Process device controlled state evaluation method, apparatus and storage medium

Abstract

The application discloses a process device controlled state evaluation method and device and a storage medium. Firstly, time series data of each monitoring point is acquired, and a sliding window method is used to calculate a correlation coefficient matrix. Then, eigenvalue calculation is performed on the correlation coefficient matrix to obtain an eigenvalue vector at a current moment. Then, clustering analysis is performed on historical eigenvalues to determine the number of cluster centers, and clustering judgment is performed on the current eigenvalue vector. Finally, the eigenvalue vector of the current window, the clustering change vector of the trend window and the clustering count of the historical window are taken as inputs, an isolation forest algorithm is used for abnormal value judgment, and the current controlled state of the device is output. The application realizes comprehensive analysis and evaluation of real-time, trend and historical information, can more accurately judge the controlled state of the device compared with a traditional method, and improves the accuracy of state evaluation.

Description

Technical Field

[0001] This invention belongs to the field of process equipment application and relates to equipment safety risk assessment technology, specifically a method, equipment and storage medium for assessing the controlled state of a process equipment. Background Technology

[0002] The chemical industry is a high-risk industry. In the process of chemical production, most of the chemical substances in the production equipment are flammable, explosive, toxic and harmful, which usually have great potential dangers. Once a safety accident occurs, it will inevitably cause serious consequences.

[0003] To ensure the safety of chemical production processes, it is essential to have a thorough understanding of the causes of hazards in chemical process equipment and to appropriately assess and manage the risks of chemical production equipment. Only in this way can corresponding preventive and mitigation measures be taken for hazardous scenarios, thereby reducing the accident rate or the severity of casualties in the event of an accident, and promoting the smooth, safe, and reliable operation of hazardous chemical process production. Therefore, designing controlled-state assessment methods for hazardous chemical process equipment, and thus implementing safety measures for hazardous process production equipment, is of paramount practical significance for improving the inherent safety of the entire chemical industry. Summary of the Invention

[0004] To address this, the present invention proposes a method, equipment, and storage medium for assessing the controlled status of process equipment. The present invention can accurately assess whether hazardous chemical process equipment is under control, reduce the intensity of manual inspections, and improve the accuracy of equipment operating condition assessment.

[0005] In a first aspect, the present invention provides a method for evaluating the controlled state of a process apparatus, comprising the following steps:

[0006] Step 1: The signal acquisition device interfaces with the subsystem under test to acquire time-series data from various monitoring points related to the operation of the subsystem under test.

[0007] Step 2: Calculate the correlation of the time-series data from the above monitoring points using the sliding window method to obtain the correlation coefficient matrix of each monitoring point at a certain time.

[0008] Step 3: Analyze the above correlation coefficient matrix. Perform eigenvalue calculation to obtain the eigenvalue vector at the current time.

[0009] Step 4: Cluster the historical feature values ​​to obtain K cluster centers. And perform clustering judgment on the feature value vector at the current time;

[0010] Step 5: Based on the observed trend window, use Step 3 to calculate T feature values ​​of the observation window with a window width of W, and obtain T feature value vectors. After clustering statistics, obtain the clustering change vector of the feature vector in the trend window.

[0011] Step 6: Based on the historical observation window, perform clustering and counting of feature vectors for the H observation windows with a width of W using the method from Step 3;

[0012] Step 7: Use the feature value vector of the current window obtained in Step 3, the cluster change vector of the trend window obtained in Step 5, and the cluster count of the historical window obtained in Step 6 as input parameters for the outlier detection algorithm. The output of the outlier detection algorithm is the result of whether the device is currently under control.

[0013] Secondly, the present invention provides a storage medium that stores instructions and data for implementing the aforementioned method for evaluating the controlled state of a process apparatus.

[0014] Thirdly, the present invention provides a controlled state assessment device for a process apparatus, comprising: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the controlled state assessment method for the process apparatus.

[0015] The beneficial effects of this invention are as follows: This invention acquires time-series data of key equipment locations in real time, generates a correlation coefficient matrix using the sliding window method to analyze real-time correlation, and proposes a comprehensive evaluation method that combines the changes in eigenvalue vectors within the trend window and the K-means clustering results of historical data. The elbow method is used to determine the number of cluster centers, realizing the comprehensive analysis and evaluation of real-time, trend, and historical information. Compared with traditional methods, it can more accurately determine the controlled state of the equipment and improve the accuracy of state evaluation. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Some specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings in an exemplary and non-limiting manner. The same reference numerals in the drawings indicate the same or similar parts or components. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0017] Figure 1 A conceptual diagram of the system under test;

[0018] Figure 2 This is a schematic diagram of the observation window;

[0019] Figure 3 This is a trend window plot of the feature vectors. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] The main technical concept of this application is as follows: First, time-series data from each monitoring point is acquired through a signal acquisition device, and the correlation coefficient matrix is ​​calculated using the sliding window method. Next, principal component analysis is used to calculate the eigenvalues ​​of the correlation coefficient matrix, obtaining the eigenvalue vector at the current moment. Then, cluster analysis is performed on historical eigenvalues, the number of cluster centers is determined using the elbow method, and clustering is judged on the current eigenvalue vector. Within the trend window, the eigenvalue vectors of multiple observation windows are calculated and cluster statistics are performed to obtain the cluster change vector; within the historical window, the cluster count of the eigenvector is calculated. Finally, the eigenvalue vector of the current window, the cluster change vector of the trend window, and the cluster count of the historical window are used as inputs, and the isolated forest algorithm is used to judge outliers, outputting the current controlled state of the device. By comprehensively analyzing the real-time, trend, and historical information of the equipment, accurate state evaluation results are provided.

[0022] This application provides a method for assessing the controlled state of a process unit, including the following steps:

[0023] Step 1: The signal acquisition device interfaces with the subsystem under test to acquire time-series data from various monitoring points related to the operation of the subsystem under test.

[0024] Step 2: Calculate the correlation of the time-series data from the above monitoring points using the sliding window method to obtain the correlation coefficient matrix of each monitoring point at a certain time.

[0025] Step 3: Analyze the above correlation coefficient matrix. Perform eigenvalue calculation to obtain the eigenvalue vector at the current time.

[0026] Step 4: Cluster the historical feature values ​​obtained based on the above method to obtain K cluster centers. And perform clustering judgment on the current feature vector;

[0027] Step 5: Based on the observed trend window, use Step 3 to calculate T feature values ​​of the observation window with a window width of W, and obtain T feature value vectors. After clustering statistics, obtain the clustering change vector of the feature vector in the trend window.

[0028] Step 6: Based on the historical observation window, perform clustering and counting of feature vectors for the H observation windows with a width of W using the method from Step 3;

[0029] Step 7: Use the feature value vector of the current window obtained in Step 3, the cluster change vector of the trend window obtained in Step 5, and the cluster count of the historical window obtained in Step 6 as input parameters for the outlier detection algorithm. The output of the outlier detection algorithm is the result of whether the device is currently under control.

[0030] In one embodiment, step 1 specifically refers to acquiring time-series data of the tested subsystem within the time period 0 to t, as shown in the example below. Figure 1 The tested subsystems include enterprise automation control systems, enterprise security risk monitoring and early warning systems, etc., and the aforementioned time-series signals This includes various forms of time-series data such as vibration, temperature, flow rate, and pressure.

[0031] In one embodiment, step 2 specifically includes the following steps:

[0032] Step 2-1: Calculate the correlation using the sliding window method. For example... Figure 2 As shown in the schematic diagram of the observation window, for obtaining time-series signals Select a monitoring point t M (t M £t), with a sliding window width of W, calculate the Pearson correlation coefficient between each signal under this width, as shown in formula (1).

[0033]

[0034] Step 2-2: Correlation Matrix. As shown in Table 1, the Pearson correlation coefficients calculated between the time series data are arranged in order to obtain a skew-symmetric correlation matrix.

[0035] Table 1 Correlation Coefficient Matrix

[0036]

[0037] In one embodiment, the eigenvalue calculation method in step 3 is principal component analysis (PCA), which ultimately yields the eigenvectors of each time series data at the current time. Compared with traditional direct eigenvalue decomposition and singular value decomposition, this method not only provides eigenvalues ​​but also directly explains the variance and principal components of the data, which is very useful for subsequent cluster analysis and anomaly detection.

[0038] In one embodiment, step 4 is a processing procedure for the long-term results of steps 1 to 3 above, specifically including the following steps:

[0039] Step 4-1: After a period of time, the above steps can yield feature vectors corresponding to many time points. and the eigenvalues ​​in these eigenvectors K-means clustering analysis was performed on these historical feature values, and the number of cluster centers was determined by the elbow method, resulting in K cluster centers.

[0040] Step 4-2: Perform clustering judgment on the above historical feature value vectors and input them into the K-means model in step 4-1. The model will calculate the Euclidean distance from each feature vector to all cluster centers and finally classify it to the center with the closest distance, as shown in formula (2). Here, ∈ indicates that the two are closest. Whether it belongs to cluster center c i A state variable, where 1 indicates belonging and 0 indicates not belonging.

[0041]

[0042] In one embodiment, step 5 specifically includes the following steps:

[0043] Step 5-1: Obtain the trend window at a certain moment based on the observation window. For example... Figure 3 As shown, the trend window consists of T observation windows with a width of W that occurred before this time. After eigenvalue calculation in step 3, T eigenvectors representing this trend window are finally obtained. (Essentially just T units) ).

[0044] Step 5-2: Clustering change vector of the trend window. When a certain time t is confirmed... M Then, the cluster change statistics are performed on the T feature vectors before that time point to obtain the cluster change vector at that time point. d i =c j if

[0045] In one embodiment, step 6 is a clustering count based on a historical window:

[0046] Based on the observation window, a trend window at a certain moment is obtained, which consists of H observation windows with a width of W before that moment. Using step 3, the cluster count statistics of K cluster centers in this historical window are calculated to obtain the count vector. Where h KThis indicates that the data in the historical window belongs to cluster center c. K The number of eigenvectors.

[0047] In one embodiment, step 7 specifically includes the following steps:

[0048] Step 7-1: Take the feature value vector of the current window obtained in Step 3, the cluster change vector of the trend window obtained in Step 5, and the cluster count vector of the historical window obtained in Step 6 as input parameters of the outlier detection algorithm, and merge them into an input matrix to form a comprehensive feature vector.

[0049] Step 7-2: Employ the Isolation Forest algorithm, train the Isolation Forest model using historical data, and predict the current comprehensive feature vector after training to obtain its anomaly score A. Based on the set threshold T... h The system determines whether the score indicates an abnormal state, thereby determining whether the device is currently under control, as shown in Formula 3, where S represents the controlled state, 0 for uncontrolled and 1 for controlled.

[0050]

[0051] This application embodiment also provides a storage medium that stores instructions and data for implementing the aforementioned method for evaluating the controlled state of a process apparatus.

[0052] This application provides a controlled state assessment device for a process apparatus, comprising: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the controlled state assessment method for the process apparatus.

[0053] The storage medium described in this application embodiment can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. More specific examples of computer-readable storage media include at least (a non-exhaustive list) the following: an electrical connection (electronic device) having one or more wires, a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic device, and portable read-only memory (CDROM). Furthermore, the computer-readable storage medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0054] In this embodiment, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, input method, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wireline, optical fiber, radio frequency (RF), etc., or any suitable combination thereof.

[0055] The technical solutions of the present invention are not limited to the specific embodiments described above. Any technical modifications made in accordance with the technical solutions of the present invention fall within the protection scope of the present invention.

Claims

1. A method for assessing the controlled state of a process unit, characterized in that, The method includes the following steps: Step 1: The signal acquisition device interfaces with the subsystem under test to acquire time-series data from various monitoring points related to the operation of the subsystem under test. ; Step 2: Calculate the correlation of the time-series data from the above monitoring points using the sliding window method to obtain the correlation coefficient matrix of each monitoring point at a certain time. ; Step 3: Analyze the above correlation coefficient matrix. Perform eigenvalue calculation to obtain the eigenvalue vector at the current time. ; Step 4: Cluster the historical feature values ​​to obtain K cluster centers. And perform clustering judgment on the feature value vector at the current time; Step 5: Based on the observed trend window, use Step 3 to calculate T feature values ​​of the observation window with a window width of W, and obtain T feature value vectors. After clustering statistics, obtain the clustering change vector of the feature vector in the trend window. Step 6: Based on the historical observation window, perform clustering and counting of feature vectors for the H observation windows with a width of W using the method from Step 3; Step 7: Use the feature value vector of the current window obtained in Step 3, the cluster change vector of the trend window obtained in Step 5, and the cluster count of the historical window obtained in Step 6 as input parameters for the outlier detection algorithm. The output of the outlier detection algorithm is the result of whether the device is currently under control. The subsystem under test mentioned in step 1 includes an enterprise automation control system and an enterprise safety risk monitoring and early warning system. The time series data includes time series data of vibration, temperature, flow rate and pressure. Step 5 specifically includes the following steps: Step 5-1: Based on the observation window, obtain the trend window at a certain moment, calculate the eigenvalues, and finally obtain T eigenvectors representing the trend window. ; Step 5-2: When a certain moment is confirmed Then, the cluster change statistics are performed on the T feature vectors before that time point to obtain the cluster change vector at that time point. .

2. The method for assessing the controlled state of a process unit according to claim 1, characterized in that: Step 2 specifically includes the following steps: Step 2-1: For time series data Select a monitoring point The sliding window width is W; calculate the Pearson correlation coefficient between each signal under this width. Step 2-2: Arrange the Pearson correlation coefficients calculated between the time series data in order to obtain a skew-symmetric correlation coefficient matrix. .

3. The method for assessing the controlled state of a process unit according to claim 1, characterized in that: The eigenvalue calculation in step 3 uses principal component analysis.

4. The method for assessing the controlled state of a process unit according to claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4-1: Perform K-means cluster analysis on the historical feature values, and use the elbow method to determine the number of cluster centers, finally obtaining K cluster centers. ; Step 4-2: Perform clustering judgment on the current feature value vector, input it into the K-means model in step 4-1, calculate the Euclidean distance from each feature vector to all cluster centers, and finally classify it to the nearest center.

5. The method for evaluating the controlled state of a process unit according to claim 1, characterized in that: Step 6 specifically involves: Based on the observation window, a trend window at a certain moment is obtained. The cluster count statistics of K cluster centers in this historical window are calculated to obtain the count vector. ,in This indicates that the historical window belongs to the cluster center. The number of eigenvectors.

6. The method for assessing the controlled state of a process unit according to claim 1, characterized in that: Step 7 specifically includes the following steps: Step 7-1: Take the feature value vector of the current window obtained in Step 3, the cluster change vector of the trend window obtained in Step 5, and the cluster count vector of the historical window obtained in Step 6 as input parameters of the outlier detection algorithm, and merge them into an input matrix to form a comprehensive feature vector. Step 7-2: Employ the Isolation Forest algorithm, train the Isolation Forest model using historical data, and after training, predict the current comprehensive feature vector to obtain its anomaly score. According to the set threshold The system determines whether the score indicates an abnormal state, thereby determining whether the device is currently under control.

7. A storage medium, characterized in that: The storage medium stores instructions and data to implement the controlled state assessment method for a process apparatus as described in any one of claims 1 to 6.

8. A controlled status assessment device for a process unit, characterized in that: include: A processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the controlled state assessment method for a process apparatus as described in any one of claims 1 to 6.

Images