Control method and system of automation equipment for industrial production

Abstract

The invention discloses a control method and system of automation equipment for industrial production, and particularly relates to the technical field of automation equipment control. The method comprises the following steps: extracting diversified samples of normal values, abnormal values and hidden defects from an existing data warehouse, labeling categories, and constructing a high-quality initial data set; evaluating the stability of the model by using a layered K-fold cross validation method, and testing and evaluating the consistency of the abnormal response capability and the historical mode through a high-load condition when the model is unstable; comprehensively analyzing the stability and the abnormal response capability of the model, and evaluating the capability of the AI for identifying error modes in historical data; for inaccurate identification, the dynamic time warping algorithm is used for detecting the deviation degree of the equipment operation parameters, an alarm is triggered in time, and the production process is paused to guarantee the safety of the equipment, so that the reliability and accuracy of the AI model in a dynamic environment are remarkably improved, and meanwhile, the risks of equipment damage and production safety accidents caused by optimization errors are effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of automation equipment control, and in particular to a control method and system for automation equipment used in industrial production. Background Art

[0002] In industrial production, utilizing automated equipment to execute and optimize production processes is a technological approach to improving efficiency. Control methods for automated equipment primarily involve coordinating and managing key production steps. Through sensors and control software, the equipment can monitor and adjust parameters at each step in real time. For example, in thermoforming, control methods can precisely set heating time and temperature based on different recipes and production requirements, ensuring the desired texture and shape.

[0003] For example, in pudding's automated production system, the automated control method also incorporates intelligent features. By incorporating artificial intelligence or machine learning algorithms, the equipment can analyze historical data from the production process, optimize parameter settings, and improve production efficiency. Furthermore, this method can be integrated with pudding production management systems (such as SCADA systems) to automate the entire process, from raw material delivery to finished product packaging. For example, if sensors detect insufficient raw materials, the control system can automatically issue a refill instruction to avoid production interruption, enabling unmanned or reduced-management operation and improving the overall stability and efficiency of the production line.

[0004] The existing technology has the following shortcomings: By incorporating artificial intelligence or machine learning algorithms, equipment can analyze historical data from the pudding production process and optimize parameter settings. However, AI relies on historical data to optimize parameters. If the historical data contains hidden flaws caused by human error or equipment aging, the algorithm may use these erroneous patterns as the basis for optimization, creating long-term potential risks. For example, if poor raw material quality causes the equipment to extend the mixing time, the AI ​​may interpret this as reasonable and continue to use this setting, ultimately causing excessive wear and tear on the equipment or increased energy consumption. Furthermore, the AI ​​may interpret abnormal operating parameters caused by equipment aging or failure in the data as optimal settings and repeatedly apply them. For example, if the aging of the equipment's heating module causes the temperature sensor to falsely indicate a low temperature, the AI ​​may interpret this as a reason to increase the set temperature, causing it to operate at an excessively high temperature for a long time. This can result in the heating module overheating and burning, or even burning the raw materials within the equipment due to overheating, causing a fire or serious production safety accident. Summary of the Invention

[0005] The object of the present invention is to provide a control method and system for industrial production automation equipment to solve the shortcomings of the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a control method for industrial production automation equipment, comprising the following steps: S1: Extract diverse samples containing normal values, abnormal values, and hidden defects from the existing data warehouse, annotate the data in the extracted diverse samples, and distinguish the categories of normal values, abnormal values, and hidden defects to construct the initial dataset; S2: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations; S3: When the AI ​​model becomes unstable, adjust the equipment to high load conditions, observe the AI ​​model's response to the anomaly, and evaluate its consistency with the pattern in historical data; S4: Comprehensively analyze the stability of the AI ​​model on different sample combinations and its ability to respond to anomalies with the consistency of patterns in historical data to evaluate the accuracy of AI in identifying error patterns in historical data; S5: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns; S6: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

[0007] Preferably, in S2, a stratified K-fold cross-validation method is used to divide the initial data set into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations, specifically: Stratified K-fold cross validation stratifies the dataset according to the distribution ratio of the categories and divides the data into K equal subsets. One subset is selected as the test set, and the rest are used as the training set, and the model training and testing are repeated in a cycle. Stratify the initial dataset by category label to ensure that the proportion of each category in the divided subset is consistent with that in the original dataset; According to the stratification results, each type of data is randomly divided into K subsets to ensure that the number of samples of each type in each subset is uniform. For data set D, each type of data Split into ; Each time a subset is taken as the test set, the remaining The subsets are used as training sets for model training and testing, and repeated K times. In each iteration, the training set is used to train the model M, and the test set is used for prediction. The prediction results and the true labels are recorded, and the overall classification accuracy of the model is calculated. , the expression is: ; In the formula, TP is the number of samples correctly predicted as positive, TN is the number of samples correctly predicted as negative, FP is the number of samples wrongly predicted as positive, and FN is the number of samples wrongly predicted as negative. The recognition sensitivity of the calculation model to the correct samples is: , the expression is: ; Calculate the accuracy of the model's predictions for correct samples , the expression is: ;comprehensive and Calculate the F1 score of the model, the expression is: ; Calculate the average value of each verification result and standard deviation , the expression is: ; K is the fold, is the evaluation index of the i-th fold; ; Calculate the stability fluctuation index, the expression is: Where, is the stability fluctuation index.

[0008] Preferably, in S2, the obtained stability fluctuation index is compared with the stability fluctuation index reference threshold set under normal conditions in historical data. If the stability fluctuation index is greater than or equal to the stability fluctuation index reference threshold, it means that the AI ​​model has large fluctuations on different sample combinations and is unstable, and an unstable signal is generated at this time; if the stability fluctuation index is less than the stability fluctuation index reference threshold, it means that the AI ​​model has small fluctuations on different sample combinations and operates stably, and a stable signal is generated at this time, indicating that the model performance is in line with expectations and no adjustment is required.

[0009] Preferably, in S3, the parameter adjustment amplitude of the AI ​​model under high load conditions is compared with the parameter adjustment amplitude in historical data to generate a parameter adjustment amplitude deviation index, and the parameter adjustment amplitude deviation index is obtained by: The parameter adjustment amplitude sequence generated by the AI ​​model under high load conditions is marked as ; Mark the parameter adjustment amplitude sequence recorded in the historical data as ; The linear correlation between two data series is measured by the Pearson correlation coefficient , and its calculation formula is: Where, The i-th adjustment amplitude value generated by the AI ​​model, is the i-th adjustment amplitude value in the historical data, is the mean of the AI ​​adjustment amplitude, is the mean of the historical adjustment range, and n is the total number of data points. The parameter adjustment range deviation index is used to quantify the numerical difference between the AI ​​adjustment range and the historical adjustment range. The calculation formula is: Where, is the absolute deviation of the i-th data point, indicating the difference between the AI-adjusted value and the historical value. It is the absolute sum of historical adjustment values, which serves as the benchmark value and is used to normalize the deviation. SD is the parameter adjustment amplitude deviation index.

[0010] Preferably, in S4, the stability of the AI ​​model on different sample combinations and the AI ​​model's ability to respond to anomalies are comprehensively analyzed with the pattern consistency in the historical data to evaluate the accuracy of the AI ​​in identifying error patterns in the historical data, specifically: The stability fluctuation index and the parameter adjustment amplitude deviation index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the accuracy value label of the error pattern in the AI ​​recognition historical data as the prediction target, and takes minimizing the sum of the prediction errors of the accuracy value labels of all AI recognition historical data as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The accuracy value of the AI ​​recognition historical data error pattern is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0011] Preferably, the accuracy value of the error pattern in the historical data obtained by AI recognition is compared with the accuracy value reference threshold set according to the normal state of the historical data. If the accuracy value of the error pattern in the historical data recognized by AI is greater than or equal to the accuracy value reference threshold, it means that the accuracy of the error pattern in the historical data recognized by AI is high, and an accurate recognition signal is generated at this time, indicating that the AI ​​recognition performance is good; if the accuracy value of the error pattern in the historical data recognized by AI is less than the accuracy value reference threshold, it means that the accuracy of the error pattern in the historical data recognized by AI is low, and an inaccurate recognition signal is generated at this time, indicating that the AI ​​model needs to be further optimized.

[0012] Preferably, in S6, for inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters significantly deviate from the normal range. When it is detected that the equipment operating parameters set by the AI ​​deviate from the normal range, an alarm is automatically triggered and the production process is suspended. Specifically: Real-time equipment operating parameter time series: R ; is the xth data point of the equipment operating parameters monitored in real time; refer to the time series of historical normal operating parameters: : is the yth data point of the equipment parameters under the historical normal state; Define the distance matrix D of two time series R and N: ; Real-time parameters and normal parameters distance; construct the cumulative cost matrix C and calculate the minimum path cost from (1,1) to (x,y): : The minimum cumulative cost to reach point (i, j); the final distance of dynamic time warping is: DTW(R,N)=C(x,y): The last value of the cumulative cost matrix represents the overall matching cost of R and N; Set the DTW distance threshold DTWthreshold within the normal range: Calculate the average DTW distance μ and standard deviation σ during normal operation from historical data: ; k is the safety factor of the control range; If DTW(R,N)>DTWthreshold, an early warning signal is generated and the difference between the real-time parameters of the device and the normal parameters is recorded; if DTW(R,N)≤DTWthreshold, no early warning signal is generated, the device continues to operate normally and records status data.

[0013] The present invention also provides a control system for industrial production automation equipment, including a data construction module, a model evaluation module, a high-load response test module, an accuracy evaluation module, a model optimization module, and an equipment control module: Data construction module: Extracts diverse samples containing normal values, abnormal values, and hidden defects from the existing data warehouse, labels the data in the extracted diverse samples, and distinguishes the categories of normal values, abnormal values, and hidden defects to construct the initial data set; Model evaluation module: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations; High-load response test module: When the AI ​​model becomes unstable, the device is adjusted to high-load conditions to observe the AI ​​model's response to the anomaly and evaluate its consistency with patterns in historical data; Accuracy Assessment Module: This module comprehensively analyzes the stability of the AI ​​model on different sample combinations, its ability to respond to anomalies, and the consistency of patterns in historical data to assess the accuracy of AI in identifying erroneous patterns in historical data. Model Optimization Module: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns; Equipment control module: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

[0014] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. This invention effectively solves the problem in existing technologies that AI relies on historical data, which may cause optimization errors, form potential risks, and lead to equipment safety accidents. By extracting and labeling normal values, abnormal values, and hidden defect samples to construct an initial data set, and combining stratified K-fold cross-validation to evaluate model stability, the AI ​​model can adapt to diverse data. When the model is unstable, high-load testing is used to evaluate the model's abnormal response ability to ensure the rationality of AI optimization in accordance with historical patterns. The stability fluctuation index and parameter adjustment amplitude deviation index are further used to comprehensively evaluate the accuracy of the model in identifying error patterns. Accurate models are optimized in real time, and inaccurate models are dynamically detected and warned of anomalies to ensure production safety.

[0015] 2. This invention significantly improves the classification accuracy and dynamic adaptability of the AI ​​model for normal and abnormal patterns, preventing equipment damage and safety hazards caused by hidden defects or misleading optimization due to abnormal data; at the same time, the system accurately predicts equipment parameter deviations through a dynamic time warping algorithm, triggers alarms in real time, and suspends the production process, effectively reducing equipment operation risks and the incidence of production accidents, and ensuring the reliability and production efficiency of automated equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0017] Figure 1 Flow chart of the method of the present invention.

[0018] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Example 1, please refer to Figure 1 and Figure 2As shown, the control method of an industrial production automation equipment described in this embodiment includes the following steps: S1: Extract diverse samples containing normal values, abnormal values, and hidden defects from the existing data warehouse, annotate the data in the extracted diverse samples, and distinguish the categories of normal values, abnormal values, and hidden defects to construct the initial dataset; S2: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations; S3: When the AI ​​model becomes unstable, adjust the equipment to high load conditions, observe the AI ​​model's response to the anomaly, and evaluate its consistency with the pattern in historical data; S4: Comprehensively analyze the stability of the AI ​​model on different sample combinations and its ability to respond to anomalies with the consistency of patterns in historical data to evaluate the accuracy of AI in identifying error patterns in historical data; S5: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns; S6: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

[0021] In S1, a diverse sample containing normal values, abnormal values, and hidden defects is extracted from the existing data warehouse. The data in the extracted diverse samples is labeled and the categories of normal values, abnormal values, and hidden defects are distinguished to construct the initial data set. Specifically: Extract key data recorded during the production process from the existing data warehouse, ensuring that it contains the following three types of samples: Normal value: Routine operating data during the production process, such as temperature, pressure, stirring speed, etc. under stable conditions, reflecting the normal working conditions of the equipment.

[0022] Outliers: Abnormal operating data caused by unexpected situations (such as equipment failure and sensor false alarms), such as extreme fluctuations in temperature or pressure.

[0023] Hidden defects: Data anomalies that are difficult to detect and are caused by equipment aging or gradual accumulation of problems, such as small but persistent temperature deviations or a trend of declining equipment efficiency.

[0024] Data extraction can be performed in the following ways: Use SQL queries to extract historical data records that meet specific criteria. Filter production records by time period to ensure that normal, abnormal, and extreme operating data are included. Invoke the API of the equipment monitoring system or IoT platform to obtain raw data.

[0025] Label the extracted diverse samples and clarify the category of each record: Category definition: Normal value: Parameters that are within the standard range of production process. Outlier: Data points that exceed the production process standards or have mutations. Hidden defects: Problems where the parameters appear normal but show a gradual deviation or hidden trend. Labeling method: Manual labeling: Analyze and classify the data based on production records and manual experience, such as referring to product quality inspection reports or equipment maintenance records. Mark outliers based on threshold rules, such as temperature exceeding the set range by ±10% or pressure short-term fluctuations exceeding the standard. Use anomaly detection algorithms (such as isolation forest or DBSCAN) to pre-label data, and combine manual review to correct the classification results.

[0026] To ensure a balanced distribution of sample categories so the model can correctly identify each type of data: If the proportion of normal values ​​is too high, oversample outliers and hidden defects (using a SMOTE algorithm, for example) or synthesize data. Control the ratio of samples in each category, for example, normal values: outliers: hidden defects = 70:20:10. Ensure that the extracted data covers different production time periods (high load, low load, and before and after maintenance).

[0027] Organize the labeled data into a standardized format to construct an initial dataset: including timestamps, key parameters (temperature, pressure, speed, etc.), and category labels (normal, abnormal, hidden defect). Store the data in CSV, JSON, or database tables to facilitate subsequent model training. Split the data into training and validation sets in a 7:3 or 8:2 ratio, reserving a portion for later testing.

[0028] S2: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations.

[0029] Stratified K-fold cross-validation stratifies the dataset according to the distribution ratio of categories (such as normal values, outliers, and hidden defects), and divides the data into K equal subsets. One subset is selected as the test set, and the rest are used as the training set, and the model training and testing are repeated.

[0030] The initial dataset is stratified by category labels (such as normal values, outliers, and hidden defects) to ensure that the proportion of each category in the divided subset is consistent with that in the original dataset.

[0031] According to the stratification results, each type of data is randomly divided into K subsets to ensure that the number of samples of each type in each subset is uniform. For example, for data set D, each type of data Split into .

[0032] Each time a subset is taken as the test set, the remaining The subsets are used as training sets for model training and testing, and repeated K times. In each iteration, the training set is used to train the model M, and the test set is used for prediction. The prediction results and the true labels are recorded, and the overall classification accuracy of the model is calculated. , the expression is: ; In the formula, TP is the number of samples correctly predicted as positive, TN is the number of samples correctly predicted as negative, FP is the number of samples wrongly predicted as positive, and FN is the number of samples wrongly predicted as negative. The recognition sensitivity of the calculation model to the correct samples is: , the expression is: ; Calculate the accuracy of the model's predictions for correct samples , the expression is: ;comprehensive and Calculate the F1 score of the model, the expression is: ; Calculate the average value of each verification result and standard deviation , the expression is: ; K is the fold, is the evaluation indicator of the i-th fold (such as accuracy, F1 score, etc.); ; Calculate the stability fluctuation index, the expression is: Where, is the stability fluctuation index.

[0033] The obtained stability fluctuation index is compared with the stability fluctuation index reference threshold set under normal conditions in historical data. If the stability fluctuation index is greater than or equal to the stability fluctuation index reference threshold, it means that the AI ​​model fluctuates too much on different sample combinations and is unstable. At this time, an unstable signal is generated, indicating that the model needs to be further optimized or the sample distribution needs to be adjusted; if the stability fluctuation index is less than the stability fluctuation index reference threshold, it means that the AI ​​model fluctuates little on different sample combinations and runs stably. At this time, a stable signal is generated, indicating that the model performance is in line with expectations and no adjustment is required.

[0034] S3: When the AI ​​model becomes unstable, adjust the device to high load conditions, observe the AI ​​model's response to the anomaly, and evaluate its consistency with the pattern in historical data.

[0035] Adjust equipment operating parameters to high loads close to or exceeding normal production ranges to simulate extreme operating conditions that might occur in real life. For example, increase the mixer speed to the maximum design value or beyond the normal operating range. Raise the heating module temperature to near the upper limit of the safety threshold. Increase the raw material input to simulate full load or even overload operating conditions.

[0036] Record real-time operating data of the equipment under high load conditions, including key parameters (such as temperature, pressure, and stirring speed) and the AI ​​model's response recommendations (such as optimizing parameter adjustments). Collect historical data from the equipment under high load conditions as a baseline for comparison. Compare the AI ​​model's response recommendations under high load conditions with the response patterns in historical data (such as the direction and magnitude of parameter adjustments). Determine whether the AI ​​model can correctly identify anomalies and generate reasonable adjustment recommendations.

[0037] The parameter adjustment amplitude deviation index is generated by comparing the parameter adjustment amplitude of the AI ​​model under high load conditions with the parameter adjustment amplitude in historical data. The parameter adjustment amplitude deviation index is obtained as follows: Mark the parameter adjustment amplitude sequence (such as temperature adjustment value, pressure adjustment value) generated by the AI ​​model under high load conditions as ; Mark the parameter adjustment amplitude sequence recorded in the historical data (the actual adjustment value of the same dimension) as ; The linear correlation between two data series is measured by the Pearson correlation coefficient, and its calculation formula is: Where, The i-th adjustment amplitude value generated by the AI ​​model, is the i-th adjustment amplitude value in the historical data, is the mean of the AI ​​adjustment amplitude, is the mean of the historical adjustment range, and n is the total number of data points. r = 1: perfect positive correlation, the AI ​​adjustment range is most consistent with the historical range. r = −1: perfect negative correlation, the AI ​​adjustment range is completely opposite to the historical range. r = 0: no correlation, the AI ​​adjustment range has no correlation with the historical range. The parameter adjustment range deviation index is used to quantify the numerical difference between the AI ​​adjustment range and the historical adjustment range. The calculation formula is: Where, is the absolute deviation of the i-th data point, indicating the difference between the AI-adjusted value and the historical value. It is the absolute sum of historical adjustment values, which serves as the benchmark value and is used to normalize the deviation. SD is the parameter adjustment amplitude deviation index.

[0038] A larger parameter adjustment deviation index indicates a lower consistency between the AI ​​model's response to anomalies and the patterns in historical data. This indicates that the AI's adjustments to high load or abnormal conditions deviate from the empirical patterns in historical data, potentially leading to inappropriate parameter settings, such as over-optimization or under-adjustment, which in turn reduces the model's reliability and adaptability in production. This deviation may stem from the model's inadequate recognition of abnormal data, imbalanced training data, or inadequate consideration of the physical limitations of equipment operation, necessitating optimization and adjustment of the model algorithm or data samples.

[0039] The smaller the parameter adjustment deviation index, the more consistent the AI ​​model's response to anomalies is with patterns in historical data. This means the AI ​​can make adjustments consistent with historical patterns under high load conditions, responding appropriately to anomalies and closely matching optimization strategies based on human experience. High consistency generally indicates that the AI ​​model has good generalization capabilities and can operate smoothly in complex environments while ensuring production efficiency and equipment safety. This performance verifies the model's stability and effectiveness, making it suitable for optimizing and controlling real-world production processes.

[0040] S4: Comprehensively analyze the stability of the AI ​​model on different sample combinations and the AI ​​model's ability to respond to anomalies with the pattern consistency in historical data to evaluate the accuracy of AI in identifying error patterns in historical data.

[0041] The stability fluctuation index and the parameter adjustment amplitude deviation index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the accuracy value label of the error pattern in the AI ​​recognition historical data as the prediction target, and takes minimizing the sum of the prediction errors of the accuracy value labels of all AI recognition historical data as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The accuracy value of the AI ​​recognition historical data error pattern is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0042] The method for obtaining the accuracy value of AI recognition of error patterns in historical data is to obtain the corresponding function expression from the comprehensive feature vector training data of the trained machine learning model: Where, is the output function of the model, is the stability fluctuation index, SD is the parameter adjustment amplitude deviation index, A value for the AI's accuracy in identifying erroneous patterns in historical data.

[0043] The obtained accuracy value of the error pattern in the AI ​​recognition historical data is compared with the accuracy value reference threshold set according to the normal state of the historical data. If the accuracy value of the error pattern in the AI ​​recognition historical data is greater than or equal to the accuracy value reference threshold, it means that the accuracy of the error pattern in the AI ​​recognition historical data is high. At this time, an accurate recognition signal is generated, indicating that the AI ​​recognition performance is good; if the accuracy value of the error pattern in the AI ​​recognition historical data is less than the accuracy value reference threshold, it means that the accuracy of the error pattern in the AI ​​recognition historical data is low. At this time, an inaccurate recognition signal is generated, indicating that the AI ​​model needs to be further optimized.

[0044] S5: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns.

[0045] For accurate identification, the introduction of real-time feedback data and dynamic learning mechanisms allows for continuous optimization of AI models, ensuring more accurate classification of normal and abnormal patterns. Real-time feedback data provides the latest information on equipment status and parameter changes during the production process, which can be used to dynamically update the model: Continuously monitor equipment operating status to obtain real-time parameters (such as temperature, pressure, and adjustment range) and production results (such as product quality). Compare the AI ​​model's classification results with actual production data to identify misclassifications or improper adjustments. Based on real-time errors, the model's weights or parameters are updated to promptly correct classification errors for abnormal patterns and improve the model's adaptability to the latest data.

[0046] Dynamic learning mechanisms enhance AI models' ability to classify complex environments and rare anomalies by continuously learning from new data. Real-time feedback data is gradually added to the model training set to prevent performance degradation caused by changes in data distribution (such as concept drift). Online learning algorithms (such as stochastic gradient descent or reinforcement learning) are introduced to automatically adjust model hyperparameters based on the real-time environment, improving classification accuracy. Rare anomaly samples are added through sampling or simulation techniques to strengthen the model's ability to recognize uncommon abnormal patterns.

[0047] S6: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

[0048] Real-time equipment operating parameter time series: R ; The xth data point of the real-time monitored equipment operating parameters (such as temperature, pressure, speed); refer to the time series of historical normal operating parameters: : is the yth data point of the equipment parameters under historical normal conditions.

[0049] Define the distance matrix D of two time series R and N: ; Real-time parameters and normal parameters Absolute distance; construct the cumulative cost matrix C and calculate the minimum path cost from (1,1) to (x,y): : The minimum cumulative cost to reach point (i, j); the final distance of dynamic time warping is: DTW(R,N)=C(x,y): The last value of the cumulative cost matrix, which represents the overall matching cost of R and N.

[0050] Set the DTW distance threshold DTWthreshold within the normal range: Calculate the average DTW distance μ and standard deviation σ during normal operation from historical data: ; k is the safety factor of the control range (usually k=2).

[0051] If DTW(R,N)>DTWthreshold, an early warning signal is generated and the degree of difference between the real-time parameters of the equipment and the normal parameters and the possible reasons are recorded; if DTW(R,N)≤DTWthreshold, no early warning signal is generated, the equipment continues to operate normally and records status data.

[0052] In this embodiment, first, a diverse set of samples containing normal values, abnormal values, and hidden defects are extracted from the existing data warehouse, and the data are labeled to construct an initial data set; then, stratified K-fold cross-validation is used to evaluate the stability of the model on different sample combinations; when the model becomes unstable, the equipment is adjusted to high load conditions to observe its response to anomalies and evaluate its consistency with historical patterns; then, the stability and abnormal response capabilities of the model are comprehensively analyzed to evaluate its accuracy in recognizing error patterns; for accurate identification, real-time feedback data and dynamic learning mechanisms are used to optimize the model to ensure its classification accuracy; if inaccurate identification is detected, the degree of deviation of the operating parameters is predicted through a dynamic time warping algorithm, and if necessary, an alarm is triggered and production is suspended to ensure system safety and stability.

[0053] Example 2: A control system for industrial production automation equipment described in this example includes a data construction module, a model evaluation module, a high-load response test module, an accuracy evaluation module, a model optimization module, and an equipment control module: Data construction module: Extracts diverse samples containing normal values, abnormal values, and hidden defects from the existing data warehouse, labels the data in the extracted diverse samples, and distinguishes the categories of normal values, abnormal values, and hidden defects to construct the initial data set; Model evaluation module: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations; High-load response test module: When the AI ​​model becomes unstable, the device is adjusted to high-load conditions to observe the AI ​​model's response to the anomaly and evaluate its consistency with patterns in historical data; Accuracy Assessment Module: This module comprehensively analyzes the stability of the AI ​​model on different sample combinations, its ability to respond to anomalies, and the consistency of patterns in historical data to assess the accuracy of AI in identifying erroneous patterns in historical data. Model Optimization Module: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns; Equipment control module: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

[0054] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0055] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0056] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0057] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A control method for industrial production automation equipment, characterized by: The following steps are involved: S1: Extract diverse samples containing normal values, abnormal values, and hidden defects from the existing data warehouse, annotate the data in the extracted diverse samples, and distinguish the categories of normal values, abnormal values, and hidden defects to construct the initial dataset; S2: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations; S3: When the AI ​​model becomes unstable, adjust the equipment to high load conditions, observe the AI ​​model's response to the anomaly, and evaluate its consistency with the pattern in historical data; S4: Comprehensively analyze the stability of the AI ​​model on different sample combinations and its ability to respond to anomalies with the consistency of patterns in historical data to evaluate the accuracy of AI in identifying error patterns in historical data; S5: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns; S6: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

2. The control method for industrial production automation equipment according to claim 1, characterized in that: In S2, the stratified K-fold cross-validation method is used to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations. Specifically: Stratified K-fold cross validation stratifies the dataset according to the distribution ratio of the categories and divides the data into K equal subsets. One subset is selected as the test set, and the rest are used as the training set, and the model training and testing are repeated in a cycle. Stratify the initial dataset by category label to ensure that the proportion of each category in the divided subset is consistent with that in the original dataset; According to the stratification results, each type of data is randomly divided into K subsets to ensure that the number of samples of each type in each subset is uniform. For data set D, each type of data Split into ; Each time a subset is taken as the test set, the remaining The subsets are used as training sets for model training and testing, and repeated K times. In each iteration, the training set is used to train the model M, and the test set is used for prediction. The prediction results and the true labels are recorded, and the overall classification accuracy of the model is calculated. , the expression is: ; In the formula, TP is the number of samples correctly predicted as positive, TN is the number of samples correctly predicted as negative, FP is the number of samples wrongly predicted as positive, and FN is the number of samples wrongly predicted as negative. The recognition sensitivity of the calculation model to the correct samples is: , the expression is: ; Calculate the accuracy of the model's predictions for correct samples , the expression is: ;comprehensive and Calculate the F1 score of the model, the expression is: ; Calculate the average value of each verification result and standard deviation , the expression is: ; K is the fold, is the evaluation index of the i-th fold; ; Calculate the stability fluctuation index, the expression is: Where, is the stability fluctuation index.

3. The control method for industrial production automation equipment according to claim 2, characterized in that: In S2, the obtained stability fluctuation index is compared with the stability fluctuation index reference threshold set under normal conditions in historical data. If the stability fluctuation index is greater than or equal to the stability fluctuation index reference threshold, it means that the AI ​​model has large fluctuations on different sample combinations and is unstable, and an unstable signal is generated at this time; if the stability fluctuation index is less than the stability fluctuation index reference threshold, it means that the AI ​​model has small fluctuations on different sample combinations and operates stably, and a stable signal is generated at this time, indicating that the model performance is in line with expectations and no adjustment is required.

4. The control method for industrial production automation equipment according to claim 3, characterized in that: In S3, the parameter adjustment amplitude of the AI ​​model under high load conditions is compared with the parameter adjustment amplitude in historical data to generate a parameter adjustment amplitude deviation index. The parameter adjustment amplitude deviation index is obtained as follows: The parameter adjustment amplitude sequence generated by the AI ​​model under high load conditions is marked as ; Mark the parameter adjustment amplitude sequence recorded in the historical data as ; The linear correlation between two data series is measured by the Pearson correlation coefficient , and its calculation formula is: Where, The i-th adjustment amplitude value generated by the AI ​​model, is the i-th adjustment amplitude value in the historical data, is the mean of the AI ​​adjustment amplitude, is the mean of the historical adjustment range, and n is the total number of data points. The parameter adjustment range deviation index is used to quantify the numerical difference between the AI ​​adjustment range and the historical adjustment range. The calculation formula is: Where, is the absolute deviation of the i-th data point, indicating the difference between the AI-adjusted value and the historical value. It is the absolute sum of historical adjustment values, which serves as the benchmark value and is used to normalize the deviation. SD is the parameter adjustment amplitude deviation index.

5. The control method for industrial production automation equipment according to claim 4, characterized in that: In S4, the stability of the AI ​​model on different sample combinations and its ability to respond to anomalies are comprehensively analyzed with the pattern consistency in historical data to evaluate the accuracy of AI in identifying error patterns in historical data. Specifically: The stability fluctuation index and the parameter adjustment amplitude deviation index are converted into comprehensive feature vectors, and the comprehensive feature vectors are used as the input of the machine learning model. The machine learning model uses each set of comprehensive feature vectors to predict the accuracy value label of the error pattern in the AI ​​recognition historical data as the prediction target, and takes minimizing the sum of the prediction errors of the accuracy value labels of all AI recognition historical data as the training target. The machine learning model is trained until the sum of the prediction errors reaches convergence, and the model training is stopped. The accuracy value of the AI ​​recognition historical data error pattern is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

6. The control method for industrial production automation equipment according to claim 5, characterized in that: The obtained accuracy value of the error pattern in the AI ​​recognition historical data is compared with the accuracy value reference threshold set according to the normal state of the historical data. If the accuracy value of the error pattern in the AI ​​recognition historical data is greater than or equal to the accuracy value reference threshold, it means that the accuracy of the error pattern in the AI ​​recognition historical data is high. At this time, an accurate recognition signal is generated, indicating that the AI ​​recognition performance is good; if the accuracy value of the error pattern in the AI ​​recognition historical data is less than the accuracy value reference threshold, it means that the accuracy of the error pattern in the AI ​​recognition historical data is low. At this time, an inaccurate recognition signal is generated, indicating that the AI ​​model needs to be further optimized.

7. The control method for industrial production automation equipment according to claim 1, characterized in that: In S6, for inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When the AI-set equipment operating parameters are detected to deviate from the normal range, an alarm is automatically triggered and the production process is suspended. Specifically: Real-time equipment operating parameter time series: R ; is the xth data point of the equipment operating parameters monitored in real time; refer to the time series of historical normal operating parameters: : is the yth data point of the equipment parameters under the historical normal state; Define the distance matrix D of two time series R and N: ; Real-time parameters and normal parameters distance; construct the cumulative cost matrix C and calculate the minimum path cost from (1,1) to (x,y): : The minimum cumulative cost to reach point (i, j); the final distance of dynamic time warping is: DTW(R,N)=C(x,y): The last value of the cumulative cost matrix represents the overall matching cost of R and N; Set the DTW distance threshold DTWthreshold within the normal range: Calculate the average DTW distance μ and standard deviation σ during normal operation from historical data: ; k is the safety factor of the control range; If DTW(R,N)>DTWthreshold, an early warning signal is generated and the difference between the real-time parameters of the device and the normal parameters is recorded; if DTW(R,N)≤DTWthreshold, no early warning signal is generated, the device continues to operate normally and records status data.

8. A control system for industrial production automation equipment, used to implement the control method for industrial production automation equipment according to any one of claims 1 to 7, characterized in that: It includes data construction module, model evaluation module, high-load response test module, accuracy evaluation module, model optimization module and equipment control module: Data construction module: Extracts diverse samples containing normal values, abnormal values, and hidden defects from the existing data warehouse, labels the data in the extracted diverse samples, and distinguishes the categories of normal values, abnormal values, and hidden defects to construct the initial data set; Model evaluation module: Use the stratified K-fold cross-validation method to divide the initial dataset into a training set and a test set to evaluate the stability of the AI ​​model on different sample combinations; High-load response test module: When the AI ​​model becomes unstable, the device is adjusted to high-load conditions to observe the AI ​​model's response to the anomaly and evaluate its consistency with patterns in historical data; Accuracy Assessment Module: This module comprehensively analyzes the stability of the AI ​​model on different sample combinations, its ability to respond to anomalies, and the consistency of patterns in historical data to assess the accuracy of AI in identifying error patterns in historical data. Model Optimization Module: For accurate identification, the AI ​​model is continuously optimized through real-time feedback data and dynamic learning mechanisms to ensure that the AI ​​model accurately classifies normal and abnormal patterns; Equipment control module: For inaccurate identification, the dynamic time warping algorithm is used to predict whether the equipment operating parameters deviate significantly from the normal range. When it is detected that the equipment operating parameters set by AI deviate from the normal range, an alarm is automatically triggered and the production process is suspended.

Images