Multi-source data fusion driven security risk management and control method and system

Abstract

The invention is applied to the field of risk management and control, and provides a security risk management and control method and system driven by multi-source data fusion, and the method comprises the steps: collecting the multi-source data of each unit, and carrying out the preprocessing of the collected multi-source data; the method comprises the following steps: preprocessing multi-source data, standardizing the preprocessed multi-source data to obtain standardized data, setting sliding windows for the standardized data, independently calculating an index weight corresponding to each index for the data in each sliding window, and dynamically updating the index weight to realize adaptive adjustment of the index weight based on a time sequence; performing dynamic risk assessment on each unit according to the standardized data of each unit and the dynamically updated index weight to obtain a risk score of each unit, and obtaining a safety decision corresponding to the corresponding unit; the dynamic updating of the weight and the real-time adjustment of the risk score are realized by applying the time sequence sliding window algorithm, and compared with a static weight method, the method has the advantage that the real-time performance and the accuracy of risk assessment are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of risk management and control, and more particularly relates to a multi-source data fusion driven safety risk management and control method and system. BACKGROUND

[0002] Current industrial and trade enterprises still face problems such as strong risk stubbornness and dynamic risk response lag in safety production management, and need to further break through the limitations of the conventional control mode. The existing mainstream safety risk management and control methods can be summarized into the following three categories: (1) Traditional offline safety management and control method. Enterprises rely on manual work to carry out hidden danger investigation, safety education and training, and on-site supervision, and the control efficiency is highly dependent on the personnel scale and professional quality, and the results lack quantitative feedback. For example, after recording hidden dangers on paper checklists, it is difficult to systematically analyze risk trends, and the rectification effect cannot form data support for continuous optimization.

[0003] (2) Primary electronic control method. Enterprises have initially introduced electronic tools and single-function systems to replace paper records, realizing the digital storage of part of the data. This semi-digital control method is insufficient in data sorting and gathering, data analysis, and data early warning.

[0004] (3) Information-based safety management and control method. Although enterprises have established functional modules (such as hidden danger investigation systems and education and training platforms), the data between systems are isolated, and are mostly stored as electronic archives, lacking functional synergy and having poor correlation between different types of data. When performing risk assessment, it is not possible to comprehensively assess various types of data.

[0005] Therefore, it is an urgent problem in the technical field to overcome the defects of the prior art. SUMMARY

[0006] The present application needs to solve the problem of how to comprehensively correlate and assess various types of data to improve the real-time and accuracy of risk assessment. The time series sliding window algorithm is introduced to solve the problem of insufficient real-time caused by static weights in the prior art.

[0007] In a first aspect, a multi-source data fusion driven safety risk management and control method is provided, comprising: Collecting multi-source data of each unit and preprocessing the collected multi-source data; Standardizing the preprocessed multi-source data to obtain standardized data, setting a sliding window for the standardized data, independently calculating the index weight corresponding to each index for the data in each sliding window, and dynamically updating the index weight, thereby realizing adaptive adjustment of the index weight based on time series; According to the standardized data of each unit and the dynamically updated index weight, dynamic risk assessment is performed on each unit to obtain a risk score of each unit. According to the risk score of the corresponding unit, a safety decision corresponding to the corresponding unit is obtained.

[0008] Preferably, the setting of the sliding window for the standardized data comprises: independently calculating the index weight corresponding to each index for the data in each sliding window, and dynamically updating the index weight, so as to realize adaptive adjustment of the index weight based on time series, which specifically comprises: Setting a sliding window size S and a sliding step L; Real-time monitoring of data flow, when new data is collected, the sliding window moves with a step L, and the index weight corresponding to each index in the current sliding window is calculated; According to the information entropy change rate of the data in the current sliding window, S and L are dynamically adjusted: according to the size of the information entropy change rate, the sliding window size S is adjusted; The expression of the index weight is:

[0009] Wherein, Wj(t) is the weight of the jth index at time t, a is the learning rate, and ΔEj is the information entropy change of the jth index.

[0010] Preferably, the dynamic risk assessment of each unit according to the standardized data of each unit and the dynamic weight comprises: According to the standardized data of each unit, the index proportion corresponding to each index in the corresponding unit is obtained, the information entropy of the corresponding index is obtained according to the index proportion corresponding to each index, and then the normalized weight corresponding to each index is obtained; When the sliding window slides, the index weight is recalculated according to the standardized data in the new sliding window to realize dynamic updating, and the risk score corresponding to each unit is obtained; For each sliding window, the expression of the index proportion is: ; The expression of the information entropy is: ; The expression of the normalized weight is: ; Wherein, is the index proportion of the jth index of the ith unit, is the standardized data of the jth index of the ith unit, and m is the number of all units, is the information entropy of the jth index, and K=1 / , is an index weight of the jth index in the initial sliding window, is an information utility value of the jth index, =1- .

[0011] Preferably, the pre-processed multi-source data is standardized to obtain standardized data, specifically including: obtaining positive type data and negative type data in the pre-processed multi-source data; standardizing the positive type data, the expression being: ; standardizing the negative type data, the expression being: ; wherein, is standardized data of the jth index of the ith unit, is the maximum data in the jth index, is the minimum data in the jth index.

[0012] Preferably, the multi-source data of each unit is collected, and the collected multi-source data is pre-processed, specifically including: collecting real-time collected sensor data, personnel training data and historical hidden danger data as the multi-source data; performing data review on the multi-source data, performing data format unification on the multi-source data, obtaining invalid data, abnormal data and missing data in the multi-source data; deleting the invalid data, filling the missing data, and correcting the abnormal data.

[0013] Preferably, the safety decision corresponding to the corresponding unit is obtained according to the risk score of the corresponding unit, specifically including: storing all safety decision data; storing the risk score data of the corresponding unit, triggering a hierarchical response mechanism when the risk score data is warehoused, determining the warning level triggered by the corresponding unit according to the size of the risk score of the corresponding unit; calling corresponding safety decision data according to the warning level of the corresponding unit.

[0014] Preferably, the multi-source data fusion driven safety risk control method further includes: obtaining a whole risk score according to the risk scores of all units, and calling corresponding whole safety decision according to the whole risk score.

[0015] In a second aspect, a multi-source data fusion driven safety risk management and control system is provided for implementing the multi-source data fusion driven safety risk management and control method, comprising a data acquisition module, a data preprocessing module, a multi-source data dynamic evaluation module, and a safety decision module, wherein: The data acquisition module is configured to collect multi-source data of each unit, and the data preprocessing module is configured to preprocess the collected multi-source data. The multi-source data dynamic evaluation module is configured to standardize the preprocessed multi-source data to obtain standardized data, and dynamically evaluate the risk of each unit according to the standardized data of each unit to obtain a risk score of each unit. The safety decision module is configured to obtain a corresponding safety decision of the corresponding unit according to the risk score of the corresponding unit.

[0016] Preferably, the data acquisition module comprises a monitoring data acquisition unit, a training data acquisition unit, and a hidden danger data acquisition unit, wherein: The monitoring data acquisition unit is configured to collect real-time sensing data, the training data acquisition unit is configured to collect personnel training data, and the hidden danger data acquisition unit is configured to collect historical hidden danger data. The collected sensing data, personnel training data, and historical hidden danger data are used as the multi-source data.

[0017] Preferably, the safety decision module comprises a safety decision resource library unit, a data storage unit, and an early warning disposal unit, wherein: The safety decision resource library unit is configured to preset all safety decision data. The data storage unit is configured to receive and store risk score data of all units from the multi-source data dynamic evaluation module, and to receive and store all safety decision data from the safety decision resource library unit. The early warning disposal unit is configured to trigger a hierarchical response mechanism when the risk score data is stored in the data storage unit, to determine the early warning level triggered by the corresponding unit according to the risk score of the corresponding unit, and to call the corresponding safety decision data according to the risk score of the corresponding unit.

[0018] Compared with the prior art, the present application has at least the following advantages: By the above method, the multi-dimensional data of multiple units are integrated, the comprehensive dynamic evaluation is carried out according to the multi-dimensional data, the risk score corresponding to each unit is obtained, and the safety decision corresponding to different risk scores is preset. When each unit triggers a risk warning, the corresponding safety decision is obtained according to the different risk scores. Compared with the existing periodic manual evaluation, the method improves the real-time performance and accuracy of the overall risk assessment. Moreover, the time series sliding window algorithm is applied to realize dynamic updating of the weight and real-time adjustment of the risk score. Compared with the static weight method, the method significantly improves the real-time performance and accuracy of the risk assessment. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly introduced as follows. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0020] Figure 1 is a method flowchart of a multi-source data fusion driven safety risk management and control method provided by the present embodiment; Figure 2 is a method flowchart of data preprocessing in a multi-source data fusion driven safety risk management and control method provided by the present embodiment; Figure 3 is a method flowchart of risk score acquisition in a multi-source data fusion driven safety risk management and control method provided by the present embodiment; Figure 4 is a method flowchart of acquisition of each index weight in a multi-source data fusion driven safety risk management and control method provided by the present embodiment; Figure 5 is a method flowchart of acquisition of safety decisions corresponding to each index in a multi-source data fusion driven safety risk management and control method provided by the present embodiment; Figure 6 is a schematic diagram of a multi-source data fusion driven safety risk management and control system provided by the present embodiment; Figure 7 is a schematic diagram of another multi-source data fusion driven safety risk management and control system provided by the present embodiment; Figure 8 is a risk assessment index system schematic diagram of a multi-source data fusion driven safety risk management and control system provided by the present embodiment. DETAILED DESCRIPTION

[0021] In order to make the objects, technical solutions and advantages of the present application clearer, the following further describes the present application with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the present application and not to limit the present application.

[0022] Unless otherwise required by context, the term "comprises" in the specification and claims is to be construed as open-ended, i.e., to the effect that "comprises but does not limited to". In the description of the specification, the terms "one embodiment", "some embodiments", "exemplary embodiment", "example", "specific example" or "some examples" are intended to mean that a specific feature, structure, material or characteristic is included in at least one embodiment or example of the present disclosure. The illustrative representations of the above terms do not necessarily refer to the same embodiment or example. In addition, the specific features, structures, materials or characteristics described can be included in any suitable manner in any one or more embodiments or examples, i.e., although they are carried by the embodiments or examples of the above terms due to the order of appearance and location, they are not limited to being carried by one embodiment or example in a combined manner.

[0023] In the description of the present application, the terms "first", "second" are only used for description purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features limited by "first", "second" can be explicitly or implicitly included in one or more features. In the description of the embodiments of the present disclosure, unless otherwise specified, the meaning of "multiple" is two or more. In addition, for example, in the description, the same type of nouns can also be described as two independent individuals by adding "A", "B" at the end, in which case the features limited by "A", "B" are only used for the purpose of distinguishing the same type of individual description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features.

[0024] In the description of the present application, the expression "A and / or B" (where A and B represent specific feature content) can be used, and the corresponding expression includes the following three combinations: only A, only B, and a combination of A and B.

[0025] In the present application, "about", "approximately" or "approximately" includes the value stated and the average value within an acceptable deviation range of the specific value, wherein the acceptable deviation range is determined by considering the measurement being discussed and the error related to the measurement of the specific quantity (i.e., the limitations of the measurement system) by a person of ordinary skill in the art.

[0026] In addition, the technical features involved in each embodiment of the application described below can be combined with each other as long as they do not conflict with each other.

[0027] Embodiment 1 The embodiment provides a safety risk management and control method driven by multi-source data fusion. Figure 1 As shown in Figure 1 As shown in In step 101, the multi-source data of each unit is collected, and the collected multi-source data is preprocessed.

[0028] The application scenario of the method provided in the embodiment can be multiple workshops of a manufacturing and trading enterprise, the unit can be a workshop, and the multi-source data can be related data of multiple dimensions of indicators of each unit, wherein the multi-source data includes device data, personnel training data and historical hidden danger data; wherein the device data can be obtained by enterprise data monitoring and analysis, that is, by collecting sensors arranged on the device, the device data includes temperature, pressure, flow, wind speed, wind pressure and valve switch quantity data of multiple devices, in the embodiment, multiple device data can be collected in real time by an Internet of Things monitoring sensor, and high-frequency data capture is supported, the principle is based on physical quantity transmission and signal conversion; the personnel training data can be obtained by collecting historical personnel operation training records, or by integrating data in an education and training platform, such as employee training content, employee training participation rate, employee training completion rate and employee training qualification rate; the historical hidden danger data can be obtained by collecting historical hidden danger investigation and management accident records, and the historical hidden danger data includes hidden danger quantity, hidden danger type, hidden danger grade, hidden danger position, hidden danger rectification situation and hidden danger acceptance situation.

[0029] The preprocessing is used to review the collected multi-source data, and according to the review result, format unification, invalid data deletion, effective data filling and data correction are performed, so that the effectiveness and reference of the multi-source data are improved, thereby improving the accuracy of subsequent risk assessment.

[0030] The method can monitor multiple dimensions of indicators of multiple units in real time, dynamically assess the units according to multiple indicators of different dimensions, and provide corresponding safety decisions, so that the safety risk management and control is no longer limited to the assessment and judgment of a single dimension, and does not need to rely on periodic manual investigation, and the timeliness and accuracy are higher.

[0031] In step 102, the pre-processed multi-source data is standardized to obtain standardized data, a sliding window is set for the standardized data, for the data in each sliding window, the index weight corresponding to each index is independently calculated, and the index weight is dynamically updated, so that the adaptive adjustment of the index weight based on the time sequence is realized.

[0032] In the embodiment, because different index types are different, the data forms are greatly different. In order to combine the data of all indexes to perform risk assessment at the same time, all pre-processed multi-source data is standardized to obtain standardized data in the embodiment, and the standardized data is used for subsequent weight calculation.

[0033] Further, the time sequence sliding window algorithm is applied to process the standardized data, the sliding window and the sliding step are set in the time dimension, in each sliding window, the index weight of the current sliding window is obtained through the index weight of the previous sliding window and the standardized data in the current sliding window, the dynamic update of the index weight is realized, the dynamic adjustment of the index weight is realized, the real-time performance of the index weight is improved, and thus the real-time performance and the accuracy of the risk assessment are improved.

[0034] In step 103, the dynamic risk assessment of each unit is performed according to the standardized data of each unit and the dynamically updated index weight, and the risk score of each unit is obtained.

[0035] In the embodiment, the corresponding safety decision is set in advance according to different levels of risk scores, when the risk score of the corresponding unit or enterprise is obtained, different levels of early warnings are triggered according to the risk score, and thus the corresponding safety decision is called.

[0036] In the embodiment, the multi-dimensional data of multiple units is integrated by the above method, the comprehensive dynamic assessment is performed according to the multi-dimensional data, the risk score corresponding to each unit is obtained, and the safety decision corresponding to different risk scores is preset. When the risk early warning of each unit is triggered, the corresponding safety decision is obtained according to the different risk scores. Compared with the existing periodic manual inspection and assessment, the real-time performance and the accuracy of the overall risk assessment are improved by the risk assessment method in the embodiment.

[0037] Further, in the embodiment, before the risk assessment is performed according to the collected data of each index, it is considered that the collected data may have invalid data, abnormal data and data loss, and the formats of the data may be different, and the data cannot be directly calculated. If the risk assessment is directly performed according to the collected data, the calculation result of the assessment may be wrong or inaccurate. In order to improve the correctness of the subsequent risk assessment, all data needs to be processed, and the pre-processing adjustment is performed according to each data type in the data. Therefore, the embodiment also relates to the following design.

[0038] The process involves collecting multi-source data from each unit and preprocessing the collected multi-source data, such as... Figure 2 As shown, the corresponding method flow includes the following.

[0039] In step 201, the real-time collected sensor data, personnel training data, and historical hazard data are used as the multi-source data.

[0040] In this embodiment, the sensor data includes information such as temperature, pressure, flow rate, wind speed, air pressure, and valve on / off status from multiple devices. This embodiment utilizes IoT monitoring sensors to collect data from multiple devices in real time and supports high-frequency data capture. The personnel training data includes employee training content, participation rate, completion rate, and pass rate. This data can be obtained by collecting historical personnel operation and training records or by integrating data from an education and training platform. The historical hazard data includes the number of hazards, hazard type, hazard level, hazard location, hazard rectification status, and hazard acceptance status. This historical hazard data can be obtained by collecting historical accident records.

[0041] In step 202, the multi-source data is reviewed, the data format is standardized, and invalid, abnormal, and missing data in the multi-source data are identified.

[0042] In this embodiment, through data review, the timestamps and units of all data are obtained and the data formats of all data are unified (for example, the pressure unit is converted to kPa and the data of all units are aligned to a unified time base). Invalid data (such as sensor disconnection records) and abnormal data (such as outliers with a false alarm rate greater than 15%) are also obtained from all data, as well as missing data (such as missing training frequencies, which can be replaced by the mean).

[0043] In step 203, the invalid data is deleted, the missing data is filled in, and the abnormal data is corrected.

[0044] In this embodiment, invalid data can be deleted according to business logic, abnormal data can be corrected using the interquartile range (ICM) or Z-score method, and missing data can be filled in according to business logic. These methods ensure that the collected data meets the requirements of subsequent steps, guaranteeing the accuracy of the subsequent risk assessment.

[0045] Further, in the embodiment, considering the need to comprehensively evaluate dynamic risks according to multiple different types of indicators, all data need to be standardized in form so that each data is unified into a consistent form, facilitating subsequent obtaining of weights under a unified standard according to each data and further performing subsequent risk evaluation. Therefore, the embodiment also relates to the following design.

[0046] The pre-processed multi-source data is standardized to obtain standardized data, specifically including: obtaining positive type data and negative type data in the pre-processed multi-source data.

[0047] The positive type data is standardized, and the expression is: The negative type data is standardized, and the expression is: wherein, is the standardized data of the jth indicator of the ith unit, is the maximum data in the jth indicator, is the minimum data in the jth indicator.

[0048] In the embodiment, the greater the value of the positive type data is, the better, for example, training completion rate and alarm response rate, and the smaller the value of the negative type data is, the better, for example, false alarm rate and repeated hidden danger rate. When judging that the data type is a negative type, the negative type data needs to be translated and standardized to a non-negative interval.

[0049] Further, since the time series sliding window algorithm needs to be applied to process the standardized data, a sliding window and a sliding step length are set in the time dimension to calculate and dynamically update the indicator weight, realize dynamic adjustment of the indicator weight, and improve the real-time performance of the indicator weight, thereby improving the real-time performance and accuracy of the risk evaluation. In the embodiment, the design of the time series sliding window algorithm and the calculation and dynamic adjustment of the indicator weight specifically include: setting the sliding window of the standardized data, for the data in each sliding window, independently calculating the indicator weight corresponding to each indicator, and dynamically updating the indicator weight, thereby realizing adaptive adjustment of the indicator weight based on the time series, specifically including: The sliding window size S and the sliding step length L are set. The data stream is monitored in real time, when new data is collected, the sliding window moves with the step length L, and the indicator weight corresponding to each indicator in the current sliding window is calculated. S and L are dynamically adjusted according to the information entropy change rate of the data in the current sliding window: the sliding window size S is adjusted according to the size of the information entropy change rate.

[0050] ​​Wherein S and L can be determined based on the mean value of the sampling frequency of historical data. If the information entropy change rate exceeds a preset threshold, S is reduced to increase sensitivity; otherwise, S is maintained or increased to maintain stability; in the embodiment, the preset threshold is set by a person skilled in the art according to the actual situation.

[0051] The expression of the index weight is:

[0052] Wherein, Wj(t) is the weight of the jth index at time t; alpha is the learning rate, used to control the index weight change speed, when the data flow is unstable, which makes the information entropy change rate large, a smaller alpha value (for example, alpha can be 0.1 to 0.3) can be taken, to suppress the excessive adjustment of the index weight due to data mutation, and smooth the risk score; when the data flow is stable, which makes the information entropy change rate small, a larger alpha value (for example, alpha can be 0.7 to 1.0) is taken, so that the index weight can follow the small trend change of the data faster, and the real-time performance of risk perception is improved. Delta Ej is the information entropy change amount of the jth index, .

[0053] Further, in the above application of time series sliding window algorithm to process the standardized data, since the information entropy used for calculating the index weight in each sliding window needs to be calculated according to the standardized data corresponding to the current sliding window, the calculation method of information entropy is needed, on the other hand, since there is no historical sliding window index weight as a basis for calculation in the initial sliding window, the index weight of the initial sliding window also needs to provide a corresponding calculation method, in view of the above, the embodiment also relates to, the corresponding design is as follows: In step 301, the index proportion corresponding to each index in the corresponding unit is obtained according to the standardized data of each unit, the information entropy of the corresponding index is obtained according to the index proportion corresponding to each index, and then the normalized weight corresponding to each index is obtained.

[0054] In step 302, when the sliding window slides, the index weight is recalculated according to the standardized data in the new sliding window, the dynamic update is realized, and the risk score corresponding to each unit is obtained.

[0055] For each sliding window, the expression of the index proportion is: ; The expression of the information entropy is: ; The expression of the normalized weight is: ; Wherein, the index proportion of the jth index of the ith unit, the standardized data of the jth index of the ith unit, m is the number of all units, the information entropy of the jth index, K = 1 / , the information utility value of the jth index, =1- .

[0056] As shown in Figure 4 , it is a flow chart for obtaining the index weight result.

[0057] In step 303, according to the standardized data of each unit on each index and the weight corresponding to each index, the risk score corresponding to each unit is obtained.

[0058] In this embodiment, the expression of the risk score corresponding to each unit is as follows.

[0059] ; wherein, the risk score of the ith unit, n is the total number of indexes, the weight of the jth index, the standardized data of the jth index of the ith unit.

[0060] Further, in this embodiment, after obtaining the risk score of each unit, the corresponding security decision that is preset can be found according to the risk score. In an actual system, the above steps need to be performed in a corresponding early warning and disposal closed loop, that is, the corresponding level of early warning is triggered according to the risk score, and the corresponding security decision is called according to the corresponding level of early warning. Therefore, this embodiment also relates to the following steps, as shown in Figure 5 .

[0061] In step 401, all security decision data is stored.

[0062] In step 402, the risk score data of the corresponding unit is stored, and when the risk score data is stored in the database, a hierarchical response mechanism is triggered, and the risk score of the corresponding unit is used to determine the early warning level triggered by the corresponding unit.

[0063] In this embodiment, the security decision data and the risk score data can be stored in a distributed data storage module. The hierarchical response mechanism is that different intervals of risk scores correspond to different levels of early warning, and when the risk score falls into the corresponding interval, the early warning of the corresponding level is triggered.

[0064] In step 403, the corresponding security decision data is called according to the early warning level of the corresponding unit.

[0065] In this embodiment, the following hierarchical response mechanism is taken as an example to more clearly show the scheme of this embodiment.

[0066] Unit early warning mechanism: when the risk score is less than 60 points, trigger the unit level early warning, that is, the major risk early warning (which can be a red early warning), the sub-plant leader leads to establish a special group, and formulates a rectification plan within 12 hours, and eliminates or reduces the risk within 24 hours. When the risk score is greater than or equal to 60 points and less than 80 points, trigger the unit level early warning, that is, the larger risk (which can be an orange early warning), the sub-plant safety department leads to establish a special group, and formulates a rectification plan within 24 hours, and eliminates or reduces the risk within 48 hours. When the risk score is greater than or equal to 80 points and less than 90 points, trigger the unit level early warning, that is, the general risk (which can be a yellow early warning), the workshop director implements risk inspection, implements risk project rectification measure plan, and eliminates or reduces the risk.

[0067] Further, after obtaining the risk scores of all units, the risk score of the whole (which can be an enterprise) can also be obtained according to the risk scores of all units, and a hierarchical early warning mechanism of the whole is preset, the corresponding early warning level is triggered according to the risk score of the whole, and the corresponding safety decision data is called, and the corresponding design is as follows.

[0068] According to the risk score of the whole, the corresponding whole safety decision is called according to the risk score of the whole.

[0069] In this embodiment, the whole risk score can be the arithmetic mean of the risk scores of all units, and the following whole hierarchical early warning mechanism is taken as an example: whole early warning mechanism: when the whole risk score is less than 60 points, trigger the whole level early warning, that is, the major risk (which can be a red early warning), the whole person in charge leads to establish an emergency command department, coordinates resources to support workshop rectification, formulates a rectification plan within 12 hours, and eliminates or reduces the risk within 24 hours. When the whole risk score is greater than or equal to 60 points and less than 80 points, trigger the whole level early warning, that is, the larger risk (which can be an orange early warning), the whole safety management department leads to establish an emergency command department, coordinates resources to support workshop rectification, formulates a rectification plan within 24 hours, and eliminates or reduces the risk within 48 hours. When the whole risk score is greater than or equal to 80 points and less than 90 points, trigger the whole level early warning, that is, the general risk (which can be a yellow early warning), and the sub-plant executes the respective early warning response mechanism.

[0070] Embodiment 2: On the basis of embodiment 1, the present embodiment provides a safety risk management and control system driven by multi-source data fusion, which is used for applying the safety risk management and control method driven by multi-source data fusion as shown in Figure 6 The safety risk management and control system driven by multi-source data fusion comprises a data acquisition module, a data preprocessing module, a multi-source data dynamic evaluation module and a safety decision module, wherein: The data acquisition module is configured to acquire multi-source data of each unit, and the data preprocessing module is configured to preprocess the acquired multi-source data.

[0071] As shown in Figure 7 The data acquisition module includes a monitoring data acquisition unit, a training data acquisition unit, and a hidden danger data acquisition unit, wherein the monitoring data acquisition unit is configured to acquire real-time sensing data, the training data acquisition unit is configured to acquire personnel training data, and the hidden danger data acquisition unit is configured to acquire historical hidden danger data; the real-time acquired sensing data, personnel training data, and historical hidden danger data are used as the multi-source data.

[0072] In this embodiment, the monitoring data acquisition module is configured to acquire real-time equipment data through an Internet of Things sensor, supports high-frequency data capture, and is based on physical quantity transmission and signal conversion. The training data acquisition module is configured to integrate education and training platform data, and includes a database interface and a user behavior log analyzer. The hidden danger data acquisition module is configured to acquire historical hidden danger records, and uses an API interface and a data cleaning engine to ensure data consistency. The data preprocessing module is configured to standardize input data, and involves data cleaning algorithms.

[0073] The multi-source data dynamic evaluation module is configured to standardize the preprocessed multi-source data to obtain standardized data, and dynamically evaluate the risk of each unit according to the standardized data of each unit to obtain a risk score of each unit.

[0074] The multi-source data dynamic evaluation module is a core evaluation engine, and includes a weighted fusion algorithm and an entropy model, and is based on dynamic calculation of index weights.

[0075] The safety decision module is configured to obtain corresponding safety decisions of each unit according to the risk score of each unit.

[0076] As shown in Figure 7 The safety decision module specifically includes a safety decision resource library unit, a data storage unit, and an early warning disposal unit, wherein the safety decision resource library unit is configured to preset all safety decision data; the data storage unit is configured to receive and store risk score data of all units from the multi-source data dynamic evaluation module, and is also configured to receive and store all safety decision data from the safety decision resource library unit; the early warning disposal unit is configured to trigger a hierarchical response mechanism when the risk score data is stored in the data storage unit, determine the early warning level triggered by each unit according to the risk score of each unit, and call corresponding safety decision data according to the risk score of each unit.

[0077] The security decision repository module is used to store preset decision schemes, and is structured as a distributed database to support fast query. The data storage module is: distributed storage is adopted, and the principle is based on Hadoop or similar framework to ensure that the data throughput reaches billions of levels. The early warning and disposal module is: a hierarchical response is triggered, and the structure includes an alarm engine and a workflow management system.

[0078] Embodiment 3 Further, this embodiment illustrates the application of the method and system of Embodiment 1 and Embodiment 2 by examples in actual scenarios.

[0079] As shown in Figure 8 , the risk assessment index system provided by this embodiment is calculated and obtained according to the risk assessment index system shown in Figure 8 . The collected multi-source data is shown in Table 1.

[0080] Table 1 Raw data

[0081] The collected multi-source data is standardized, and the data matrix is formed after the data is obtained from the data source. First, the data is standardized, where i is the unit number of the subplant, and j is the index number. The specific formula is as follows: The positive type data is positively indexed, and the expression is: ; The negative type data is positively indexed, and the expression is: ; Wherein, is the standardized data of the jth index of the ith unit, is the maximum data in the jth index, is the minimum data in the jth index.

[0082] The data after data standardization is shown in Table 2. Each row in the table is the index data of a unit, and each column in the table is the data of each unit under a certain index.

[0083] Table 2 Multi-source data after data standardization

[0084] Further, after the data standardization processing, a new standardized data matrix is formed, and the index proportion calculation is performed to form a proportion calculation result matrix.

[0085] The expression of the index proportion is: ; The index proportion of each index corresponding to each unit is shown in Table 3. Each row in the table represents the index proportion of each index of a unit, and each column in the table represents the index proportion of each index under a certain index.

[0086] Table 3 Index proportion of each index corresponding to each unit

[0087] Further, the information entropy corresponding to each index is calculated according to the index proportion of each index corresponding to each unit, and the determinant of the information entropy is obtained. The expression of the information entropy is as follows.

[0088] The information entropy corresponding to each index is shown in Table 4.

[0089] Table 4 Information entropy corresponding to each index

[0090] When Ej approaches 1, it indicates that the data difference is small and the index differentiation is low, and when Ej approaches 0, it indicates that the data difference is large and the index differentiation is high.

[0091] Further, the information utility value corresponding to each index is obtained according to the information entropy corresponding to each index, which is used for subsequent weight calculation. The information utility value corresponding to each index is: The information utility value corresponding to each index is shown in Table 5.

[0092] Table 5 Information utility value corresponding to each index

[0093] Continuous time series data is collected, and a sliding window (W=60 minutes, S=12 minutes) is applied. In each window, the index weight is calculated, but the weight is updated with the window sliding: Time t=0: Window data is the previous 60 minutes, and weight W is calculated j (0); t=12 minutes: Window sliding, new data added, and W is recalculated j (12); If the entropy change rate ΔE j >0.1, then W is automatically reduced to 30 minutes to improve sensitivity.

[0094] Further, the weight corresponding to each index is obtained according to the information utility value corresponding to each index, and the expression is: The weight corresponding to each index is shown in Table 6.

[0095] Table 6 Weight corresponding to each index ​

[0096] According to the weight corresponding to each index and the standardized data of each unit on each index, the risk score of each unit can be obtained, and the expression is: .

[0097] In this embodiment, the comprehensive scores of each unit (i.e. workshop) are as follows.

[0098] Workshop A: (0.1101x0.5+0.1101x0.5+...+0.1192x1)x100≈49.2 points Workshop B: (0.1101x1+0.1101x1+...+0.1192x0)x100≈79.2 points Workshop C: (0.1101x0+0.1101x0+...+0.1192x0.4)x100≈24.6 points Ranking: B>A>C (Workshop B is the best in safety) The score of workshop A is 49.2, triggering a first-level warning of the unit (<60 points), a major risk (red), and a special group led by the plant leader is established to develop a rectification plan within 12 hours and eliminate or reduce the risk within 24 hours.

[0099] The score of workshop B is 79.2, triggering a second-level warning of the unit (60-80 points), a major risk (orange), and a special group led by the plant safety department is established to develop a rectification plan within 24 hours and eliminate or reduce the risk within 48 hours.

[0100] The score of workshop C is 24.6, triggering a first-level warning of the unit (<60 points), a major risk (red), and a special group led by the plant leader is established to develop a rectification plan within 12 hours and eliminate or reduce the risk within 24 hours.

[0101] Further, according to the risk score of each unit, the overall risk score (i.e. the comprehensive score of the enterprise) is obtained, and the arithmetic mean of the scores of workshops ABC is the comprehensive score of the enterprise (49.2+79.2+24.6) / 3=51 points, triggering a first-level warning of the enterprise (<60 points), a major risk (red), and an emergency command department led by the main responsible person of the enterprise is established to coordinate resources to support the rectification of the workshop, develop a rectification plan within 12 hours, and eliminate or reduce the risk within 24 hours.

[0102] Those skilled in the art will readily understand that the above description is only a preferred embodiment of the present application and is not intended to limit the present application, and any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for managing and controlling security risks driven by multi-source data fusion, characterized in that, The method comprises the following steps: Collecting multi-source data of each unit and preprocessing the collected multi-source data; Standardizing the preprocessed multi-source data to obtain standardized data, setting a sliding window for the standardized data, independently calculating the index weight corresponding to each index for the data in each sliding window, and dynamically updating the index weight, so as to realize adaptive adjustment of the index weight based on time series; Performing dynamic risk assessment on each unit according to the standardized data and the dynamically updated index weight of each unit to obtain the risk score of each unit; Obtaining the safety decision corresponding to the corresponding unit according to the risk score of the corresponding unit. 2.The multi-source data fusion driven safety risk management and control method according to claim 1, characterized in that, The setting of the sliding window for the standardized data, the independent calculation of the index weight corresponding to each index for the data in each sliding window, and the dynamic updating of the index weight to realize adaptive adjustment of the index weight based on time series specifically comprise: Setting the size S and the sliding step L of the sliding window; Real-time monitoring of data flow, when new data is collected, the sliding window moves with a step L, and the index weight corresponding to each index in the current sliding window is calculated; Dynamically adjusting S and L according to the information entropy change rate of the data in the current sliding window: adjusting the size S of the sliding window according to the size of the information entropy change rate. The expression of the index weight is: Wherein, Wj(t) is the weight of the jth index at time t, α is the learning rate, and ΔEj is the information entropy change of the jth index. 3.The multi-source data fusion driven safety risk management and control method according to claim 2, characterized in that, The dynamic risk assessment on each unit according to the standardized data and the dynamic weight of each unit specifically comprises: Obtaining the index proportion corresponding to each index in the corresponding unit according to the standardized data of each unit, obtaining the information entropy of the corresponding index according to the index proportion corresponding to each index, and then obtaining the normalized weight corresponding to each index; After the sliding window slides, the index weight is recalculated according to the standardized data in the new sliding window to realize dynamic updating, and the risk score corresponding to each unit is obtained; For each sliding window, the expression of the index proportion is: ； The expression of the information entropy is: ； The expression of the normalized weight is: ； in, The weight of the j-th indicator in the i-th unit. Let m be the standardized data for the j-th indicator of the i-th unit, and m be the total number of units. Let K be the information entropy of the j-th indicator, and K = 1 / , Let the weight of the j-th indicator in the initial sliding window be denoted as . Let j be the information utility value of the j-th indicator. =1- .

4. The multi-source data fusion driven safety risk management and control method according to claim 1, characterized in that, The standardization of the preprocessed multi-source data to obtain standardized data specifically comprises: Obtaining the positive type data and the negative type data in the preprocessed multi-source data; Standardizing the positive type data, the expression is: ； Standardizing the negative type data, the expression is: ； wherein, is the standardized data for the jth indicator of the ith unit, is the maximum data in the jth indicator, is the minimum data in the jth indicator.

5. The multi-source data fusion driven safety risk management and control method according to claim 1, characterized in that, The collection of multi-source data of each unit and the preprocessing of the collected multi-source data specifically comprises: Collecting real-time collected sensor data, personnel training data and historical hidden danger data as the multi-source data; Data review is performed on the multi-source data, the data format of the multi-source data is unified, invalid data, abnormal data and missing data in the multi-source data are obtained; The invalid data is deleted, the missing data is filled, and the abnormal data is corrected.

6. The multi-source data fusion driven safety risk management and control method according to claim 1, characterized in that, The safety decision corresponding to the corresponding unit is obtained according to the risk score of the corresponding unit, specifically comprising: Storing all safety decision data; The risk score data of the corresponding unit is stored, and a grading response mechanism is triggered when the risk score data is stored in the database, and the warning level triggered by the corresponding unit is determined according to the risk score of the corresponding unit; According to the risk score of the corresponding unit, the corresponding safety decision data is called.

7. The multi-source data fusion driven safety risk management and control method according to claim 1, characterized in that, The multi-source data fusion driven safety risk management and control method further comprises: According to the risk score of all units, the overall risk score is obtained, and the corresponding overall safety decision is called according to the overall risk score.

8. A multi-source data fusion driven safety risk management and control system for implementing the multi-source data fusion driven safety risk management and control method according to any one of claims 1-7, characterized in that, It comprises: Data acquisition module, data preprocessing module, multi-source data dynamic evaluation module and safety decision module, wherein: The data acquisition module is used for collecting multi-source data of each unit, and the data preprocessing module is used for preprocessing the collected multi-source data; The multi-source data dynamic evaluation module is used for standardizing the preprocessed multi-source data to obtain standardized data, and dynamically evaluating the risk of each unit according to the standardized data of each unit to obtain the risk score of each unit; The safety decision module is used for obtaining the corresponding safety decision of the corresponding unit according to the risk score of the corresponding unit. 9.The multi-source data fusion driven safety risk management and control system according to claim 8, characterized in that, The data acquisition module comprises a monitoring data acquisition unit, a training data acquisition unit and a hidden danger data acquisition unit, wherein: The monitoring data acquisition unit is used for real-time acquisition of sensing data, the training data acquisition unit is used for acquisition of personnel training data, and the hidden danger data acquisition unit is used for acquisition of historical hidden danger data; The collected sensing data, personnel training data and historical hidden danger data are used as the multi-source data.

10. The multi-source data fusion driven safety risk management and control system according to claim 8, characterized in that, The safety decision module specifically comprises a safety decision resource library unit, a data storage unit and an early warning disposal unit, wherein: The safety decision resource library unit is used for prepositioning all safety decision data; The data storage unit is used for receiving and storing the risk score data of all units from the multi-source data dynamic evaluation module, and is also used for receiving and storing all safety decision data from the safety decision resource library unit; The early warning disposal unit is used for triggering a grading response mechanism when the risk score data is stored in the data storage unit, determining the warning level triggered by the corresponding unit according to the risk score of the corresponding unit, and calling the corresponding safety decision data according to the risk score of the corresponding unit.

Images