New material production whole process monitoring method, system, equipment and medium

Abstract

The invention discloses a new material production whole process monitoring method, system and device and a medium. The method comprises the following steps: collecting monitoring data of a plurality of process links in a new material production process and preprocessing the monitoring data; performing correlation analysis on the monitoring data of different process links, identifying a coupling relationship between the previous and later process links, and establishing a correlation model between the process stages; fusing the monitoring data of the plurality of process links to generate a dynamic key monitoring index under the current production condition; performing anomaly detection on the real-time monitoring data based on the association model and the dynamic key monitoring index, and judging whether abnormal data deviating from an expected range exists in the production process or not; and pushing the detected abnormity or potential risk to an operator, and / or generating a parameter optimization and adjustment suggestion of the production process, and sending the parameter optimization and adjustment suggestion to a production control system and the operator. According to the monitoring method provided by the invention, the monitoring precision can be enhanced, and the fine management level of the production process is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of production process data processing, and in particular relates to a new material production full-process monitoring method, system, equipment and medium. Background Art

[0002] With the rapid development of industrial technology, the research and development and production of new materials have occupied an important position in the manufacturing industry. However, the production process of new materials usually involves complex process flows, including raw material processing, equipment operation, environmental control, and multi-stage production processes. There is a high degree of coupling and dynamics between these links, and a slight deviation may lead to a decline in product performance or even production failure. Therefore, higher requirements are placed on the monitoring and management of the entire process of new material production.

[0003] Existing production monitoring methods often focus on a single link or static monitoring, which makes it difficult to achieve global and dynamic analysis of the entire production process. For example, traditional monitoring systems are usually based on preset parameter thresholds or simple statistical analysis, which cannot effectively capture the complex dependencies between variables in the production process, and it is difficult to detect potential risks in a timely manner and provide optimization suggestions. In addition, data synchronization, fusion and key indicator extraction of various process links under dynamic conditions are still facing challenges, which limits the further development of intelligent production.

[0004] The Chinese invention patent application with publication number CN118350619A discloses an intelligent full-process production management method and its management system, which relates to the technical field of intelligent full-process production management, including: obtaining a modified feature set based on the initial feature set of the output product of each process at the current moment; obtaining the quality coefficient of the output product of each process at the current moment based on the modified feature set; obtaining the quality index of each process at the current moment based on the basic information of each process at the current moment and the quality coefficient of the output product; obtaining the energy consumption index and output index of each process at the current moment based on the basic information of each process at the current moment, and constructing the process analysis matrix at the current moment based on the energy consumption index, output index and quality index of each process at the current moment; obtaining the process management correction result of each process in the whole production process based on the process analysis matrix.

[0005] The process analysis matrix constructed by this solution is based on process information at fixed moments, which makes it difficult to adapt to real-time dynamic changes in the production process. It does not consider the dynamic coupling relationship between the previous and subsequent processes and its impact on the entire process, and is unable to identify and handle anomalies in the production process in a timely and accurate manner. Summary of the invention

[0006] The present invention provides a method, system, equipment and medium for monitoring the entire process of new material production, aiming to solve the problem that the existing technical solutions are unable to timely and accurately identify and process abnormal data when monitoring the entire process of new material production.

[0007] In order to solve the above technical problems, the first aspect of the present invention provides a new material production full process monitoring method, comprising the following steps: Collect monitoring data of multiple process links in the new material production process, including raw material data, equipment operation data, environmental data and production process data, and pre-process the monitoring data; Conduct correlation analysis on monitoring data of different process links, identify the coupling relationship between previous and subsequent process links, and establish a correlation model between process stages; Integrate monitoring data from multiple process links to generate dynamic key monitoring indicators under current production conditions; Based on the correlation model and dynamic key monitoring indicators, perform anomaly detection on real-time monitoring data to determine whether there is abnormal data that deviates from the expected range in the production process; The detected anomalies or potential risks are pushed to the operators, and / or the parameter optimization adjustment suggestions for the production process are generated and sent to the production control system and operators.

[0008] Preferably, the method of association analysis is specifically: Normalize data of different dimensions and synchronize data of different links according to timestamps; Selecting variables that affect the process and extracting statistical characteristics of the variables; The nonlinear correlation and time correlation between the statistical features are calculated, and the comprehensive correlation is calculated according to the set weight coefficient.

[0009] Preferably, the calculation method of the comprehensive correlation is:

[0010] In the formula, Represents statistical characteristics The comprehensive correlation of is the set weight coefficient, is the normalized statistical feature Nonlinear correlation, For statistical characteristics The time correlation of express The absolute maximum value of and The calculation method is as follows:

[0011]

[0012]

[0013] In the formula, is the nonlinear correlation before normalization, Respectively and The entropy of express At the same time, the value is The joint probability distribution of Respectively The marginal probability distribution of Respectively indicate time time The observed value of Respectively The average value of is the time lag, used to represent relatively The offset of is the total number of time series data points.

[0014] Preferably, the association model is a graph-based Bayesian network association model, and a directed acyclic graph is established based on the comprehensive correlation data and the time lag data in the time correlation.

[0015] Preferably, the method for generating the dynamic key monitoring indicators is: Performing time alignment and space alignment on the monitoring data, wherein the time alignment synchronizes the monitoring data in the time dimension by interpolation or resampling, and the space alignment normalizes the monitoring data from different sources; Use sliding windows to divide the aligned monitoring data into time series. For each time point, extract the data of several past time points as features. The divided monitoring data are integrated and characteristic components are extracted through principal component analysis or linear discriminant analysis; The extreme gradient boosting tree is used to obtain the feature importance scores of the feature components, and the feature components with the greatest influence are identified as dynamic key monitoring indicators.

[0016] Preferably, the anomaly detection method is specifically: Calculate the conditional probability of each parameter in the dynamic key monitoring indicator under the current conditions through the Bayesian network, and mark the nodes whose conditional probability is lower than the first set value; For each parameter in the dynamic key monitoring indicator, a standard score is calculated, and nodes whose standard scores are greater than a second set value are marked; Calculate the distance between various parameters in the dynamic key monitoring indicators, and mark the nodes whose distance is greater than the third set value; Traverse all nodes and issue alarms and / or display nodes that meet the preset marking rules.

[0017] Preferably, the distance is calculated as follows:

[0018] In the formula, Represents the vector of various parameters in the dynamic key monitoring indicators at the current moment, is the mean vector of each monitoring indicator under the historical normal state, represents the corresponding covariance matrix.

[0019] In a second aspect of the present invention, a new material production full-process monitoring system is further provided, wherein the system is used to implement the method described in the first aspect of the present invention, comprising: The data acquisition module is used to collect monitoring data of multiple process links in the production process of new materials, including raw material data, equipment operation data, environmental data and production process data; The data preprocessing module is used to preprocess the collected monitoring data, including data normalization, time synchronization and space alignment; The correlation analysis module is used to perform correlation analysis on the monitoring data of different process links, identify the coupling relationship between the previous and next process links, and establish a correlation model between process stages based on the comprehensive correlation calculation method and time correlation; Dynamic key monitoring indicator generation module, which is used to integrate the monitoring data of multiple process links to generate dynamic key monitoring indicators under current production conditions; The anomaly detection module is used to detect anomalies in real-time monitoring data based on the association model and dynamic key monitoring indicators, calculate conditional probabilities, standard scores, and parameter distances, and issue alarms or display nodes that meet the preset marking rules; The optimization suggestion module is used to generate parameter optimization adjustment suggestions for the production process based on the abnormality detection results, and push the suggestions to the production control system and operators; The user interaction module is used to display the abnormal detection results, dynamic key monitoring indicators and optimization adjustment suggestions during the production monitoring process, and receive feedback from operators.

[0020] In the third aspect of the present invention, an electronic device is also proposed, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein when the processor executes the computer program, the full-process monitoring method as described in the first aspect of the present invention is implemented.

[0021] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the full-process monitoring method as described in the first aspect of the present invention can be implemented.

[0022] Compared with the prior art, the present invention has the following technical effects: 1. The full-process monitoring method proposed in the present invention uniformly collects, preprocesses and integrates and analyzes raw material data, equipment operation data, environmental data and production process data, eliminating the inconsistency and isolation between different data sources, thereby enhancing monitoring accuracy, improving the level of refined management of the production process, automatically analyzing the operating conditions and correlations of process links, reducing manual intervention, and improving production efficiency. It provides an efficient, accurate and sustainable full-process monitoring and optimization solution for the new material production process, effectively improving production efficiency, ensuring product quality, reducing production energy consumption and costs, and enhancing the intelligence and adaptability of the system.

[0023] 2. The full-process monitoring method proposed in the present invention generates dynamic key monitoring indicators under current production conditions, and subsequently uses the dynamic key monitoring indicators for calculations, which can effectively reduce the calculation complexity, reduce the demand for computing resources, improve calculation efficiency, and thereby improve the overall full-process monitoring efficiency.

[0024] 3. The full-process monitoring method proposed in the present invention identifies the potential relationship between the previous and next process links through the calculation of nonlinear correlation and time correlation, and constructs a correlation model between process stages. The dynamic coupling relationship between the previous and next processes and its impact on the whole process are considered, so that the abnormal identification results are more timely and accurate.

[0025] 4. The generation method of dynamic monitoring indicators in the full-process monitoring method proposed in the present invention can be updated in real time as production conditions change, ensuring that the monitored indicators have the highest sensitivity to the current production status.

[0026] 5. The full-process monitoring method proposed in the present invention combines methods such as Bayesian network conditional probability, standard score and parameter distance to identify anomalies and potential risks from multiple dimensions, thereby enhancing the comprehensiveness and accuracy of anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of the full-process monitoring method of the present invention. DETAILED DESCRIPTION

[0028] In order to make the objectives, technical solutions and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in combination with specific embodiments of the present application and with reference to the accompanying drawings.

[0029] Embodiment 1 This embodiment is a new material production process monitoring method. Figure 1 As shown, the steps include steps one to five: Step 1: Collect monitoring data of multiple process links in the new material production process, including raw material data, equipment operation data, environmental data and production process data, and pre-process the monitoring data to eliminate noise and redundant information.

[0030] The data collection of this embodiment is achieved by deploying multiple types of sensors at necessary links of the production line, such as temperature, pressure, humidity, chemical composition analyzer and other sensors. The collected raw material data includes composition, purity, particle size, humidity, etc.; the equipment operation data includes temperature, pressure, speed, current, etc.; the environmental data includes ambient temperature, humidity, gas concentration, temperature difference, etc.; the production process data includes reaction time, energy consumption, waste generation, etc.

[0031] In some other embodiments of the present invention, it also includes building an Internet of Things architecture to connect the collection equipment to a unified data platform; and configuring edge computing equipment for real-time data preprocessing and transmission, completing data denoising, completion and formatting at the edge computing node to reduce data transmission delays.

[0032] Step 2: Perform correlation analysis on the monitoring data of different process links, identify the coupling relationship between the previous and next process links, and establish a correlation model between process stages. The correlation model is a graph-based Bayesian network correlation model, which establishes a directed acyclic graph (DAG) based on the comprehensive correlation data and the time lag data in the time correlation.

[0033] In this embodiment, the association analysis method is specifically as follows: S21: Normalize data of different dimensions (such as temperature, pressure, and humidity), and synchronize data of different links according to timestamps to ensure the time consistency of data.

[0034] The specific normalization can be achieved through the following formula:

[0035] In the formula, is the normalized data, is the original data before normalization, is the minimum value of all data in this dimension. It is the maximum value of all data in this dimension.

[0036] S22: Select variables that affect the process, and extract statistical features of the variables. In this embodiment, the extracted statistical features include mean, standard deviation, extreme value, rate of change, etc.

[0037] S23: Calculate the nonlinear correlation and time correlation between the statistical features, and calculate the comprehensive correlation according to the set weight coefficient.

[0038] Specifically, the calculation method of the comprehensive correlation is:

[0039] In the formula, Represents statistical characteristics (or variables) The comprehensive correlation of is the set weight coefficient, is the normalized statistical feature Nonlinear correlation, For statistical characteristics The time correlation of express The absolute maximum value of , which represent parameters of different dimensions, such as ambient temperature and reaction time, reactor temperature and pressure, raw material particle size and reaction time, etc.

[0040] Weight coefficient It is used to control the contribution of mutual information and cross-correlation in the comprehensive correlation. If you are more interested in static correlation, you can set a larger weight coefficient value (such as 0.7); if you are more interested in time dependence, set a smaller weight coefficient value (such as 0.3). Comprehensive Correlation The closer it is to 1, the and Through this comprehensive calculation, the static and dynamic correlations between variables can be evaluated more comprehensively, providing a basis for subsequent model construction.

[0041] and The calculation method is as follows:

[0042]

[0043]

[0044] In the formula, is the nonlinear correlation before normalization, Respectively and The entropy of express At the same time, the value is The joint probability distribution of Respectively The marginal probability distribution of Respectively indicate time time The observed value of Respectively The average value of is the time lag, used to represent relatively The offset of is the total number of time series data points.

[0045] like ,illustrate and Completely independent and without any connection; if ,illustrate and There is information correlation, and the larger the value, the stronger the correlation. , indicating two and Variables lagged in time If , indicating two and Variables lagged in time If , indicating two and Variables lagged in time There is no correlation under the condition of lag; by testing different lag values, the time dependence between variables can be found.

[0046] The graph-based Bayesian network association model is represented by a directed acyclic graph (DAG), in which nodes represent variables and edges represent conditional dependencies between variables. The initial structure of the directed acyclic graph is determined based on the comprehensive correlation data. Higher, it can be preliminarily considered and There may be direct conditional dependencies between them; such relationships can be used to connect the edges of the initial graph, for example or . Lagged Data in Time Correlation Can be used to determine the direction of causality, e.g. The hysteresis value significantly affects , then there may be For variable pairs with low comprehensive correlation, they can be excluded from the search space of structure learning to reduce computational complexity.

[0047] Specifically, when constructing a graph-based Bayesian network association model, an initial graph is generated based on a comprehensive correlation matrix, and the variable pairs with higher correlation are connected, and the direction is initially based on the time correlation result. Then a correlation threshold is set to remove low-correlation variable pairs and reduce unnecessary edges. A constraint-based method or a scoring method is used to optimize the initial graph to ensure that the generated graph is a directed acyclic graph; in this embodiment, the constraint-based method can use the PC (Peter-Clark) algorithm, and the scoring method can be used as the K2 algorithm. The conditional probability distribution is calculated using actual data through maximum likelihood estimation or Bayesian estimation.

[0048] Comprehensive correlation data is an important reference for establishing Bayesian network association models. It can effectively narrow the search space, optimize the structural learning process, improve the reliability and interpretability of the model, and significantly improve modeling efficiency and effectiveness.

[0049] Step 3: Integrate the monitoring data of multiple process links to generate dynamic key monitoring indicators under current production conditions.

[0050] In this embodiment, the method for generating the dynamic key monitoring indicators is: S31: Time-aligning and space-aligning the monitoring data. The time alignment synchronizes the monitoring data in the time dimension by interpolation or resampling, and the space alignment normalizes the monitoring data from different sources.

[0051] Among them, time alignment ensures the synchronization of all data sources in the time dimension. Interpolation or resampling is used to solve the problem of inconsistent data frequency. High-frequency data (such as equipment operation data) can be down-converted to keep its frequency consistent with low-frequency data (such as environmental data). The interpolation formula is as follows:

[0052] In the formula, The value of the moment is missing. Indicates the value to be inserted into the missing position. They are The previous time point and the next time point that are adjacent to each other, for A time point before the moment The value of for A time point after the moment The value of .

[0053] The normalization method for spatial alignment may use the same normalization algorithm as in step S21 .

[0054] S32: Use a sliding window to divide the aligned monitoring data into time series, and for each time point, extract data from several past time points as features.

[0055] S33: Fuse the divided monitoring data and extract feature components through principal component analysis (PCA) or linear discriminant analysis (LDA). Specifically, feature fusion can be achieved through PCA or clustering tools provided by the sklearn library of Python.

[0056] S34: Use extreme gradient boosting tree to obtain the feature importance score of the feature component, identify the feature component with the greatest influence, and use it as a dynamic key monitoring indicator. Specifically, it is implemented through Python's xgboost library or sklearn library, with specific components such as GradientBoostingClassifier and GradientBoostingRegressor. Configure initial parameters such as learning rate, maximum tree depth, subsampling rate, number of iterations, etc. After training, it can be used to obtain the importance score of the real-time feature component. A threshold can be set to select features with feature importance scores greater than the threshold as dynamic key monitoring indicators, or features with a certain proportion of the cumulative importance score can be selected as dynamic key monitoring indicators. Map the index of the selected feature component to the actual meaning (such as temperature, pressure, energy consumption, etc.) and store it as a dynamic key monitoring indicator.

[0057] Step 4: Based on the association model and dynamic key monitoring indicators, perform anomaly detection on the real-time monitoring data to determine whether there is abnormal data that deviates from the expected range in the production process.

[0058] The method of anomaly detection is specifically as follows: S41: Calculate the conditional probability of each parameter in the dynamic key monitoring indicator under the current conditions through the Bayesian network, and mark the nodes whose conditional probability is lower than the first set value.

[0059] S42: For each parameter in the dynamic key monitoring indicator, calculate the standard score, and mark the node whose standard score is greater than the second set value. The calculation method of the standard score is as follows:

[0060] in, For the current moment The variable value of Variables in normal state The historical average For variables The historical standard deviation is used to measure the fluctuation range of the variable. When it is greater than the second set value, the variable value can be considered abnormal and the variable is marked.

[0061] S43: Calculate the distance between various parameters in the dynamic key monitoring indicator, and mark the nodes whose distance is greater than the third set value.

[0062] In this step, the distance is calculated as follows:

[0063] In the formula, Represents the vector of various parameters in the dynamic key monitoring indicators at the current moment, is the mean vector of each monitoring indicator under the historical normal state, Represents the corresponding covariance matrix, which is used to measure the joint abnormality of multiple indicators.

[0064] S44: traverse all nodes, and issue an alarm and / or display for nodes that meet the preset marking rules.

[0065] The specific preset marking rules include the combination of the and relationship and the or relationship of the three markings described in the above steps S41-S43, which can be specifically set according to the implementation requirements. For example, if any two markings exist, the node is extracted for alarm and / or display, or only the nodes with three markings at the same time are extracted for the next step.

[0066] In addition, multi-level alarms can be set as needed: Level 1, single indicator exceeds the limit; Level 2, multiple indicators are abnormal; Level 3, key path node abnormalities in the association model. The indicators, association paths and impact range of the abnormal points are displayed in real time, and high-risk nodes in the association model and their upstream and downstream dependencies are dynamically displayed.

[0067] In some other embodiments of the present invention, the distance in step S43 is the Euclidean distance.

[0068] Step 5: Push the detected anomalies or potential risks to the operators, and / or generate parameter optimization adjustment suggestions for the production process and send them to the production control system and operators.

[0069] Parameter optimization adjustment suggestions can be generated based on preset process rules, such as setting the upper and lower limits, optimal value range and adjustment step of important parameters in the production process. After detecting deviations, optimization suggestions are generated according to the rules. It is also possible to combine domain expert knowledge, embed empirical adjustment strategies into the system, and build a set of rule bases. The rule base design includes trigger conditions and corresponding suggestions. When anomalies are detected, the corresponding adjustment rules are obtained from the rule base for output.

[0070] Embodiment 2 This embodiment is a new material production full-process monitoring system, which is used to implement the method described in the first embodiment, including: The data acquisition module is used to collect monitoring data of multiple process links in the new material production process, including raw material data, equipment operation data, environmental data and production process data.

[0071] The data preprocessing module is used to preprocess the collected monitoring data, including data normalization, time synchronization and space alignment.

[0072] The correlation analysis module is used to perform correlation analysis on the monitoring data of different process links, identify the coupling relationship between the previous and subsequent process links, and establish a correlation model between process stages based on the comprehensive correlation calculation method and time correlation.

[0073] The dynamic key monitoring indicator generation module is used to integrate the monitoring data of multiple process links to generate dynamic key monitoring indicators under current production conditions.

[0074] The anomaly detection module is used to detect anomalies in real-time monitoring data based on association models and dynamic key monitoring indicators, calculate conditional probabilities, standard scores, and parameter distances, and issue alarms or displays for nodes that meet preset marking rules.

[0075] The optimization suggestion module is used to generate parameter optimization adjustment suggestions for the production process based on the anomaly detection results, and push the suggestions to the production control system and operators.

[0076] The user interaction module is used to display the abnormal detection results, dynamic key monitoring indicators and optimization adjustment suggestions during the production monitoring process, and receive feedback from operators.

[0077] Embodiment 3 This embodiment is an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the full-process monitoring method described in Example 1 is implemented.

[0078] Embodiment 4 This embodiment is a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it can implement the full-process monitoring method described in Example 1.

[0079] The above is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A new material production full process monitoring method, characterized in that: The following steps are involved: Collect monitoring data of multiple process links in the new material production process, including raw material data, equipment operation data, environmental data and production process data, and pre-process the monitoring data; Conduct correlation analysis on monitoring data of different process links, identify the coupling relationship between previous and subsequent process links, and establish a correlation model between process stages; Integrate monitoring data from multiple process links to generate dynamic key monitoring indicators under current production conditions; Based on the correlation model and dynamic key monitoring indicators, perform anomaly detection on real-time monitoring data to determine whether there is abnormal data that deviates from the expected range in the production process; The detected anomalies or potential risks are pushed to the operators, and / or the parameter optimization adjustment suggestions for the production process are generated and sent to the production control system and operators.

2. A new material production full-process monitoring method according to claim 1, characterized in that: The method of the association analysis is specifically as follows: Normalize data of different dimensions and synchronize data of different links according to timestamps; Selecting variables that affect the process and extracting statistical characteristics of the variables; The nonlinear correlation and time correlation between the statistical features are calculated, and the comprehensive correlation is calculated according to the set weight coefficient.

3. A new material production full-process monitoring method according to claim 2, characterized in that: The calculation method of the comprehensive correlation is: In the formula, Represents statistical characteristics The comprehensive correlation of is the set weight coefficient, is the normalized statistical feature Nonlinear correlation, For statistical characteristics The time correlation of express The absolute maximum value of and The calculation method is as follows: In the formula, is the nonlinear correlation before normalization, Respectively and The entropy of express At the same time, the value is The joint probability distribution of Respectively The marginal probability distribution of Respectively indicate time time The observed value of Respectively The average value of is the time lag, used to represent relatively The offset of is the total number of time series data points.

4. A new material production full-process monitoring method according to claim 3, characterized in that: The association model is a graph-based Bayesian network association model, which establishes a directed acyclic graph based on comprehensive correlation data and time lag data in time correlation.

5. A new material production full-process monitoring method according to claim 1, characterized in that: The method for generating the dynamic key monitoring indicators is: Performing time alignment and space alignment on the monitoring data, wherein the time alignment synchronizes the monitoring data in the time dimension by interpolation or resampling, and the space alignment normalizes the monitoring data from different sources; Use sliding windows to divide the aligned monitoring data into time series. For each time point, extract the data of several past time points as features. The divided monitoring data are integrated and characteristic components are extracted through principal component analysis or linear discriminant analysis; The extreme gradient boosting tree is used to obtain the feature importance scores of the feature components, and the feature components with the greatest influence are identified as dynamic key monitoring indicators.

6. A new material production full-process monitoring method according to claim 1, characterized in that: The method of anomaly detection is specifically as follows: Calculate the conditional probability of each parameter in the dynamic key monitoring indicator under the current conditions through the Bayesian network, and mark the nodes whose conditional probability is lower than the first set value; For each parameter in the dynamic key monitoring indicator, a standard score is calculated, and nodes whose standard scores are greater than a second set value are marked; Calculate the distance between various parameters in the dynamic key monitoring indicators, and mark the nodes whose distance is greater than the third set value; Traverse all nodes and issue alarms and / or display nodes that meet the preset marking rules.

7. A new material production full-process monitoring method according to claim 6, characterized in that: The distance is calculated as follows: In the formula, Represents the vector of various parameters in the dynamic key monitoring indicators at the current moment, is the mean vector of each monitoring indicator under the historical normal state, represents the corresponding covariance matrix.

8. A new material production full process monitoring system, characterized in that: The system is used to implement the full-process monitoring method according to any one of claims 1 to 7, comprising: The data acquisition module is used to collect monitoring data of multiple process links in the production process of new materials, including raw material data, equipment operation data, environmental data and production process data; The data preprocessing module is used to preprocess the collected monitoring data, including data normalization, time synchronization and space alignment; The correlation analysis module is used to perform correlation analysis on the monitoring data of different process links, identify the coupling relationship between the previous and next process links, and establish a correlation model between process stages based on the comprehensive correlation calculation method and time correlation; Dynamic key monitoring indicator generation module, which is used to integrate the monitoring data of multiple process links to generate dynamic key monitoring indicators under current production conditions; The anomaly detection module is used to detect anomalies in real-time monitoring data based on the association model and dynamic key monitoring indicators, calculate conditional probabilities, standard scores, and parameter distances, and issue alarms or display nodes that meet the preset marking rules; The optimization suggestion module is used to generate parameter optimization adjustment suggestions for the production process based on the abnormality detection results, and push the suggestions to the production control system and operators; The user interaction module is used to display the abnormal detection results, dynamic key monitoring indicators and optimization adjustment suggestions during the production monitoring process, and receive feedback from operators.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor executes the computer program, it implements the full-process monitoring method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it can implement the full-process monitoring method as described in any one of claims 1-7.

Images