Chemical safety production management method and system based on big data

Abstract

The invention relates to the technical field of data safety management, in particular to a chemical safety production management method and system based on big data. The method comprises the following steps: acquiring data points including chemical safety parameter values of a plurality of production monitoring dimensions at each moment in a chemical production process; determining the importance of each production monitoring dimension; determining a weighted distance between the data points, and clustering the data points; determining a time domain feature of each cluster for evaluating the stability of the production rhythm; acquiring neighborhood parameters, used for identifying abnormal working conditions and risk early warning in chemical safety production, of each cluster; and on the basis of the neighborhood parameters, obtaining an abnormal score of each data point by using an LOF algorithm so as to realize chemical safety production management based on big data. According to the method, the abnormal score of the data point is obtained by using the LOF algorithm based on the neighborhood parameter, so that the problem of large result deviation of a traditional abnormal detection algorithm is avoided, and abnormal conditions in the production process can be found in time.

Description

Technical Field

[0001] The present invention relates to the technical field of data security management, and particularly to a chemical production safety management method and system based on big data. Background Art

[0002] In the field of chemical production, safety management is a key link to ensure the stable operation of production and avoid accidents. The traditional safety management mode relies on manual inspections and judgments based on past experience. Manual inspections not only consume a large amount of manpower, material resources and time, with extremely low efficiency, but also are inevitably prone to omissions by humans, greatly reducing the accuracy. In addition, experience-based judgments are highly subjective, and the judgment criteria and ability levels of different personnel vary greatly, which undoubtedly poses potential hazards to chemical production safety.

[0003] With the development of information technology, big data technology has emerged, bringing new opportunities for the safety management of chemical enterprises. More and more chemical enterprises have begun to try to use data analysis technology to manage and monitor safety data in the production process. By collecting various sensor data information, potential safety risks can be identified in order to take corresponding measures in advance.

[0004] However, in the actual application process, the existing big data analysis methods still face many challenges in the chemical production safety scenario. For example, the data production monitoring has a high dimension, low accuracy and insufficient real-time performance, etc. In chemical production safety management, chemical production data is extremely complex, covering many different types of parameters, with a wide range of data sources and being interrelated; therefore, the high dimension of data production monitoring is a significant challenge. When traditional data analysis and anomaly detection algorithms process high-dimensional data, due to the characteristic that the distances between data points tend to be equal in high-dimensional space, the calculation accuracy decreases, and then the anomaly detection results deviate greatly, making it impossible to accurately identify potential safety risks. Summary of the Invention

[0005] In order to solve the problem that when traditional data analysis and anomaly detection algorithms process high-dimensional data, due to the fact that in high-dimensional space, the distances between data points tend to be equal, the calculation accuracy decreases, and then the anomaly detection results are inaccurate, affecting the safety of chemical production, the present invention provides a chemical production safety management method and system based on big data.

[0006] In the first aspect, the present invention provides a chemical production safety management method based on big data, adopting the following technical solutions: A chemical industrial safety production management method based on big data, comprising: obtaining data points containing chemical safety parameter values of multiple production monitoring dimensions at each moment during the chemical production process; determining the importance of each production monitoring dimension according to the information entropy of the chemical safety parameter values of all data points in each production monitoring dimension, the chemical safety parameter values of each data point in each production monitoring dimension, the mean and standard deviation of the chemical safety parameter values of all data points within the data segment to which each data point belongs in each production monitoring dimension; for any two data points, determining the weighted distance between the two data points according to the difference in the chemical safety parameter values of the two data points in each production monitoring dimension and the importance; clustering all data points based on the weighted distance and sorting them in time sequence to obtain multiple clustering clusters; obtaining the mean and standard deviation of the intervals corresponding to the times of all adjacent two data points within each clustering cluster, and determining the time domain characteristics of each clustering cluster for evaluating the stability of the production rhythm; correcting the preset initial neighborhood parameter according to the time domain characteristics to obtain the neighborhood parameter of each clustering cluster for identifying abnormal working conditions and risk warnings in chemical industrial safety production; based on the neighborhood parameter, obtaining the anomaly score of each data point by using the LOF algorithm, so as to realize chemical industrial safety production management based on big data.

[0007] By calculating the importance weights of each production monitoring dimension, the present invention enhances the influence of key production monitoring dimensions on distance calculation, reduces the interference of irrelevant or noisy production monitoring dimensions, improves the accuracy of distance measurement, and overcomes the problem of distance convergence in high-dimensional spaces in traditional methods; the weighted distance based on the production monitoring dimension weights more reasonably reflects the actual differences between data points, making the clustering results more in line with the internal structure of the data, thereby improving the accuracy of anomaly detection and being able to accurately identify potential safety risks; dynamically correcting the initial neighborhood parameter according to the mean and standard deviation of the time intervals of the data points within each clustering cluster makes the neighborhood division of the LOF algorithm more in line with the time sequence characteristics of the data, enhancing the timeliness and pertinence of anomaly detection; by integrating information entropy statistics, multi-dimensional weighted distance, time domain characteristics and LOF anomaly scoring, comprehensively analyzing chemical production data from multiple angles, and improving the processing ability of the safety management system for complex and diverse data; through more accurate and timely anomaly detection, it helps operators and management to discover potential safety hazards, take corresponding measures in time, ensure production safety, and reduce the accident rate.

[0008] Further, the multiple production monitoring dimensions include a temperature dimension, a humidity dimension, a pressure dimension, and a gas concentration dimension.

[0009] Further, the importance satisfies: ; where is the importance of the th production monitoring dimension, is the information entropy of the chemical safety parameter values of all data points in the th production monitoring dimension, is the number of data points, is the th data point's chemical safety parameter value in the th production monitoring dimension, and are respectively the mean and standard deviation of the chemical safety parameter values of all data points within the data segment to which the th data point belongs in the th production monitoring dimension, is the standard normalization function.

[0010] By combining the information entropy to measure the dispersion and variability of the production monitoring dimension with the normalized average deviation, the present invention can simultaneously reflect the complexity and local fluctuation characteristics of the data in this production monitoring dimension, and avoid the deviation caused by evaluating the importance of the production monitoring dimension with a single indicator; by performing standard normalization on the information entropy and the normalized average deviation respectively, the differences in dimension and numerical range are eliminated, enabling the two indicators to have an equal impact in the product operation, and enhancing the scientificity and stability of the evaluation of the importance of the production monitoring dimension.

[0011] Furthermore, the weighted distance satisfies: ; where is the weighted distance between the data point and the data point , is the number of production monitoring dimensions, is the importance of the th production monitoring dimension, is the chemical safety parameter value of the data point in the th production monitoring dimension, is the chemical safety parameter value of the data point in the th production monitoring dimension, is the absolute value symbol.

[0012] By normalizing the importance of the production monitoring dimension as a weighted coefficient, the present invention makes the production monitoring dimension with high importance contribute more when calculating the distance between data points, weakens the interference of the production monitoring dimension with low importance, and ensures that the distance metric is more in line with the actual data characteristics.

[0013] Furthermore, the clustering uses the iterative self-organizing clustering algorithm.

[0014] Furthermore, the clustering uses the K-means clustering algorithm.

[0015] Furthermore, the time domain features satisfy: ; where is the time domain feature of the th clustering cluster, is the mean of the intervals between the corresponding times of all adjacent two data points within the th clustering cluster, is the standard deviation of the intervals between the corresponding times of all adjacent two data points within the th clustering cluster, is the standard normalization function.

[0016] By calculating the mean and standard deviation of the time intervals between adjacent data points within the clustering cluster and performing standard normalization, the present invention can objectively reflect the timing distribution law of the data in this cluster, revealing the concentration degree and fluctuation of the data on the time axis; using the time domain features as the basis for modifying the neighborhood parameters enables the neighborhood range to adapt to the timing characteristics of different clustering clusters, avoiding the timing mismatch problem caused by static neighborhood parameters, thereby improving the accuracy of local outlier detection.

[0017] Furthermore, the neighborhood parameters satisfy: ; where is the neighborhood parameter of the th clustering cluster, is the initial neighborhood parameter, is the time domain feature of the th clustering cluster, is a hyperparameter, is the ceiling symbol.

[0018] By dynamically calculating the neighborhood parameters based on the initial neighborhood parameter, combining the time domain features and adjusting the hyperparameters, the present invention enables the neighborhood range to flexibly change according to the timing characteristics of different clustering clusters, avoiding the deficiencies caused by fixed neighborhood parameters; the adaptive neighborhood parameters can more reasonably delimit the neighborhood range of each clustering cluster, enabling the LOF algorithm to effectively identify outliers in clusters with different time distribution densities, improving the reliability and discrimination of the outlier scores.

[0019] Furthermore, obtaining the outlier score of each data point using the LOF algorithm to implement chemical industrial safety production management based on big data includes: for any data point, using the LOF algorithm to take the neighborhood parameter of the clustering cluster where the data point is located as the neighborhood parameter of the data point, and obtaining the outlier score of the data point; in response to the outlier score being greater than a preset outlier threshold, determining that the data point is abnormal, storing the data point separately, and sending a warning prompt, thus completing chemical industrial safety production management based on big data.

[0020] In a second aspect, the present invention provides a chemical industrial safety production management system based on big data, adopting the following technical solution: A chemical industrial safety production management system based on big data, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned chemical industrial safety production management method based on big data is implemented.

[0021] By adopting the above technical solution, the above-mentioned chemical industrial safety production management method based on big data is generated into a computer program and stored in the memory, so as to be loaded and executed by the processor, and thus a terminal device is manufactured according to the memory and the processor, which is convenient to use.

[0022] The present invention has the following technical effects: (1) Aiming at the problems of high production monitoring dimensions and strong complexity of chemical production data, the present invention determines the importance of each production monitoring dimension by comprehensively considering the information entropy of the chemical safety parameter values of all data points in each production monitoring dimension, the chemical safety parameter values of each data point in each production monitoring dimension, the mean value and standard deviation of the chemical safety parameter values of the data segment to which each data point belongs in each production monitoring dimension, which can fully mine the potential information of each production monitoring dimension in high-dimensional data, avoid the problem of decreased calculation accuracy caused by the characteristic that the distances between data points in the high-dimensional space tend to be equal in the traditional algorithm, and effectively improve the processing ability of high-dimensional data.

[0023] (2) When determining the distance between data points, the weighted distance is calculated according to the difference between the chemical safety parameter values of two data points in each production monitoring dimension and the importance of each production monitoring dimension, making the calculation of the distance more reasonable, being able to more accurately reflect the actual similarity degree between data points, providing a more reliable basis for subsequent clustering analysis, and helping to more accurately divide data categories.

[0024] (3) The mean value and standard deviation of the corresponding time interval of adjacent data points within each clustering cluster are obtained to determine the time-domain characteristics of the clustering cluster, and the preset initial neighborhood parameter is corrected according to this characteristic, fully considering the time series characteristics of the data, enabling the neighborhood parameter to better adapt to the data distribution of different clustering clusters, and improving the rationality and effectiveness of the parameter.

[0025] (4) Based on the corrected neighborhood parameter, the LOF algorithm is used to obtain the anomaly score of each data point, effectively avoiding the problem of large deviation in the results of traditional anomaly detection algorithms, being able to more accurately identify potential safety risk data points in the chemical production process, timely discover abnormal situations in the production process, providing more powerful guarantee for chemical industrial safety production, helping to reduce the occurrence of safety accidents, and reducing production risks and losses. Brief Description of the Drawings

[0026] Figure 1 It is the flowchart of a chemical industrial safety production management method based on big data in an embodiment of the present invention. Detailed Embodiment

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0028] An embodiment of the present invention discloses a chemical industrial safety production management method based on big data. Refer to Figure 1 , which includes steps S1 - S6: S1: Obtain data points of chemical safety parameter values including multiple production monitoring dimensions at each moment during the chemical production process.

[0029] It should be noted that the data points at each moment during the chemical production process refer to various information and data related to safety production during the chemical production process, including but not limited to equipment production process parameters, environmental data, equipment operating status, etc. By collecting and analyzing these data points, real-time monitoring of the production process can be achieved, potential risks can be discovered in a timely manner, and the system can respond in a timely manner.

[0030] Use corresponding sensors to obtain data such as temperature values, humidity values, pressure values, gas concentration values, etc., and take the chemical safety parameter values in all production monitoring dimensions collected at the same moment as a data point.

[0031] Implementers can set the collection frequency according to the specific implementation situation. For example, 20Hz.

[0032] Specifically, the multiple production monitoring dimensions include a temperature dimension, a humidity dimension, a pressure dimension, and a gas concentration dimension.

[0033] S2: Determine the importance of each production monitoring dimension.

[0034] It should be noted that in the chemical production process, since there are many characteristic production monitoring dimensions involved, if all the collected data is directly input into the anomaly monitoring model, not only will the computational complexity increase significantly, but also the LOF algorithm realizes the anomaly monitoring of data points based on the Euclidean distance, and the Euclidean distance between different production monitoring dimensions will affect the anomaly detection result. In order to reduce the influence of the distance between different production monitoring dimensions on the anomaly detection result, the present invention calculates the importance of any one production monitoring dimension for subsequent further analysis based on the importance result.

[0035] Determine the importance of each production monitoring dimension according to the information entropy of the chemical safety parameter values of all data points in each production monitoring dimension, the chemical safety parameter values of each data point in each production monitoring dimension, and the mean and standard deviation of the chemical safety parameter values of all data points in each production monitoring dimension within the data segment to which each data point belongs.

[0036] The implementer can set the length of the data segment according to the specific implementation situation. For example, 21, including the data point and the previous 20 data points.

[0037] Specifically, the importance satisfies: ; In the formula, is the importance of the th production monitoring dimension, is the information entropy of the chemical safety parameter values of all data points in the th production monitoring dimension, is the number of data points, is the th data point in the th production monitoring dimension of the chemical safety parameter value, and are respectively the mean and standard deviation of the chemical safety parameter values of all data points within the data segment to which the th data point belongs in the th production monitoring dimension, is the standard normalization function.

[0038] Among them, in chemical production, for any one production monitoring dimension, during normal operation, it is stable within a certain range, but in abnormal situations, there will be violent fluctuations, that is, there may be potential safety anomalies. Therefore, for any one production monitoring dimension, the greater the data fluctuation of this production monitoring dimension, the more attention needs to be paid to this production monitoring dimension, that is, the higher the importance of this production monitoring dimension. Indicates the distribution of all data in this production monitoring dimension. The larger this value, the larger the data distribution range of this production monitoring dimension, and the more uniform the data value distribution. Then, the greater the data fluctuation of this production monitoring dimension, the more likely there are potential safety anomalies, and the higher the importance. However, information entropy cannot take into account the local changes in the data of this production monitoring dimension. Therefore, is introduced. By calculating the differences between all data points and the data values within their respective data segments, the local difference situation is obtained. The larger this value, the more obvious the local data fluctuation of this production monitoring dimension, the more likely there are potential safety anomalies, and the higher the importance.

[0039] S3: Determine the weighted distance between data points and cluster the data points.

[0040] It should be noted that since the amount of data for directly performing anomaly detection on all data is huge, the computational complexity is too high, and the time consumption is too long. Therefore, in the present invention, by clustering the data, it is convenient to analyze each clustering cluster separately. When clustering, both the processing efficiency can be improved, and the weight of the production monitoring dimension with higher importance can be increased, thereby improving the accuracy. Since different production monitoring dimensions have different degrees of influence on chemical industrial safety production management, the production monitoring dimension with higher importance is more crucial for judging the similarity between data points. Therefore, the weighted distance between data points is obtained by combining the importance, and the Euclidean distance in conventional clustering is replaced with the weighted distance to improve the accuracy of the clustering result.

[0041] For any two data points, based on the differences in the chemical safety parameter values of the two data points in each production monitoring dimension and the importance, the weighted distance between the two data points is determined.

[0042] Specifically, the weighted distance satisfies: ; In the formula, is the weighted distance between data point and data point , is the number of production monitoring dimensions, is the importance of the th production monitoring dimension, is the chemical safety parameter value of data point in the th production monitoring dimension, is the chemical safety parameter value of data point in the th production monitoring dimension, is the absolute value symbol, .

[0043] Based on the weighted distance, all data points are clustered and sorted in time sequence to obtain multiple clusters.

[0044] Specifically, the clustering uses the iterative self-organizing clustering algorithm.

[0045] In another embodiment, the clustering uses the K-means clustering algorithm.

[0046] S4: Determine the time-domain features of each cluster for evaluating the stability of the production rhythm.

[0047] It should be noted that in order to improve the efficiency of anomaly detection, the present invention clusters the collected data to obtain multiple clusters, and uses the LOF algorithm for anomaly detection for each cluster separately. The traditional LOF algorithm directly presets a fixed distance parameter, that is, the neighborhood parameter, and completes the anomaly detection according to this parameter. In the present invention, since the clustering is completed according to the weighted distance between data points in different clusters, the time interval factor is not considered. For example, if the time interval in a certain cluster is far and the distance between data points is large, if a fixed distance parameter is used, the anomaly score of the data points in this cluster will be relatively high. Therefore, by calculating the time interval in the cluster to obtain the eigenvalue of the cluster, that is, the time-domain feature, and then selecting an appropriate distance parameter according to the eigenvalue.

[0048] Obtain the mean and standard deviation of the intervals corresponding to the moments of all adjacent two data points in each cluster, and determine the time-domain features of each cluster for evaluating the stability of the production rhythm.

[0049] Specifically, the time-domain features satisfy: ; In the formula, is the time-domain feature of the th cluster, is the mean of the intervals corresponding to the moments of all adjacent two data points in the th cluster, is the standard deviation of the intervals corresponding to the moments of all adjacent two data points in the th cluster, is the standard normalization function.

[0050] Among them, The larger the value of, the wider the distribution range of the data points in this cluster from the time series perspective, that is, the stronger the time-domain feature. On the contrary, the smaller its value, the narrower the distribution range of the data points in this cluster from the time series perspective, that is, the weaker the time-domain feature; reflects the distribution uniformity of all time intervals in this cluster. The smaller this value, the more uniform the time interval distribution, and the stronger the time-domain feature, and vice versa.

[0051] S5: Obtain the neighborhood parameters for each cluster to identify abnormal working conditions and risk warnings in chemical production safety.

[0052] It should be noted that for any cluster, if the time-domain feature of the cluster is stronger, it means that the data points in the cluster are more discrete in the time series, and a larger range of neighborhood parameters needs to be selected to avoid the influence of local noise caused by too small a parameter range.

[0053] Modify the preset initial neighborhood parameters according to the time-domain features to obtain the neighborhood parameters for each cluster to identify abnormal working conditions and risk warnings in chemical production safety.

[0054] Specifically, the neighborhood parameters satisfy: ; In the formula, is the neighborhood parameter of the th cluster, is the initial neighborhood parameter, is the time-domain feature of the th cluster, is the hyperparameter, is the ceiling symbol.

[0055] Among them, represents the time-domain feature reference value. If the time-domain feature of the cluster is greater than , that is, , it means that the time-domain feature is too large, and the neighborhood parameters need to be appropriately enlarged. Otherwise, the neighborhood parameters need to be appropriately reduced.

[0056] Implementers can set the initial neighborhood parameters and hyperparameters according to the specific implementation situation. For example, the initial neighborhood parameter is 10 and the hyperparameter is 0.5.

[0057] S6: Based on the neighborhood parameters, use the LOF algorithm to obtain the anomaly score of each data point to achieve chemical production safety management based on big data.

[0058] Specifically, the use of the LOF algorithm to obtain the anomaly score of each data point to achieve chemical production safety management based on big data includes: For any data point, use the LOF algorithm to take the neighborhood parameter of the cluster where the data point is located as the neighborhood parameter of the data point, and obtain the anomaly score of the data point; In response to the anomaly score being greater than the preset anomaly threshold, determine that the data point is abnormal, store the data point separately, and issue a warning prompt to complete chemical production safety management based on big data.

[0059] Implementers can set the anomaly threshold according to the specific implementation situation. For example, 1.

[0060] An embodiment of the present invention also discloses a chemical industrial safety production management system based on big data, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a chemical industrial safety production management method based on big data according to the present invention is implemented.

[0061] The above system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be elaborated here.

[0062] The above are all preferred embodiments of the present invention. The protection scope of the present invention is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A chemical production safety management method based on big data, characterized in that: include: Obtain data points of chemical safety parameter values ​​containing multiple production monitoring dimensions at each moment in the chemical production process; Determine the importance of each production monitoring dimension based on the information entropy of the chemical safety parameter values ​​of all data points in each production monitoring dimension, the chemical safety parameter value of each data point in each production monitoring dimension, and the mean and standard deviation of the chemical safety parameter values ​​of all data points in the data segment to which each data point belongs in each production monitoring dimension; For any two data points, a weighted distance between the two data points is determined according to the difference in the chemical safety parameter values ​​of the two data points in each production monitoring dimension and the importance; based on the weighted distance, all data points are clustered and sorted in time sequence to obtain multiple clusters; Obtain the mean and standard deviation of the intervals between all two adjacent data points in each cluster, and determine the time domain characteristics of each cluster for evaluating the stability of the production rhythm; Modifying the preset initial neighborhood parameters according to the time domain characteristics to obtain neighborhood parameters for each cluster for identifying abnormal conditions and risk warnings in chemical production safety; Based on the neighborhood parameters, the LOF algorithm is used to obtain the anomaly score of each data point to achieve chemical production safety management based on big data.

2. A chemical production safety management method based on big data according to claim 1, characterized in that: The multiple production monitoring dimensions include temperature dimension, humidity dimension, pressure dimension and gas concentration dimension.

3. According to a chemical production safety management method based on big data according to claim 1, it is characterized in that: The importance is satisfied: ； In the formula, For the The importance of each production monitoring dimension, For all data points in The information entropy of chemical safety parameter values ​​in the production monitoring dimension, is the number of data points, For the The data point in Chemical safety parameter values ​​in production monitoring dimensions, and Respectively All data points in the data segment to which the data point belongs are in The mean and standard deviation of chemical safety parameter values ​​in each production monitoring dimension, is the standard normalization function.

4. A chemical production safety management method based on big data according to claim 1, characterized in that: The weighted distance satisfies: ； In the formula, For data points With data points The weighted distance between The number of production monitoring dimensions, For the The importance of each production monitoring dimension, For data points In the Chemical safety parameter values ​​in production monitoring dimensions, For data points In the Chemical safety parameter values ​​in production monitoring dimensions, is the absolute value symbol.

5. The chemical production safety management method based on big data according to claim 1 is characterized in that: The clustering adopts an iterative self-organizing clustering algorithm.

6. A chemical production safety management method based on big data according to claim 1, characterized in that: The clustering adopts K-means clustering algorithm.

7. A chemical production safety management method based on big data according to claim 1, characterized in that: The time domain characteristics satisfy: ； In the formula, For the The time domain characteristics of clusters, For the The mean of the intervals between the corresponding moments of all two adjacent data points in a cluster, For the The standard deviation of the intervals between all two adjacent data points in a cluster. is the standard normalization function.

8. A chemical production safety management method based on big data according to claim 1, characterized in that: The neighborhood parameters satisfy: ； In the formula, For the The neighborhood parameters of the clusters, is the initial neighborhood parameter, For the The time domain characteristics of clusters, is a hyperparameter, The round-up symbol.

9. A chemical production safety management method based on big data according to claim 1, characterized in that: The method of using the LOF algorithm to obtain the abnormal score of each data point to realize chemical production safety management based on big data includes: For any data point, the LOF algorithm is used to take the neighborhood parameters of the cluster where the data point is located as the neighborhood parameters of the data point to obtain the abnormal score of the data point; In response to the anomaly score being greater than a preset anomaly threshold, the data point is identified as abnormal, the data point is stored separately, and an early warning prompt is issued to complete chemical production safety management based on big data.

10. A chemical production safety management system based on big data, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a chemical production safety management method based on big data according to any one of claims 1 to 9 is implemented.

Images