholds of the corresponding parameters in the threshold standard library in real time to verify whether the current experimental data exceeds the upper or lower limit of the statistical threshold, thus obtaining the single-parameter verification result. The formula for calculating the Pearson correlation coefficient is:

[0056] Where r is the Pearson correlation coefficient between the two parameters, and n is the sample size. Let X be the i-th sample value. Let X be the sample mean of the parameter. Let Y be the i-th sample value. The sample mean of parameter Y; The formula for determining single-parameter over-limit is:

[0057] in, These are the current experimental parameter values. This is the lower limit of the statistical threshold for this parameter. Set the upper limit of the statistical threshold for this parameter.

[0058] A sliding window algorithm is used to train on historical qualified data. The sliding window is selected from the most recent fifty sets of qualified experimental data. After training, a stable correlation interval between parameters is obtained. The formatted parameter correlation coefficients are compared with the stable correlation interval in real time, and the deviation of the parameter correlation coefficients is calculated to obtain the multi-parameter correlation analysis results. The deviation calculation formula is:

[0059] Where D represents the deviation of the correlation coefficient. This represents the correlation coefficient between the current parameters. This represents the mean of the correlation stability interval. denoted as the standard deviation of the correlation stability interval.

[0060] Pre-defined anomaly detection rules are used to match single-parameter validation results and multi-parameter correlation analysis results with these rules, thereby completing a comprehensive anomaly determination of the experimental data. Simultaneously, anomaly types are precisely labeled, including data exceeding limits and correlation fluctuations. The comprehensive anomaly determination formula is as follows:

[0061] Where Y represents the anomaly determination result, Y=1 indicates that it is determined to be abnormal, and Y=0 indicates that it is determined to be normal; For anomalies determined to be of the data exceedance type, this layer synchronously calculates the extent of the data exceedance to quantitatively classify the degree of anomaly. The formula for calculating the extent of the data exceedance is as follows:

[0062] Where P represents the data exceeding the limit, the degree of abnormality is divided into three categories: mild, moderate, and severe, based on the numerical range of P and the deviation degree D.

[0063] The model outputs anomaly determination results, anomaly types, anomaly parameter lists, and anomaly severity. The anomaly determination results include two categories: normal and abnormal. The anomaly parameter list includes single parameters or related parameter groups that are determined to be abnormal. Mild anomalies correspond to data exceeding the limit by less than or equal to 10% or a deviation of 1.5 to 2.0. Moderate anomalies correspond to data exceeding the limit by 10% to 20% or a deviation of 2.0 to 3.0. Severe anomalies correspond to data exceeding the limit by more than 20% or a deviation of more than or equal to 3.0.

[0064] Once an anomaly is detected, an early warning process is automatically triggered. The location of the abnormal data and the corresponding experimental node are accurately marked with prominent identification through the dynamic visualization model in step S3. Early warning information is generated simultaneously. The early warning information includes, but is not limited to, the anomaly type, the name of the abnormal parameter, the abnormal value, the corresponding threshold range and the possible cause. The information is fed back to the R&D personnel's terminal through pop-up windows, message push and other means to remind them to carry out troubleshooting in a timely manner.

[0065] Furthermore, the prominent identification includes a red flashing border and special marking symbols; the anomaly types include data exceeding limits and correlation fluctuations; the possible causes include equipment accuracy drift, operational process deviation, and environmental interference; and the troubleshooting includes experimental equipment calibration, operational standardization verification, and environmental factor detection.

[0066] Step S42: Construct an intelligent suggestion model, integrating the statistical analysis results of step S2, historical data iteration patterns, and the shortcomings of the current experimental data to generate targeted optimization suggestions. The specific execution process is as follows: The intelligent suggestion model receives the multi-dimensional experimental data statistical analysis results from step S2, historical experimental data from the hierarchical database, current experimental data output by the anomaly intelligent judgment model in step S41, and the R&D goals of IC substrate materials. The statistical analysis results from step S2 include key parameter screening results, parameter-performance correlation models, and process window ranges. The historical experimental data includes historical optimization measures and performance changes after the implementation of optimization measures. The current experimental data includes unmet performance indicators and details of abnormal parameters.

[0067] All types of data are uniformly received and structured to form a standardized data format, ensuring the orderly conduct of subsequent calculations. Feature engineering methods are used to extract features from the processed current experimental data to obtain the bottleneck features of the current experiment. These bottleneck features include the name of the unmet performance indicator, the difference between the unmet performance indicator and the R&D target, the influence weight of abnormal parameters, and the degree of deviation between the current parameter combination and the optimal parameter combination. The influence weight of abnormal parameters is determined based on the variance analysis results of step S2. The formula for calculating the performance indicator bottleneck difference is:

[0068] in, For performance indicators, the shortcomings are the main difference. Develop target values ​​for performance indicators. These are the current measured values ​​for the performance indicators.

[0069] The K-nearest neighbor algorithm is used, with the weakest link feature as the matching dimension, to filter similarity from the organized historical experimental data. Historical experimental cases were used, with K set to 10. These selected historical experimental cases were analyzed to extract effective optimization measures and corresponding performance improvement results. The formula for calculating the similarity of the bottleneck features is:

[0070] Where S represents the similarity of the shortest feature, and m represents the number of dimensions of the shortest feature. Let j be the characteristic value of the shortest component in the current experiment. Let be the characteristic value of the j-th dimension of the historical case.

[0071] Based on the parameter-performance correlation model and process window range from step S2, and with the constraints of achieving R&D goals in performance indicators and ensuring process parameters meet actual production feasibility, combined with extracted effective optimization measures, a parameter adjustment scheme is calculated; the parameter sensitivity calculation formula is:

[0072] in, For parameters Sensitivity to performance index P This is the partial derivative of the performance index with respect to this parameter; For key formulation parameters, prioritize them according to the significance of their impact on performance indicators, and output the parameter objects for formulation adjustment and the specific adjustment range. The adjustment range is determined based on the sensitivity analysis results of the parameter-performance correlation model. For key process parameters, combine the upper and lower limits of the process window range, and output the parameter objects for process optimization, the adjustment direction, and the specific adjustment range. Based on the performance variation patterns of the compiled historical experimental data, and combined with the parameter adjustment scheme obtained from the parameter optimization calculation layer, a random forest regression algorithm is used to quantify and predict the performance improvement effect after the parameter adjustment scheme is implemented, clarifying the prediction range and probability of achieving each performance indicator; the performance indicator prediction formula is:

[0073] in, The performance metric prediction result is given, where k is the number of decision trees in the random forest. Let be the weight coefficients of the t-th decision tree. For the t-th decision tree based on the adjusted parameters The predicted performance value.

[0074] Using a standardized template, integrate parameter adjustment schemes, adjustment priority rankings, and expected performance improvement results to generate a standardized optimization suggestion report. The report also clearly defines the key performance indicators to be monitored after the scheme is implemented.

[0075] Output a standardized optimization suggestion report, which includes suggestions for formula adjustment, process optimization, expected performance improvement, and implementation precautions. The formula adjustment suggestions include the parameter objects, adjustment range, and priority ranking; the process optimization suggestions include the parameter objects, adjustment direction, adjustment range, and priority ranking; the expected performance improvement includes the estimated range and probability of achieving each performance indicator; and the implementation precautions include the performance indicators that need to be monitored after the implementation of the solution.

[0076] For performance indicators that do not meet the standards or parameter configurations that need optimization in the current experimental data, suggestions are generated from the dimensions of formula adjustment and process optimization. These suggestions include the objects of parameter adjustment, the specific adjustment range, and the priority ranking of adjustments based on the significance of the parameter's impact on performance. Based on the performance change patterns of historical data and statistical model extrapolation, the expected performance improvement effect after the optimization suggestions are implemented is quantitatively predicted, and the estimated range of performance indicators is clarified. The optimization suggestions are presented in the form of a standardized report, which is linked and stored with the corresponding experimental data, statistical results, and visualization charts through R&D traceability codes. It also supports online viewing and is compatible with exporting in PNG, PDF, and Excel formats, providing direct data support for R&D personnel to formulate subsequent experimental plans.

[0077] Through the implementation of steps S41 to S42, step S4 finally outputs the exception handling record and optimization suggestions, including: The anomaly handling record includes basic anomaly information, fault investigation process record, and handling results; the optimization suggestions include formula adjustment suggestions and process optimization suggestions, specifically including parameter adjustment objects, specific adjustment ranges, adjustment priority ranking, and expected performance improvement effects. The above anomaly handling record and optimization suggestions provide core data support for the data accumulation and R&D system iteration in the subsequent step S5.

[0078] Furthermore, step S5 specifically includes the following steps: Step S51: Construct a data tagging and asset repository optimization model to achieve standardized data asset accumulation and dynamic updates. The specific execution process is as follows: Receive the statistical analysis results from step S2, the visualization charts from step S3, the optimization suggestions and anomaly handling records from step S4, and the preset classification dimension rules; the preset classification dimension rules include specific classification standards for material type, R&D direction, and performance indicators. Material types include ABF, BT resin, PI film, and copper foil. R&D directions include peel strength optimization, thermal conductivity improvement, and moisture and heat resistance improvement. Performance indicators include thermal conductivity, peel strength, insulation resistance, and water absorption rate.

[0079] All input data is uniformly received and format-validated, and data with incorrect formats is removed to form a standardized data set, ensuring consistency in subsequent tagging processing. A rule-based matching algorithm is used to match corresponding tags to each data asset based on preset classification dimension rules. The tag matching formula is as follows:

[0080] Where L is the final set of labels for the data assets, and k is the number of classification dimensions. Given a subset of tags corresponding to data asset X under the d-th category dimension, output tags that conform to the data characteristics by traversing the rule base of each category dimension; For cross-dimensional data assets, multiple corresponding tags are automatically associated to ensure complete tag coverage; the tagged data assets are supplemented to the corresponding database according to the hierarchical architecture of "basic database - experimental database - result database", and the relationship between new data assets and original data is established through R&D traceability codes, and integrated into a standardized R&D data asset library to achieve centralized storage and unified management of multi-source data assets. A data validity assessment algorithm is used to verify the data asset database at preset intervals. The data validity assessment formula is as follows:

[0081] in, The result of the validity assessment of data asset X. =1 indicates that it is valid. =0 indicates invalid. Let X be the set of features of data asset X. It is a set of invalid data characteristics, including duplicate record characteristics, erroneous data characteristics, and redundant data characteristics; Score the R&D value of data asset X. A preset value threshold; Based on the judgment results, invalid data is eliminated, while the reuse value of high-quality data is retained and enhanced to ensure the accuracy and effectiveness of the data asset library.

[0082] The data tagging and asset library optimization model outputs a standardized R&D data asset library, a data tag mapping table, and an invalid data removal log. The data tag mapping table records the association between each data asset and its corresponding tag, and the invalid data removal log contains the identifier, type, and removal time of the invalid data.

[0083] Step S52: Construct an algorithm and model iterative optimization model to achieve continuous upgrading of statistical analysis algorithms and visualization models. The specific execution process is as follows: Receive historical experimental data from the standardized R&D data asset library, actual implementation effect data of the optimization suggestions in step S4, statistical analysis algorithm parameters in step S2, and visualization model configuration parameters in step S3; among which, actual implementation effect data includes performance improvement data after parameter adjustment and experimental results after process optimization, statistical analysis algorithm parameters include the initial weights of each statistical indicator, and visualization model configuration parameters include presentation logic rules.

[0084] The input data is structured and organized, and parameter-performance correlation features from historical experimental data and optimization effectiveness features from implementation effect data are extracted to form an iterative optimization dataset. The gradient boosting tree algorithm is used to mine the influence of statistical indicators on performance evaluation results, the matching characteristics between process window thresholds and data distribution, and the correlation characteristics between visualization logic and data interpretation efficiency in historical data. The formula for calculating feature importance is as follows:

[0085] in, Let f be the importance score for feature f, m be the number of decision trees in the gradient boosting tree, and Δ be the value of feature f. Let f be the amount of loss function reduction brought about by feature f in the t-th decision tree; Based on the feature importance score, the weights of the multi-dimensional statistical indicators in step S2 are dynamically adjusted. The weight adjustment formula is as follows:

[0086] in, Let be the adjusted weight of the i-th statistical indicator. Let n be the feature importance score corresponding to the i-th statistical indicator, and n be the total number of statistical indicators. This is the sum of weights; Based on the distribution characteristics of the newly added qualified data, the process window threshold in step S2 is optimized. The threshold optimization formula is as follows:

[0087] in, To optimize the threshold range of the post-process window, This represents the average of the newly added qualified data. The standard deviation of the newly added qualified data is given by k, which is an adjustment coefficient obtained based on the data distribution fitting. Based on the data interpretation efficiency feedback data, update the presentation logic of the standardized visualization model in step S3, including the dynamic update frequency of the parameter trend model, the correlation strength labeling rules of the correlation analysis model, the regional division accuracy of the process window model, and the time axis scale density of the traceability visualization model. Simultaneously, the actual implementation effect data of the optimization suggestions in step S4 is fed back to the standardized R&D data asset library, forming a closed-loop R&D system of "experimental data collection - multi-dimensional statistical analysis - dynamic visualization presentation - anomaly warning and optimization suggestions - implementation effect feedback - data accumulation and model iteration - guidance for new experiments".

[0088] The final output includes the adjusted statistical indicator weight configuration, the optimized process window threshold, the updated visualization logic, and the iterative optimization report. The iterative optimization report includes a comparison of parameters before and after optimization and effect verification data. Through this closed-loop mechanism, the R&D process can be continuously optimized, and the efficiency and results conversion rate of IC substrate material R&D can be continuously improved.

[0089] Example 2: This embodiment also provides an experimental data statistics and visualization system for IC substrate material research and development, such as... Figure 2 As shown, it includes: Data acquisition and storage module 1 is used to deploy multi-source data acquisition terminals to collect experimental data related to the entire process of IC substrate material research and development. After preprocessing the collected data, a unique research and development traceability code is assigned to each data and associated with metadata. A hierarchical database is built to realize data linkage and full life cycle traceability. Multidimensional statistical analysis module 2 is used to establish a multidimensional statistical indicator system, dynamically adjust the indicator weights for different IC substrate material types, use multi-factor interactive analysis methods to explore the correlation between formula parameters, process parameters and performance indicators and screen key parameters, establish an iterative statistical mechanism based on historical data in a hierarchical database, and output statistical analysis results. Dynamic visualization module 3 is used to build standardized visualization models to present data in layers, and configure multi-dimensional interactive functions to support data adjustment, filtering, drill-down and result export. The anomaly warning and optimization suggestion module 4 is used to establish an anomaly warning mechanism based on statistical analysis results and visualization models. It sets parameter statistical thresholds, and automatically triggers warnings and pushes relevant information when data exceeds the thresholds or when parameter correlation fluctuates abnormally. It also generates optimization suggestions by combining multi-factor statistical analysis results and historical data iteration patterns. The data accumulation and iteration module 5 is used to store statistical analysis results, visualization charts, anomaly handling records and optimization suggestions in a tagged manner according to preset dimensions, supplement them to a hierarchical database to form a standardized R&D data asset library, and iteratively optimize statistical analysis algorithms and visualization models based on historical data in the data asset library.

[0090] Furthermore, the data acquisition and storage module 1, through the deployment of multi-source data acquisition terminals, comprehensively covers core experimental stages such as formula preparation, manufacturing process, performance testing, and environmental control. It is compatible with the communication interfaces of commonly used experimental equipment such as electronic balances, hot presses, universal testing machines, and impedance analyzers, and supports data transmission via multiple protocols such as RS485, USB, and Ethernet. It can automatically acquire key data such as formula ratio, molding temperature, pressure, time, performance indicators, and ambient temperature and humidity in real time. The acquisition frequency can be flexibly set within the range of 1-10Hz according to R&D needs, effectively avoiding errors caused by manual input.

[0091] After data collection, the data acquisition and storage module 1 will perform preprocessing according to preset standardized specifications, including converting formula ratios to mass fractions and standardizing temperature units to degrees Celsius, etc. The system employs a dual-validation approach, combining criteria and experimental logic, to remove outliers and a method for filling in missing values ​​using interpolation or the mean of similar experiments. Furthermore, each pre-processed experimental data point is assigned a unique R&D traceability code, linking it to comprehensive metadata such as experimental batch, material type, R&D stage, operator, and experimental equipment, thus achieving precise data-driven experimental context.

[0092] By constructing a three-tiered storage architecture of "basic database - experimental database - results database", the database stores static data such as material standard parameters and experimental specifications, standardized real-time collected data, metadata and experimental process records, as well as subsequent statistical analysis results and performance evaluation reports. The three-tiered database is linked by the R&D traceability code to ensure the traceability of data throughout its entire lifecycle.

[0093] Furthermore, the multi-dimensional statistical analysis module 2, based on the core requirements of IC substrate material research and development, establishes a multi-dimensional statistical indicator system including basic statistical indicators, correlation indicators, process window indicators, and performance optimization indicators. Among them, the basic statistical indicators include mean, variance, standard deviation, and range; the correlation indicators include Pearson correlation coefficient and Spearman rank correlation coefficient; the process window indicators include CPK value and parameter fluctuation range; and the performance optimization indicators include signal-to-noise ratio, orthogonal experimental range, and variance.

[0094] For different IC substrate materials such as resin, copper foil, and film, the multi-dimensional statistical analysis module dynamically adjusts the weight allocation of various statistical indicators. For example, for resin materials, it focuses on indicators related to curing degree and viscosity, while for copper foil materials, it focuses on indicators related to tensile strength and surface roughness, ensuring that statistical analysis is accurately matched with R&D goals.

[0095] In the statistical analysis process, the multi-dimensional statistical analysis module 2 employs multi-factor interaction analysis methods such as orthogonal experimental analysis and response surface methodology (RSM) to deeply explore the interaction effects and nonlinear correlations between formulation parameters, process parameters, and performance indicators. For example, a three-dimensional statistical model of multiple parameters and performance indicators is constructed using response surface methodology to accurately calculate the optimal combination of parameters and the process window range. Then, the significance level is set through analysis of variance (ANOVA). This process involves identifying key parameters that significantly impact material performance, eliminating ineffective parameters, and simplifying the direction of research and optimization. Simultaneously, a standardized iterative statistical mechanism is established based on historical experimental data from a hierarchical database. Historical and current experimental data are compared and statistically analyzed according to material type and experimental purpose. The impact trend of parameter adjustments on performance is analyzed, and key indicators such as performance improvement rate and parameter optimization magnitude are calculated. Clustering algorithms are also used to integrate similar experimental data to form a standardized experimental dataset, reducing redundant trial and error and providing solid data support for subsequent experimental design. Finally, complete statistical analysis results are output.

[0096] Furthermore, the dynamic visualization module 3 constructs four standardized visualization models—parameter trend model, correlation analysis model, process window model, and traceability visualization model—to meet different R&D needs, thereby achieving layered data presentation.

[0097] The parameter trend model uses dynamic line charts or bar charts to clearly show the changing trends of single or multiple parameters with experimental batches and time, and supports local data magnification and outlier annotation. The correlation analysis model visually presents the strength of the correlation between formulation, process and performance parameters through heat maps or scatter plot matrices, with red indicating strong correlation and blue indicating weak correlation. The process window model uses contour plots or 3D surface plots to clearly present the performance ranges corresponding to the combination of key parameters, with green markings for qualified process windows and red markings for unqualified areas. The traceability visualization model uses a timeline graph to fully display the data trajectory of a single batch of experiments from formula preparation to performance testing, and associates and marks abnormal nodes with handling measures.

[0098] To enhance usability, the dynamic visualization module 3 is also equipped with a wealth of multi-dimensional interactive functions, allowing R&D personnel to adjust visualization dimensions, filter data ranges, and switch visualization models as needed; it provides a data drill-down function, allowing users to click on any data point in the visualization chart to view the corresponding original experimental data, metadata, and complete experimental records; it also supports exporting visualization results in multiple formats such as PNG, PDF, and Excel, and generates data interpretation notes simultaneously, providing powerful assistance for R&D reporting and decision-making.

[0099] Furthermore, the anomaly warning and optimization suggestion module 4 establishes a standardized data anomaly warning mechanism based on the statistical analysis results output by the multi-dimensional statistical analysis module and the visualization model of the dynamic visualization presentation module.

[0100] For each experimental parameter, combined with historical qualified data in the hierarchical database The scope and corresponding industry standards and R&D design requirements for IC substrate materials are defined, and a unique and clear statistical threshold is set to form a threshold standard library that is precisely associated with material type and R&D stage.

[0101] The anomaly warning and optimization suggestion module 4 monitors the changes in the correlation between current experimental data and parameters in real time. When a single experimental data point exceeds the corresponding statistical threshold, or when the correlation coefficient between parameters deviates from the historical stable range and exhibits unexpected abnormal fluctuations, the warning process is automatically triggered. The location of the abnormal data and the corresponding experimental node are accurately marked with prominent identification through a visual chart. At the same time, warning information including the anomaly type, abnormal parameter name, abnormal value, corresponding threshold range, and possible causes is pushed to remind R&D personnel to carry out troubleshooting work such as equipment calibration and operation standardization review in a timely manner.

[0102] In terms of generating optimization suggestions, the anomaly warning and optimization suggestion module 4 constructs an intelligent suggestion model, which integrates the results of multi-factor interactive statistical analysis, the iterative patterns of historical data, and the shortcomings of current experimental data. It generates targeted suggestions from two core dimensions: formula adjustment and process optimization, clarifies the parameter adjustment objects, specific adjustment ranges, and adjustment priority rankings, and quantitatively predicts the expected performance improvement effect after the optimization suggestions are implemented based on the performance change patterns of historical data and statistical model deduction, and clarifies the estimated range of performance indicators.

[0103] The optimization suggestions are presented in a standardized report format, which is linked to the corresponding experimental data, statistical results, and visualization charts through R&D traceability codes. It supports online viewing and export in multiple formats, providing direct data support for R&D personnel to formulate subsequent experimental plans.

[0104] Furthermore, the data accumulation and iteration module 5 stores the statistical analysis results output by the multi-dimensional statistical analysis module, the visualization charts generated by the dynamic visualization presentation module, and the anomaly handling records and optimization suggestions formed by the anomaly warning and optimization suggestion module in a tagged and classified manner according to preset dimensions such as material type, R&D direction, and performance indicators, and supplements them to the hierarchical database to build a standardized R&D data asset library.

[0105] To ensure data asset quality, the data accumulation and iteration module establishes a regular cleaning and update mechanism. It checks the data asset repository according to a preset cycle, removing duplicate records, erroneous data, and redundant and invalid data with no R&D value, retaining high-quality data to enhance its reusability. Based on this, the module continuously iterates and optimizes statistical analysis algorithms and visualization models using historically accumulated data in the data asset repository. This includes dynamically adjusting the weights of statistical indicators, optimizing process window thresholds, and updating visualization logic, thereby improving the accuracy and adaptability of the system's data statistics and visualization.

[0106] Meanwhile, the data accumulation and iteration module 5 will feed back the implementation effect of the optimization suggestions to the standardized R&D data asset library, providing real and effective data basis for the next round of iteration and optimization, promoting the continuous upgrading of system support capabilities, and helping to continuously improve the R&D efficiency of IC substrate materials.

[0107] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The methods disclosed in the embodiments are described simply because they correspond to the systems disclosed in the embodiments; relevant details can be found in the method section.

[0108] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0109] In the embodiments provided by this invention, it should be understood that the disclosed systems, methods, and approaches can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0110] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0111] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit.

[0112] Similarly, in the various embodiments of the present invention, each processing unit can be integrated into a functional module, or each processing unit can exist physically, or two or more processing units can be integrated into a functional module.

[0113] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0114] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to include non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0115] The above-disclosed embodiments are merely preferred embodiments of the present invention, but the present invention is not limited thereto. Any non-creative variations that can be conceived by those skilled in the art, as well as any improvements and modifications made without departing from the principles of the present invention, should fall within the protection scope of the present invention.

Claims

1. A method for statistical analysis and visualization of experimental data for IC substrate material research and development, characterized in that, Includes the following steps: Step S1: Deploy multi-source data acquisition terminals to collect experimental data related to the core aspects of IC substrate material research and development. After preprocessing the collected data, assign a unique R&D traceability code and associate it with metadata to build a hierarchical database. Step S2: Establish a multi-dimensional statistical indicator system, dynamically adjust the indicator weights for different IC substrate material types; use a multi-factor interactive analysis method to mine parameter correlations and screen key parameters, establish an iterative statistical mechanism based on a hierarchical database; output the statistical analysis results. Step S3: Construct a standardized visualization model to present the data; Configure multi-dimensional interactive functions; Step S4: Establish an anomaly warning mechanism based on statistical analysis results and visualization models, set parameter statistical thresholds, and automatically trigger warnings and push relevant information when data exceeds the thresholds or parameter correlations fluctuate abnormally; combine multi-factor statistical analysis results and historical data iteration patterns to generate optimization suggestions. Step S5: Store the relevant data by labeling it according to preset dimensions and supplement it to the hierarchical database to form a standardized R&D data asset library; The statistical analysis algorithm and visualization model are iteratively optimized based on historical data from the data asset library.

2. The method according to claim 1, characterized in that, In step S1, the hierarchical database is a three-tier storage architecture of "basic database - experimental database - result database"; The basic database stores static data; the experimental database stores preprocessed real-time acquired data, metadata, and experimental process records; and the results database stores statistically analyzed data, performance evaluation reports, and optimization suggestions. A data linkage mechanism between three layers of databases was established by developing traceability codes.

3. The method according to claim 2, characterized in that, In step S1, the preprocessing includes data format unification, outlier removal, and missing value completion. The data format standardization operation includes converting the formula ratio into a mass fraction format, standardizing the temperature unit to degrees Celsius, and converting all other types of data according to the corresponding unified standard format. The outlier removal operation is based on The criteria and experimental logic are double-checked to eliminate abnormal data caused by equipment failure or operational errors; The missing value completion operation uses interpolation or the mean of similar experiments to complete the missing data, and marks the completed data with clear completion labels.

4. The method according to claim 3, characterized in that, In step S2, the multi-dimensional statistical indicator system includes basic statistical indicators, correlation indicators, process window indicators, and performance optimization indicators; The basic statistical indicators include mean, variance, standard deviation, and range; the correlation indicators include Pearson correlation coefficient and Spearman rank correlation coefficient; the process window indicators include CPK value and parameter fluctuation range; and the performance optimization indicators include signal-to-noise ratio, orthogonal experiment range, and orthogonal experiment variance.

5. The method according to claim 4, characterized in that, In step S2, the multi-factor interaction analysis method includes orthogonal experimental analysis, response surface methodology, and analysis of variance. A three-dimensional statistical model of multiple parameters and performance indicators was constructed using the response surface methodology to calculate the optimal combination of parameters and the process window range. A significance level was set using analysis of variance, and key parameters that significantly affect material properties were screened based on the significance level, while invalid parameters were removed.

6. The method according to claim 5, characterized in that, In step S3, the standardized visualization model includes a parameter trend model, a correlation analysis model, a process window model, and a traceability visualization model; Among them, the parameter trend model uses dynamic line charts or bar charts; the correlation analysis model uses heat maps or scatter plot matrices; the process window model uses contour plots or 3D surface plots; and the traceability visualization model uses time axis plots.

7. The method according to claim 6, characterized in that, In step S5, the iterative optimization statistical analysis algorithm and visualization model include: dynamically adjusting the weight of statistical indicators, optimizing the process window threshold, updating the visualization presentation logic, and feeding back the implementation effect of the optimization suggestions to the standardized R&D data asset library to form a closed-loop R&D system.

8. The method according to claim 7, characterized in that, The standardized R&D data asset library is constructed by tagging data assets according to preset classification dimensions and then integrating the tagged data assets into a hierarchical database. Furthermore, a regular cleaning and updating mechanism is established to check the data asset database and remove invalid data according to a preset cycle.

9. A statistical and visualization system for experimental data in IC substrate material research and development, characterized in that, include: The data acquisition and storage module is used to deploy multi-source data acquisition terminals to collect experimental data related to the entire process of IC substrate material research and development. After preprocessing the collected data, a unique R&D traceability code is assigned to each data point and associated with metadata. A hierarchical database is built to realize data linkage and full life cycle traceability. The multi-dimensional statistical analysis module is used to establish a multi-dimensional statistical indicator system, dynamically adjust the indicator weights for different IC substrate material types, use multi-factor interactive analysis methods to explore the correlation between formula parameters, process parameters and performance indicators and screen key parameters, establish an iterative statistical mechanism based on historical data in a hierarchical database, and output statistical analysis results. The dynamic visualization module is used to build standardized visualization models to present data in layers, and configure multi-dimensional interactive functions to support data adjustment, filtering, drill-down and result export; The anomaly warning and optimization suggestion module is used to establish an anomaly warning mechanism based on statistical analysis results and visualization models. It sets parameter statistical thresholds, and automatically triggers warnings and pushes relevant information when data exceeds the thresholds or when parameter correlation fluctuates abnormally. It also generates optimization suggestions by combining multi-factor statistical analysis results and historical data iteration patterns. The data accumulation and iteration module is used to store statistical analysis results, visualization charts, anomaly handling records, and optimization suggestions in a tagged manner according to preset dimensions, supplementing them to a hierarchical database to form a standardized R&D data asset library. Based on the historical data in the data asset library, the statistical analysis algorithm and visualization model are iteratively optimized.

10. The system according to claim 9, characterized in that, The standardized visualization models constructed by the dynamic visualization presentation module include parameter trend models, correlation analysis models, process window models, and traceability visualization models. The parameter trend model uses dynamic line charts or bar charts, the correlation analysis model uses heat maps or scatter plot matrices, the process window model uses contour plots or 3D surface plots, and the traceability visualization model uses time axis plots.

Images