A process optimization method based on MES data analysis

Abstract

The application provides a process optimization method based on MES data analysis, comprising: obtaining historical optimization data and current dynamic factor information from a manufacturing execution system to generate a comprehensive data set; grouping and processing historical optimization trajectories according to the comprehensive data set to determine a trend change mode; generating a predicted optimization direction vector according to the trend change mode and a current production plan; generating an optimized parameter adjustment trajectory according to the initial solution set distribution and the dynamic factor information; judging whether to update the trend change mode according to the parameter adjustment trajectory and real-time production data; training a model to generate an enhanced prediction accuracy index according to the updated trend change mode; and determining a final process parameter combination according to the prediction accuracy index and the updated solution set distribution.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a process optimization method based on MES data analysis. Background Technology

[0002] Manufacturing is a core sector driving industrial development, and its process optimization directly impacts production efficiency, product quality, and enterprise competitiveness. In the context of intelligent manufacturing, Manufacturing Execution Systems (MES) provide crucial support for process optimization by collecting and storing production data in real time.

[0003] However, existing methods for process optimization using MES data often face challenges due to data complexity and dynamism. Current solutions typically rely on fixed rules or single models to analyze historical data, making it difficult to adapt to the complex, multi-variable, and highly dynamic scenarios in production environments. This results in inaccurate optimization results that fail to meet actual production needs.

[0004] In process optimization, the core challenge lies in effectively mining the patterns in historical optimization data accumulated in the MES system and applying them to dynamic production scenarios. Historical optimization data contains optimization trajectories and solution set distribution information under different combinations of process parameters, which hides the trend changes in process optimization. However, due to the complexity of the data, which involves multiple dynamic factors such as production plans, equipment status, and raw material batches, traditional methods struggle to accurately identify these trends. For example, on an automotive parts production line, adjusting process parameters requires considering equipment wear, raw material batch differences, and production cycle changes simultaneously. However, existing methods often fail to integrate these factors and predict the optimization direction, leading to delayed parameter adjustments that impact production efficiency and product quality. Furthermore, the lack of dynamic analysis of optimization trajectories makes it difficult to quickly converge to high-quality solutions in multi-objective optimization, resulting in low optimization efficiency.

[0005] Therefore, how to mine trend change patterns based on historical optimization data of MES, and combine them with dynamic factors such as production plan, equipment status and raw material batches to accurately predict the optimization direction and provide heuristic guidance for multi-objective evolutionary algorithms has become a key issue in process optimization. Summary of the Invention

[0006] This invention provides a process optimization method based on MES data analysis, mainly including:

[0007] Historical optimization data and current dynamic factor information are acquired from the Manufacturing Execution System (MES) to generate a comprehensive dataset. Historical optimization trajectories are grouped based on this comprehensive dataset to determine trend change patterns. Based on these trend change patterns and the current production plan, a predicted optimization direction vector is generated. A population is initialized using the optimization direction vector and integrated with the trend change patterns to generate an initial solution set distribution. Based on the initial solution set distribution and the dynamic factor information, an optimized parameter adjustment trajectory is generated. Based on the parameter adjustment trajectory and real-time production data, it is determined whether to update the trend change pattern. Based on the updated trend change pattern, the model is trained to generate an enhanced prediction accuracy index. Based on the prediction accuracy index and the updated solution set distribution, the final combination of process parameters is determined. Furthermore, the step of acquiring historical optimization data and current dynamic factor information from the Manufacturing Execution System (MES) to generate a comprehensive dataset includes: acquiring historical optimization data containing process parameter combination optimization trajectories from the MES; acquiring current dynamic factor information containing production plans, equipment status, and raw material batch details; integrating the historical optimization data and the dynamic factor information to generate a comprehensive dataset containing process parameter combination optimization trajectories, production plans, equipment status, and raw material batch details; and preprocessing the comprehensive dataset to generate a standardized dataset format. Furthermore, the step of grouping historical optimization trajectories based on the comprehensive dataset to determine trend change patterns includes: using a clustering algorithm to group the historical optimization trajectories in the comprehensive dataset; extracting feature data from each group of historical optimization trajectories based on the grouping results; analyzing the trend change patterns of each group of historical optimization trajectories based on the feature data; and generating pattern parameters characterizing the change patterns of similar trajectories based on the trend change patterns. Furthermore, the step of generating a predicted optimization direction vector based on the trend change pattern and the current production plan includes: determining whether the trend change pattern matches the current production plan; if they match, obtaining the equipment status and raw material batch variables in the current production plan; inputting the equipment status and raw material batch variables into a pre-trained neural network model; and generating the predicted optimization direction vector based on the output of the neural network model. Furthermore, the step of initializing the population based on the optimization direction vector and incorporating the trend change pattern to generate an initial solution set distribution includes: initializing the population using a genetic algorithm based on the optimization direction vector; incorporating the trend change pattern as a heuristic constraint into the genetic algorithm; performing multi-objective optimization through the genetic algorithm to generate an initial solution set distribution; and evaluating the initial solution set distribution to generate distribution characteristic parameters of the initial solution set.Furthermore, generating an optimized parameter adjustment trajectory based on the initial solution set distribution and the dynamic factor information includes: determining whether the initial solution set distribution deviates from the trend change pattern; if it deviates, adjusting the crossover and mutation operators of the genetic algorithm based on the dynamic factor information; iteratively optimizing the adjusted genetic algorithm to generate an optimized parameter adjustment trajectory; verifying the parameter adjustment trajectory to generate a trajectory optimization result. Furthermore, determining whether to update the trend change pattern based on the parameter adjustment trajectory and real-time production data includes: obtaining real-time production data feedback; comparing the real-time production data with the parameter adjustment trajectory and the optimized direction vector; determining whether to update the trend change pattern based on the comparison result; if an update is needed, generating new trend change pattern parameters based on the real-time production data. Furthermore, training the model to generate an enhanced prediction accuracy index based on the updated trend change pattern includes: generating a training dataset based on the updated trend change pattern; retraining the neural network model using the training dataset and updating the model weights; generating a prediction result based on the updated neural network model; and generating an enhanced prediction accuracy index based on a comparison of the prediction result with actual production data.

[0008] Furthermore, the step of determining the final process parameter combination based on the predicted accuracy index and the updated solution set distribution includes: comparing the predicted accuracy index with a preset threshold; if the predicted accuracy index exceeds the preset threshold, proceeding to the next selection process; otherwise, returning to the iterative update trend change mode; selecting high-quality solutions from the updated solution set distribution; and selecting the solution with the highest fitness from the high-quality solutions as the final process parameter combination.

[0009] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0010] This invention discloses a process optimization method based on MES data analysis. Addressing the lack of adaptability and predictive accuracy in process parameter optimization within Manufacturing Execution Systems (MES), this method integrates historical optimization trajectories with current production plans, equipment status, and raw material batch information to generate a comprehensive dataset. It then utilizes clustering algorithms to extract trend change patterns and combines neural networks and genetic algorithms to generate predictive optimization direction vectors and initial solution set distributions, dynamically adjusting the parameter trajectory. This invention updates trend change patterns through real-time production data feedback, retrains the neural network to improve predictive accuracy, and ultimately outputs the optimal combination of process parameters. This invention solves the problem of insufficient adaptability of traditional optimization methods to dynamic production environments, significantly improving the intelligence level and production efficiency of process parameter optimization, and achieving adaptive and precise manufacturing process optimization. Attached Figure Description

[0011] Figure 1This is a flowchart illustrating a process optimization method based on MES data analysis according to the present invention. Detailed implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0013] Reference Figure 1 This embodiment of a process optimization method based on MES data analysis may specifically include:

[0014] Step S1: Obtain historical optimization data and current dynamic factor information from the Manufacturing Execution System (MAS) to generate a comprehensive dataset, including: obtaining historical optimization data containing process parameter combination optimization trajectories from the MAS; obtaining current dynamic factor information containing production plans, equipment status, and raw material batch details; integrating the historical optimization data and the dynamic factor information to generate a comprehensive dataset containing process parameter combination optimization trajectories, production plans, equipment status, and raw material batch details; and preprocessing the comprehensive dataset to generate a standardized dataset format.

[0015] In one embodiment, historical optimization data and current dynamic factor information are obtained from the manufacturing execution system to generate a comprehensive dataset, including the following steps:

[0016] Step S101: Obtain historical optimization data containing the optimization trajectory of process parameter combinations from the manufacturing execution system.

[0017] Specifically, by querying the historical database through the system interface, the system can extract the sequence of process parameter changes recorded in past production cycles, such as the optimized path data of temperature, pressure and time combinations. This data reflects the trajectory evolution of the previous optimization process.

[0018] Based on the extracted historical optimization data, preliminary filtering is performed to remove outliers to ensure data reliability.

[0019] Step S102: Obtain current dynamic factor information including production plan, equipment status and raw material batch details.

[0020] Specifically, the system reads the scheduling details of the current production plan, the operating status of equipment such as load and maintenance records, and the specific attributes of raw material batches such as batch number and quality indicators from the real-time monitoring module.

[0021] During the acquisition process, the timestamps of dynamic factors are updated synchronously to match the time sequence of historical data.

[0022] Step S103: Based on the historical optimization data and the dynamic factor information, a comprehensive dataset containing the process parameter combination optimization trajectory, production plan, equipment status and raw material batch details is generated.

[0023] Specifically, historical optimization trajectories are associated and mapped with current dynamic factors. For example, production plans are linked to similar trajectory segments through key-value matching to form a multidimensional dataset, in which each data point contains trajectory coordinates, production plan labels, equipment status vectors, and raw material batch attributes.

[0024] Key nodes in historical trajectories are identified and the current device state is injected as a weighting factor to calculate the integrated similarity score, thereby optimizing the coherence of the dataset.

[0025] Using data fusion algorithms such as weighted averaging, dynamic factors are embedded into trajectory data, such as applying influence coefficients to equipment state variables, to ensure that the comprehensive dataset captures the potential impact of real-time changes on the optimized trajectory.

[0026] Verify the integrity of the integrated dataset by checking the proportion of missing values; if it exceeds a threshold, backtrack and supplement the data.

[0027] Step S104: Preprocess the comprehensive dataset to generate a standardized dataset format.

[0028] Specifically, normalization techniques are applied to scale numerical features to a uniform range. For example, min-max normalization is used to process process parameter values. At the same time, categorical features such as raw material batches are individually encoded, and the output is a dataset in matrix format, which is convenient for subsequent clustering or neural network input.

[0029] The choice of preprocessing parameters is based on dataset statistics, such as calculating the mean and standard deviation to adjust the normalization boundary.

[0030] For example, in the electronic component manufacturing scenario, the historical optimization data obtained in step S101 may include the trajectory change of the welding temperature from 150 degrees to 180 degrees, reflecting the path of production optimization; the integration in step S103 combines this trajectory with the current production plan, such as the target of 1,000 pieces per day, equipment status such as machine load of 80% and raw material batch purity of 95%, to generate a comprehensive dataset, which helps to improve the optimization accuracy by up to 15%.

[0031] In one embodiment, regarding the integration process in step S103, it is understood that by aligning multi-source data, the seamless integration of historical trajectories and dynamic factors is ensured. For example, in chemical production, the integration of trajectory data can reduce the scrap rate caused by parameter deviation by 5%.

[0032] For example, the preprocessing in step S104 can further include anomaly detection, such as using box plot methods to remove points that exceed 1.5 times the interquartile range. After generating a standardized dataset, the robustness of subsequent algorithms can be improved by 10%.

[0033] In one embodiment, when calculating the similarity score in step S103, the Euclidean distance formula is used to measure the proximity between the trajectory point and the dynamic factor. For example, a distance of less than 0.1 is considered a high match, which has the beneficial effect of enhancing the predictive relevance of the dataset.

[0034] For example, in the extension of electronic manufacturing, the trajectory data extracted in step S101 covers multiple batches, and the raw material details obtained in step S102 include supplier variations. The standard dataset generated after integration and preprocessing can be directly input into the clustering algorithm, which optimizes the trajectory grouping efficiency by 20%.

[0035] Step S2: Grouping historical optimized trajectories according to the comprehensive dataset to determine trend change patterns, including: grouping historical optimized trajectories in the comprehensive dataset using a clustering algorithm; extracting feature data of each group of historical optimized trajectories based on the grouping results; analyzing the trend change patterns of each group of historical optimized trajectories based on the feature data; and generating pattern parameters representing the change patterns of similar trajectories based on the trend change patterns.

[0036] In one embodiment, a clustering algorithm is used to group the historical optimization trajectories in the comprehensive dataset. Specifically, step S201 involves obtaining the historical optimization trajectories in the comprehensive dataset. These trajectories include the optimization path of the process parameter combination and the corresponding production plan, equipment status, and raw material batch details.

[0037] From historical optimization data extracted from the MES system, trajectory sequences similar to current dynamic factors are selected, such as the combined change paths of temperature, pressure, and time parameters.

[0038] These trajectory sequences are grouped using the K-means clustering algorithm. The K-means clustering algorithm iteratively calculates the Euclidean distance from data points to cluster centers and assigns similar trajectories to the same group until the cluster centers are stable. The K value is preset to 5 to 10 depending on the number of trajectories to cover different optimization scenarios.

[0039] In one embodiment, feature data of historical optimized trajectories of each group are extracted based on the grouping results. Specifically, step S202 involves calculating statistical features of the trajectory from each cluster group, such as the average rate of change and standard deviation. These feature data reflect the fluctuation range of the trajectory during the parameter adjustment process.

[0040] For the trajectory within each group, the vector difference between the parameter combinations from the initial values ​​to the optimized values ​​is calculated and used as part of the feature data.

[0041] Extract common features of trajectories within a group, such as equipment status variables like runtime and raw material batch variables like the average purity level, to ensure that the feature data matches the production plan.

[0042] In one embodiment, the trend change pattern of each group of historical optimized trajectories is analyzed based on the feature data. Specifically, step S203 involves using a linear regression method to fit a trend line to the feature data of each group. Linear regression is the best-fit line for feature data points obtained by solving the least squares method, and the trend change pattern is represented by the slope and intercept.

[0043] Time series analysis is performed on the rate of change of characteristic data within a group to identify upward, downward, or stable trends. For example, in the temperature parameter trajectory, if the average rate of change is greater than 0, the pattern is optimized towards higher values.

[0044] Compare the trend differences between groups to determine the mode that matches the current production plan. For example, if the current plan involves high-purity raw materials, select the mode of the corresponding group, where the mode includes acceleration or deceleration phases of parameter adjustment.

[0045] To verify the stability of the trend pattern, the variance of the trajectory within the group is calculated. If the variance is less than 0.05, the pattern is considered reliable. This analysis process ensures that the trend change pattern can reflect the similarity of historical optimization trajectories, which is beneficial to improving the accuracy of predicting the optimization direction. For example, in actual processes, this pattern analysis can reduce ineffective parameter adjustments and shorten the optimization cycle.

[0046] In one embodiment, based on the trend change pattern, a pattern parameter representing the change law of similar trajectories is generated. Specifically, step S204 involves combining the slope, intercept, and variance values ​​of the trend change pattern into a vector form of pattern parameters. These parameters represent the law of the trajectory from the initial state to the optimized state. For example, the pattern parameter vector is [slope value, intercept value, variance value].

[0047] Normalize the mode parameters so that they range from 0 to 1 to facilitate subsequent neural network inputs, for example, by processing slope values ​​using a min-max normalization formula.

[0048] The generated pattern parameters are stored in the database and associated with the grouping results to form a queryable pattern representation. For example, in scenarios with different raw material batches, the pattern parameters can be used to quickly match similar trajectories.

[0049] In one embodiment, for the trend change pattern analysis in step S203, in process parameter optimization, such as temperature control trajectory, if the feature data shows an average change rate of 0.2 per minute, then the slope of the fitted linear regression represents an upward trend pattern. This pattern is beneficial for predicting the direction of parameter adjustment under equipment conditions and reducing energy consumption.

[0050] In one embodiment, regarding the generation of model parameters in step S204, if the variance in the high-pressure parameter trajectory group is 0.03, the generated model parameter vector [0.15, 1.2, 0.03] represents the stable change law. This parameter can support the initial solution set of the genetic algorithm and effectively improve the optimization efficiency.

[0051] In one embodiment, steps S203 and S204 are used in combination. For example, in a scenario where the purity of the raw material batch is 95%, the trend pattern obtained by analysis is a parameter-increasing type. The generated pattern parameters are used for iterative updates to ensure the continuity of closed-loop optimization cycle data, which is beneficial for real-time applications in the MES system.

[0052] Step S3: Generate a predicted optimization direction vector based on the trend change pattern and the current production plan, including: determining whether the trend change pattern matches the current production plan; if it matches, obtaining the equipment status and raw material batch variables in the current production plan; inputting the equipment status and raw material batch variables through a pre-trained neural network model; and generating a predicted optimization direction vector based on the output of the neural network model.

[0053] In one embodiment, generating a predicted optimization direction vector based on the trend change pattern and the current production plan includes the following steps:

[0054] Step S301: Determine whether the trend change pattern matches the current production plan.

[0055] Key features in trend change patterns are extracted, such as the trajectory change rate of process parameter combinations and the similarity of dynamic factors in production plans. The degree of matching is quantified by calculating Euclidean distance, which is defined as the square root of the sum of the squares of the differences in each dimension between two vectors. If the distance is less than a preset threshold, it is considered a match.

[0056] If the calculated Euclidean distance is less than the threshold, a match is confirmed; otherwise, the system switches to standby mode to handle mismatches.

[0057] For example, in chemical production scenarios, trend change patterns are derived from grouping historical optimization trajectories. The current production plan involves the status of specific equipment. If the rate of change of parameters in the pattern is similar to the rate of change of raw material batch details in the plan, the Euclidean distance can be used to quickly determine whether the historical pattern is applicable, thereby improving prediction efficiency and avoiding resource waste caused by invalid calculations.

[0058] In one embodiment, the judgment process in step S301 can improve the response speed of the overall optimization loop because, through quantitative matching, irrelevant historical data can be directly filtered out, ensuring that subsequent steps are based on a reliable foundation.

[0059] Step S302: If a match is found, obtain the equipment status and raw material batch variables in the current production plan.

[0060] For example, equipment status includes runtime and maintenance records, and raw material batch variables cover batch purity and supply time. These data are extracted directly from the MES system to form a standardized input vector.

[0061] Step S303: Input the equipment status and raw material batch variables through a pre-trained neural network model.

[0062] A neural network model is constructed, which adopts a multilayer perceptron structure, including an input layer, a hidden layer, and an output layer. The input layer receives device states such as temperature and pressure values, as well as raw material batch variables such as component ratios. The hidden layer uses the ReLU activation function to handle nonlinear relationships, and the output layer generates preliminary direction vector components.

[0063] The training process uses historical optimization data as samples and adjusts the weights through the backpropagation algorithm. The backpropagation algorithm involves calculating the gradient of the loss function and updating the parameters to minimize the prediction error. For example, the loss function can be the mean squared error, which is defined as the squared average of the differences between the actual optimization direction and the predicted direction.

[0064] After normalizing the acquired equipment status and raw material batch variables, they are input into the model. Normalization is achieved by subtracting the mean and dividing by the standard deviation to ensure that the input data is within the range of 0 to 1.

[0065] For example, in metal processing and production, if the equipment status shows high-temperature operation and the raw material batch is a high-purity alloy, the hidden layer after model input captures the interaction between variables, such as the influence of temperature on batch purity. The ReLU function is used to avoid the gradient vanishing problem, thereby outputting more accurate vector components. This method is beneficial for handling the uncertainty brought about by dynamic factors and improving the robustness of prediction.

[0066] In one embodiment, the neural network model training in step S303 is based on historical optimization trajectories and dynamic factors. When extended, it can be inferred that batch gradient descent is used to optimize the weights. Batch gradient descent is the process of calculating the gradient and updating the parameters of all training samples at once. This helps to converge efficiently on large datasets such as historical MES data, resulting in higher prediction accuracy.

[0067] For example, from multiple perspectives, if the production scenario involves different batches of raw materials, the model can learn the influence of batch variables on the optimization direction. For instance, when the batch purity is high, the direction vector tends to reduce energy consumption; if the equipment status is after maintenance, the direction vector tends to increase output. This multi-faceted support ensures that the model is highly adaptable and is beneficial to the stability of the closed-loop optimization cycle.

[0068] Step S304: Generate a predicted optimized direction vector based on the output of the neural network model.

[0069] Post-processing of the model output, such as vector normalization and threshold filtering, ensures that each component represents the relative strength of the optimization direction.

[0070] The combined components form a complete vector, which is used for subsequent genetic algorithm initialization.

[0071] For example, in pharmaceutical manufacturing, the model output may be a three-dimensional vector representing the adjustment direction of temperature, pressure, and time parameters. By using a post-processing noise filtering component, the generated direction vector can directly guide the parameter optimization trajectory, which helps reduce the number of iterations and improve production efficiency.

[0072] In one embodiment, the generation process of step S304 closely follows the goal of predicting the optimization direction vector. By comparing it with the optimization direction vector, it ensures that the generated vector incorporates the trend pattern as a constraint, thereby forming a closed-loop process optimization cycle data.

[0073] Step S4: Initialize the population according to the optimization direction vector and incorporate the trend change pattern to generate an initial solution set distribution, including: initializing the population according to the optimization direction vector using a genetic algorithm; incorporating the trend change pattern as a heuristic constraint into the genetic algorithm; performing multi-objective optimization through the genetic algorithm to generate an initial solution set distribution; evaluating the initial solution set distribution to generate distribution characteristic parameters of the initial solution set.

[0074] In one embodiment, step S401, which uses a genetic algorithm to initialize the population based on the optimized direction vector, specifically includes calculating the initial position coordinates of each individual in the population based on the vector component values ​​in the optimized direction vector. The optimized direction vector is obtained by prediction through a neural network model and includes the adjustment direction and magnitude of the process parameters.

[0075] The vector components are mapped to the solution space of the genetic algorithm to generate the chromosome codes of the initial population.

[0076] Specifically, this implementation method can be applied in the process optimization scenario of the MES system, using optimization direction vectors obtained from historical data to guide population initialization.

[0077] For example, in a chemical production batch, the optimization direction vector represents the predicted adjustment direction of temperature and pressure parameters. If the vector value is (0.8, -0.5), it means the temperature needs to increase by 0.8 units and the pressure needs to decrease by 0.5 units. Therefore, when initializing the population, individual codes closer to this direction are preferentially generated. This initialization method allows the population to be biased towards historical trends from the beginning, avoiding the inefficiency of search caused by random initialization and facilitating rapid convergence to a high-quality solution set.

[0078] In one embodiment, step S402, incorporating the trend change pattern as a heuristic constraint into the genetic algorithm, specifically includes extracting key trajectory points in the trend change pattern, such as inflection points and slope values ​​of parameter changes, and converting them into fitness function constraints of the genetic algorithm. Here, the trend change pattern is a similar pattern obtained from the grouping and processing of historical optimized trajectories by a clustering algorithm.

[0079] Adjusting the selection probability during population evolution based on constraints reduces the fitness of individuals that deviate from the trend pattern.

[0080] Incorporating pattern constraints into cross operations ensures that offspring individuals inherit the characteristics of the trend trajectory.

[0081] Specifically, this implementation method is applicable to scenarios where equipment status and raw material batches change dynamically in production planning. By using trend patterns as heuristic constraints, it can guide the genetic algorithm to avoid invalid search paths.

[0082] For example, in a metal processing technology, if the trend pattern shows that the temperature parameter follows a linear upward trajectory under specific equipment conditions, then constraints will penalize individuals with drastic temperature fluctuations, thus making the evolutionary process more closely aligned with historical optimization experience. This integration improves the algorithm's adaptability to dynamic factors, reduces the number of iterations, and enhances the stability of the optimization trajectory.

[0083] In one possible implementation, the trend pattern can be further subdivided into sub-patterns for different batches of raw materials. For example, the pattern for high batch purity emphasizes slow adjustment, while the pattern for low batch purity allows for greater variation. This allows for dynamic switching of constraints in the genetic algorithm, enhancing prediction accuracy and reducing production deviations.

[0084] Step S403 involves performing multi-objective optimization using the genetic algorithm to generate an initial solution set distribution. Specifically, this includes using the initialized population and the incorporated constraints to perform multiple generations of evolution. The optimization objectives include minimizing production costs and maximizing output. The multi-objective optimization uses the Pareto front method to generate a non-dominated solution set as the distribution.

[0085] Specifically, this step directly uses the aforementioned initial population and constraints to iterate and generate a distribution for subsequent evaluation.

[0086] Step S404: Evaluate the initial solution set distribution and generate distribution characteristic parameters of the initial solution set. Specifically, this includes calculating indicators such as the center position, dispersion, and Pareto front coverage of the solution set distribution. The distribution characteristic parameters are used to determine whether the distribution deviates from the historical trend pattern.

[0087] If the distribution center deviates from the optimization direction vector by more than a threshold, it is marked as a parameter that needs to be adjusted.

[0088] Specifically, this implementation method is applied in the optimization of process parameter combinations. By evaluating and generating characteristic parameters, such as a dispersion of 0.2 indicating a solution set, it is beneficial for subsequent mutation adjustments in the genetic algorithm.

[0089] For example, in an electronic component manufacturing scenario, if the evaluation shows a distribution coverage rate of 85%, then the characteristic parameters can be directly input into the adjustment step to ensure that the optimized trajectory matches the current production plan. This evaluation process forms a closed loop from distribution generation to parameter feedback, improving the robustness of the overall process optimization.

[0090] Step S5: Generate an optimized parameter adjustment trajectory based on the initial solution set distribution and the dynamic factor information, including: determining whether the initial solution set distribution deviates from the trend change pattern; if it deviates, adjusting the crossover and mutation operators of the genetic algorithm based on the dynamic factor information; performing iterative optimization through the adjusted genetic algorithm to generate an optimized parameter adjustment trajectory; verifying the parameter adjustment trajectory to generate a trajectory optimization result.

[0091] In one embodiment, determining whether the initial solution set distribution deviates from the trend change pattern includes: step S501, determining the degree of deviation by calculating the deviation value between the statistical index of the initial solution set distribution and the corresponding index of the trend change pattern, wherein the deviation value is calculated based on Euclidean distance, and if the deviation value exceeds a preset threshold, it is determined to be a deviation.

[0092] Step S502: If there is a deviation, the crossover and mutation operator of the genetic algorithm is adjusted according to the dynamic factor information, including: calculating the adjustment coefficient of the crossover probability according to the dynamic factor information such as equipment status and raw material batch details, wherein the initial value of the crossover probability is 0.8, and if the dynamic factors show that the equipment load is high, the coefficient is multiplied by 1.2 to increase diversity.

[0093] Adjust the mutation operator by associating the mutation rate with the fluctuation range of dynamic factors. For example, if the base value of the mutation rate is 0.1, the mutation rate is increased to 0.15 if the change rate of raw material batches exceeds 10%, ensuring that the algorithm adapts to the current production environment.

[0094] For example, in process parameter optimization, assuming the trend change pattern shows that the temperature parameter tends to stabilize in the 200-220 degree Celsius range in the historical trajectory, if the initial solution set distribution's temperature solutions are biased towards 180-200 degrees Celsius, then the calculated deviation value, such as an average difference of 15 degrees Celsius, is considered a deviation if it exceeds the threshold of 10 degrees Celsius. In this case, adjusting the crossover and mutation operator based on dynamic factors, such as a 5% increase in the moisture content of the raw material batch, can make the algorithm converge more quickly to adapt to the temperature combination of the new batch, thereby improving optimization efficiency.

[0095] In one embodiment, step S503, which involves iterative optimization using an adjusted genetic algorithm to generate an optimized parameter adjustment trajectory, includes: applying the adjusted crossover and mutation operator to the population iteration, wherein the fitness function is evaluated in each generation based on multiple objectives of yield and quality, and after 50 iterations, the solution sequence on the Pareto optimal front is selected as the trajectory.

[0096] Specifically, this iterative optimization can incorporate trend change patterns as constraints. For example, when initializing the population, individuals that are close to the historical trajectory are selected first. The adjusted operators ensure that the trajectory gradually moves closer to the predicted optimization direction vector, which helps to reduce invalid searches and improve the stability of parameter adjustment.

[0097] For example, in scenarios where the equipment is under high load, the generated trajectory after iteration may show that the speed parameter is gradually adjusted from the initial 15 m / min to 18 m / min, and the trajectory length covers 10 adjustment points. Each point verifies a 2% increase in output, which can support the accuracy of subsequent verification.

[0098] In one embodiment, step S504 verifies the parameter adjustment trajectory and generates a trajectory optimization result, including: comparing the matching degree of each parameter point in the trajectory with the real-time production data feedback; if the matching degree is higher than 90%, the trajectory is confirmed as an optimization result; otherwise, iterative optimization is returned.

[0099] For example, in the verification, if the pressure parameter predicted by the trajectory is 500 Pa and the real-time data is 510 Pa, and the matching degree is calculated to be 98%, then the generated result is used for MES system application to improve the prediction accuracy of process closed loop.

[0100] Step S6: Based on the parameter adjustment trajectory and real-time production data, determine whether to update the trend change pattern, including: obtaining real-time production data feedback; comparing the real-time production data with the parameter adjustment trajectory and the optimized direction vector; determining whether to update the trend change pattern based on the comparison result; if an update is required, generating new trend change pattern parameters based on the real-time production data.

[0101] In one embodiment, the step of adjusting the trajectory and real-time production data based on the parameters and determining whether to update the trend change pattern includes the following steps:

[0102] Step S601: Obtain real-time production data feedback.

[0103] Specifically, key indicators during the production process, such as equipment operating status, raw material batch changes, and process parameter values, are collected in real time through the MES system interface to form a feedback dataset. This dataset is directly used for subsequent comparisons to ensure data timeliness.

[0104] Step S602: Compare the real-time production data with the parameter adjustment trajectory and the optimized direction vector.

[0105] For example, after acquiring the feedback dataset, the Euclidean distance between the real-time data and the parameter adjustment trajectory is calculated and compared with the cosine similarity of the optimized direction vector. The Euclidean distance quantifies numerical deviation, while the cosine similarity measures directional consistency. If the distance exceeds a preset threshold or the similarity is below the threshold, it is marked as a significant deviation. This comparison process incorporates historical optimized trajectories to ensure that dynamic factors such as equipment status are included in the evaluation.

[0106] The deviation index is calculated based on the comparison results, for example, by using a weighted average deviation value, where the weights are based on the importance of the raw material batch.

[0107] Use deviation indicators to generate comparison reports to determine update requirements.

[0108] In one possible implementation, this step is applied to chemical production scenarios, such as polymerization processes. Real-time data, including temperature and pressure values, is compared with the adjustment trajectory. If deviations cause fluctuations in output, this comparison identifies the trend patterns that need optimization, thereby improving the accuracy of predictions.

[0109] Step S603: Determine whether to update the trend change pattern based on the comparison results.

[0110] Specifically, if the deviation index exceeds a preset threshold, such as 0.1, an update is required; otherwise, the original mode is maintained. The preset threshold is determined through historical data statistics to ensure the robustness of the judgment.

[0111] If the deviation index is within the threshold range, a sustain signal is output.

[0112] If the limit is exceeded, an update process will be triggered.

[0113] For example, in electronics manufacturing, this judgment can respond promptly when changes in equipment status cause parameter drift, avoiding production interruptions and improving the efficiency of the overall optimization cycle.

[0114] Step S604: If an update is required, generate new trend change pattern parameters based on the real-time production data.

[0115] For example, clustering algorithms can be used to group real-time data and extract trend features, such as slope and inflection points, to form new parameters. Specifically, the K-means clustering algorithm first initializes cluster centers, then iteratively allocates data points and updates the centers until convergence, thereby generating pattern parameters such as the average rate of change.

[0116] Preprocess real-time data and remove noise using a moving average filter.

[0117] The K-means algorithm is applied to group the data, and the trend parameters for each cluster are calculated, such as the slope fitted by linear regression.

[0118] By integrating historical patterns, the final new parameters are generated, ensuring alignment with the optimized direction vector.

[0119] In one embodiment, this step is applied in food processing scenarios, such as baking processes. Real-time data feedback of temperature trajectory, if deviating significantly from the predicted vector, generates new mode parameters such as a heating rate adjustment value of 0.5 degrees / minute, which can reduce energy waste and improve product quality consistency.

[0120] Specifically, in pharmaceutical production, the process of generating new parameters can be extended to multi-objective optimization, such as simultaneously considering output and energy consumption. The number of clusters in the K-means algorithm is set to 3, corresponding to low, medium and high change patterns respectively. The parameters include peak value and cycle length. In this way, the updated pattern can better match the current dynamic factors, resulting in higher prediction accuracy and production stability.

[0121] For example, on another level, when raw material batches mutate, the generation parameters emphasize the weight of batch details, and the calculation results show that the slope of the new pattern increases from 0.2 to 0.3, which supports the subsequent initialization of the genetic algorithm and improves the adaptability of the optimized trajectory.

[0122] Step S7: Based on the updated trend change pattern, train the model to generate an enhanced prediction accuracy index, including: generating a training dataset based on the updated trend change pattern; retraining the neural network model using the training dataset and updating the model weights; generating a prediction result based on the updated neural network model; and generating an enhanced prediction accuracy index based on a comparison of the prediction result with actual production data.

[0123] In one embodiment, step S701, generating a training dataset based on the updated trend change pattern, specifically includes extracting feature data from the updated trend change pattern that includes process parameter combination optimization trajectory, production plan equipment status, and raw material batch details. These feature data originate from historical optimization data and current dynamic factor information in the MES system, and are similar trajectory trends formed after grouping and processing by a clustering algorithm.

[0124] The extracted feature data is organized into input-output pairs to form a training dataset for subsequent model training.

[0125] Specifically, this implementation method can quickly generate datasets in a production environment. For example, in a chemical production scenario, after the trend change pattern is updated, the combination of parameters such as temperature and pressure of the past optimized trajectory can be directly extracted from the pattern as input, and the corresponding equipment status and raw material batches can be used as labels, thereby ensuring that the dataset reflects the real dynamic changes and is beneficial to improving the model's adaptability to similar production plans.

[0126] In one embodiment, step S702, retraining the neural network model using the training dataset and updating the model weights, specifically includes inputting the training dataset into the neural network model, wherein the neural network model adopts a multilayer perceptron structure, including an input layer, a hidden layer and an output layer. The input layer receives combinations of process parameters and dynamic factors, the hidden layer processes nonlinear relationships through activation functions such as ReLU, and the output layer generates a predicted optimization direction vector.

[0127] The loss function, such as mean squared error, is calculated using the backpropagation algorithm. The model weights are iteratively adjusted based on the input-output pairs of the training dataset, where each iteration optimizes the weight values ​​through gradient descent until the loss function converges to below a preset threshold.

[0128] Save the updated weights to form a retrained model for subsequent predictions.

[0129] Specifically, this implementation method is applicable to process optimization in the manufacturing field. For example, on an electronic product assembly line, when the trend change pattern is updated to include a trajectory with new equipment status, the model weights can be adjusted through retraining to better capture the impact of raw material batch changes on parameters. This can improve the robustness of prediction, reduce production deviations, and ensure that the weights reflect the latest trends through multiple iterations, which is beneficial to achieving a closed-loop optimization cycle.

[0130] For example, in one possible implementation, consider a specific production batch. The training dataset includes 100 historical trajectory samples, each with 10 process parameter dimensions. The retraining process sets the learning rate to 0.01. After 50 iterations, the weight update improves the model's prediction accuracy for similar batches from 80% to 95%, demonstrating the benefit of improving the model's generalization ability through dynamic updates.

[0131] In one embodiment, step S703, generating a prediction result based on the updated neural network model, specifically includes taking the equipment status and raw material batch variables of the current production plan as inputs, feeding them into the updated neural network model, calculating the predicted optimization direction vector through forward propagation, and outputting the result.

[0132] Specifically, this step ensures that the prediction results are derived directly from the retrained model, which is beneficial for rapid response to production changes.

[0133] In one embodiment, step S704 involves generating an enhanced prediction accuracy index based on the comparison between the prediction result and the actual production data. Specifically, this includes acquiring real-time production data, such as the actual process parameter adjustment trajectory and optimization effect, fed back from the MES system.

[0134] Calculate the similarity between the predicted result and the actual data. For example, use the cosine similarity formula to compare and optimize the direction vectors. The cosine similarity is defined as the dot product of two vectors divided by the product of their magnitudes. The closer the value is to 1, the higher the similarity.

[0135] An accuracy metric is generated based on the similarity value. For example, the similarity is converted into a percentage and compared with historical metrics. If the improvement exceeds 10%, it is marked as an enhanced predictive accuracy metric.

[0136] The generated indicators are recorded in the system to determine whether further iterations and updates to the trend change pattern are needed.

[0137] Specifically, this implementation method plays a key role in production optimization. For example, in the food processing scenario, after comparing the predicted temperature adjustment vector with the actual data, the accuracy index is calculated to be 92%, which is an improvement compared to 85% in the previous model. This helps to confirm the effectiveness of the model and guides the genetic algorithm to select high-quality solutions through the index, ultimately determining the combination of process parameters and forming a closed-loop process optimization cycle data.

[0138] For example, in one possible implementation, for a metal processing batch, the predicted result vector is [0.8, 0.6, 0.9], the actual data vector is [0.75, 0.65, 0.85], the calculated cosine similarity is approximately 0.98, and the generated index is 98%. This indicates that the enhanced accuracy exceeds the threshold of 90%, which is beneficial to reducing production waste and improving overall efficiency. Through this comparison, the influence of dynamic factors such as raw material batches on the trajectory can also be revealed, further supporting the iterative update of trend patterns.

[0139] Step S8: Determine the final combination of process parameters based on the predicted accuracy index and the updated solution set distribution, including: comparing the predicted accuracy index with the preset threshold; if the predicted accuracy index exceeds the preset threshold, proceed to the next selection process; otherwise, return to the iterative update trend change mode; select high-quality solutions from the updated solution set distribution, and select the solution with the highest fitness from the high-quality solutions as the final combination of process parameters.

[0140] In one embodiment, determining the final combination of process parameters based on the prediction accuracy index and the updated solution set distribution includes the following sub-steps:

[0141] Step S801: Compare the prediction accuracy index with the preset threshold.

[0142] The preset threshold is the average accuracy obtained through multiple iterations in the historical optimization trajectory. For example, setting it to 0.85 means that the model's prediction and the actual optimization direction need to match more than 85%.

[0143] If the prediction accuracy index exceeds the preset threshold, proceed to the next selection process; otherwise, return to the iterative update trend change mode.

[0144] Step S802: Select high-quality solutions from the updated solution set distribution.

[0145] The updated solution set distribution is a set of parameter combinations obtained after optimization by the genetic algorithm. Each solution represents a combination of process parameters, including variables such as temperature, pressure, and time.

[0146] The selection of high-quality solutions is based on a fitness function, which calculates the multi-objective value of each solution, such as maximizing output and minimizing energy consumption.

[0147] The specific calculation process of the fitness function is as follows: for each solution, calculate the weighted sum of the output target value and the energy consumption target value. The weighting coefficient is dynamically adjusted according to the production plan, for example, the output weight is 0.6 and the energy consumption weight is 0.4.

[0148] Step S803: Determine the final combination of process parameters.

[0149] The solution with the highest fitness among the high-quality solutions is selected as the final combination of process parameters.

[0150] For example, in a chemical production scenario, if the solution set distribution contains multiple temperature-pressure combinations, the one with the highest fitness is selected, such as a temperature of 150 degrees and a pressure of 2.5 MPa.

[0151] In one embodiment, in step S802, high-quality solutions are selected from the updated solution set distribution. Specifically, the genetic algorithm is an optimization algorithm that simulates natural evolution by iterating the population through selection, crossover, and mutation operations.

[0152] Here, the population of the genetic algorithm evolves from the initial solution set distribution, and each individual is encoded as a vector of process parameters, such as [temperature, pressure, time].

[0153] The selection process uses a roulette wheel method, where probability is proportional to fitness, ensuring that high-quality individuals are retained.

[0154] The crossover operation swaps the parameter vectors of two parent individuals at random points to generate offspring.

[0155] Mutation operations randomly change the values ​​in a vector with a small probability, such as increasing or decreasing the temperature value by 5 degrees.

[0156] These operations are used to select solutions with fitness values ​​exceeding the average from the updated solution set distribution as high-quality solutions.

[0157] The benefit of doing this is that it improves the global optimization capability of parameter combinations, avoids getting trapped in local optima, and ensures more stable process parameters in complex production environments.

[0158] For example, in semiconductor manufacturing in the same field, the solution set distribution may include a combination of wafer processing time and chemical concentration. After iterative genetic algorithm selection, the solution with the highest fitness can be selected, which can reduce the defect rate by up to 10%.

[0159] In one embodiment, when comparing the prediction accuracy index with a preset threshold in step S801, if the index is exactly equal to the threshold, it can be considered as exceeding the threshold, in order to accommodate boundary situations.

[0160] The prediction accuracy index is calculated based on the deviation between the output of the neural network model and the real-time feedback; the smaller the deviation, the higher the index.

[0161] A neural network model is a multilayer perceptron, consisting of an input layer, a hidden layer, and an output layer. The inputs are equipment status and raw material batch variables, and the output is an optimization direction vector.

[0162] The training process uses the backpropagation algorithm to adjust the weights, and the loss function is the mean squared error.

[0163] The updated solution set distribution incorporates trend change patterns as constraints to ensure that the solution set conforms to historical trajectories.

[0164] The beneficial effect is that it improves the robustness of decision-making, and can still accurately determine the parameter combination even in scenarios with large batch variations of raw materials.

[0165] For example, assuming the preset threshold is 0.85 and the calculated index is 0.87, then a solution is selected from the solution set distribution; if it is 0.83, then the mode is iteratively updated.

[0166] In one embodiment, after determining the final combination of process parameters in step S803, its matching degree with the production plan can be further verified.

[0167] The matching degree is obtained by calculating the similarity between the parameter combination and the planned objective. If the similarity is high, it is applied directly.

[0168] This closed-loop optimization can continuously improve process efficiency.

[0169] For example, in pharmaceutical production, a final combination such as a stirring speed of 200 rpm and a temperature of 37 degrees Celsius can increase yield by 15% and reduce energy consumption.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A process optimization method based on MES data analysis, characterized in that, include: Historical optimization data and current dynamic factor information are obtained from the manufacturing execution system to generate a comprehensive dataset; Based on the comprehensive dataset, historical optimized trajectories are grouped to determine trend change patterns; the method also includes: using a clustering algorithm to group historical optimized trajectories in the comprehensive dataset. Extract feature data of historical optimized trajectories for each group based on the grouping results; Analyze the trend change patterns of each group of historical optimization trajectories based on the aforementioned feature data; Based on the trend change pattern, generate pattern parameters that characterize the changing patterns of similar trajectories; Based on the trend change pattern and the current production plan, a predicted optimization direction vector is generated; the method also includes: determining whether the trend change pattern matches the current production plan. If a match is found, the equipment status and raw material batch variables in the current production plan are retrieved. The equipment status and raw material batch variables are input through a pre-trained neural network model; Based on the output of the neural network model, a predicted optimization direction vector is generated; The population is initialized based on the optimized direction vector and incorporated into the trend change pattern to generate an initial solution set distribution; the method also includes: initializing the population based on the optimized direction vector using a genetic algorithm. The aforementioned trend change pattern is incorporated into the genetic algorithm as a heuristic constraint. The genetic algorithm is used to perform multi-objective optimization and generate an initial solution set distribution. The initial solution set distribution is evaluated to generate distribution characteristic parameters of the initial solution set; Based on the initial solution set distribution and the dynamic factor information, an optimized parameter adjustment trajectory is generated; Based on the parameters, adjust the trajectory and real-time production data to determine whether to update the trend change pattern; Based on the updated trend change pattern, the trained model generates an enhanced predictive accuracy metric. Based on the prediction accuracy index and the updated solution set distribution, the final combination of process parameters is determined.

2. The method as described in claim 1, characterized in that, The process of acquiring historical optimization data and current dynamic factor information from the manufacturing execution system to generate a comprehensive dataset includes: Obtain historical optimization data containing process parameter combination optimization trajectories from the manufacturing execution system; Obtain current dynamic factor information including production plans, equipment status, and raw material batch details; Based on the historical optimization data and the dynamic factor information, a comprehensive dataset is generated that includes the optimization trajectory of process parameter combinations, production plan, equipment status and raw material batch details; The comprehensive dataset is preprocessed to generate a standardized dataset format.

3. The method as described in claim 1, characterized in that, The step of generating an optimized parameter adjustment trajectory based on the initial solution set distribution and the dynamic factor information includes: Determine whether the initial solution set distribution deviates from the trend change pattern; If the deviation occurs, the crossover and mutation operators of the genetic algorithm are adjusted according to the dynamic factor information. The optimized parameter adjustment trajectory is generated by iterative optimization using the adjusted genetic algorithm. The trajectory of the parameter adjustment is verified, and the trajectory optimization result is generated.

4. The method as described in claim 1, characterized in that, The step of adjusting the trajectory and real-time production data based on the parameters, and determining whether to update the trend change pattern, includes: Obtain real-time production data feedback; The real-time production data is compared with the parameter adjustment trajectory and the optimized direction vector; Determine whether to update the trend change pattern based on the comparison results; If an update is required, new trend change pattern parameters are generated based on the real-time production data.

5. The method as described in claim 1, characterized in that, The step of training the model to generate enhanced prediction accuracy metrics based on the updated trend change pattern includes: A training dataset is generated based on the updated trend change pattern. The neural network model is retrained using the training dataset, and the model weights are updated. Predictions are generated based on the updated neural network model; Based on the comparison between the predicted results and actual production data, an enhanced prediction accuracy index is generated.

6. The method as described in claim 1, characterized in that, The step of determining the final combination of process parameters based on the prediction accuracy index and the updated solution set distribution includes: Compare the prediction accuracy index with the preset threshold. If the prediction accuracy index exceeds the preset threshold, proceed to the next selection process; otherwise, return to the iterative update trend change mode. Select high-quality solutions from the updated solution set distribution, and choose the solution with the highest fitness from the high-quality solutions as the final combination of process parameters.

Images