Data management method and system for silicone process formulation parameters

Abstract

This invention belongs to the field of parameter data management technology. It provides a data management method and system for silicone process formulation parameters, including: real-time acquisition of multiple state variable data affecting the post-processing quality of silicone products, wherein the state variable data includes inherent equipment state variables and environmental state variables; based on the acquired state variable data, a compensation rule base is invoked to calculate hierarchical and multi-factor coupled compensation values; wherein the multi-level includes: a main effect compensation layer, a coupling effect compensation layer, and an environmental fine-tuning layer. By constructing a three-level compensation architecture of main effect—coupling effect—environmental fine-tuning, it can not only handle the impact of independent changes in a single state variable on the process, but also identify and quantify the nonlinear interactions between multiple key variables, thereby improving the compensation accuracy under complex and dynamic operating conditions and reducing problems of neglecting coupling relationships between variables, leading to inaccurate compensation or overlooking certain aspects.

Description

Technical Field

[0001] This invention belongs to the field of parameter data management technology, specifically a data management method and system for silicone process formulation parameters. Background Technology

[0002] Silicone products, especially high-performance liquid silicone rubber (LSR) products, are highly dependent on precise process formulations and stable production process control. With the development of intelligent manufacturing, systematic and digital data management of silicone process formulation parameters has become an industry trend. Traditional and existing data management methods focus on static recording and single-point control of formulation components and basic process parameters (such as vulcanization temperature and time), and attempt to compensate for process deviations by establishing an empirical rule base, aiming to improve production consistency and quality stability. However, existing technical solutions have shortcomings when dealing with highly complex post-processing processes such as plasma treatment and secondary sulfurization: the quality of these processes is affected by multiple dynamic variables, such as equipment status (electrode wear, gas purity), environmental parameters (temperature, humidity, air pressure) and the results of previous processes. Existing rule-based compensation methods mostly adopt "if-then" single-factor independent compensation logic, that is, a fixed compensation value is set for a single variable (such as "if the electrode is used for too long, then increase the processing time"). This method ignores the objective fact that multiple state variables change simultaneously and have nonlinear interactions in the actual production process. For example, when electrode wear and gas purity decrease at the same time, the negative impact on the treatment effect is not simply superimposed, but may produce a multiplication or cancellation effect, which leads to a sharp drop in the accuracy of the compensation rule under complex working conditions. The compensation result often "takes one thing and loses another", or even introduces new fluctuations. To this end, the present invention provides a data management method and system for silicone process formulation parameters. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0004] The technical solution adopted by this invention to solve its technical problem is: a data management method for silicone process formulation parameters, including: Real-time acquisition of multiple state variable data affecting the quality of post-processing of silicone products, including inherent equipment state variables and environmental state variables; Based on the acquired state variable data, the compensation rule library is invoked to calculate the compensation value in a hierarchical and multi-factor coupled manner. The multi-layer structure includes: main effect compensation layer, coupling effect compensation layer, and environmental fine-tuning layer. By integrating the basic compensation amount calculated by the main effect compensation layer, the interaction compensation amount calculated by the coupling effect compensation layer, and the fine-tuning compensation amount calculated by the environmental fine-tuning layer, a comprehensive compensation scheme for the current post-processing process is generated. The comprehensive compensation scheme is output to the production control system or equipment controller to complete the adaptive adjustment of the post-processing parameters of silicone products.

[0005] A data management system for silicone process formulation parameters, comprising: State variable data acquisition module: Real-time acquisition of multiple state variable data that affect the quality of post-processing of silicone products, including inherent equipment state variables and environmental state variables; Process parameter compensation analysis module: Based on the acquired state variable data, it calls the compensation rule library to calculate compensation values ​​in a hierarchical and multi-factor coupled manner, specifically including: The multi-layer structure includes: main effect compensation layer, coupling effect compensation layer, and environmental fine-tuning layer; The main effect compensation layer includes: calculating the basic compensation amount for a single state variable based on the skewness of the single state variable to the standard value; The coupling effect compensation layer includes: calculating the compensation amount of the interaction caused by the common deviation of at least two state variables based on multivariate interaction; The environmental fine-tuning layer includes: calculating the fine-tuning compensation amount based on real-time environmental fluctuations; Process parameter adjustment module: Integrates basic compensation amount, interaction compensation amount and fine-tuning compensation amount to generate a comprehensive compensation scheme for the current process, and outputs the comprehensive compensation scheme to the production control system or equipment controller to complete the adaptive adjustment of the post-processing parameters of silicone products.

[0006] The beneficial effects of this invention are as follows: This invention constructs a three-level compensation architecture of main effect, coupling effect, and environmental fine-tuning. It can not only handle the impact of independent changes in a single state variable on the process, but also identify and quantify the nonlinear interaction between multiple key variables, thereby improving the compensation accuracy under complex and dynamic conditions and reducing the problems of neglecting one aspect or inaccurate compensation caused by ignoring the coupling relationship between variables. This invention collects multi-source data such as equipment status and environmental parameters in real time, and performs intelligent analysis and compensation calculations based on a dynamic rule base. This enables the automatic and real-time adjustment of key process parameters, ensuring that process conditions always approach the optimal settings. It reduces product quality fluctuations caused by factors such as equipment aging, material fluctuations, and environmental changes, and improves the consistency of post-processing quality and batch-to-batch stability of silicone products. This invention avoids over-processing or under-processing caused by process deviations through precise real-time compensation, thereby reducing energy waste and material loss. Furthermore, automated compensation reduces manual intervention and machine setup time, which helps improve overall equipment efficiency and production line utilization. Attached Figure Description

[0007] The invention will now be further described with reference to the accompanying drawings.

[0008] Figure 1 This is a flowchart of the steps of the data management method for silicone process formulation parameters of the present invention; Figure 2 This is a flowchart of step two of the data management method for silicone process formulation parameters of the present invention; Figure 3 This is a module architecture diagram of the data management system for silicone process formulation parameters of this invention. Detailed Implementation

[0009] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments. Example

[0010] Please see Figure 1 and Figure 2 As shown in the embodiment of the present invention, the data management method for silicone process formulation parameters includes the following steps: Step 1: Real-time acquisition of data on multiple state variables that affect the quality of post-processing of silicone products, including inherent equipment state variables and environmental state variables; In step one, for the target post-processing technology, the key variables that need to be monitored are identified. These inherent state variables of the equipment include, but are not limited to: Cumulative working time of electrodes / devices (e.g., total discharge time of plasma processors); Loss indicators of core components (such as electrode morphology parameters and lamp luminous intensity attenuation coefficient). Working medium condition (e.g., purity of the gas being processed, temperature and flow rate of the coolant, vacuum level); Core direct parameters of the process (such as the intensity of characteristic peaks in plasma emission spectrum, actual output power and setting deviation). Preceding process parameters (such as primary sulfidation degree, surface silane residual rate); Environmental state variables include: Production microenvironment parameters: real-time temperature and relative humidity of the processing area; Atmospheric state parameters: workshop atmospheric pressure on the current day; Time-related parameters: material storage time, batch production period; It should be noted that the aforementioned inherent variables of the equipment and environmental state variables are acquired through a deployed data acquisition network. Specifically, the data acquisition network consists of: connecting sensors to the process equipment, including at least: timers, optical sensors, gas analyzers, pressure sensors, temperature and humidity sensors, and spectrometers; for intelligent devices with data interfaces (such as modern plasma processing machines and vulcanizing ovens), internal state parameters are directly read through device protocols (such as Modbus, OPC UA); and establishing edge computing nodes or data acquisition gateways to uniformly receive, standardize, and initially filter multi-source, heterogeneous sensor signals.

[0011] Step Two: Based on the acquired state variable data, the compensation rule library is invoked to calculate the compensation value in a hierarchical and multi-factor coupled manner, specifically including: The multi-layer structure includes: main effect compensation layer, coupling effect compensation layer, and environmental fine-tuning layer; The main effect compensation layer includes: calculating the basic compensation amount for a single state variable based on the skewness of the single state variable to the standard value; In the main effect compensation layer, for any process parameter, the dominant state variable that is strongly correlated with the process to be compensated is obtained; Furthermore, the process for obtaining a single dominant state variable with strong correlation is as follows: Based on historical production data, the statistical correlation between each state variable and the actual optimal set value of the target process parameter is calculated. Specifically, by calculating the mutual information coefficient, state variables whose absolute value of the mutual information coefficient is greater than a threshold are extracted as a set of candidate state variables that are related to the target process parameters; Partial least squares regression analysis was used to identify the single or multiple dominant state variables that have the highest explanatory power for the variation of process parameters from the set of candidate state variables; Construct the independent variable matrix X and dependent variable matrix Y based on historical production data: X: Dimension is n×p, containing the observations of p candidate state variables in n historical batches; Y: The dimension is n×q, containing the actual optimal set values ​​of the corresponding q target process parameters; X and Y are centered and standardized to eliminate the influence of dimensions; A latent variable model between X and Y is constructed using a nonlinear iterative partial least squares algorithm. Using K-fold cross-validation (e.g., K=10), we iterate through the number of latent variables and calculate the sum of squared predicted residuals PRESS for each validation. The value of L when PRESS first reaches its minimum or when the change tends to level off is determined as the optimal number of latent variables L. We then calculate the sum of squared predicted residuals of the model under different numbers of latent variables. Choose the number of latent variables that minimizes the sum of squared predicted residuals to avoid overfitting and underfitting; Establish the final relationship model between X and Y: Y = XB + E, where B is the regression coefficient matrix and E is the residual matrix; Calculate the variable importance projection value for each state variable; Optionally, the formula for calculating the variable importance projection value is as follows:

[0012] Where p is the number of candidate state variables, and L is the number of optimal latent variables. The sum of squares of variation in Y explained by the l-th latent variable. Let be the weight of the j-th candidate state variable among the l-th latent variables; If the variable importance projection value is greater than 1.0, it indicates that the corresponding candidate state variable is significantly important in explaining the variation of Y. If the projected value of the variable importance is greater than 0.8 and less than or equal to 1.0, it indicates that the corresponding candidate state variable has moderate importance in explaining the variation of Y. If the variable importance projection value is less than or equal to 0.8, it means that the corresponding candidate state variable contributes little to the explanation of the variation in Y. Standardized regression coefficients are extracted from the final relational model. For each target process parameter, the regression coefficients of each candidate state variable are evaluated. Absolute value of coefficient: reflects the intensity of the influence of candidate state variables on target process parameters; The sign of the coefficient indicates the direction of influence (positive correlation / negative correlation); The dominant state variables are selected according to the priority rules. The first priority is to select candidate state variables in which the projected importance value of all target process parameters is greater than 1.0 and the regression coefficients are in the same direction. Second priority: If no variable satisfies the first priority, then select independently for each target process parameter: For each target process parameter, the state variable with the highest variable importance projection value is selected as the dominant state variable. If a candidate state variable is the dominant variable for multiple target process parameters, the direction of its regression coefficients among different target process parameters is verified. If the regression coefficients are in the same direction, they are identified as the common dominant state variables of these target process parameters; If the regression coefficients conflict in direction, the direction of influence on the core target process parameter shall be prioritized for determination, and for the other conflicting target process parameters, the state variable with the second highest variable importance projection value and the same regression coefficient direction shall be reselected as the dominant state variable. It is important to note that when competing variables have similar projected importance values ​​(the difference in projected importance values ​​is less than 0.1), stability analysis should be introduced. Use the Bootstrap resampling method (e.g., 500 samples); Calculate the confidence intervals and coefficients of variation for the projected importance values ​​of each variable; Choose variables with narrow confidence intervals and small coefficients of variation to ensure the stability of the selection results; For each selected dominant state variable, a standard value is determined based on historical stable production data; The standard value is the moving average of n stable batch observations under optimal process conditions. Real-time acquisition of actual observed values ​​of the dominant state variable in the current production batch; Calculate the normalized deviation between the actual observed value and the standard value; Optionally, the normalized deviation calculation process is as follows: first, calculate the absolute value of the difference between the actual observed value and the standard value, then calculate the percentage of the absolute value of the difference to the standard value, and normalize the obtained percentage to obtain the normalized deviation. For each pair of dominant state variables and target process parameters, the compensation rule base contains a corresponding compensation function. The compensation function defines the adjustment amount to be made to the target process parameters when the state variables deviate from the standard value. The function form may include, but is not limited to, linear functions, piecewise functions, or empirical curves fitted based on historical data. For example, the compensation function can be a linear function. For instance, for the dominant state variable, the cumulative working time of the electrode, and the target process parameter, the plasma treatment time, the compensation function is: Where k is the compensation coefficient, determined through regression analysis of historical data, and m is the current value of the cumulative working time of the electrode. This is the standard value for the cumulative working time of the electrode. Basic compensation amount; Substituting the normalized deviation into the corresponding compensation function yields the basic compensation amount under the independent action of the dominant state variables. If there are multiple target process parameters, calculate the basic compensation amount corresponding to each target process parameter to form a preliminary set of basic compensation amounts; The basic compensation values ​​of all target process parameters are summarized to generate the main effect compensation vector.

[0013] The coupling effect compensation layer includes: calculating the compensation amount of the interaction caused by the common deviation of at least two state variables based on multivariate interaction; It should be noted that in the coupling effect compensation layer, if for a certain target process parameter, the number of dominant state variables identified by the main effect compensation layer is less than two, the system determines that there is no significant interaction that needs to be calculated, the interaction compensation amount of the parameter is set to zero, and the comprehensive compensation scheme is directly generated by integrating the basic compensation amount and the fine-tuning compensation amount. In the coupling effect compensation layer, if for a certain target process parameter, the number of dominant state variables identified by the main effect compensation layer is no less than two; Then, the normalized deviation of the dominant state variable is calculated based on the aforementioned main effect compensation layer; Extract the dominant state variables whose normalized deviation is greater than a preset threshold, and use them as the dominant state variables that simultaneously exhibit significant deviations. From all the dominant state variables that are in a significantly deviated state, filter out the combination of variables that have an interaction model predefined in the compensation rule base; For example, the rule base predefines variable combinations for key interaction models such as "electrode loss-gas purity" and "ambient temperature-humidity"; For each predefined combination of variables in an interaction model, check whether all the state variables included are simultaneously significantly deviating from the dominant state variables in the current batch. If so, mark the combination as an active interaction combination. For each active combination of interactions, the corresponding interaction compensation model is called from the compensation rule base to calculate the interaction compensation amount; Furthermore, the interaction compensation model can be: a linear interaction model, a nonlinear response surface model, or a machine learning-based model (such as a neural network or a support vector machine). Optionally, the interaction compensation model can be: Linear interaction model: For two dominant state variables A and B, the compensation amount of their interaction. It can be represented as: ;

[0014] The interaction term coefficients were determined through regression analysis of historical data. , These are the normalized deviations of variables A and B, respectively; Nonlinear response surface model: for the two dominant state variables A and B, their interaction compensation amount. It can be represented as a model containing quadratic terms and interaction terms, for example: a, b, c, d, and e are model coefficients, obtained by fitting historical experimental data (such as Design of Experiments (DOE)). The machine learning-based model takes the normalized deviation of the aforementioned multiple state variables as input features, and directly outputs the interaction compensation amount after model calculation. Substitute the normalized deviation of each state variable within the active interaction combination in the current batch into the corresponding interaction compensation model to calculate the interaction compensation amount. It should be noted that the amount of interaction compensation is not equal to the simple algebraic sum of the compensation amounts of the main effects of each state variable. It quantifies the additional adjustment required by the synergistic (effect enhancement) or antagonistic (effect cancellation) effects between state variables. When there are multiple monthly active interaction combinations, the compensation amount for the generated interaction effects is integrated; The compensation amounts calculated from the combination of each independent interaction are summed by vector to obtain the total interaction compensation vector. If a state variable participates in multiple interaction combinations simultaneously (for example, state variable A exists in both combination AB and combination AC), it is necessary to check whether a higher-order interaction compensation model (such as ABC ternary interaction) is defined in the rule base. If it is defined, the higher-order model should be called first for calculation to avoid duplicate calculations and error amplification. If different interaction compensation models compensate for the same process parameter in opposite directions, then the conflict resolution mechanism is activated: The weighted average is calculated based on the confidence weights assigned to each model in the rule base (the weights are set based on the model's historical prediction accuracy), or arbitration is conducted based on the frequency and effect of various interactions in historical data. All calculated and integrated interaction compensation quantities are summarized to generate the interaction compensation layer output vector.

[0015] The coupling effect compensation layer completes the quantitative modeling and compensation calculation of the complex interaction between multiple state variables, making up for the shortcomings of the main effect compensation layer which only considers the independent effect of a single variable. It is a key link in improving the accuracy of process parameter adjustment under complex working conditions.

[0016] The environmental fine-tuning layer includes: calculating the fine-tuning compensation amount based on real-time environmental fluctuations; In the environment fine-tuning layer, instantaneous observations of environmental state variables are obtained, typically including: Production microenvironment parameters: instantaneous observations of temperature and relative humidity in the treatment area; Atmospheric state parameters: instantaneous observations of the workshop's atmospheric pressure on that day; The difference between the instantaneous observed value of the environmental state variable and the standard value of the environmental condition is calculated as the instantaneous deviation value; For example, the difference between an instantaneous observation of temperature and a standard temperature value is used as the instantaneous temperature deviation. For each target process parameter, the environmental fine-tuning model is called from the compensation rule base. The environmental fine-tuning model defines the small adjustment amount of the process parameter caused by a unit environmental deviation. Optionally, the environment fine-tuning model is typically in the form of a linear mapping or lookup table, for example: a linear fine-tuning function:

[0017] ;in, Let be the fine-tuning compensation amount for the i-th process parameter. , , These are the fine-tuning coefficients for the process parameters, namely temperature, relative humidity, and air pressure, obtained through statistical analysis of historical data. , , These are the instantaneous deviations of temperature, relative humidity, and air pressure, respectively. The lookup table method is used to establish a correspondence between the environmental deviation range and the fine-tuning compensation amount, for example: The instantaneous temperature deviation range is [-2.0, -1.0), and the corresponding fine-tuning compensation is +0.5. The instantaneous temperature deviation range is [-1.0, +1.0), and the corresponding fine-tuning compensation is 0. The instantaneous temperature deviation range is [+1.0, +2.0), and the corresponding fine-tuning compensation is -0.3. Substitute the calculated instantaneous temperature deviation, instantaneous relative humidity deviation, and instantaneous air pressure deviation into the environmental fine-tuning model to calculate the fine-tuning environmental quantity for each target process parameter; The fine-tuning compensation amounts of all target process parameters are summarized to generate an environmental fine-tuning compensation vector.

[0018] It should be noted that the compensation rule base is the core knowledge carrier and intelligent engine of the data management method. It is not a static empirical parameter table, but a dynamic model library that integrates process physics knowledge, historical data patterns and adaptive learning capabilities. Specifically, the rule base systematically stores and manages the core knowledge modules that support the three-layer calculation of "main effect-coupling effect-environmental fine-tuning": including a set of main effect compensation functions for quantifying the independent influence of single variables, a compensation model for characterizing and quantifying the nonlinear interaction between multiple variables, and meta-rules for guiding model selection and conflict resolution. The rule base has the ability to incrementally learn and periodically optimize based on production feedback data, thereby ensuring that the compensation decision of the silicone post-processing process can continuously and intelligently adapt to dynamic production factors such as equipment aging, raw material fluctuations and environmental changes. Before the system is put into use, an initial compensation rule base needs to be built. The building process includes: Data collection phase: Under controlled or standard production conditions, run N production batches (e.g., N≥50) and synchronously collect all defined state variable data (X) and the corresponding actual set values ​​of process parameters (Y) that have been confirmed to be optimal through quality inspection through the data acquisition grid. Main effect compensation function: For each identified pair of (dominant state variable, target process parameter), a specific compensation function is fitted based on the collected (X, Y) data pairs using methods such as linear regression and multinomial regression. Interaction compensation model: Through factorial experimental design, the levels of two or more key state variables are actively changed, and their impact on process results is measured. The collected data is used to fit a response surface model (such as a quadratic polynomial model), or the collected historical data is used to train a machine learning model (such as a random forest or gradient boosting tree) to quantify the interaction between variables. Environmental fine-tuning model: Analyze the correlation data between environmental variable fluctuations and small adjustments to process parameters under stable equipment conditions, and determine the fine-tuning coefficients or establish a lookup table through statistical regression; Rule entry stage: The specific functions, model coefficients, lookup tables, etc. obtained from the above training are stored in the compensation rule base according to the predetermined data structure and interface format, and the system initialization is completed.

[0019] For example, the parameter compensation process for the plasma treatment of silicone medical catheters: Main effect compensation layer execution: First, it is determined that for the process parameter of plasma treatment time, the strongly correlated dominant state variables are the cumulative working time of the electrode (reflecting equipment aging) and the purity of the process gas (argon) (reflecting the reaction atmosphere). The standard working time of the electrode is 200 hours, and it has actually been working for 230 hours. The calculated normalized deviation is "positive 15%"; The standard purity value for argon gas is 99.999%, and the current detection value is 99.99%. The calculated normalized deviation is "-0.009%" (purity decrease). Call the linear compensation functions for the two dominant variables respectively from the compensation rule base; According to the function "cumulative electrode working time - processing time": for every 1% positive deviation, the processing time needs to be increased by 0.4 seconds. Currently, it is 15% positive, so the calculated basic compensation amount A is "an increase of 6 seconds". According to the "gas purity - processing time" function: for every negative deviation of 0.001% in purity, the processing time needs to be increased by 0.1 seconds. Currently, it is negative 0.009%, so the calculated basic compensation amount B is "an increase of 0.9 seconds". Coupling effect compensation layer execution: The system determines that both the current cumulative working time of the electrode and the gas purity have deviated significantly, and the rule base predefines the interaction model of these two variables ("electrode loss-gas purity" combination). By calling the corresponding nonlinear response surface model, it is revealed that when electrode loss and gas purity decrease simultaneously, the weakening effect on the treatment effect is far greater than the sum of the individual effects of the two. Substituting the electrode deviation (positive 15%) and gas purity deviation (negative 0.009%) into the model, it is calculated that due to this synergistic deterioration effect, an additional interaction compensation amount C is required on top of the basic compensation, which is "an increase of 3.1 seconds". This value is not equal to (6 seconds + 0.9 seconds), reflecting the special quantification of the coupling effect. Execution of the environment fine-tuning layer: Real-time sensor data revealed that the current workshop temperature was 1.5 degrees Celsius higher than the standard process temperature. Using the "ambient temperature - processing time" micro-survey model, it is stipulated that for every 1 degree Celsius increase in temperature, the processing time should be reduced by 0.2 seconds to avoid overheating. The current increase is 1.5 degrees Celsius, so the calculated environmental fine-tuning compensation amount D is "reduced by 0.3 seconds".

[0020] The purpose of step two is to transform the real-time acquired multi-state data into precise and executable process parameter compensation instructions. By constructing a three-tiered compensation architecture of main effect, coupling effect, and environmental fine-tuning, process deviations from different sources are decomposed and quantified: the main effect layer handles the independent and systematic drift of key equipment state variables; the coupling effect layer characterizes the nonlinear interaction between multiple key variables; and the environmental fine-tuning layer addresses high-frequency, small-amplitude random environmental fluctuations. This makes the compensation results closer to the physicochemical reality under complex operating conditions where multiple variables such as electrode loss, gas purity, and ambient temperature and humidity change simultaneously. It effectively avoids the problems of neglecting one aspect or over-compensation / under-compensation, thereby improving the quality consistency and stability of silicone product post-processing processes (such as plasma treatment and secondary vulcanization).

[0021] Step 3: Integrate the basic compensation amount, the interaction compensation amount, and the fine-tuning compensation amount to generate a comprehensive compensation scheme for the current process. Output the comprehensive compensation scheme to the production control system or equipment controller to complete the adaptive adjustment of the post-processing parameters of silicone products. In step three, for any target process parameter, the basic compensation amount, the interaction compensation amount, and the fine-tuning compensation amount are added together to obtain the comprehensive compensation amount; It should be noted that the interaction compensation is zero when there is no active interaction. The comprehensive compensation vector is obtained by summing up the comprehensive compensation amounts of all target process parameters. The comprehensive compensation vector is added to the standard setpoint vector of each process parameter to obtain the final process parameter setpoint vector. The process parameter setpoint vector and its metadata (including batch number, calculation timestamp, main state variables that trigger compensation, and deviation, etc.) are encapsulated into a standardized instruction set; The instruction set can be sent to the production control system or directly to the corresponding equipment controller through a predefined application programming interface. The output methods may include: Write the production work order parameter table into the Manufacturing Execution System; Data can be sent directly to programmable logic controllers or process equipment via industrial communication protocols (such as OPC UA, Modbus TCP). Waiting to receive a confirmation signal from the production control system or equipment controller confirming successful parameter setting.

[0022] The core optimization of data management in this embodiment lies in transforming the management of silicone process parameters from the traditional static recording and post-event traceability paradigm to a proactive intelligent paradigm of real-time perception, intelligent analysis, dynamic decision-making, and closed-loop execution. By constructing a dynamic knowledge system with a "compensation rule base" as the intelligent core and integrating process physics and data science, it not only realizes the collection and integration of multi-source heterogeneous production data (equipment status, environmental variables), but more importantly, through a three-layer analysis architecture of "main effect - coupling effect - environmental fine-tuning," it deeply mines and releases the coupled value of the data, quantifies the impact of single variables and the nonlinear interaction of multiple variables, and finally transforms data insights into automatically executable process parameter optimization instructions in real time. This solves the problems of data silos, knowledge loss, and inaccurate compensation under complex working conditions in traditional methods, and realizes a fundamental transformation of data from passive recording to driving adaptive optimization of the process. Example

[0023] Based on the same inventive concept as the data management method for silicone process formulation parameters in the foregoing embodiments, such as Figure 3 As shown, this application provides a data management system for silicone process formulation parameters, wherein the system specifically includes: State variable data acquisition module: Real-time acquisition of multiple state variable data that affect the quality of post-processing of silicone products, including inherent equipment state variables and environmental state variables; Process parameter compensation analysis module: Based on the acquired state variable data, it calls the compensation rule library to calculate compensation values ​​in a hierarchical and multi-factor coupled manner, specifically including: The multi-layer structure includes: main effect compensation layer, coupling effect compensation layer, and environmental fine-tuning layer; The main effect compensation layer includes: calculating the basic compensation amount for a single state variable based on the skewness of the single state variable to the standard value; The coupling effect compensation layer includes: calculating the compensation amount of the interaction caused by the common deviation of at least two state variables based on multivariate interaction; The environmental fine-tuning layer includes: calculating the fine-tuning compensation amount based on real-time environmental fluctuations; Process parameter adjustment module: Integrates basic compensation amount, interaction compensation amount and fine-tuning compensation amount to generate a comprehensive compensation scheme for the current process, and outputs the comprehensive compensation scheme to the production control system or equipment controller to complete the adaptive adjustment of the post-processing parameters of silicone products.

[0024] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A data management method for silicone process formulation parameters, characterized in that: include: Real-time acquisition of multiple state variable data affecting the quality of post-processing of silicone products, including inherent equipment state variables and environmental state variables; Based on the acquired state variable data, the compensation rule library is invoked to calculate the compensation value in a hierarchical and multi-factor coupled manner. The multi-layer structure includes: main effect compensation layer, coupling effect compensation layer, and environmental fine-tuning layer. By integrating the basic compensation amount calculated by the main effect compensation layer, the interaction compensation amount calculated by the coupling effect compensation layer, and the fine-tuning compensation amount calculated by the environmental fine-tuning layer, a comprehensive compensation scheme for the current post-processing process is generated. The comprehensive compensation scheme is output to the production control system or equipment controller to complete the adaptive adjustment of the post-processing parameters of silicone products.

2. The data management method for silicone process formulation parameters according to claim 1, characterized in that: The process of calculating compensation values ​​by calling the compensation rule base in a hierarchical and multi-factor coupled manner is as follows: In the main effect compensation layer, the basic compensation amount for the independent action of the state variable on the target process parameter is calculated based on the deviation of a single state variable from the standard value. In the coupling effect compensation layer, the interaction compensation amount is calculated based on the interaction caused by at least two state variables deviating from the standard value. In the environmental fine-tuning layer, the fine-tuning compensation amount is calculated based on the fluctuation of real-time environmental state variables relative to the standard values ​​of environmental conditions.

3. The data management method for silicone process formulation parameters according to claim 2, characterized in that: The main effect compensation layer includes: For the target process parameters, at least one dominant state variable is selected from multiple state variables; Calculate the normalized deviation between the real-time observed value and the corresponding standard value of each dominant state variable; The basic compensation amount is calculated by calling the pre-stored compensation function corresponding to each dominant state variable and target process parameter from the compensation rule base, substituting the corresponding normalized deviation into the compensation function.

4. The data management method for silicone process formulation parameters according to claim 3, characterized in that: Select at least one dominant state variable from multiple state variables, including: Based on historical production data, by calculating the mutual information coefficient, state variables with a mutual information coefficient absolute value greater than a threshold are extracted as a set of candidate state variables that are associated with the target process parameters. Partial least squares regression analysis was used to calculate the variable importance projection values ​​for each state variable; Based on the variable importance projection values ​​and the corresponding standardized regression coefficients, the dominant state variables are selected according to the preset priority rules.

5. The data management method for silicone process formulation parameters according to claim 4, characterized in that: Priority rules include: First priority: Select candidate state variables whose projected importance values ​​are all greater than the first threshold and whose standardized regression coefficients are in the same direction as all target process parameters as dominant state variables; Second priority: If there is no candidate state variable that satisfies the first priority, then for each target process parameter, select the candidate state variable with the highest variable importance projection value as the dominant state variable. If a candidate state variable is selected as the dominant state variable for multiple target process parameters, then the direction of the regression coefficients between different target process parameters is verified: If the regression coefficients are in the same direction, they are identified as the common dominant state variables of these target process parameters; If the regression coefficients conflict in direction, the direction of influence on the core target process parameter should be prioritized for determination. For the remaining conflicting target process parameters, the state variable with the second highest variable importance projection value and consistent regression coefficient direction should be reselected as the dominant state variable.

6. The data management method for silicone process formulation parameters according to claim 1, characterized in that: The coupling effect compensation layer includes: Identify at least two dominant state variables that simultaneously deviate significantly from each other in the current batch; The active interaction combination is determined from the compensation rule base by identifying a combination of variables that predefines the interaction model and includes all dominant state variables that simultaneously deviate significantly. For each active combination of interactions, the corresponding interaction compensation model in the compensation rule base is invoked, and the normalized deviation of the relevant state variables is input into the interaction compensation model to calculate the interaction compensation amount.

7. The data management method for silicone process formulation parameters according to claim 6, characterized in that: Obtaining the dominant state variables that simultaneously deviate significantly: Normalized deviation of the dominant state variable based on computation; Extract the dominant state variables whose normalized deviation is greater than a preset threshold, and use them as the dominant state variables that simultaneously exhibit significant deviation.

8. The data management method for silicone process formulation parameters according to claim 1, characterized in that: The environmental fine-tuning layer includes: Obtain instantaneous observations of environmental state variables and calculate the instantaneous deviations from the corresponding environmental condition standard values; For each target process parameter, the environmental fine-tuning model is called from the compensation rule library; Substitute the instantaneous deviation value into the environmental fine-tuning model to calculate the fine-tuning compensation amount; The environmental fine-tuning model is a linear mapping function or a lookup table.

9. The data management method for silicone process formulation parameters according to claim 1, characterized in that: Generate a comprehensive compensation scheme for the current process, including: For any target process parameter, the basic compensation amount, the interaction compensation amount, and the fine-tuning compensation amount are added together to obtain the comprehensive compensation amount; It should be noted that the interaction compensation is zero when there is no active interaction. The comprehensive compensation vector is obtained by summing up the comprehensive compensation amounts of all target process parameters. The comprehensive compensation vector is added to the standard setpoint vector of each process parameter to obtain the final process parameter setpoint vector.

10. A data management system for silicone process formulation parameters, characterized in that, The system is used to perform the method according to any one of claims 1-9, the system comprising: State variable data acquisition module: Real-time acquisition of multiple state variable data that affect the quality of post-processing of silicone products, including inherent equipment state variables and environmental state variables; Process parameter compensation analysis module: Based on the acquired state variable data, it calls the compensation rule library to calculate compensation values ​​in a hierarchical and multi-factor coupled manner, specifically including: The multi-layer structure includes: main effect compensation layer, coupling effect compensation layer, and environmental fine-tuning layer; The main effect compensation layer includes: calculating the basic compensation amount for a single state variable based on the skewness of the single state variable to the standard value; The coupling effect compensation layer includes: calculating the compensation amount of the interaction caused by the common deviation of at least two state variables based on multivariate interaction; The environmental fine-tuning layer includes: calculating the fine-tuning compensation amount based on real-time environmental fluctuations; Process parameter adjustment module: Integrates basic compensation amount, interaction compensation amount and fine-tuning compensation amount to generate a comprehensive compensation scheme for the current process, and outputs the comprehensive compensation scheme to the production control system or equipment controller to complete the adaptive adjustment of the post-processing parameters of silicone products.

Images