A chemical plating process simulation optimization system and method for metal processing

Abstract

The present invention discloses a chemical plating process simulation optimization system and method for metal processing, which specifically relates to the technical field of metal processing simulation process, and comprises the following steps: before simulation, multiple data dimensions are collected to obtain an initial data set, and data dimensions with hidden dangers of insufficient data volume are identified through preliminary analysis; for these dimensions, whether to resample is determined through deductive analysis, and collection is performed according to the simulation sampling volume; when resampling is required, the collected data replaces the corresponding dimensional data in the initial data set to generate a simulation input data set; if resampling is not required, the initial data set is directly used for simulation, and finally the simulation results are output and applied to actual operation; the present invention can automatically adjust the data collection volume of data dimensions by introducing a dynamic data source optimization mechanism, ensure that the data volume of each parameter is sufficient, and ensure the stability and reliability of the simulation results.

Description

Technical Field

[0001] The present invention relates to the technical field of metal processing simulation technology, and more specifically, to a chemical plating process simulation optimization system and method for metal processing. Background Art

[0002] In industrial manufacturing, especially in the field of metal processing, processing parameters such as temperature, pressure, time, etc. have an important impact on the final quality of the product. Traditional processing methods usually rely on experience or multiple experiments to determine processing parameters. This method is not only time-consuming and labor-intensive, but also prone to large errors. In addition, different combinations of processing parameters have different effects on multiple goals such as energy consumption, efficiency, and processing quality, which makes it difficult to quickly find the optimal processing solution in the actual production process.

[0003] In the prior art, although some simulation technologies have been used to simulate the processing process, there is often a lack of targeted data source optimization mechanisms. For different production environments and different types of metal processing production, the accuracy and reliability of the simulation results are difficult to guarantee. Traditional simulation systems are usually based on static data sources or a single experimental data set, and it is difficult to fully consider the dynamic data changes in multiple dimensions and their comprehensive impact on the processing process. Moreover, most of the existing simulations are based on the analysis and deduction of the initially collected data, without fully considering the sufficiency of the data volume and the differences in the data of each dimension, resulting in insufficient data for certain key parameters or too much redundant data, which further affects the optimization ability of the simulation model. With the increase of data dimensions and data volume, it is also easy to cause problems such as data redundancy and excessive computational complexity. Therefore, a simulation optimization system and method for chemical plating process for metal processing is proposed. Summary of the invention

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for simulating and optimizing a chemical plating process for metal processing comprises the following steps:

[0006] Before the simulation, data collection is performed, i.e., multiple data dimensions are collected separately to obtain the initial data set for the simulation, and then a preliminary analysis is performed to identify the data dimension types that have the potential risk of insufficient data volume;

[0007] For data dimension types that have the potential for insufficient data volume, a deductive analysis is performed, and based on the deductive analysis results, a decision is made as to whether to resample the data dimension type, and when resampling of the data dimension type is required, a simulation sampling amount for the data dimension type is determined;

[0008] When resampling is required, all data dimension types that need to be resampled are resampled according to the determined simulation sampling amount, and then the data of all resampled data dimension types are used to replace the corresponding data dimension type data in the simulation initial data set to obtain the simulation input data set. When resampling is not required, the simulation initial data set is used as the simulation input data set;

[0009] Perform simulations based on the simulation input data set, output the simulation results and apply them.

[0010] In a preferred embodiment, a pre-production, i.e., experimental processing simulation, is first performed, and various parameters of the chemical plating process for metal processing are simulated through simulation to obtain a simulation result, i.e., the optimized parameter combination of each processing parameter, and finally the optimized parameter combination is used for large-scale actual production.

[0011] In a preferred embodiment, performing a preliminary analysis refers to:

[0012] Calculate the correlation between each input data dimension and the simulation result error respectively to obtain the correlation coefficient ; Indicates the input data dimension type, Indicates the simulation result error;

[0013] Then determine the influence coefficient of input data dimension type on the accuracy of simulation results:

[0014] ; represents the correlation coefficient between the j-th input data dimension type and the simulation result error, m is the total number of input data dimension types, Indicates the input data dimension type The corresponding influence coefficient;

[0015] The basic data volume D is preset, according to the basic data volume D and the input data dimension type The corresponding influence coefficient Determine the risk threshold for this data dimension type:

[0016] ; Indicates the input data dimension type threshold of hidden danger.

[0017] In a preferred embodiment, the logic for identifying data dimension types that have a risk of insufficient data volume is:

[0018] The total amount of data collection of the target data dimension type is obtained from the initial data set of the simulation, and then the hidden danger threshold of the target data dimension type obtained by preliminary analysis is obtained. The total amount of data collection of the target data dimension type is compared with the corresponding hidden danger threshold. If the total amount of data collection of the target data dimension type is greater than or equal to the corresponding hidden danger threshold, a normal signal is generated. If the total amount of data collection of the target data dimension type is less than the corresponding hidden danger threshold, a hidden danger signal is generated. When the hidden danger signal is generated, it indicates that the target data dimension type has the hidden danger of insufficient data volume.

[0019] In a preferred embodiment, the inductive analysis refers to:

[0020] Use K-means clustering algorithm to cluster data of all dimensional types. The goal of clustering is to assign similar data dimensions to the same cluster to maximize the similarity within the cluster and minimize the difference between clusters.

[0021] After clustering, check the characteristics of each cluster center, perform pairwise correlation analysis on the dimension type data classified into the same cluster, and obtain the correlation coefficient between the paired data dimension types. ; i and j both represent the numbers corresponding to the data dimension types. Indicates the data dimension type Data dimension type The correlation coefficient between

[0022] Filter out the data dimension types with insufficient data volume in the target cluster, obtain the data volume of each non-risk data dimension type in the target cluster and mark it as , and then generate derived weights for each non-hidden danger data dimension type , derive the weight The calculation formula is:

[0023] ; is the number of non-hidden danger data dimension types in the target cluster, and then the correlation coefficient between the target data dimension type with insufficient data volume in the target cluster and each non-hidden danger data dimension type in the target cluster is obtained and marked as , calculate the derivation index of the target data dimension type that has the risk of insufficient data volume:

[0024] ; The data dimension type that has the risk of insufficient data volume The derived index.

[0025] In a preferred embodiment, determining whether to resample the data dimension type based on the inductive analysis result refers to:

[0026] Get the data dimension type that has the risk of insufficient data volume The derived index , and then compare it with the preset derivation threshold. If the derivation index If the derivation index is greater than or equal to the preset derivation threshold, there is no need to resample the data dimension type. If it is smaller than the preset derivation threshold, the data dimension type needs to be resampled.

[0027] In a preferred embodiment, the determination logic of the simulation sampling amount is:

[0028] Get the standard deviation of all sampled data of target data dimension type i in the initial data set of the simulation , set the following calculation logic:

[0029] ; is the scale factor, is the error corresponding to the data distribution concentration of target data dimension type i; the proportional factor With standard deviation The preset standard deviation range falls into a one-to-one correspondence. The preset standard deviation area range is obtained by the corresponding scale factor ;

[0030] Set a valid numerical standard interval, and count the proportion of all sampled data that fall into the valid numerical standard interval as the success rate ;

[0031] Using fuzzy logic reasoning, the standard deviation As the input variable, the confidence level Zi is taken as the output variable, and the confidence level Zi is obtained by reasoning according to the set fuzzy rules;

[0032] The simulation sampling amount is determined by:

[0033] ; is the preset benchmark standard deviation value, is the simulation sampling volume.

[0034] In a preferred embodiment, a chemical plating process simulation and optimization system for metal processing includes:

[0035] The initial sampling module is used to perform data collection operations before simulation, that is, to collect multiple data dimensions respectively to obtain an initial data set for simulation;

[0036] The preliminary analysis module is used to conduct preliminary analysis on the initial data set of the simulation and identify the data dimension types that have the potential risk of insufficient data volume;

[0037] The inductive analysis module is used to perform inductive analysis on data dimension types that have the potential risk of insufficient data volume, and decide whether to resample the data dimension type based on the inductive analysis results;

[0038] A resampling module is used to determine the simulation sampling amount of the data dimension type when it is necessary to resample the data dimension type. When resampling is required, all data dimension types that need to be resampled are resampled according to the determined simulation sampling amount, and then the data of all resampled data dimension types are used to replace the corresponding data dimension type data in the simulation initial data set to obtain the simulation input data set. When resampling is not required, the simulation initial data set is used as the simulation input data set;

[0039] The simulation module is used to perform simulation according to the simulation input data set, output the simulation results, and obtain the optimized parameter combination of each processing parameter.

[0040] Technical effects and advantages of the present invention:

[0041] The present invention collects data from multiple dimensions and combines it with inductive analysis to automatically identify dimensions with insufficient data volume and issue warnings in a timely manner, indicating the need for further sampling. This dynamic collection mechanism avoids unnecessary sampling waste, optimizes data collection resources, ensures that each data dimension has enough data to support the simulation model, and improves the accuracy and reliability of the simulation results.

[0042] By introducing a dynamic data source optimization mechanism, the present invention can automatically adjust the amount of data collected for data dimensions to ensure that the amount of data for each parameter is sufficient. Through deductive analysis, the present invention can effectively utilize existing data dimensions to determine whether data can be derived through other dimensions, thereby reducing the need for direct sampling. This intelligent derivation mechanism reduces the risk of simulation result deviation caused by insufficient data while ensuring data accuracy, thus ensuring the stability and reliability of the simulation results.

[0043] The present invention obtains the optimized parameter combination of each processing parameter through simulation, and finally applies it to actual production, realizing seamless connection from pre-production experiment to large-scale production. By pre-simulating the optimized parameter combination, it is possible to improve efficiency in actual production, reduce trial and error costs, and ensure the stability of product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0045] Figure 1 This is a schematic diagram of a method for simulating and optimizing a chemical plating process for metal processing in the present invention.

[0046] Figure 2 This is a schematic diagram of a chemical plating process simulation and optimization system for metal processing in the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Reference Figure 1 - Figure 2 The following embodiments are obtained:

[0049] Embodiment 1: A method for simulating and optimizing a chemical plating process for metal processing, comprising the following steps:

[0050] Before the simulation, data collection is performed, that is, multiple data dimensions are collected separately to obtain the initial data set for the simulation, and then a preliminary analysis is performed to identify the data dimension types with the hidden danger of insufficient data volume; preliminary analysis refers to:

[0051] Calculate the correlation between each input data dimension and the simulation result error respectively to obtain the correlation coefficient ; Indicates the input data dimension type. For example, a simulation model involves three input dimensions (temperature, pressure, and time). Indicates the simulation result error. The simulation result error can be measured by one of the following indicators, which can be selected by the user according to the actual situation:

[0052] Error rate: the relative error between the simulation results and the actual experimental results.

[0053] Mean Squared Error (MSE): The average squared error between the simulated output and the true value.

[0054] Mean absolute error (MAE): The mean absolute error between the simulated output and the true value.

[0055] Then determine the influence coefficient of input data dimension type on the accuracy of simulation results:

[0056] ; It represents the correlation coefficient between the j-th input data dimension type and the simulation result error, that is, the correlation coefficient measures the linear relationship between the input dimension and the simulation result error. The larger the absolute value of the correlation coefficient, the greater the impact of the dimension on the simulation result error. You can use the Pearson correlation coefficient or the Spearman correlation coefficient, which is selected by the user according to the actual situation. m is the total number of input data dimension types. Indicates the input data dimension type The corresponding influence coefficient represents the influence weight of the dimension on the simulation results in the overall input dimension. By dividing the correlation of the input dimension by the sum of the correlations of all dimensions, Normalize so that the sum is 1. Reflects its relative importance. If the input data dimension has a greater impact on the simulation result error, more data sampling may be needed to improve its accuracy. If it is smaller, it means that the input data dimension has little influence on the simulation result error, and the sampling amount can be appropriately reduced.

[0057] The basic data volume D is preset, according to the basic data volume D and the input data dimension type The corresponding influence coefficient Determine the risk threshold for this data dimension type:

[0058] ; Indicates the input data dimension type The hidden danger threshold reflects the risk level of insufficient data in this dimension under the current sampling data. D is the basic data volume, which is used to measure the overall demand for data volume. By multiplying the basic data volume with the impact coefficient of the dimension, the hidden danger value of the dimension can be determined. The higher the hidden danger value, the more sampling is needed for this dimension. A larger value indicates that this dimension has a greater impact on the simulation results, and the current data volume is insufficient, so further sampling may be required. A smaller value indicates that the dimension has less impact on the simulation results and the current data volume is sufficient.

[0059] The basic data volume D can be obtained from historical simulation data. By analyzing the accuracy of historical simulation results and the relevant data volume, the average data volume required to achieve a specific simulation accuracy can be calculated. This method is suitable for situations where there is sufficient historical data, and can ensure that the data volume is set based on actual operating experience.

[0060] Simulation accuracy requirements: The determination of the amount of basic data should also take into account the accuracy requirements of the current simulation model. For example, if the simulation model is used in a high-precision technical field, such as the manufacture of aviation components, a larger amount of basic data may be required to ensure the accuracy and reliability of the simulation results. The specific value can be set through expert evaluation or preliminary small-scale experiments.

[0061] Statistical inference: Statistical methods can be used to estimate the appropriate amount of basic data. For example, statistical tools can be used to determine the minimum sample size required to achieve stable simulation results at a specified confidence level. This method is suitable for scenarios where there is a preliminary understanding of the data distribution.

[0062] Parameter sensitivity analysis: Parameter sensitivity analysis is performed to evaluate the impact of different data amounts on simulation results. By changing the amount of input data and observing the changes in simulation output, a "safety threshold" is determined, that is, above this amount of data, the simulation results do not change much, ensuring the stability and repeatability of the simulation.

[0063] The logic for identifying data dimension types that have the risk of insufficient data volume is:

[0064] The total amount of data collection of the target data dimension type is obtained from the initial data set of the simulation, and then the hidden danger threshold of the target data dimension type obtained by preliminary analysis is obtained. The total amount of data collection of the target data dimension type is compared with the corresponding hidden danger threshold. If the total amount of data collection of the target data dimension type is greater than or equal to the corresponding hidden danger threshold, a normal signal is generated. If the total amount of data collection of the target data dimension type is less than the corresponding hidden danger threshold, a hidden danger signal is generated. When the hidden danger signal is generated, it indicates that the target data dimension type has the hidden danger of insufficient data volume.

[0065] By comparing the total amount of data collected and the hidden danger threshold, it is determined whether the data is sufficient to support the calculation of the simulation model. Insufficient data will lead to deviation or inaccuracy in the simulation results, affecting the accuracy and reliability of the model. When the amount of data is insufficient, a hidden danger signal is generated to promptly warn that more data collection is needed to avoid the simulation model being built on insufficient data and ensure the credibility of the results.

[0066] During the data collection process, the system automatically monitors whether the data dimension collection meets the requirements by dynamically comparing the collection volume with the hidden danger threshold. The system automatically identifies dimensions with insufficient data, reduces the need for manual inspection and intervention, and improves the efficiency of collection and simulation.

[0067] Perform a hidden danger analysis on each data dimension to determine whether the data is sufficient to support the simulation. For insufficient data dimensions, a hidden danger signal is issued to prompt further sampling; for dimensions with sufficient data collection, a normal signal is issued to avoid oversampling. Through this logic, data collection resources can be optimized, unnecessary sampling can be avoided, and the focus can be placed on dimensions with insufficient data collection, thereby improving the utilization efficiency of simulation resources.

[0068] The rationality of the amount of data collected is measured by the threshold to ensure that each dimension of the model has sufficient data support, avoiding the deviation or instability of the simulation model caused by insufficient data in some dimensions. The stability and consistency of the model are improved, so that the simulation results have higher reliability in multiple runs.

[0069] The generation of hidden danger signals is equivalent to early warning of data collection risks, indicating which data dimensions are insufficient and intervening in advance to ensure that subsequent simulation analysis can proceed smoothly. It can prevent the adjustment and rework of the simulation model in the later stage due to insufficient data volume, reduce delays in project progress, and improve overall efficiency.

[0070] For data dimension types that have the potential for insufficient data volume, a deductive analysis is performed, and based on the deductive analysis results, a decision is made as to whether to resample the data dimension type, and when resampling of the data dimension type is required, a simulation sampling amount for the data dimension type is determined;

[0071] Deductive analysis refers to:

[0072] Use the K-means clustering algorithm to cluster data of all dimensional types. The goal of clustering is to assign similar data dimensions to the same cluster to maximize the similarity within the cluster and minimize the difference between clusters. The K-means clustering algorithm is an existing technology and will not be elaborated on here. The purpose of this step is to use K-means clustering to group all data dimension types and determine which dimensions have similar characteristics. After clustering, several clusters will be obtained. For each cluster, the data dimensions in it are further analyzed, and the correlation between each dimension and other dimensions is calculated. Specifically: pairwise correlation analysis is performed on the data dimensions in each cluster to understand whether some dimensions can be derived from other dimensions. Data dimensions with strong correlation can be derived from each other, which can reduce the need for direct sampling of dimensions with insufficient data. The size of the correlation value: The larger the value, the stronger the correlation between the dimensions and the better the derivation. A correlation close to 1 indicates a strong correlation, and a correlation close to 0 indicates a weak correlation or irrelevance.

[0073] By clustering data dimensions, we can identify which data dimensions have the risk of insufficient data. Dimensions with insufficient data may not be able to be derived from other dimensions, so more data collection is required. After determining the dimensions with insufficient data, we can evaluate the derivation ability of these dimensions by analyzing the number of other data dimensions in the same cluster and their correlation with the target dimension. The level of derivation ability here depends on the degree of influence of other dimensions on this dimension: dimensions with large influence can reduce direct sampling of insufficient dimensions through derivation.

[0074] The derivability index is used to quantify whether a data dimension can be derived from other dimensions. The higher the derivability index, the easier it is for the dimension to be derived from other dimensions, so additional sampling may not be required. When the derivability index exceeds the preset threshold, it can be determined that the data dimension can be derived from other dimensions, thereby reducing or avoiding the need for additional sampling of the dimension.

[0075] The specific logic of obtaining the inference index is as follows: after clustering, check the characteristics of each cluster center, perform pairwise correlation analysis on the dimension type data classified into the same cluster, and obtain the correlation coefficient between the paired data dimension types. ; i and j both represent the numbers corresponding to the data dimension types. Indicates the data dimension type Data dimension type The correlation coefficient between .

[0076] If the total amount of data collected of the target data dimension type is greater than or equal to the corresponding hidden danger threshold, a normal signal is generated. If the total amount of data collected of the target data dimension type is less than the corresponding hidden danger threshold, a hidden danger signal is generated. When a hidden danger signal is generated, it indicates that the target data dimension type has a hidden danger of insufficient data volume. When a normal signal is generated, it indicates that the target data dimension type does not have a hidden danger of insufficient data volume, and the target data dimension type is a non-hidden danger data dimension type.

[0077] Filter out the data dimension types with insufficient data volume in the target cluster, obtain the data volume of each non-risk data dimension type in the target cluster and mark it as , and then generate derived weights for each non-hidden danger data dimension type , derive the weight The calculation formula is:

[0078] ; is the number of non-hidden danger data dimension types in the target cluster. In each cluster, first find the dimensions that do not have the hidden danger of insufficient data volume, and calculate the total amount of these dimensions. This total amount reflects the data resources of all other dimensions in the same cluster that can be provided to the target data dimension for derivation. The larger the total amount, the more non-hidden danger data dimensions there are in the same cluster, which provides more support for subsequent derivation. If the total amount is small, it means that there are fewer dimensions available for derivation and the derivation ability is limited.

[0079] Assign a derivation weight to each non-hidden danger data dimension. The derivation weight is generated by the ratio of the amount of collected data for each non-hidden danger data dimension type relative to the amount of data for other non-hidden danger dimensions. In other words, dimensions with higher weights contribute more to the derivation. The purpose of generating derivation weights is to measure the contribution of each non-hidden danger data dimension to the derivation of the target dimension. The larger the derivation weight, the larger the amount of data for the non-hidden danger data dimension relative to other dimensions, which means that it plays a more important role in the derivation process and contributes more. Dimensions with smaller derivation weights have weaker support for the derivation of the target dimension and may not play as important a role in the derivation process as dimensions with larger weights.

[0080] Then, the correlation coefficient between the target data dimension type with insufficient data volume risk in the target cluster and each non-risk data dimension type in the target cluster is obtained and marked as , calculate the derivation index of the target data dimension type that has the risk of insufficient data volume:

[0081] ; The data dimension type that has the risk of insufficient data volume The derivation weights of all non-hidden danger dimensions are used to comprehensively calculate the derivation index of the target dimension.

[0082] The correlation coefficient is used to measure the strength of the correlation between the target data dimension and each non-hidden danger data dimension. This correlation coefficient is used to determine the extent to which one dimension can be explained or derived by another dimension. The derivation weight measures the contribution of the data volume of each non-hidden danger data dimension to the derivation. The larger the derivation weight, the more important role the non-hidden danger dimension plays in the derivation process. The higher its data volume, the stronger its derivation ability. Multiplying the correlation coefficient of each non-hidden danger dimension by its derivation weight indicates the comprehensive derivation ability of the non-hidden danger dimension to the target data dimension. Dimensions with high correlation coefficients and important derivation weights have the greatest derivation ability. If a non-hidden danger dimension has a high correlation coefficient and a large derivation weight, the value of the product will be high, indicating that the dimension has a very strong derivation ability to the target data dimension. If the correlation coefficient or derivation weight is small, then their product value will also be small, indicating that the dimension has a limited role in the derivation.

[0083] The derivability index is the sum of the product of the correlation coefficients of all non-hidden danger dimensions and the derivation weights. It represents the total derivation capability that the target data dimension can obtain from other dimensions. This index can help determine whether additional data collection of the target data dimension is required. The higher the derivability index, the more information support the target data dimension can obtain from other non-hidden danger dimensions, thereby reducing the need for data collection. If the derivability index is low, it means that the target data dimension is difficult to derive from other dimensions, and more data collection may be required.

[0084] The final derivability index determines whether the target data dimension needs additional data collection. If the derivability index reaches a certain preset threshold, it can be considered that the derivation capability of this dimension through other non-hidden danger dimensions is sufficient and no additional data collection is required.

[0085] The preset threshold can be determined by:

[0086] By statistically analyzing historical simulation data, we can find out the specific impact of different levels of derivability index on the accuracy of simulation results. Based on these data, we can set a threshold value so that when the derivability index is lower than this threshold value, the error of the simulation result increases significantly, and when it is higher than this threshold value, the error is within an acceptable range.

[0087] Rely on the experience of industry experts to set the threshold. Experts can estimate a reasonable threshold of the derivation index based on their understanding of specific processing and material characteristics to ensure that the data quality meets the standards of production and R&D.

[0088] The threshold is determined through actual experimental verification. The simulation model is run under different derivation indexes, the performance of the model and the accuracy of the output results are recorded and analyzed, and the derivation index value at which the model performance begins to significantly decline is found as the preset threshold.

[0089] If the inference index exceeds the preset threshold, a signal can be generated, indicating that the target data dimension does not need additional sampling. If the inference index is lower than the threshold, a signal of insufficient data volume will be generated, indicating that further data collection is required to ensure the accuracy of the simulation results. Therefore, deciding whether to resample the data dimension type based on the inference analysis results means: obtaining the data dimension type with insufficient data volume risk The derived index , and then compare it with the preset derivation threshold. If the derivation index If the derivation index is greater than or equal to the preset derivation threshold, there is no need to resample the data dimension type. If it is smaller than the preset derivation threshold, the data dimension type needs to be resampled.

[0090] When resampling is required, all data dimension types that need to be resampled are resampled according to the determined simulation sampling amount, and then the data of all resampled data dimension types are used to replace the corresponding data dimension type data in the simulation initial data set to obtain the simulation input data set. When resampling is not required, the simulation initial data set is used as the simulation input data set.

[0091] The standard deviation is a key indicator for measuring the fluctuation of data distribution and indicates the degree of data dispersion. The larger the standard deviation, the greater the data fluctuation and the wider the data distribution range; the smaller the standard deviation, the more concentrated the data and the smaller the difference between the data. In the process of calculating the sampling amount, the standard deviation determines the size of the data error, which affects the final sampling amount. Therefore, the logic for determining the simulation sampling amount is: obtain the standard deviation of all sampled data of the target data dimension type i in the initial data set of the simulation , set the following calculation logic:

[0092] ; is the scale factor, is the error corresponding to the data distribution concentration of target data dimension type i; the proportional factor With standard deviation The preset standard deviation range falls into a one-to-one correspondence relationship. The preset standard deviation area range is obtained by the corresponding scale factor ; Calculate the error of data distribution in the target data dimension type. This error is corrected by a scale factor to correct the standard deviation. The scale factor reflects the degree to which the actual data deviates from the ideal standard deviation. The error is the deviation between the actual standard deviation and the ideal standard deviation. The larger the error, the greater the degree to which the data deviates from the ideal standard deviation. The simulation sampling volume may need to be increased to improve the accuracy of the data. When the error is small, it means that the data is closer to the ideal state, the sampling volume can be appropriately reduced, and the scale factor Usually the value is 1 or 2, and a cutoff value can be set. When it falls on one side of the cutoff value, for example, the standard deviation of the temperature sampling during the chemical plating process is 1, then the proportional factor The value 1 indicates a small error, and the standard deviation When it falls on the other side of the cutoff value, for example, the standard deviation of the temperature sampling during the chemical plating process is 3, then the proportional factor A value of 2 indicates a large error.

[0093] The cutoff value of a scaling factor is usually a numerical boundary that is used to distinguish different levels of data processing strategies or weight assignments. For example, in a simulation model, the scaling factor may be used to weight the contribution of different data dimensions, and the cutoff value is used to determine when to increase or decrease the weight of a specific data dimension.

[0094] Methods for determining cut-off values ​​usually rely on the following:

[0095] Statistical analysis: Statistical analysis is performed on historical data to observe the impact of different scaling factor values ​​on simulation results. By analyzing data sensitivity and output changes, a scaling factor cutoff value that optimizes the results can be found.

[0096] Experimental verification: Set different scaling factors, conduct a series of experiments or simulations, and record the results of each setting. By comparing the performance or accuracy under different settings, determine the optimal scaling factor cutoff value.

[0097] Error tolerance: Set the cutoff value based on the error tolerance of the application scenario. If the simulation model is very sensitive to errors, a lower cutoff value may be required to maintain high accuracy; conversely, if the model can tolerate larger errors, the cutoff value can be relatively high.

[0098] Expertise: Use the experience and knowledge of industry experts to set a reasonable cutoff value based on the needs of the specific application area and previous case studies.

[0099] Set a valid numerical standard interval, and count the proportion of all sampled data that fall into the valid numerical standard interval as the success rate ; Set the standard interval of valid values, and count the proportion of all sampled data that fall within the valid interval. This proportion value indicates the validity of the data, which is usually measured by the success rate. The larger the proportion value, the more data falls within the valid range, the higher the validity of the data, and the sampling volume can be appropriately reduced. Conversely, the smaller the proportion value, the wider the data distribution or deviates from the expected range, and the sampling volume needs to be increased to ensure the simulation accuracy. For example, the standard interval of valid values ​​for temperature sampling in the chemical plating process is 25 to 50°C. The 45°C data in the collected data sequence is valid data, and the 80°C data is invalid data.

[0100] The valid numerical standard interval is a method used in data analysis, simulation or other technical applications to determine whether the data meets a specific standard or range. This interval defines the acceptable range of data to ensure the accuracy and reliability of the analysis or simulation. The method of determining this interval depends on the needs of the specific application, the characteristics of the data and industry standards. The valid numerical standard interval is a numerical range used to define the range of data values ​​that are considered valid or acceptable when performing technical processing or analysis. When data falls within this range, it is generally considered to be in line with expectations, reliable quality or meet specific operational requirements.

[0101] First, the purpose of setting the interval must be clear, such as to ensure product quality, meet safety standards, or ensure the accuracy of data analysis. The analysis involves the characteristics of the data, such as distribution type, volatility, extreme value conditions, etc.

[0102] Reference industry standards: Consult relevant industry standards or specifications, which may already provide the numerical range that the data should meet.

[0103] Historical data analysis: Use historical data to determine a reasonable range. Determine the interval in which most data points fall through statistical analysis such as mean, standard deviation, quantiles, etc. Analyze outliers in historical data to determine whether they should be included in the valid interval.

[0104] Experiments and tests: Experiments are conducted to verify the performance or effect of data within different numerical ranges, and the effective numerical standard interval is adjusted according to the results. For example, in materials science, testing of material strength can help determine the effective strength range used in structural safety analysis.

[0105] Assume that in the chemical production process, it is crucial to monitor the temperature of a chemical reaction. Through historical data analysis, it is found that the product quality is optimal when the reaction temperature is in the range of 200°C to 250°C. Therefore, this range can be set as the standard interval of valid values. Further experiments show that exceeding this range may lead to safety risks or reduced product quality, thus verifying the applicability of this range.

[0106] Using fuzzy logic reasoning, the standard deviation As the input variable, the confidence level Zi is taken as the output variable, and reasoning is performed according to the set fuzzy rules to obtain the output confidence level Zi.

[0107] Define the standard deviation of the initial data collection as the input variable of the fuzzy logic system. You can divide the standard deviation into several fuzzy sets to indicate the degree of data fluctuation. For example: Small standard deviation: indicates that the data fluctuation is small and the data is concentrated. Medium standard deviation: indicates that the data fluctuation is moderate and the data is dispersed. Large standard deviation: indicates that the data fluctuation is large and the data is dispersed. The specific fuzzy set definition can be set according to the data situation, and the range of standard deviation can be determined by expert experience or historical data.

[0108] Define fuzzy output: Confidence level: Use the confidence level Zi as the output variable of the fuzzy logic system. The confidence level can be divided into several fuzzy sets as needed, such as: Low confidence level: corresponds to a lower confidence level, such as 90% (Zi=1.645). Medium confidence level: corresponds to a medium confidence level, such as 95% (Zi=1.96). High confidence level: corresponds to a higher confidence level, such as 99% (Zi=2.576).

[0109] Define fuzzy inference rules: These rules determine the relationship between the size of the standard deviation and the confidence level. For example: Rule 1: If the standard deviation is "small", the confidence level is "high". Rule 2: If the standard deviation is "medium", the confidence level is "medium". Rule 3: If the standard deviation is "large", the confidence level is "low". These rules can be set flexibly to infer the appropriate confidence level according to different ranges of standard deviation.

[0110] Using the fuzzy logic system, input the standard deviation data, fuzzify it, and then perform reasoning based on the set fuzzy rules to output a fuzzified confidence level. Finally, the specific confidence coefficient is obtained through the defuzzification process (the commonly used method is the weighted average method). For example, when the standard deviation is 0.5, the fuzzy logic reasoning may output a value between 95% and 99% confidence level (such as Zi=1.9).

[0111] Fuzzy logic reasoning: Calculation of standard deviation: By analyzing the initial data set, the standard deviation of each data dimension is calculated to measure the volatility and uncertainty of the data. Determination of confidence level: Using fuzzy logic reasoning, the confidence level of the data is determined according to the size of the standard deviation. For example, if the standard deviation is less than 0.5, the confidence level obtained by reasoning is between 95% and 99% (assuming the confidence level is 1.9).

[0112] Success rate evaluation: Success rate calculation: The success rate is defined according to the proportion of data falling within the preset valid range, which helps to evaluate the validity of the sampling data and the reliability of the simulation model.

[0113] The simulation sampling amount is determined by: ; is the preset benchmark standard deviation value, is the simulation sampling volume. The sampling volume required for simulation is calculated by the formula. The size of the sampling volume depends on the combined effect of the previous error, proportion value, success rate, confidence level, standard deviation, and benchmark standard deviation value. This sampling volume will be used to determine the amount of data that needs to be collected in subsequent simulations to ensure the accuracy of the data. The larger the sampling volume, the wider the data distribution or the larger the deviation, and the simulation process requires more data support; when the sampling volume is small, it means that the data is more concentrated, and not too much data is needed to meet the simulation requirements.

[0114] Perform simulations based on the simulation input data set, output the simulation results and apply them.

[0115] First, pre-production, i.e. experimental processing simulation, is carried out. Through simulation, various parameters of the chemical plating process for metal processing are simulated to obtain the simulation result, i.e. the optimized parameter combination of each processing parameter. Finally, the optimized parameter combination is used for large-scale actual production.

[0116] Embodiment 2: A chemical plating process simulation optimization system for metal processing, comprising:

[0117] The initial sampling module is used to perform data collection operations before simulation, that is, to collect multiple data dimensions respectively to obtain an initial data set for simulation;

[0118] The preliminary analysis module is used to conduct preliminary analysis on the initial data set of the simulation and identify the data dimension types that have the potential risk of insufficient data volume;

[0119] The inductive analysis module is used to perform inductive analysis on data dimension types that have the potential risk of insufficient data volume, and decide whether to resample the data dimension type based on the inductive analysis results;

[0120] A resampling module is used to determine the simulation sampling amount of the data dimension type when it is necessary to resample the data dimension type. When resampling is required, all data dimension types that need to be resampled are resampled according to the determined simulation sampling amount, and then the data of all resampled data dimension types are used to replace the corresponding data dimension type data in the simulation initial data set to obtain the simulation input data set. When resampling is not required, the simulation initial data set is used as the simulation input data set;

[0121] The simulation module is used to perform simulation according to the simulation input data set, output the simulation results, and obtain the optimized parameter combination of each processing parameter.

[0122] It should be noted that the simulation module is used to perform simulations according to the preset simulation strategy. The specific simulation strategy is not specified. The simulation results are output to obtain the optimized parameter combination of each processing parameter. Here is a simulation strategy that can be implemented:

[0123] The goal of the parameter optimization simulation strategy for metal processing is to optimize the key processing parameters (such as temperature, pressure and processing time) in metal processing through simulation to improve product quality and production efficiency.

[0124] Preset simulation strategies: Based on past experience and criteria set by experts, a set of simulation strategies is defined in advance, including the initial range of machining parameters and optimization goals, such as minimizing energy consumption and improving production efficiency.

[0125] Step 1: Set parameter range: Set the initial range of each processing parameter. For example, the temperature is set between 200 and 400 degrees, the pressure is set between 50 and 150, and the processing time is set between 5 and 20 minutes. These ranges are preset based on expert experience and can be adjusted according to different processing requirements.

[0126] Step 2: Random sampling and simulation: Use the random sampling method to select different values ​​from the preset range of each parameter for simulation. For example, select multiple temperature values ​​randomly between 200 and 400 degrees, select multiple pressure values ​​between 50 and 150 degrees, and select multiple time values ​​between 5 and 20 minutes. According to the random sampling method, it is guaranteed that each parameter within the set range has sufficient coverage to ensure data diversity.

[0127] Step 3: Simulate and calculate and output results: Simulate and calculate each set of randomly selected processing parameters to obtain the corresponding processing results. For example, the system obtains product quality, processing time, energy consumption and other indicators through simulation. According to the preset simulation strategy, the system automatically simulates the processing effect of each set of parameters without manual adjustment.

[0128] Simulation strategy example: Example 1: Simulation strategy based on grid search, purpose: within the range of each processing parameter, use the grid search method to gradually increase the value of each parameter to perform a comprehensive simulation. For example, the temperature gradually increases from 200 to 400, and is simulated every 50 degrees; the pressure gradually increases from 50 to 150, and is simulated every 20 minutes; the time gradually increases from 5 minutes to 20 minutes, and is simulated every 5 minutes. This preset simulation strategy can ensure comprehensive coverage of all parameter combinations, and ultimately select the optimal parameter combination from the simulation results.

[0129] Example 2: Simulation strategy based on optimization algorithm, purpose: use optimization algorithm (such as simulated annealing or genetic algorithm) to optimize parameters. For example, through simulated annealing algorithm, a set of processing parameters are randomly selected initially, and the parameters are gradually adjusted so that the processing results gradually approach the preset goals (such as minimum energy consumption or highest quality). This type of simulation strategy is preset, and the system will automatically adjust the parameters according to the steps of the algorithm until the optimal solution is found. The optimized parameter combination can significantly reduce the simulation time.

[0130] Simulation result output: According to the preset simulation strategy, the simulation result will output the optimal combination of each processing parameter. For example, after simulation, the system concluded that the processing effect is best when the temperature is 350 degrees, the pressure is 120, and the processing time is 10 minutes. Whether using random sampling, grid search, or optimization algorithm, these strategies are preset before the simulation starts, and automatically simulated according to the target, and finally the optimized parameter combination is obtained.

[0131] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0132] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean 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 application.

[0133] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0134] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0135] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for simulating and optimizing a chemical plating process for metal processing, characterized in that: The following steps are involved: Before the simulation, data collection is performed, i.e., multiple data dimensions are collected separately to obtain the initial data set for the simulation, and then a preliminary analysis is performed to identify the data dimension types that have the potential risk of insufficient data volume; For data dimension types that have the potential risk of insufficient data volume, a deductive analysis is performed, and based on the deductive analysis results, a decision is made as to whether to resample the data dimension type, and when resampling of the data dimension type is required, the simulation sampling amount of the data dimension type is determined; When resampling is required, all data dimension types that need to be resampled are resampled according to the determined simulation sampling amount, and then the data of all resampled data dimension types are used to replace the corresponding data dimension type data in the simulation initial data set to obtain the simulation input data set. When resampling is not required, the simulation initial data set is used as the simulation input data set; Perform simulations based on simulation input data sets, output simulation results and apply them; Conducting a preliminary analysis means: Calculate the correlation between each input data dimension and the simulation result error respectively to obtain the correlation coefficient ; Indicates the input data dimension type, Indicates the simulation result error; Then determine the influence coefficient of input data dimension type on the accuracy of simulation results: ; represents the correlation coefficient between the j-th input data dimension type and the simulation result error, m is the total number of input data dimension types, Indicates the input data dimension type The corresponding influence coefficient; The basic data volume D is preset, according to the basic data volume D and the input data dimension type The corresponding influence coefficient Determine the risk threshold for this data dimension type: ; Indicates the input data dimension type The hidden danger threshold; The logic for identifying data dimension types that have the risk of insufficient data volume is: The total amount of data collection of the target data dimension type is obtained from the initial data set of the simulation, and then the hidden danger threshold of the target data dimension type obtained by preliminary analysis is obtained, and the total amount of data collection of the target data dimension type is compared with the corresponding hidden danger threshold. If the total amount of data collection of the target data dimension type is greater than or equal to the corresponding hidden danger threshold, a normal signal is generated. If the total amount of data collection of the target data dimension type is less than the corresponding hidden danger threshold, a hidden danger signal is generated. When the hidden danger signal is generated, it indicates that the target data dimension type has a hidden danger of insufficient data volume; Deductive analysis refers to: Use K-means clustering algorithm to cluster data of all dimensional types. The goal of clustering is to assign similar data dimensions to the same cluster to maximize the similarity within the cluster and minimize the difference between clusters. After clustering, check the characteristics of each cluster center, perform pairwise correlation analysis on the dimension type data classified into the same cluster, and obtain the correlation coefficient between the paired data dimension types. ; i and j both represent the numbers corresponding to the data dimension types. Indicates the data dimension type Data dimension type The correlation coefficient between Filter out the data dimension types with insufficient data volume in the target cluster, obtain the data volume of each non-risk data dimension type in the target cluster and mark it as , and then generate derived weights for each non-hidden danger data dimension type , derive the weight The calculation formula is: ; is the number of non-hidden danger data dimension types in the target cluster, and then the correlation coefficient between the target data dimension type with insufficient data volume in the target cluster and each non-hidden danger data dimension type in the target cluster is obtained and marked as , calculate the derivation index of the target data dimension type that has the risk of insufficient data volume: ; The data dimension type that has the risk of insufficient data volume The derived index of Deciding whether to resample the data dimension type based on the results of the inferential analysis means: Get the data dimension type that has the risk of insufficient data volume The derived index , and then compare it with the preset derivation threshold. If the derivation index If the derivation index is greater than or equal to the preset derivation threshold, there is no need to resample the data dimension type. If it is smaller than the preset derivation threshold, the data dimension type needs to be resampled.

2. The method for simulating and optimizing chemical plating process for metal processing according to claim 1, characterized in that: First, pre-production, i.e. experimental processing simulation, is carried out. Through simulation, various parameters of the chemical plating process for metal processing are simulated to obtain the simulation result, i.e. the optimized parameter combination of each processing parameter. Finally, the optimized parameter combination is used for large-scale actual production.

3. The method for simulating and optimizing chemical plating process for metal processing according to claim 2, characterized in that: The logic for determining the simulation sampling amount is: Get the standard deviation of all sampled data of target data dimension type i in the initial data set of the simulation , set the following calculation logic: ; is the scale factor, is the error corresponding to the data distribution concentration degree of target data dimension type i; Scale Factor With standard deviation The preset standard deviation range falls into a one-to-one correspondence. The preset standard deviation area range is obtained by the corresponding scale factor ; Set a valid numerical standard interval, and count the proportion of all sampled data that fall into the valid numerical standard interval as the success rate ; Using fuzzy logic reasoning, the standard deviation As the input variable, the confidence level Zi is taken as the output variable, and the confidence level Zi is obtained by reasoning according to the set fuzzy rules; The simulation sampling amount is determined by: ; is the preset benchmark standard deviation value, is the simulation sampling volume.

4. A metal processing chemical plating process simulation optimization system, which is implemented based on a metal processing chemical plating process simulation optimization method according to any one of claims 1 to 3, characterized in that: include: The initial sampling module is used to perform data collection operations before simulation, that is, to collect multiple data dimensions respectively to obtain an initial data set for simulation; The preliminary analysis module is used to conduct preliminary analysis on the initial data set of the simulation and identify the data dimension types that have the potential risk of insufficient data volume; The inductive analysis module is used to perform inductive analysis on data dimension types that have the potential risk of insufficient data volume, and decide whether to resample the data dimension type based on the inductive analysis results; A resampling module is used to determine the simulation sampling amount of the data dimension type when it is necessary to resample the data dimension type. When resampling is needed, all data dimension types that need to be resampled are resampled according to the determined simulation sampling amount, and then the data of all resampled data dimension types are used to replace the corresponding data dimension type data in the simulation initial data set to obtain the simulation input data set. When resampling is not needed, the simulation initial data set is used as the simulation input data set; The simulation module is used to perform simulation according to the simulation input data set, output the simulation results, and obtain the optimized parameter combination of each processing parameter.

Images