Intelligent early warning method for colorless tea drink production based on machine learning

Abstract

The invention discloses an intelligent early warning method for colorless tea drink production based on machine learning. The method comprises the following steps: acquiring data of colorless tea drink production; calculating the similarity between the production data value corresponding to the colorless tea drink production data and all production data values in the historical safety production data corresponding to the production data; establishing an intelligent early warning model of colorless tea drink production by using the obtained historical safety production data, and verifying the intelligent early warning model; an intelligent early warning model is adopted to carry out online detection on colorless tea drink production; through principal component analysis, partial least squares regression and support vector machine classifier machine learning technologies, the influence of single production parameter change on production safety and the potential safety hazard probability in the whole production process can be predicted more accurately; according to the method, the limitation of dependence on manual monitoring and experience judgment traditionally is overcome, potential risks are recognized in advance through data analysis and model prediction, and the safety of the production process is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of food safety, and in particular to an intelligent early warning method for colorless tea beverage production based on machine learning. Background Art

[0002] In the preparation process of colorless tea, multiple steps are required, including raw material processing, extraction, filtration, blending, etc. These processes not only involve complex chemical reactions, but also require precise control of parameters such as temperature, pressure, and time. Mistakes in any link may lead to a decline in product quality or even a safety accident. Traditional safety management methods mostly rely on manual monitoring and experience judgment, and lack effective prediction methods, which to a certain extent limits production efficiency and safety. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent early warning method for colorless tea production based on machine learning, combining advanced machine learning technology and intelligent data analysis methods to achieve early identification and early warning of potential risks in the production process of colorless tea, thereby significantly improving production safety and efficiency.

[0004] The purpose of the present invention can be achieved through the following technical solutions: The present application provides an intelligent early warning method for colorless tea production based on machine learning, comprising the following steps: Acquire the data of the colorless tea production; calculate the similarity between the production data value corresponding to the colorless tea production data and all the production data values ​​in the historical safety production data corresponding to the production data; Using the acquired historical production safety data, an intelligent early warning model for the production of the colorless tea beverage is established, and the intelligent early warning model is verified; The intelligent early warning model includes a first model and a second model. The first model is used to predict the impact of changes in a single production data parameter on production safety; the second model is used to predict the probability of safety hazards in the overall production process. The intelligent early warning model is used to perform online detection on the production of the colorless tea beverage.

[0005] Furthermore, before obtaining the historical production safety data of the colorless tea beverage production, it also includes: collecting the historical production safety data of the colorless tea beverage production, and also includes collecting real-time monitoring data generated during the operation of all equipment for colorless tea beverage production.

[0006] Further, calculating the similarity between the production data value corresponding to the colorless tea production data and all the production data values ​​in the historical safety production data corresponding to the production data specifically includes: Extract all production data values ​​corresponding to the production data parameter from the historical safety production data; The cosine distance between the production data value corresponding to the production data and all production data values ​​corresponding to the production data is calculated.

[0007] Furthermore, the acquired historical production safety data is input into a pre-set first algorithm model to obtain the first model; the acquired historical production safety data is input into a pre-set second algorithm model to obtain the second model; wherein, the first algorithm model includes principal component analysis PCA and partial least squares regression PLSR, and the second algorithm model includes support vector machine SVM classifier.

[0008] Furthermore, the first algorithm model uses principal component analysis to perform dimensionality reduction processing, and then applies partial least squares regression to establish a first model to predict the impact of a single parameter change on production safety. The specific principal component analysis method is expressed as: ,in is the ith principal component, is the eigenvector, is the original feature; the specific partial least squares regression method is expressed as: , where Y is the response variable, X is the input matrix after PCA processing, B is the regression coefficient, and F is the residual.

[0009] The second algorithm model uses a support vector machine classifier to establish a second model, and the support vector machine classifier is expressed as: ,in, is the Lagrange multiplier, is the category label, is the kernel function and b is the intercept.

[0010] Furthermore, a support vector machine (SVM) classifier is used to build a model to predict the probability of safety hazards in the overall production process. The output value of the SVM decision function is mapped to probability through Platt scaling, including: Use the training data to train the SVM model, and obtain the output value f(x) of the SVM decision function based on the training instance; Use the output f(x) of SVM as the input feature and the actual category label as the target variable to train the logistic regression model; the logistic regression model is expressed as: ,in, is the probability that a sample belongs to the positive class (usually labeled 1) given the feature vector x, A and B are parameters that control the shape and position of the sigmoid curve. It is the decision function output of SVM, which calculates the distance between sample x and the decision boundary.

[0011] f(x) is obtained through the SVM model and then input into the logistic regression model to obtain the probability estimate.

[0012] Further, according to the established first model and second model, the first model is trained using historical production safety data, and the second model is trained using a genetic algorithm to find the best model parameter combination; Then, based on the real-time monitoring data collected during the production process, the data is input into the trained first model and the second model for real-time analysis. If the output probability exceeds the preset threshold, the early warning mechanism is triggered and the production personnel are notified in time.

[0013] Furthermore, the intelligent early warning model is verified, specifically including: using data independent of the training set to test whether the accuracy, precision, recall rate and F1 value of the first model and the second model meet the preset conditions; if not, returning to rebuild the model.

[0014] Furthermore, the early warning mechanism includes: When the first model predicts that the impact of a change in a single production data parameter on production safety is greater than or equal to a first threshold, a low-level alarm is triggered to indicate that the production data parameter is abnormal; When the probability of potential safety hazards in the overall production process predicted by the second model is greater than or equal to a second threshold, a high-level alarm is triggered and production is stopped.

[0015] Furthermore, based on the prediction results of the intelligent early warning model, the error between each actual production accident and the prediction results of the intelligent early warning model is recorded; and the first model or the second model is adjusted based on the results of multiple error statistical analyses.

[0016] The beneficial effects of the present invention are: (1) By adopting principal component analysis (PCA), partial least squares regression (PLSR) and support vector machine (SVM) classifier machine learning technology, the present invention can more accurately predict the impact of changes in a single production parameter on production safety and the probability of safety hazards in the overall production process. This method overcomes the limitations of traditional reliance on manual monitoring and empirical judgment, and through data analysis and model prediction, it can identify potential risks in advance, thereby significantly improving the safety of the production process; (2) By using a combination of weighted cosine distance and dynamic time warping (DTW) to analyze the similarity between production data and historical safety production data, the present invention can accurately identify potential risks in the production process. Compared with traditional empirical judgment, this method can foresee abnormal situations earlier, thereby achieving early warning. This early warning gives operators more time to deal with possible safety hazards, thereby avoiding accidents and significantly improving the safety of the production process. (3) The present invention also plays a role in the early warning stage. It can achieve continuous improvement of the model by recording and analyzing the errors between actual accidents and predicted results. This feedback-based iterative optimization process enables the early warning process to be continuously adjusted and improved over time, ensuring that it can maintain a high degree of adaptability and effectiveness under different production stages and conditions. In this way, the level of production safety management can be continuously improved, accidents can be reduced, and the safety and health of production personnel can be guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0018] Figure 1 A schematic diagram of a process flow of an intelligent early warning method for colorless tea production based on machine learning provided in Example 1 of the present application; Figure 2 A schematic diagram of a flow chart of calculating the similarity between all production data values ​​of an intelligent early warning method for colorless tea beverage production based on machine learning provided in Example 1 of the present application; Figure 3 A schematic diagram of a process flow for predicting the probability of potential safety hazards in the overall production process of an intelligent early warning method for colorless tea production based on machine learning provided in Example 1 of the present application; Figure 4 A flow chart of the early warning mechanism of an intelligent early warning method for colorless tea production based on machine learning provided in Example 1 of the present application. DETAILED DESCRIPTION

[0019] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0020] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0021] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0022] Example 1 See also Figure 1-Figure 4 This embodiment provides an intelligent early warning method for colorless tea production based on machine learning, which combines advanced machine learning technology and intelligent data analysis methods to achieve early identification and early warning of potential risks in the production process of colorless tea, significantly improving production safety and efficiency.

[0023] The present invention provides an intelligent early warning method for colorless tea production based on machine learning, comprising the following steps: S1. Obtain the data of the colorless tea production; calculate the similarity between the production data value corresponding to the colorless tea production data and all the production data values ​​in the historical safety production data corresponding to the production data; Before obtaining the historical production safety data of the colorless tea beverage production, it also includes: collecting the historical production safety data of the colorless tea beverage production, and also includes collecting real-time monitoring data generated during the operation of all equipment for colorless tea beverage production.

[0024] The historical production safety data includes at least one of the following: production data of each process of the colorless tea production, environmental factors, equipment operation conditions, personnel operation conditions, material usage conditions, product test results, equipment failure records, accident records, hidden danger investigation records, emergency plan drill records, and emergency rescue records. The real-time monitoring data includes at least one of the following: production status information, production equipment information, production environment information, production quality information, and production safety information.

[0025] Specifically, the historical production safety data of colorless tea beverage production and the real-time monitoring data generated during the operation of all equipment for colorless tea beverage production are cleaned to remove outliers and missing values; the data is then analyzed to select features that are closely related to production safety and remove redundant features.

[0026] Specifically, the features closely related to production safety are selected, including: the change trend of physical parameters such as the operating status of the equipment, temperature, pressure, and the periodic characteristics in the time series. Each parameter in the production data has a different impact on production safety. Therefore, the weight is set based on the importance of each parameter to production safety. The similarity is calculated by weighted cosine distance, and the contribution of different parameters to the final similarity is adjusted. The weighted cosine distance calculation is specifically expressed as: ,in, is the weight of the ith parameter, and are the i-th parameter values ​​in production data and historical data, respectively; Then, dynamic time warping (DTW) is used to calculate the similarity between time series, which is specifically expressed as: , where X and Y are two time series, Represents the distance between two sequences at time i and j. By combining the weighted cosine distance and dynamic time warping (DTW) method, it is possible to more accurately evaluate the similarity of key parameters such as equipment operating status, temperature, and pressure in the production process of colorless tea beverages, and timely identify abnormal trends similar to historical safety accidents, thereby effectively improving production safety and the accuracy of early warning.

[0027] The specific selection of features that are closely related to production safety includes: When analyzing the production data of colorless tea drinks to select features that are closely related to production safety and remove redundant features, statistical analysis will be performed first to calculate the correlation coefficient between each feature and the production safety index, and identify highly correlated features. Next, machine learning models, such as random forests, are used to evaluate the importance of features and exclude those that contribute the least to the model. At the same time, principal component analysis (PCA) is used to reduce data dimensions and eliminate features with small variance and low information content. In addition, cluster analysis is performed to identify features that provide similar information, and the knowledge of domain experts is combined to ensure that the selected features are of practical significance in actual production.

[0028] Specifically, by obtaining historical safety data and real-time monitoring data of colorless tea production, and performing data cleaning and feature selection, the problems of data clutter, outliers, and excessive redundant information in the production process were effectively solved. Through machine learning methods such as statistical analysis, random forests, and principal component analysis (PCA), key features closely related to production safety were selected, thereby improving the accuracy and efficiency of the prediction model. This method can effectively warn of potential safety hazards, ensure a safe and smooth production process, reduce accidents, and improve production efficiency and product quality.

[0029] Further, calculating the similarity between the production data value corresponding to the colorless tea production data and all the production data values ​​in the historical safety production data corresponding to the production data specifically includes: S11. extracting all production data values ​​corresponding to the production data parameter from the historical production safety data; S12. Calculate the cosine distance between the production data value corresponding to the production data and all production data values ​​corresponding to the production data.

[0030] Specifically, the cosine distance is used as the similarity between the production data value corresponding to the production data and all the production data values ​​in the historical safety production data corresponding to the production data parameter; by calculating the similarity between the colorless tea production data and the historical safety production data (using the cosine distance), it is possible to effectively evaluate whether the current production status is consistent with the historical safety production data. This similarity calculation helps to quickly identify abnormal situations, especially when the production process deviates from the past safety mode. This method can detect potential safety hazards at an early stage. S2. Using the acquired historical production safety data, establish an intelligent early warning model for the production of the colorless tea beverage, and verify the intelligent early warning model; Among them, the intelligent early warning model includes a first model and a second model. The first model is used to predict the impact of changes in a single production data parameter on production safety; the second model is used to predict the probability of safety hazards in the overall production process.

[0031] Furthermore, the acquired historical production safety data is input into a pre-set first algorithm model to obtain the first model; the acquired historical production safety data is input into a pre-set second algorithm model to obtain the second model; wherein, the first algorithm model includes principal component analysis PCA and partial least squares regression PLSR, and the second algorithm model includes support vector machine SVM classifier.

[0032] Specifically, the first algorithm model uses principal component analysis to perform dimensionality reduction processing, and then applies partial least squares regression to establish a first model to predict the impact of a single parameter change on production safety. The specific principal component analysis method is expressed as: ,in is the ith principal component, is the eigenvector, is the original feature; the specific partial least squares regression method is expressed as: , where Y is the response variable, X is the input matrix after PCA processing, B is the regression coefficient, and F is the residual.

[0033] Specifically, the second algorithm model uses a support vector machine (SVM) classifier to establish a model, and the specific support vector machine classifier is expressed as: ,in, is the Lagrange multiplier, is the category label, is the kernel function and b is the intercept.

[0034] By inputting historical production safety data into the first algorithm model (PCA and PLSR) and the second algorithm model (SVM classifier), the system can effectively analyze and predict potential safety hazards that may occur during the production process. The first model uses principal component analysis (PCA) dimensionality reduction processing and partial least squares regression (PLSR) method to help identify the impact of changes in individual production parameters on safety, ensuring that potential risks are discovered at an early stage. The second model predicts the probability of safety hazards in the overall production process through the support vector machine (SVM) classifier, combined with Platt scaling conversion to probability output, further improving the accuracy of safety hazard prediction. This multi-model provides a more comprehensive and accurate intelligent early warning mechanism for production safety.

[0035] Furthermore, a support vector machine (SVM) classifier is used to build a model to predict the probability of safety hazards in the overall production process. The output value of the SVM decision function is mapped to probability through Platt scaling, including: S21, training the SVM model using the training data, and obtaining the output value f(x) of the decision function of the SVM according to the training instance; S22. Use the output f(x) of SVM as the input feature and the actual category label as the target variable to train a logistic regression model; wherein the logistic regression model is expressed as: ,in, is the probability that a sample belongs to the positive class (usually labeled 1) given the feature vector x, A and B are parameters that control the shape and position of the sigmoid curve. It is the decision function output of SVM, which calculates the distance between sample x and the decision boundary.

[0036] S23. Obtain f(x) through the SVM model, and then input it into the logistic regression model to obtain a probability estimate.

[0037] Specifically, by combining the support vector machine (SVM) classifier with Platt scaling, the probability of safety hazards in the overall production process can be predicted more accurately. The SVM model first classifies the training data and calculates the output value f(x) of the decision function, but the SVM originally outputs the classification result rather than the probability. By using these output values ​​as input features and training a logistic regression model, the SVM decision results are mapped to the probability, and finally a more intuitive safety hazard probability estimate is obtained. This method greatly improves the prediction accuracy of the early warning system for potential risks, can identify and quantify possible safety hazards in the production process in advance, and provide a basis for further risk management.

[0038] Further, according to the established first model and second model, the first model is trained using historical production safety data, and the second model is trained using a genetic algorithm to find the best model parameter combination; Then, based on the real-time monitoring data collected during the production process, the data is input into the trained first model and the second model for real-time analysis. If the output probability exceeds the preset threshold, the early warning mechanism is triggered and the production personnel are notified in time.

[0039] Furthermore, the intelligent early warning model is verified, specifically including: using data independent of the training set to test whether the accuracy, precision, recall rate and F1 value of the first model and the second model meet the preset conditions; if not, returning to rebuild the model.

[0040] Specifically, by using the established first model and the second model, using historical production safety data to train the first model, and using a genetic algorithm to optimize the parameter combination of the second model, the model can effectively predict safety hazards when real-time monitoring data input is input. When the predicted probability exceeds a preset threshold, the early warning mechanism is automatically triggered to notify production personnel in a timely manner, thereby avoiding potential safety accidents; In order to ensure the accuracy and reliability of the intelligent early warning model, the model is further verified using data independent of the training set. By testing the accuracy, precision, recall rate and F1 value of the first model and the second model, the performance of the model in different aspects can be comprehensively evaluated. If these indicators do not meet the preset conditions, the system will automatically return and rebuild the model to optimize model performance. This closed-loop verification mechanism ensures the efficiency and stability of the early warning model, which not only reduces the false alarm rate, but also improves the early warning capability of safety hazards and ensures the continuous safety of the production process.

[0041] S3. Use the intelligent early warning model to perform online detection on the production of the colorless tea beverage.

[0042] Furthermore, the early warning mechanism includes: S31. When the first model predicts that the impact of a change in a single production data parameter on production safety is greater than or equal to a first threshold, a low-level alarm is triggered to indicate that the production data parameter is abnormal; S32: When the probability of potential safety hazards in the overall production process predicted by the second model is greater than or equal to a second threshold, a high-level alarm is triggered and production is stopped.

[0043] Specifically, when only the output of the first model exceeds the first threshold, it can be judged as a local abnormality, and the system issues a low-level warning. This warning will not directly lead to production suspension or emergency response, but will prompt the operator to pay attention to the parameter and conduct an inspection; If multiple parameters of the first model exceed the first threshold at the same time, and there is a certain correlation between these parameters (such as temperature increase leading to pressure increase), an intermediate warning is issued. At this time, the operator should take certain measures to avoid the risk from expanding; When the overall process evaluation result of the second model exceeds the second threshold, the system will issue an advanced warning even if the individual parameters are not obviously abnormal. This is because the overall trend indicates that the system may be in a dangerous state and requires immediate action (such as stopping production, system inspection, etc.).

[0044] The first threshold and the second threshold are different and are key risk assessment parameters that determine when to take action to ensure production safety. The first threshold is for changes in individual production data parameters. When the impact of these changes on production safety reaches or exceeds the first threshold, the system will issue a specific alarm and prompt the operator to check and handle abnormal conditions of the relevant parameters; the second threshold is used to assess the probability of safety hazards in the overall production process. When the predicted probability reaches or exceeds this threshold, the system will trigger a comprehensive alarm and automatically stop production to prevent possible safety accidents; the setting of these two thresholds needs to be based on in-depth data analysis, historical event records and expert experience to ensure that they can effectively warn of potential risks without causing unnecessary production interruptions due to excessive sensitivity.

[0045] Furthermore, based on the prediction results of the intelligent early warning model, the error between each actual production accident and the prediction results of the intelligent early warning model is recorded; and the first model or the second model is adjusted based on the results of multiple error statistical analyses.

[0046] The core of the intelligent early warning method of the present invention is to predict the potential safety hazards in the production of colorless tea drinks through a machine learning model based on historical production safety data and real-time monitoring data, and to provide timely early warnings. The specific implementation process is as follows: First, the historical safety data and real-time monitoring data of colorless tea production were obtained, and outliers and redundant features were removed through data cleaning. Then, statistical analysis and machine learning methods (such as random forests, principal component analysis PCA, etc.) were used to select features closely related to production safety, and two models were built based on these features. Among them, the first model (using PCA and partial least squares regression) is used to predict the impact of changes in a single production data parameter on production safety; the second model (based on the SVM classifier) ​​predicts the probability of safety hazards in the entire production process.

[0047] In the real-time monitoring process, real-time data is input into the trained model for analysis. If the predicted risk probability or parameter impact exceeds the preset threshold, the system will issue an alarm to prompt production personnel to conduct inspections or stop production. These thresholds are set based on historical events and expert experience to ensure timely warnings without excessive disruption to production. At the same time, the accuracy and performance of the model can be continuously optimized through error analysis between the predicted results and actual accidents.

[0048] Example 2 This implementation provides an intelligent early warning method for colorless tea production based on machine learning, which is applied to the production process of colorless tea. The intelligent early warning not only improves the safety of colorless tea production, but also improves production efficiency and product quality, ensuring the continuity and stability of the production process.

[0049] Specific process: First, production-related information is collected from multiple data sources, including historical safety records and real-time equipment monitoring data. These production data cover all aspects of the production process, such as process data, environmental conditions, equipment status, personnel operations, material usage, product quality test results, equipment failures, accident records, hidden danger inspections, emergency plan drills, and emergency rescue records.

[0050] The obtained production data is thoroughly cleaned to eliminate outliers and missing values ​​to ensure the quality and accuracy of the data. Then, through statistical analysis and machine learning techniques, such as random forest and principal component analysis (PCA), key features closely related to production safety are extracted from the data, and two intelligent early warning models are constructed. The first model is used to evaluate the potential impact of changes in a single production parameter on safety, while the second model predicts possible safety hazards in the overall production process.

[0051] During the model training phase, genetic algorithms are used to optimize the parameters of the second model to improve its prediction accuracy. After training, real-time monitoring data is input into the two models for continuous safety assessment. When the probability of potential safety hazards predicted by the model exceeds the preset first threshold, or the impact of a single parameter change exceeds the second threshold, the early warning mechanism will be triggered immediately, automatically stopping production and notifying production personnel so that timely measures can be taken to prevent potential safety accidents. In addition, the error between each early warning and the actual accident will be recorded, and these data will be used to continuously adjust and optimize the model to ensure the long-term effectiveness and reliability of the early warning system.

[0052] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent early warning method for colorless tea production based on machine learning, characterized in that: The method comprises the following steps: obtaining the production data of the colorless tea beverage; Calculate the similarity between the production data value corresponding to the colorless tea production data and all the production data values ​​in the historical safety production data corresponding to the production data; Using the acquired historical production safety data, an intelligent early warning model for the production of the colorless tea beverage is established, and the intelligent early warning model is verified; The intelligent early warning model includes a first model and a second model. The first model is used to predict the impact of changes in a single production data parameter on production safety; the second model is used to predict the probability of safety hazards in the overall production process. The intelligent early warning model is used to perform online detection on the production of the colorless tea beverage.

2. The intelligent early warning method for colorless tea production based on machine learning according to claim 1, characterized in that: Before obtaining the historical production safety data of the colorless tea beverage production, it also includes: collecting the historical production safety data of the colorless tea beverage production, and also includes collecting real-time monitoring data generated during the operation of all equipment for colorless tea beverage production.

3. The intelligent early warning method for colorless tea production based on machine learning according to claim 1, characterized in that: Calculate the similarity between the production data value corresponding to the colorless tea production data and all the production data values ​​in the historical safety production data corresponding to the production data, specifically including: Extract all production data values ​​corresponding to the production data parameter from the historical safety production data; The cosine distance between the production data value corresponding to the production data and all production data values ​​corresponding to the production data is calculated.

4. The intelligent early warning method for colorless tea production based on machine learning according to claim 2, characterized in that: The acquired historical production safety data is input into a pre-set first algorithm model to obtain the first model; the acquired historical production safety data is input into a pre-set second algorithm model to obtain the second model; wherein, the first algorithm model includes principal component analysis PCA and partial least squares regression PLSR, and the second algorithm model includes support vector machine SVM classifier.

5. The intelligent early warning method for colorless tea production based on machine learning according to claim 4, characterized in that: The first algorithm model uses principal component analysis to perform dimensionality reduction processing, and then applies partial least squares regression to establish the first model to predict the impact of a single parameter change on production safety. The specific principal component analysis method is expressed as: ,in is the ith principal component, is the eigenvector, is the original feature; the specific partial least squares regression method is expressed as: , where Y is the response variable, X is the input matrix after PCA processing, B is the regression coefficient, and F is the residual; The second algorithm model uses a support vector machine classifier to establish a second model, and the support vector machine classifier is expressed as: ,in, is the Lagrange multiplier, is the category label, is the kernel function and b is the intercept.

6. The intelligent early warning method for colorless tea production based on machine learning according to claim 5, characterized in that: The second model is established using a support vector machine classifier to predict the probability of safety hazards in the overall production process. The output value of the SVM decision function is mapped to probability through Platt scaling, including: Use the training data to train the SVM model, and obtain the output value f(x) of the SVM decision function based on the training instance; Use the output f(x) of SVM as the input feature and the actual category label as the target variable to train the logistic regression model; the logistic regression model is expressed as: ,in, is the probability that a sample belongs to the positive class (usually labeled 1) given the feature vector x, A and B are parameters that control the shape and position of the sigmoid curve. It is the decision function output of SVM, which calculates the distance between sample x and the decision boundary; f(x) is obtained through the SVM model and then input into the logistic regression model to obtain the probability estimate.

7. The intelligent early warning method for colorless tea production based on machine learning according to claim 6, characterized in that: According to the established first model and second model, the first model is trained using historical production safety data, and the second model is trained using a genetic algorithm to find the best model parameter combination; Then, based on the real-time monitoring data collected during the production process, the data is input into the trained first model and the second model for real-time analysis. If the output probability exceeds the preset threshold, the early warning mechanism is triggered and the production personnel are notified in time.

8. The intelligent early warning method for colorless tea production based on machine learning according to claim 1, characterized in that: Verifying the intelligent early warning model specifically includes: using data independent of the training set to test whether the accuracy, precision, recall rate and F1 value of the first model and the second model meet the preset conditions; if not, returning to rebuild the model.

9. The intelligent early warning method for colorless tea production based on machine learning according to claim 8, characterized in that: The early warning mechanism includes: When the first model predicts that the impact of a change in a single production data parameter on production safety is greater than or equal to a first threshold, a low-level alarm is triggered to indicate that the production data parameter is abnormal; When the probability of potential safety hazards in the overall production process predicted by the second model is greater than or equal to a second threshold, a high-level alarm is triggered and production is stopped.

10. The intelligent early warning method for colorless tea production based on machine learning according to claim 1, characterized in that: According to the prediction results of the intelligent early warning model, record the error between each actual production accident and the prediction results of the intelligent early warning model; The first model or the second model is adjusted based on the results of multiple error statistical analyses.

Images