A hierarchical early warning method for production safety based on big data
Through big data analysis and principal component analysis, the value of safety production warnings is quantified, and the problem of low accuracy of safety production warnings in the existing technology is solved, achieving more efficient and objective safety hierarchical warnings.
Patent Information
- Application Number
- CN202311426911.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-10-30
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2043-10-30
AI Technical Summary
The existing safety production early warning methods rely on manual threshold setting and are easily subjectively affected, resulting in low warning accuracy and equipment updates lead to threshold failure, making it impossible to adapt to production changes in real time.
The safety production hierarchical warning method is adopted based on big data. By obtaining and analyzing data of various categories of safety impact parameters, performing principal component analysis and weight assignment, calculating the influence change indicators and membership, quantifying the safety warning values, and achieving safety hierarchical warnings.
It improves the accuracy and objectivity of production safety warnings, can adapt to production changes in real time, quantify multiple production safety-related indicators, and enhances the reliability of the warning results.
Smart Images

Figure CN117495169B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of safety data processing, and in particular to a safety production graded early warning method based on big data. Background Art
[0002] Big data has a wide range of applications. For example, big data can be used to provide early warnings for production safety. Since production safety often affects the health and safety of workers and consumers and is closely related to people's lives, early warnings for production safety are crucial. At present, the usual method for early warnings for production safety is to use preset thresholds to provide early warnings for production safety.
[0003] However, when the above method is adopted, the following technical problems often occur:
[0004] Since the preset thresholds are often set through manual experience and are often subject to subjective influences of human factors, the rationality of the threshold settings is often low, which often leads to low accuracy of production safety warnings. In addition, since some equipment is often updated during the production process, the originally set preset thresholds may no longer be applicable to actual production conditions, which often leads to low accuracy of production safety warnings. Summary of the invention
[0005] The content of the present invention is used to introduce the concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of the present invention is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.
[0006] In order to solve the technical problem of low accuracy of production safety early warning, the present invention proposes a production safety graded early warning method based on big data.
[0007] The present invention provides a production safety hierarchical early warning method based on big data, the method comprising:
[0008] Obtain each safety impact data corresponding to each safety impact parameter in each target category;
[0009] Performing principal component analysis on all safety impact data corresponding to all safety impact parameters in each target category to obtain a preset number of principal components corresponding to each target category;
[0010] Determine the projection values of all safety impact parameters under each principal component corresponding to each target category according to the preset safety weights corresponding to all safety impact parameters in each target category;
[0011] According to all the safety impact data under each principal component and the projection values of all the safety impact parameters under each principal component, the impact change index corresponding to each principal component is determined;
[0012] Determine the target impact weight corresponding to each principal component based on the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, and the impact change index corresponding to each principal component;
[0013] According to all the safety impact data under each principal component, determine the initial membership degree corresponding to each principal component;
[0014] Determine the target membership corresponding to each target category according to the target influence weights and initial memberships corresponding to the preset number of principal components corresponding to each target category;
[0015] According to the target affiliation corresponding to all target categories and all preset security weights, the security warning value is determined, and a security graded warning is performed according to the security warning value.
[0016] Optionally, determining the projection values of all safety impact parameters under each principal component corresponding to each target category according to preset safety weights corresponding to all safety impact parameters in each target category includes:
[0017] According to all the safety impact data corresponding to all the safety impact parameters in the target category, construct an original matrix corresponding to the target category, wherein each row of the original matrix is composed of all the safety impact data corresponding to the same safety impact parameter, and each column of the original matrix is composed of all the safety impact data collected at the same collection time;
[0018] According to the preset safety weights corresponding to all safety impact parameters in the target category, construct a target diagonal matrix corresponding to the target category, wherein the diagonal elements in the target diagonal matrix are the preset safety weights corresponding to the safety impact parameters;
[0019] Constructing a principal component matrix corresponding to the target category according to a preset number of principal components corresponding to the target category;
[0020] The product of the original matrix, the target diagonal matrix and the principal component matrix is determined as a projection value matrix corresponding to the target category, wherein the elements in the projection value matrix are projection values of the safety impact parameters under the principal component.
[0021] Optionally, determining the impact change index corresponding to each principal component according to all safety impact data under each principal component and the projection values of all safety impact parameters under each principal component includes:
[0022] Determine the mean of all safety impact data under the principal component as the original impact index corresponding to the principal component;
[0023] Determine the mean of the projection values of all safety impact parameters under the principal component as the projection impact index corresponding to the principal component;
[0024] Determine the absolute value of the difference between the original influence index and the projected influence index as the influence difference index corresponding to the principal component;
[0025] An influence change index corresponding to the principal component is determined according to the influence difference index, wherein the influence difference index is positively correlated with the influence change index.
[0026] Optionally, the formula corresponding to the target influence weight corresponding to the principal component is:
[0027] Among them, QW it is the target influence weight corresponding to the tth principal component corresponding to the ith target category; WL i is the cumulative value of the preset safety weights corresponding to all safety impact parameters in the i-th target category; w it is the preset safety weight corresponding to the tth principal component corresponding to the ith target category; min() is the minimum value function; B it is the impact change index corresponding to the tth principal component corresponding to the i-th target category; A i is the cumulative value of the pseudo weights corresponding to all principal components corresponding to the i-th target category; N is the preset number; i is the serial number of the target category; t is the serial number of the principal component corresponding to the i-th target category.
[0028] Optionally, determining the initial membership degree corresponding to each principal component according to all safety impact data under each principal component includes:
[0029] According to all the safety impact data under the principal component, construct a target spectrum graph corresponding to the principal component;
[0030] Determining a periodicity index corresponding to the principal component according to the target frequency spectrum;
[0031] According to the periodic index and all safety impact data under the principal component, an initial membership degree corresponding to the principal component is determined.
[0032] Optionally, the formula corresponding to the periodicity indicator corresponding to the principal component is:
[0033] Among them, SL itis the periodicity index corresponding to the tth principal component corresponding to the i-th target category; exp() is an exponential function with a natural constant as the base; D it is the maximum amplitude in the target spectrum corresponding to the tth principal component of the ith target category; F it is the weighted average of the frequencies of all peaks in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the i-th target category; C it is the length of the period of maximum amplitude in the target spectrum corresponding to the tth principal component of the i-th target category; N it is the number of harmonics in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the i-th target category; || is the absolute value function; a is a preset adjustment parameter greater than 0; H it is the maximum peak value of the harmonic in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the ith target category; i is the serial number of the target category; t is the serial number of the principal component corresponding to the ith target category.
[0034] Optionally, the formula corresponding to the initial membership degree corresponding to the principal component is:
[0035] Among them, μ it is the initial membership of the tth principal component corresponding to the ith target category; n it is the number of safety impact data under the tth principal component corresponding to the i-th target category; μ itb is the membership index corresponding to the bth safety impact data under the tth principal component corresponding to the i-th target category; SL it is the periodicity index corresponding to the tth principal component corresponding to the ith target category; norm() is the normalization function; is the Euler function; x itb is the bth safety impact data under the tth principal component corresponding to the i-th target category; x it1 is the mean of all safety impact data under the tth principal component corresponding to the i-th target category; σ it is the variance of all safety impact data under the t-th principal component corresponding to the i-th target category; i is the serial number of the target category; t is the serial number of the principal component corresponding to the i-th target category; b is the serial number of the safety impact data under the t-th principal component.
[0036] Optionally, determining the target membership corresponding to each target category according to the target influence weights and initial memberships corresponding to a preset number of principal components corresponding to each target category includes:
[0037] Determine the mean of the target influence weights corresponding to a preset number of principal components corresponding to the target category as the reference weight corresponding to the target category;
[0038] The product of the target influence weight corresponding to each principal component and the initial membership degree is determined as the first modified membership degree corresponding to each principal component;
[0039] Determine the accumulated value of the first modified membership corresponding to the preset number of principal components corresponding to the target category as the second modified membership corresponding to the target category;
[0040] The ratio of the second modified membership to the reference weight is determined as the target membership corresponding to the target category.
[0041] Optionally, determining the safety warning value according to the target memberships corresponding to all target categories and all preset safety weights includes:
[0042] Determine the cumulative value of the preset safety weights corresponding to all safety impact parameters in each target category as the overall safety weight corresponding to the target category;
[0043] The product of the overall safety weight corresponding to each target category and the target membership degree is determined as the early warning indicator corresponding to the target category;
[0044] The safety warning value is determined based on the warning indicators corresponding to all target categories, where the warning indicators are positively correlated with the safety warning value.
[0045] Optionally, performing a security graded warning according to the security warning value includes:
[0046] When the safety warning value is less than or equal to the first preset safety threshold, it is determined that there is a low risk in production safety;
[0047] When the safety warning value is greater than a first preset safety threshold and less than or equal to a second preset safety threshold, it is determined that there is a general risk in production safety, wherein the second preset safety threshold is greater than the first preset safety threshold;
[0048] When the safety warning value is greater than the second preset safety threshold and less than or equal to the third preset safety threshold, it is determined that there is a greater risk in production safety, wherein the third preset safety threshold is greater than the second preset safety threshold;
[0049] When the safety warning value is greater than the third preset safety threshold, it is determined that there is a high risk in production safety.
[0050] The present invention has the following beneficial effects:
[0051] The hierarchical early warning method for safe production based on big data of the present invention realizes early warning for safe production by processing multiple safety impact data, solves the technical problem of low accuracy of early warning for safe production, and improves the accuracy of early warning for safe production. First, each safety impact data corresponding to each safety impact parameter related to safe production in each target category is obtained, which can facilitate subsequent early warning for safe production. Then, principal component analysis is performed according to all safety impact data corresponding to all safety impact parameters in each target category, which can achieve dimensionality reduction and facilitate subsequent accurate processing of each principal component. Then, since each safety impact parameter often has an impact on safe production, the preset safety weights corresponding to all safety impact parameters in each target category are comprehensively considered, and the projection values of all safety impact parameters under each principal component corresponding to each target category can be quantified. Continuing, since the greater the data change before and after the projection under the principal component, it often indicates that the principal component should be given a higher weight, so the greater the impact change index corresponding to the principal component, it often indicates that the principal component should be given a higher weight. Afterwards, the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, as well as the impact change index corresponding to each principal component are comprehensively considered, the target impact weight corresponding to each principal component can be quantified. Then, the safety warning value is quantified by quantifying the target impact weight and initial membership, thereby quantifying the target membership and the preset safety weight. Finally, a safety graded warning is performed based on the safety warning value, and a safety graded warning is realized. Compared with the subjective static threshold warning, the present invention improves the accuracy of the safety warning value quantification by real-time analysis of the safety impact data and the preset safety weight, thereby improving the accuracy of the safety graded warning. Secondly, the present invention quantifies multiple indicators related to production safety, such as projection value, impact change index, target impact weight, initial membership, target membership and safety warning value, etc., which can make the subsequent safety graded warning results relatively more objective, thereby improving the accuracy of the safety graded warning. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0053] Figure 1 The figure is a flow chart of the big data-based production safety graded early warning method of the present invention. DETAILED DESCRIPTION
[0054] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0055] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0056] The present invention provides a safety production hierarchical early warning method based on big data, which comprises the following steps:
[0057] Obtain each safety impact data corresponding to each safety impact parameter in each target category;
[0058] Performing principal component analysis on all safety impact data corresponding to all safety impact parameters in each target category to obtain a preset number of principal components corresponding to each target category;
[0059] Determine the projection values of all safety impact parameters under each principal component corresponding to each target category according to the preset safety weights corresponding to all safety impact parameters in each target category;
[0060] According to all the safety impact data under each principal component and the projection values of all the safety impact parameters under each principal component, the impact change index corresponding to each principal component is determined;
[0061] Determine the target impact weight corresponding to each principal component based on the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, and the impact change index corresponding to each principal component;
[0062] According to all the safety impact data under each principal component, determine the initial membership degree corresponding to each principal component;
[0063] Determine the target membership corresponding to each target category according to the target influence weights and initial memberships corresponding to the preset number of principal components corresponding to each target category;
[0064] According to the target affiliation corresponding to all target categories and all preset security weights, the security warning value is determined, and security graded warning is carried out according to the security warning value.
[0065] The following is a detailed explanation of each of the above steps:
[0066] refer to Figure 1, shows the process of some embodiments of the method for hierarchical early warning of production safety based on big data according to the present invention. The method for hierarchical early warning of production safety based on big data includes the following steps:
[0067] Step S1, obtaining each security impact data corresponding to each security impact parameter in each target category.
[0068] In some embodiments, each safety impact data corresponding to each safety impact parameter in each target category may be obtained.
[0069] Among them, the target category can be a category that affects the safe production of a refinery. Since the seven major risk factors that affect the safe production of a refinery can include inherent risk, production operation, field operation, double defense, LDAR, equipment operation and other factors, there can be 7 target categories, which can be inherent risk category, production operation category, field operation category, double defense category, LDAR category, equipment operation category and other factor category. The safety impact parameter can be a parameter that affects the safe production of a refinery, that is, an element that constitutes the target category. For example, the safety impact parameters included in the inherent risk category can be, but are not limited to: the total number of production units, the number of first-level major hazardous sources, the number of second-level major hazardous sources, the number of third-level major hazardous sources and the number of fourth-level major hazardous sources. Among them, the total number of production units can be the number of production units in a refinery. The first-level major hazardous source, the second-level major hazardous source, the third-level major hazardous source and the fourth-level major hazardous source can be the hazardous source levels divided according to the degree of harm to the refinery.
[0070] The safety impact parameters included in the production operation category may be, but are not limited to: the number of units that have been shut down, the number of units that are being started up, the number of units that are being shut down, the number of units in trial production, the production plan completion rate, the average unit load rate, the number of unplanned shutdowns, the number of particularly serious accidents, the number of serious accidents, the number of major accidents, the number of general accidents, and the number of abnormal events. Among them, particularly serious accidents, major accidents, major accidents, and general accidents may be accident levels classified according to the degree of accident danger.
[0071] The safety impact parameters included in the on-site operation categories may include but are not limited to: the number of special-grade hot work operations, the number of first-grade hot work operations, the number of second-grade hot work operations, the number of operations entering confined spaces, the number of blind plate plugging operations, the number of operations at heights, the number of lifting operations, the number of temporary power supply operations, the number of earth-breaking operations, the number of circuit-breaking operations, the number of inspection and maintenance operations, and the number of contractor operations.
[0072] The safety impact parameters included in the dual defense category may be, but are not limited to: the number of existing major hidden dangers, the number of existing general hidden dangers, the completion rate of risk investigation plan, the number of newly added major hidden dangers, the number of newly added general hidden dangers, the completion rate of hidden danger rectification, the timely rate of hidden danger rectification, the number of newly added internal employee violations and the number of newly added external personnel penalty orders. Major hidden dangers and general hidden dangers may be hidden danger levels divided according to the degree of hidden dangers.
[0073] The safety impact parameters included in the LDAR category may include, but are not limited to: the number of remaining dynamic seal point leakage points, the number of remaining static seal point leakage points, the number of dynamic seal points, and the number of static seal points.
[0074] The safety impact parameters included in the equipment operation category may include, but are not limited to: equipment failure rate, inspection and maintenance plan completion rate, inspection and maintenance plan accuracy rate, equipment integrity rate, and equipment operation rate.
[0075] The safety influencing parameters included in other factor categories may be, but are not limited to: the number of existing internal employees who have failed safety education and training, the number of existing internal employees with expired qualification certificates, wind speed, temperature, weather and humidity.
[0076] The safety impact data may be safety impact parameter values after preprocessing. The preprocessing may include but is not limited to: data normalization and data cleaning.
[0077] It should be noted that obtaining each safety impact data corresponding to each safety impact parameter related to production safety in each target category can facilitate subsequent production safety warning.
[0078] As an example, this step may include the following steps:
[0079] In the first step, each security impact parameter value corresponding to each security impact parameter in each target category is obtained.
[0080] The safety impact parameter value corresponding to the safety impact parameter may be time series data.
[0081] For example, each different type of safety impact parameter value at each preset moment within a preset time period can be collected from the database. The preset time period can be a preset time period. For example, the duration corresponding to the preset time period can be 1 month. The end time of the preset time period can be the current time, which can facilitate the subsequent implementation of a graded warning for safety production at the current moment. The preset moment can be a preset moment. The duration between adjacent preset moments can be 1 hour.
[0082] For example, taking temperature as an example of the safety impact parameter, a temperature sensor may be used to collect a temperature value every hour within a preset time period as a safety impact parameter value corresponding to the safety impact parameter.
[0083] In the second step, all safety impact parameter values corresponding to each safety impact parameter in each target category are normalized and cleaned, and each safety impact parameter value after data normalization and data cleaning is used as safety impact data.
[0084] It should be noted that the richer the safety impact data collected, the easier it is to conduct graded early warnings for production safety based on big data. Big data, or massive data, refers to the amount of data involved that is so large that it cannot be detected by mainstream software tools. Secondly, the production safety risks of refining and chemical enterprises often include the production safety risks of hazardous chemicals. The monitoring and early warning system for production safety risks of hazardous chemicals often accesses the real-time monitoring data and video surveillance data of the enterprise, and relies on the basic data of various safety risks within the enterprise. Through information and intelligent means, it realizes dynamic early warning, risk distribution, online inspection, safety commitment and other functions, providing support for comprehensive analysis, risk prevention, dynamic assessment of risk situation, and accident emergency response.
[0085] Step S2, performing principal component analysis on all safety impact data corresponding to all safety impact parameters in each target category to obtain a preset number of principal components corresponding to each target category.
[0086] In some embodiments, principal component analysis may be performed based on all safety impact data corresponding to all safety impact parameters in each target category to obtain a preset number of principal components corresponding to each target category.
[0087] The preset number may be a preset number, for example, 3.
[0088] It should be noted that by performing principal component analysis on all safety impact data corresponding to all safety impact parameters in each target category, dimensionality reduction can be achieved, which can facilitate subsequent accurate processing of each principal component.
[0089] As an example, this step may include the following steps:
[0090] In the first step, all safety impact data corresponding to each safety impact parameter are subjected to mean-variance normalization so that the mean of the processed data is 0 and the variance is 1, so that each safety impact parameter is on the same order of magnitude, thereby avoiding the dominance of the contribution of certain safety impact parameters to the principal component.
[0091] In the second step, based on the mean-variance standardized data in each target category, principal component analysis (PCA) is used to determine a preset number of principal components, where the principal components are also called main eigenvectors, which can characterize the safety impact parameters that have a relatively large impact on the safe production of refining enterprises.
[0092] Step S3, determining the projection values of all safety impact parameters under each principal component corresponding to each target category according to the preset safety weights corresponding to all safety impact parameters in each target category.
[0093] In some embodiments, the projection values of all safety influencing parameters under each principal component corresponding to each target category may be determined according to preset safety weights corresponding to all safety influencing parameters in each target category.
[0094] The preset safety weight may be a weight of the impact of a preset safety impact parameter on safe production. For example, the preset safety weight may be a weight assigned based on experience.
[0095] It should be noted that since each safety impact parameter often has an impact on safe production, the preset safety weights corresponding to all safety impact parameters in each target category are comprehensively considered to quantify the projection values of all safety impact parameters under each principal component corresponding to each target category.
[0096] As an example, this step may include the following steps:
[0097] In the first step, based on all the safety impact data corresponding to all the safety impact parameters in the above target categories, an original matrix corresponding to the above target categories is constructed.
[0098] Each row of the original matrix may be composed of all safety impact data corresponding to the same safety impact parameter. The number of rows of the original matrix may be equal to the number of safety impact parameters in the target category. Each column of the original matrix may be composed of all safety impact data collected at the same collection time. The number of columns of the original matrix may be equal to the number of safety impact data at the same collection time.
[0099] The second step is to construct a target diagonal matrix corresponding to the above target categories based on the preset safety weights corresponding to all safety impact parameters in the above target categories.
[0100] The diagonal elements in the target diagonal matrix may be preset safety weights corresponding to the safety impact parameters. The target diagonal matrix may be a diagonal matrix whose diagonal elements are preset safety weights corresponding to the safety impact parameters. The diagonal elements may be elements on the main diagonal line of the diagonal matrix.
[0101] The third step is to construct a principal component matrix corresponding to the above target category according to the preset number of principal components corresponding to the above target category.
[0102] In the fourth step, the product of the original matrix, the target diagonal matrix and the principal component matrix is determined as the projection value matrix corresponding to the target category.
[0103] The elements in the projection value matrix may be projection values of the safety impact parameters under the principal components.
[0104] For example, the projection value matrix corresponding to the target category can be: i =X i ×Q i ×P i ; Among them, Y i X is the projection value matrix corresponding to the i-th target category. i is the original matrix corresponding to the i-th target category. Q i is the target diagonal matrix corresponding to the i-th target category. i is the principal component matrix corresponding to the i-th target category. i is the serial number of the target category. i The element in the jth row and tth column in Y can be the projection value of the jth safety impact parameter under the tth principal component corresponding to the i-th target category. i or the serial number of the safety impact parameter under the tth principal component. i The column number of or the ordinal number of the principal component corresponding to the i-th target category.
[0105] It should be noted that, since the principal component analysis selects a preset number of principal components in each target category as the representative parameters of the target category, and the remaining safety impact parameters have a relatively small impact on the evaluation value of the entire target category, they will also have an impact on safe production. Therefore, the impact of other safety impact parameters can be transferred to the preset number of principal components. In addition, in today's safe production classification and early warning system, each safety impact parameter often has a weight assigned based on empirical values, that is, the corresponding preset safety weight. This weight characterizes the severity of the safety impact parameter, that is, the degree of impact on safe production. Therefore, when projecting, the corresponding preset safety weight can be used to perform weight assignment operations on the original matrix, so that the projection value obtained after projection can fully characterize the situation of all safety impact parameters to the greatest extent.
[0106] Step S4, determining the impact change index corresponding to each principal component according to all the safety impact data under each principal component and the projection values of all the safety impact parameters under each principal component.
[0107] In some embodiments, the impact change index corresponding to each principal component may be determined based on all safety impact data under each principal component and the projection values of all safety impact parameters under each principal component.
[0108] It should be noted that the greater the data change before and after projection under the principal component, the higher the weight that the principal component should be given. Therefore, the greater the influence change index corresponding to the principal component, the higher the weight that the principal component should be given.
[0109] As an example, this step may include the following steps:
[0110] In the first step, the mean of all safety impact data under the above principal components is determined as the original impact index corresponding to the above principal components.
[0111] In the second step, the mean of the projection values of all safety impact parameters under the above principal components is determined as the projection impact index corresponding to the above principal components.
[0112] In the third step, the absolute value of the difference between the original influence index and the projected influence index is determined as the influence difference index corresponding to the principal component.
[0113] The fourth step is to determine the impact change index corresponding to the above principal components based on the above impact difference index.
[0114] Among them, the above-mentioned impact difference index can be positively correlated with the above-mentioned impact change index.
[0115] For example, the formula for determining the impact change index corresponding to the main component can be: Among them, B it is the impact change index corresponding to the tth principal component corresponding to the ith target category. it1 is the original impact index corresponding to the tth principal component of the ith target category, that is, the mean of all safety impact data under the tth principal component. it2 is the projection impact index corresponding to the tth principal component corresponding to the i-th target category, that is, the mean of the projection values of all safety impact parameters under the tth principal component. || is the absolute value function. |x it1 -x it2 | is the impact difference index corresponding to the tth principal component corresponding to the i-th target category. max() is the maximum value function. R it1 is the total distance of all safety impact data under the tth principal component. it2 It is the full range of the projection values of all safety impact parameters under the tth principal component. γ1 is a pre-set factor greater than 0, which is mainly used to prevent the denominator from being 0. For example, γ1 can be 0.01.
[0116] It should be noted that the mean difference of the data before and after projection |x it1 -x it2 | can represent the amount of other data that is mapped to the tth principal component after the data is projected. The more data is projected on the tth principal component, the greater the mean difference is, and the more data is projected, the higher the weight of the tth principal component should be. The full range is used to measure the influencing change index with the full range, which can make the entire influencing change index related to the data change range of the projected data. The proportion of the mean difference within the change range is B it , which can more clearly reflect the specific size of the change and make the subsequent weight allocation more convenient.
[0117] Step S5, determining the target impact weight corresponding to each principal component according to the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, and the impact change index corresponding to each principal component.
[0118] In some embodiments, the target impact weight corresponding to each principal component can be determined based on the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, and the impact change index corresponding to each principal component.
[0119] It should be noted that by comprehensively considering the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, as well as the impact change index corresponding to each principal component, the target impact weight corresponding to each principal component can be quantified.
[0120] As an example, the formula for determining the target impact weight corresponding to each principal component can be:
[0121] Among them, QW it WL is the target influence weight corresponding to the tth principal component corresponding to the i-th target category. i is the cumulative value of the preset safety weights corresponding to all safety impact parameters in the i-th target category. it is the preset safety weight corresponding to the tth principal component corresponding to the ith target category. min() is the minimum value function. B it A is the impact change index corresponding to the tth principal component corresponding to the i-th target category. i is the cumulative value of the pseudo weights corresponding to all principal components corresponding to the i-th target category. N is the preset number. i is the serial number of the target category. t is the serial number of the principal component corresponding to the i-th target category.
[0122] It should be noted that the impact change index is used to modify the preset safety weight corresponding to the safety impact parameter. itIt represents not only the influence of the t-th principal component itself on safety production, but also the influence of the remaining safety influencing parameters removed by dimensionality reduction on safety production, that is, QW it It can represent the overall impact of the projection on the tth principal component, so its pseudo weight is often greater than the original preset safety weight, so it can be used represents the pseudo weight. When WL i When t is larger, it often means that the cumulative value of the preset safety weight of the target category to which the tth principal component belongs is larger, which often means that the empirical weight of the target category to which the tth principal component belongs is larger, which often means that the target category to which the tth principal component belongs has a greater impact on safe production, and which often means that the weight of the tth principal component may be relatively larger. When the value is larger, it often means that the pseudo weight corresponding to the t-th principal component accounts for a larger proportion of the total pseudo weight corresponding to the i-th target category, which often means that the influence of the t-th principal component on production safety is greater, and the weight of the t-th principal component may be relatively larger. In order to prevent the denominator from being 0, the denominator included in all ratios in the embodiment can be added with a factor greater than 0, such as 0.01.
[0123] Step S6, determining the initial membership degree corresponding to each principal component according to all the safety impact data under each principal component.
[0124] In some embodiments, the initial membership degree corresponding to each principal component may be determined based on all safety impact data under each principal component.
[0125] It should be noted that by comprehensively considering all the safety impact data under each principal component, the initial membership degree corresponding to each principal component can be quantified.
[0126] As an example, this step may include the following steps:
[0127] The first step is to construct the target spectrum graph corresponding to the above principal components based on all the safety impact data under the above principal components.
[0128] For example, based on all the safety impact data under the principal component, the Fourier transform can be used to convert the time domain data into a representation in the frequency domain, and then the result of the Fourier transform can be used to calculate the phase and amplitude of the data in different frequency domains, thereby drawing a spectrum diagram as the target spectrum diagram corresponding to the principal component.
[0129] In the second step, according to the above target spectrum, the formula corresponding to the periodic index corresponding to the above main component can be determined as follows:
[0130] Among them, SL itis the periodicity index corresponding to the tth principal component corresponding to the ith target category. exp() is an exponential function with a natural constant as the base. it is the maximum amplitude in the target spectrum corresponding to the tth principal component of the ith target category. it It is the weighted average of the frequencies of all peaks in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the i-th target category. The frequency of the peak is the number of peaks with the same value. it is the length of the period of maximum amplitude in the target spectrum corresponding to the tth principal component of the i-th target category. N it is the number of harmonics in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the i-th target category. || is the absolute value function. a is a preset adjustment parameter greater than 0, which can make the function value range more in line with the actual situation. For example, a can be 1. H it is the maximum peak value of the harmonic in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the ith target category. i is the serial number of the target category. t is the serial number of the principal component corresponding to the ith target category.
[0131] It should be noted that when C it When D is larger, it often means that the length of the maximum amplitude period in the target spectrum corresponding to the tth principal component of the i-th target category is larger, which often means that the main period in the data is longer, which often means that the possibility of periodicity in the data is lower. Among them, the main period is the period of the maximum amplitude in the target spectrum. it When N is larger, it often means that the amplitude is larger, which often means that the periodicity intensity of the data on the main cycle is higher. Harmonics represent cycles that are integer multiples of the main cycle. More harmonics often mean that the data contains more repetitive patterns of cycles. The more harmonics there are, the more obvious the periodicity of the data may be. Therefore, when N is larger, it often means that the periodicity of the data may be higher. it When H is larger, it often indicates that the periodicity of the data may be more obvious. it When it is larger, it often means that the harmonic peak value is larger, which often means that the periodicity intensity of the data in the harmonic period is higher. it It can represent the distribution characteristics of periodicity in data in different frequency bands. it The larger the value, the more balanced the periodicity distribution of the data and the more obvious the periodicity. it -N it | represents the difference between the main period and the number of harmonics. The larger its value is, the more complex the periodicity of the data is, that is, the closer it is to a state of no periodicity.
[0132] In the third step, based on the above periodic indicators and all the safety impact data under the above principal components, the formula corresponding to the initial membership degree corresponding to the above principal components can be determined as follows:
[0133] Among them, μ it is the initial membership of the tth principal component corresponding to the ith target category. it is the number of safety impact data under the tth principal component corresponding to the i-th target category. itb is the membership index corresponding to the bth safety impact data under the tth principal component corresponding to the i-th target category. it is the periodicity index corresponding to the tth principal component corresponding to the ith target category. norm() is the normalization function. is the Euler function. x itb is the bth safety impact data under the tth principal component corresponding to the i-th target category. it1 is the mean of all safety impact data under the tth principal component corresponding to the i-th target category. it is the variance of all safety impact data under the t-th principal component corresponding to the i-th target category. i is the serial number of the target category. t is the serial number of the principal component corresponding to the i-th target category. b is the serial number of the safety impact data under the t-th principal component.
[0134] It should be noted that when SL it When x is larger, it often indicates that the periodicity intensity of the data in the main period is higher, and the corresponding membership is stronger. itb <x it1 hour, The larger the value, the stronger the membership of the tth principal component corresponding to the ith target category. itb ≥x it1 hour, The larger the value, the weaker the membership of the tth principal component corresponding to the i-th target category. it The larger it is, the stronger the membership of the tth principal component corresponding to the ith target category is.
[0135] Step S7, determining the target membership corresponding to each target category according to the target influence weights and initial memberships corresponding to the preset number of principal components corresponding to each target category.
[0136] In some embodiments, the target membership corresponding to each target category may be determined based on the target influence weights and initial memberships corresponding to a preset number of principal components corresponding to each target category.
[0137] It should be noted that, by comprehensively considering the target influence weights and initial memberships corresponding to a preset number of principal components corresponding to each target category, the accuracy of determining the target membership corresponding to each target category can be improved.
[0138] As an example, this step may include the following steps:
[0139] In the first step, the mean of the target influence weights corresponding to a preset number of principal components corresponding to the above target categories is determined as the reference weight corresponding to the above target categories.
[0140] In the second step, the product of the target influence weight corresponding to each principal component and the initial membership is determined as the first modified membership corresponding to each principal component.
[0141] In the third step, the accumulated value of the first modified membership corresponding to the preset number of principal components corresponding to the target category is determined as the second modified membership corresponding to the target category.
[0142] The fourth step is to determine the target membership corresponding to the target category by taking the ratio of the second modified membership to the reference weight.
[0143] For example, the formula for determining the target membership corresponding to the target category can be: Among them, U i is the target membership corresponding to the i-th target category. N is the preset number. QW it is the target influence weight corresponding to the tth principal component corresponding to the ith target category. it is the initial membership of the tth principal component corresponding to the ith target category. i is the serial number of the target category. t is the serial number of the principal component corresponding to the ith target category. It is the reference weight corresponding to the tth principal component corresponding to the ith target category, that is, the mean of the target influence weights corresponding to the preset number of principal components corresponding to the ith target category. QW it ×μ it It is the first modified membership corresponding to the tth principal component corresponding to the ith target category. is the second modified membership corresponding to the i-th target category.
[0144] It should be noted that when μ it When it is larger, it often means that the membership of the tth principal component corresponding to the i-th target category is stronger. it It not only represents the influence of the t-th principal component itself on safety production, but also represents the influence of the remaining safety influencing parameters removed by dimensionality reduction on safety production, that is, QW it It can represent the overall impact of the projection on the tth principal component. i The larger it is, the stronger the membership of the tth principal component corresponding to the ith target category is.
[0145] Step S8, determining the safety warning value according to the target membership corresponding to all target categories and all preset safety weights, and performing safety graded warning according to the safety warning value.
[0146] In some embodiments, a safety warning value may be determined based on the target memberships corresponding to all target categories and all preset safety weights, and a safety graded warning may be performed based on the safety warning value.
[0147] It should be noted that, through the quantified target impact weight and initial membership, the quantified target membership and the preset safety weight, the safety warning value is quantified, and the safety graded warning is carried out based on the safety warning value, thereby realizing the safety graded warning.
[0148] As an example, this step may include the following steps:
[0149] In the first step, the cumulative value of the preset safety weights corresponding to all safety impact parameters in each target category is determined as the overall safety weight corresponding to the above target category.
[0150] In the second step, the product of the overall safety weight corresponding to each target category and the target membership degree is determined as the early warning indicator corresponding to the above target category.
[0151] The third step is to determine the safety warning value based on the warning indicators corresponding to all target categories.
[0152] Among them, the early warning indicators can be positively correlated with the safety warning values.
[0153] For example, the formula for determining the safety warning value may be: Where SA is the safety warning value. norm() is the normalization function. G is the number of target categories. U i is the target membership corresponding to the i-th target category. i It is the cumulative value of the preset safety weights corresponding to all safety impact parameters in the i-th target category, that is, the overall safety weight corresponding to the i-th target category. i ×U i is the warning indicator corresponding to the i-th target category. i is the sequence number of the target category.
[0154] It should be noted that when U i When WL is larger, it often means that the membership degree of the tth principal component corresponding to the i-th target category is stronger, which often means that serious problems are more likely to occur and the degree of danger of production safety is greater. iWhen SA is larger, it often means that the experience weight of the i-th target category is larger, the impact of the i-th target category on production safety is greater, the possibility of serious problems, and the greater the risk of production safety. Therefore, when SA is larger, it often means that the risk of production safety is greater.
[0155] The fourth step is to determine that there is a low risk in production safety when the above safety warning value is less than or equal to the first preset safety threshold.
[0156] The first preset safety threshold may be a preset threshold, for example, the first preset safety threshold may be 0.8.
[0157] The fifth step is to determine that there is a general risk in production safety when the above safety warning value is greater than the first preset safety threshold and less than or equal to the second preset safety threshold.
[0158] The second preset safety threshold may be a preset threshold. The second preset safety threshold may be greater than the first preset safety threshold. For example, the second preset safety threshold may be 0.875.
[0159] Step 6: When the above safety warning value is greater than the second preset safety threshold and less than or equal to the third preset safety threshold, it is determined that there is a greater risk in production safety.
[0160] The third preset safety threshold may be a preset threshold. The third preset safety threshold may be greater than the second preset safety threshold. For example, the third preset safety threshold may be 0.95.
[0161] Step 7: When the above safety warning value is greater than the third preset safety threshold, it is determined that there is a high risk in production safety.
[0162] In summary, safety graded warning is carried out based on safety warning values, and safety graded warning is realized. Compared with subjective static threshold warning, the present invention improves the accuracy of safety warning value quantification by performing real-time analysis on safety impact data and preset safety weights, thereby improving the accuracy of safety graded warning. Secondly, the present invention quantifies multiple indicators related to production safety, such as projection value, impact change index, target impact weight, initial membership, target membership and safety warning value, etc., which can make subsequent safety graded warning results relatively more objective, thereby improving the accuracy of safety graded warning.
[0163] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A production safety graded early warning method based on big data, characterized in that: The following steps are involved: Obtain each safety impact data corresponding to each safety impact parameter in each target category; Performing principal component analysis on all safety impact data corresponding to all safety impact parameters in each target category to obtain a preset number of principal components corresponding to each target category; Determine the projection values of all safety impact parameters under each principal component corresponding to each target category based on the preset safety weights corresponding to all safety impact parameters in each target category; Determine the impact change index corresponding to each principal component based on all safety impact data under each principal component and the projection values of all safety impact parameters under each principal component; Determine the target impact weight corresponding to each principal component based on the preset safety weights corresponding to all safety impact parameters in the target category to which each principal component belongs, and the impact change index corresponding to each principal component; Determine the initial membership degree of each principal component based on all safety impact data under each principal component; Determine the target membership corresponding to each target category based on the target influence weights and initial memberships corresponding to the preset number of principal components corresponding to each target category; Determine the security warning value based on the target membership corresponding to all target categories and all preset security weights, and perform security graded warning based on the security warning value; The step of determining the projection values of all safety impact parameters under each principal component corresponding to each target category according to the preset safety weights corresponding to all safety impact parameters in each target category includes: Constructing an original matrix corresponding to the target category based on all safety impact data corresponding to all safety impact parameters in the target category, wherein each row of the original matrix is composed of all safety impact data corresponding to the same safety impact parameter, and each column of the original matrix is composed of all safety impact data collected at the same collection time; Constructing a target diagonal matrix corresponding to the target category based on the preset safety weights corresponding to all safety impact parameters in the target category, wherein the diagonal elements in the target diagonal matrix are the preset safety weights corresponding to the safety impact parameters; Constructing a principal component matrix corresponding to the target category according to a preset number of principal components corresponding to the target category; Determine the product of the original matrix, the target diagonal matrix and the principal component matrix as the projection value matrix corresponding to the target category, wherein the elements in the projection value matrix are projection values of the safety impact parameters under the principal components; The formula corresponding to the target influence weight of the principal component is: Among them, QW it is the target influence weight corresponding to the tth principal component of the i-th target category; WL i is the cumulative value of the preset safety weights corresponding to all safety impact parameters in the i-th target category; w it is the preset security weight corresponding to the tth principal component corresponding to the i-th target category; min() is the minimum value function; B it is the impact change index corresponding to the tth principal component corresponding to the i-th target category; A i is the cumulative value of the pseudo weights corresponding to all principal components corresponding to the i-th target category; N is the preset number; i is the serial number of the target category; t is the serial number of the principal component corresponding to the i-th target category; Determining the initial membership degree corresponding to each principal component based on all safety impact data under each principal component includes: Constructing a target spectrum graph corresponding to the principal component according to all safety impact data under the principal component; Determining a periodicity index corresponding to the principal component according to the target spectrum; Determining an initial degree of membership corresponding to the principal component based on the periodic indicator and all safety impact data under the principal component; The formula for the periodic indicator corresponding to the principal component is: Among them, SL it is the periodicity index corresponding to the tth principal component corresponding to the i-th target category; exp() is the exponential function with a natural constant as the base; D it is the maximum amplitude in the target spectrum corresponding to the tth principal component of the i-th target category; F it is the weighted average of the frequencies of all peaks in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the i-th target category; C it is the length of the period of maximum amplitude in the target spectrum corresponding to the tth principal component of the i-th target category; N it is the number of harmonics in the period of maximum amplitude in the target spectrum corresponding to the tth principal component corresponding to the i-th target category; || is the absolute value function; a is a preset adjustment parameter greater than 0; H it is the maximum peak value of the harmonic in the period of maximum amplitude in the target spectrum corresponding to the t-th principal component corresponding to the i-th target category; i is the serial number of the target category; t is the serial number of the principal component corresponding to the i-th target category; The formula corresponding to the initial membership degree of the principal component is: Among them, μ it is the initial membership of the tth principal component corresponding to the i-th target category; n it is the number of safety impact data under the tth principal component corresponding to the i-th target category; μ itb is the membership index corresponding to the bth safety impact data under the tth principal component corresponding to the i-th target category; SL it is the periodicity index corresponding to the tth principal component corresponding to the i-th target category; norm() is the normalization function; is the Euler function; x itb is the bth safety impact data under the tth principal component corresponding to the i-th target category; x it1 is the mean of all safety impact data under the tth principal component corresponding to the i-th target category; σ it is the variance of all safety impact data under the t-th principal component corresponding to the i-th target category; i is the serial number of the target category; t is the serial number of the principal component corresponding to the i-th target category; b is the serial number of the safety impact data under the t-th principal component.
2. The production safety graded early warning method based on big data according to claim 1 is characterized in that: Determining the impact change index corresponding to each principal component based on all safety impact data under each principal component and the projection values of all safety impact parameters under each principal component includes: Determine the mean of all safety impact data under the principal component as the original impact index corresponding to the principal component; Determine the mean of the projection values of all safety impact parameters under the principal component as the projection impact index corresponding to the principal component; Determine the absolute value of the difference between the original influence index and the projected influence index as the influence difference index corresponding to the principal component; An influence change index corresponding to the principal component is determined according to the influence difference index, wherein the influence difference index is positively correlated with the influence change index.
3. The production safety graded early warning method based on big data according to claim 1 is characterized in that: The determining of the target membership corresponding to each target category according to the target influence weights and initial memberships corresponding to the preset number of principal components corresponding to each target category includes: Determine the mean of the target influence weights corresponding to a preset number of principal components corresponding to the target category as the reference weight corresponding to the target category; The product of the target influence weight corresponding to each principal component and the initial membership is determined as the first modified membership corresponding to each principal component; Determining the accumulated values of the first modified memberships corresponding to the preset number of principal components corresponding to the target category as the second modified membership corresponding to the target category; The ratio of the second modified membership to the reference weight is determined as the target membership corresponding to the target category.
4. The production safety graded early warning method based on big data according to claim 1 is characterized in that: Determining the safety warning value based on the target membership corresponding to all target categories and all preset safety weights includes: Determine the cumulative value of the preset safety weights corresponding to all safety impact parameters in each target category as the overall safety weight corresponding to the target category; The product of the overall safety weight corresponding to each target category and the target membership degree is determined as the early warning indicator corresponding to the target category; The safety warning value is determined based on the warning indicators corresponding to all target categories, among which the warning indicators are positively correlated with the safety warning value.
5. The production safety graded early warning method based on big data according to claim 1 is characterized in that: The performing of a security graded warning according to the security warning value includes: When the safety warning value is less than or equal to the first preset safety threshold, the production safety is determined to be low risk; When the safety warning value is greater than a first preset safety threshold and less than or equal to a second preset safety threshold, the production safety is determined to be a general risk, wherein the second preset safety threshold is greater than the first preset safety threshold; When the safety warning value is greater than the second preset safety threshold and less than or equal to the third preset safety threshold, it is determined that the production safety is at a greater risk, wherein the third preset safety threshold is greater than the second preset safety threshold; When the safety warning value is greater than a third preset safety threshold, the production safety is determined to be at high risk.
