A risk assessment-based industrial production safety identification system and method

Abstract

The application discloses a kind of industrial production safety identification system and method based on risk assessment, specifically includes: S1, collection multi-source operation data is completed time alignment and numerical normalization, form risk assessment input vector;S2, input vector is calculated to generate the risk assessment result vector including probability, stability, data quality, similarity and time continuity component;S3, risk assessment result vector is executed hierarchical evidence modeling and then applies evidence concentration constraint;S4, the consistency of evidence is calculated to form support evidence time series and conflict evidence time series;S5, two kinds of evidence time series are respectively executed sequential probability ratio statistic accumulation;S6, according to the decision condition to determine safety state identification result or maintain undetermined state;S7, output industrial production safety identification result.The application introduces hierarchical evidence sequential determination to realize stable identification.

Description

Technical Field

[0001] This invention relates to the field of industrial production safety risk identification technology, and in particular to an industrial production safety identification system and method based on risk assessment. Background Technology

[0002] As industrial production systems continue to expand in scale, production processes are characterized by diverse equipment types, complex operating conditions, and dispersed data sources, posing higher demands on the identification and control of production safety risks. Industrial production processes typically generate operational data from multiple data sources simultaneously, including equipment operating status, process parameters, environmental conditions, and system acquisition quality. These data exhibit significant differences in sampling periods, time precision, and numerical scales, directly impacting the stability and consistency of risk assessment results.

[0003] Existing industrial safety identification technologies mostly use single risk indicators or static thresholds to determine operational status, making it difficult to characterize the evolution of risk status over time and lacking effective mechanisms for handling inconsistencies between multi-source data. When multi-source data is missing, noisy, or conflicting, traditional methods often fuse data through simple weighting or empirical rules, which can easily lead to unstable judgment results under the influence of local abnormal data.

[0004] On the other hand, while some security identification methods based on probabilistic models introduce statistical judgment mechanisms, they typically map multi-source information uniformly to a single probability value, failing to perform hierarchical modeling of the evidence contributions from different risk components and data sources. This leads to amplification or dilution of risk evidence during the fusion process. Furthermore, existing methods often employ fixed time windows or one-off judgments during the judgment process, lacking characterization of the continuous accumulation of risk evidence over time, making it difficult to maintain a reasonable judgment state under uncertain conditions.

[0005] Therefore, there is an urgent need for an industrial production safety identification method that can uniformly model multi-source operational data, hierarchically process risk evidence, and combine time series statistical judgment mechanisms to meet the safety identification needs in complex industrial scenarios. Summary of the Invention

[0006] One objective of this invention is to propose an industrial production safety identification system and method based on risk assessment. This invention introduces a hierarchical evidence modeling mechanism and a dual-channel sequential probability ratio statistical model to perform component-level, source-level, and global-level evidence mapping and time series accumulation judgment on multi-source industrial production operation data, constructing a continuously evolving safety status identification process. It has the characteristics of consistent risk identification logic, controllable conflicting evidence, and stable judgment results.

[0007] An industrial production safety identification system and method based on risk assessment according to an embodiment of the present invention includes the following steps:

[0008] S1. Collect operational data from several data sources during industrial production, perform time alignment and numerical normalization on the operational data, and form a multi-source risk assessment input vector; S2. Perform risk assessment calculations on the multi-source risk assessment input vector to generate a risk assessment result vector containing probability components, stability components, data quality components, similarity components, and time continuity components; S3. Perform hierarchical evidence modeling on the risk assessment result vector, generating component-level evidence parameters, source-level evidence parameters, and global evidence parameters at the risk component layer, data source layer, and global layer, respectively. The evidence parameter mapping process is controlled by evidence concentration constraints; 4. Calculate the evidence consistency metric based on component-level evidence parameters and source-level evidence parameters to form supporting evidence time series and conflicting evidence time series; S5. Perform cumulative calculations of the first sequential probability ratio statistic and the second sequential probability ratio statistic on the supporting evidence time series and the conflicting evidence time series, respectively; S6. Determine the safety status identification result when the first sequential probability ratio statistic meets the support decision condition and the second sequential probability ratio statistic meets the conflict suppression decision condition; maintain the undetermined state when neither decision condition is met; S7. Output the safety status identification result or the undetermined state as the industrial production safety identification result.

[0009] Optionally, the operational data in S1 specifically includes:

[0010] Collect equipment operating status data corresponding to industrial production equipment, including speed, current, voltage, temperature, pressure, and vibration amplitude; collect environmental status data corresponding to the industrial production environment, including ambient temperature and humidity; collect process parameter data corresponding to the industrial production process, including process setpoints and process feedback values; collect data quality status data corresponding to the industrial production system, including sampling time intervals, missing data markers, and anomaly markers; arrange all types of data in a unified time index order to form the operating data in step S1.

[0011] Optionally, the execution risk assessment calculation of S2 specifically includes:

[0012] Traverse the multi-source risk assessment input vector sequence in time index order, input the input vector corresponding to each time index into the scoring weight matrix and the scoring bias vector, and generate a risk score numerical sequence arranged by time index.

[0013] Using the risk score numerical sequence as the calculation object, the probability component is obtained by performing interval mapping operation on the risk score numerical sequence, the stability component is obtained by performing variance calculation on the risk score numerical sequence within a preset window length, the data quality component is obtained by performing ratio calculation on the sampling time interval value, the data missing mark value, and the data anomaly mark value, the similarity component is obtained by performing similarity calculation on the input vector and the reference input vector set, and the temporal continuity component is obtained by performing adjacent difference calculation on the risk score numerical sequence.

[0014] The probability component, stability component, data quality component, similarity component, and time continuity component are arranged in a preset order to form a risk assessment result vector.

[0015] Optionally, S3 specifically includes:

[0016] Read the risk assessment result vector generated in step S2, and perform component splitting operation on the risk assessment result vector according to the position index of probability component, stability component, data quality component, similarity component and time continuity component to obtain a set of component-level input vectors distinguished by risk component type;

[0017] For each component-level input vector in the component-level input vector set, input the component-level evidence mapping matrix and the component-level evidence bias vector, calculate the component-level evidence parameters of the corresponding risk component, store the component-level evidence parameters in non-negative numerical form and establish a correspondence with the risk component type identifier;

[0018] Grouping of the component-level evidence parameters according to the data source identifier, and arranging the component-level evidence parameters in order of risk component type within each data source group to form a sequence of component-level evidence parameters within the data source.

[0019] Input the source-level evidence mapping matrix and the source-level evidence bias vector into the component-level evidence parameter sequence within the data source, calculate the source-level evidence parameters of the corresponding data source, and record the corresponding data source identifier while keeping the source-level evidence parameters in non-negative numerical form.

[0020] An evidence concentration constraint is imposed on the mapping process from component-level evidence parameters to source-level evidence parameters. The evidence concentration constraint is achieved by limiting the proportion of numerical participation of component-level evidence parameters in the mapping calculation.

[0021] Arrange the source-level evidence parameters in the order of the data source identifiers to form a set of source-level evidence parameters. Input the set of source-level evidence parameters into the global evidence mapping matrix and the global evidence bias vector to calculate the global evidence parameters.

[0022] An evidence concentration constraint is imposed on the mapping process from source-level evidence parameters to global evidence parameters. The evidence concentration constraint is achieved by limiting the proportion of the numerical participation of source-level evidence parameters in the mapping calculation, thus forming the global evidence parameters in step S3.

[0023] Optionally, the evidence concentration constraint control in S3 specifically includes:

[0024] S31. Read the set of component-level evidence parameters and the set of source-level evidence parameters. Read the set of mapping terms from the component-level evidence parameters to the source-level evidence parameters. The mapping term set records the component-level evidence parameter identifier, data source identifier, and initial value of the mapping coefficient corresponding to each mapping term. S32. Perform component summation on the component-level evidence parameters corresponding to each data source identifier to generate a component-level evidence concentration value. Perform component summation on the source-level evidence parameters corresponding to each data source identifier to generate a source-level evidence concentration value. S33. Read the preset concentration threshold parameter. Perform a threshold comparison operation on the component-level evidence concentration value and the source-level evidence concentration value to generate a concentration constraint marker value. S34. S35. For mapping items with concentration constraint markers as constraint trigger values, perform a scaling operation on the initial value of the mapping coefficient and the concentration threshold parameter to generate a corrected value of the mapping coefficient. For mapping items with concentration constraint markers as non-constraint trigger values, retain the initial value of the mapping coefficient. S36. For each mapping item in the mapping item set, perform a multiplication operation on the corrected value of the mapping coefficient and the corresponding component-level evidence parameter to obtain the mapping contribution value. Perform a summation operation on the mapping contribution values ​​of the same data source identifier to update the source-level evidence parameter set. S37. Read the mapping item set from the source-level evidence parameters to the global evidence parameters, perform concentration constraint calculation on the source-level evidence parameter set according to S32 to S35, and update the global evidence parameters.

[0025] Optionally, S4 specifically includes:

[0026] Read the component-level evidence parameters and source-level evidence parameters in the order of time index. Arrange the component-level evidence parameters in the order of risk component type at each time index position to form a component-level evidence vector. Then read the source-level evidence parameters at the same time index position to form a source-level evidence vector.

[0027] The component summation operation is performed on the component-level evidence vector and the source-level evidence vector respectively to obtain the component-level evidence concentration value and the source-level evidence concentration value. The division normalization operation is performed on the component-level evidence vector and the source-level evidence vector respectively to generate the normalized component-level evidence vector and the normalized source-level evidence vector. Then, the component absolute difference calculation and component summation operation are performed on the normalized component-level evidence vector and the normalized source-level evidence vector to obtain the evidence difference value.

[0028] An interval mapping operation is performed on the evidence difference values ​​to generate evidence consistency measure values ​​and evidence conflict measure values, which are then written into the supporting evidence time series and conflicting evidence time series respectively in time index order.

[0029] Optionally, S5 includes the following steps:

[0030] Read the supporting evidence time series as the observation sequence for the first sequential probability ratio statistic, and read the conflicting evidence time series as the observation sequence for the second sequential probability ratio statistic. The supporting evidence time series and the conflicting evidence time series constitute the classification input of the sequential statistic.

[0031] Read the support decision parameter set and the conflict suppression decision parameter set. The support decision parameter set includes the set of support null hypothesis probability mapping parameters, the set of support alternative hypothesis probability mapping parameters, the first upper boundary threshold and the first lower boundary threshold. The conflict suppression decision parameter set includes the set of conflict null hypothesis probability mapping parameters and the set of conflict alternative hypothesis probability mapping parameters, the second upper boundary threshold and the second lower boundary threshold. Perform initial value setting on the first sequential probability ratio statistic and the second sequential probability ratio statistic.

[0032] Traverse the supporting evidence time series in time index order, read the supporting evidence value at each time index position, input the supporting evidence value into the supporting null hypothesis probability mapping parameter set and the supporting alternative hypothesis probability mapping parameter set respectively to obtain the supporting null hypothesis probability value and the supporting alternative hypothesis probability value, perform a ratio operation on the supporting alternative hypothesis probability value and the supporting null hypothesis probability value and then perform a logarithmic mapping to obtain the first statistic increment, and perform an addition operation on the first statistic increment and the current value of the first sequential probability ratio statistic to obtain the updated value of the first sequential probability ratio statistic;

[0033] Traverse the time series of conflict evidence in chronological order by time index, read the conflict evidence value at each time index position, input the conflict evidence value into the conflict null hypothesis probability mapping parameter set and the conflict alternative hypothesis probability mapping parameter set respectively to obtain the conflict null hypothesis probability value and the conflict alternative hypothesis probability value, perform a ratio operation on the conflict alternative hypothesis probability value and the conflict null hypothesis probability value and then perform a logarithmic mapping to obtain the second statistic increment, and perform an addition operation on the second statistic increment and the current value of the second sequential probability ratio statistic to obtain the updated value of the second sequential probability ratio statistic;

[0034] Write the security state determination trigger flag value when the updated value of the first sequential probability ratio statistic and the first upper boundary threshold satisfy the support decision condition and the updated value of the second sequential probability ratio statistic and the second lower boundary threshold satisfy the conflict suppression decision condition.

[0035] Optionally, S6 specifically includes:

[0036] Read the updated values ​​of the first sequential probability ratio statistic, the second sequential probability ratio statistic, the support decision parameter set, and the conflict suppression decision parameter set; perform a threshold comparison operation on the updated value of the first sequential probability ratio statistic and the first upper boundary threshold in the support decision parameter set to generate a first decision condition marker value; perform a threshold comparison operation on the updated value of the second sequential probability ratio statistic and the second lower boundary threshold in the conflict suppression decision parameter set to generate a second decision condition marker value; when both the first and second decision condition marker values ​​are satisfied, write the security state identification result into the security state cache; when either the first or second decision condition marker value is not satisfied, maintain an undetermined state record in the security state cache.

[0037] The beneficial effects of this invention are:

[0038] (1) This invention performs unified time alignment and numerical normalization on multi-source operation data in industrial production process, and then constructs a risk assessment result vector containing probability component, stability component, data quality component, similarity component and time continuity component in the risk assessment stage. This enables a multi-dimensional characterization of risk status, avoids the dominant influence of a single indicator on the safety status identification result, and ensures that the risk assessment result maintains structural consistency in the case of fluctuation, missing or scale differences in multi-source data.

[0039] (2) This invention introduces a hierarchical evidence modeling structure of component level, source level and global level, generates corresponding evidence parameters at different levels, and controls the participation ratio of evidence in the hierarchical mapping process by evidence concentration constraint, effectively limiting the diffusion range of local abnormal evidence in the fusion process, so that the evidence fusion process remains numerically controllable and structurally stable, and adapts to the situation where there is inconsistency of multi-source evidence in complex industrial production scenarios.

[0040] (3) In the safety status determination stage, the present invention adopts a dual-channel sequential probability ratio statistic accumulation mechanism that separates supporting evidence and conflicting evidence. By performing continuous cumulative determination on the evidence time series, the dynamic identification of the safety status is realized. When the judgment conditions are not met, the undetermined state is maintained, thereby avoiding premature or one-sided safety status output, so that the industrial production safety identification results have continuity and consistency in the time dimension. Attached Figure Description

[0041] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings:

[0042] Figure 1 This is a flowchart of a risk assessment-based identification system and method for industrial production safety proposed in this invention.

[0043] Figure 2 This is a schematic diagram of the hierarchical evidence modeling and evidence concentration constraint mechanism of the industrial production safety identification system and method based on risk assessment proposed in this invention.

[0044] Figure 3 This is a schematic diagram of the dual-channel sequential probability ratio statistics for an industrial production safety identification system and method based on risk assessment proposed in this invention. Detailed Implementation

[0045] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0046] refer to Figure 1 Figure 3 A risk assessment-based identification system and method for industrial production safety includes the following steps:

[0047] S1. Collect operational data from several data sources during industrial production, perform time alignment and numerical normalization on the operational data, and form a multi-source risk assessment input vector; S2. Perform risk assessment calculations on the multi-source risk assessment input vector to generate a risk assessment result vector containing probability components, stability components, data quality components, similarity components, and time continuity components; S3. Perform hierarchical evidence modeling on the risk assessment result vector, generating component-level evidence parameters, source-level evidence parameters, and global evidence parameters at the risk component layer, data source layer, and global layer, respectively. The evidence parameter mapping process is controlled by evidence concentration constraints; 4. Calculate the evidence consistency metric based on component-level evidence parameters and source-level evidence parameters to form supporting evidence time series and conflicting evidence time series; S5. Perform cumulative calculations of the first sequential probability ratio statistic and the second sequential probability ratio statistic on the supporting evidence time series and the conflicting evidence time series, respectively; S6. Determine the safety status identification result when the first sequential probability ratio statistic meets the support decision condition and the second sequential probability ratio statistic meets the conflict suppression decision condition; maintain the undetermined state when neither decision condition is met; S7. Output the safety status identification result or the undetermined state as the industrial production safety identification result.

[0048] In this embodiment, the running data in S1 specifically includes:

[0049] Collect equipment operating status data corresponding to industrial production equipment, including speed, current, voltage, temperature, pressure, and vibration amplitude; collect environmental status data corresponding to the industrial production environment, including ambient temperature and humidity; collect process parameter data corresponding to the industrial production process, including process setpoints and process feedback values; collect data quality status data corresponding to the industrial production system, including sampling time intervals, missing data markers, and anomaly markers; arrange all types of data in a unified time index order to form the operating data in step S1.

[0050] In this embodiment, the execution risk assessment calculation of S2 specifically includes:

[0051] Traverse the multi-source risk assessment input vector sequence in time index order, input the input vector corresponding to each time index into the scoring weight matrix and the scoring bias vector, and generate a risk score numerical sequence arranged by time index.

[0052] Using the risk score numerical sequence as the calculation object, the probability component is obtained by performing interval mapping operation on the risk score numerical sequence, the stability component is obtained by performing variance calculation on the risk score numerical sequence within a preset window length, the data quality component is obtained by performing ratio calculation on the sampling time interval value, the data missing mark value, and the data anomaly mark value, the similarity component is obtained by performing similarity calculation on the input vector and the reference input vector set, and the temporal continuity component is obtained by performing adjacent difference calculation on the risk score numerical sequence.

[0053] The probability component, stability component, data quality component, similarity component, and time continuity component are arranged in a preset order to form a risk assessment result vector.

[0054] In this embodiment, S3 specifically includes:

[0055] Read the risk assessment result vector generated in step S2, and perform component splitting operation on the risk assessment result vector according to the position index of probability component, stability component, data quality component, similarity component and time continuity component to obtain a set of component-level input vectors distinguished by risk component type;

[0056] For each component-level input vector in the component-level input vector set, input the component-level evidence mapping matrix and the component-level evidence bias vector, calculate the component-level evidence parameters of the corresponding risk component, store the component-level evidence parameters in non-negative numerical form and establish a correspondence with the risk component type identifier;

[0057] Grouping of the component-level evidence parameters according to the data source identifier, and arranging the component-level evidence parameters in order of risk component type within each data source group to form a sequence of component-level evidence parameters within the data source.

[0058] Input the source-level evidence mapping matrix and the source-level evidence bias vector into the component-level evidence parameter sequence within the data source, calculate the source-level evidence parameters of the corresponding data source, and record the corresponding data source identifier while keeping the source-level evidence parameters in non-negative numerical form.

[0059] An evidence concentration constraint is imposed on the mapping process from component-level evidence parameters to source-level evidence parameters. The evidence concentration constraint is achieved by limiting the proportion of numerical participation of component-level evidence parameters in the mapping calculation.

[0060] Arrange the source-level evidence parameters in the order of the data source identifiers to form a set of source-level evidence parameters. Input the set of source-level evidence parameters into the global evidence mapping matrix and the global evidence bias vector to calculate the global evidence parameters.

[0061] An evidence concentration constraint is imposed on the mapping process from source-level evidence parameters to global evidence parameters. The evidence concentration constraint is achieved by limiting the proportion of the numerical participation of source-level evidence parameters in the mapping calculation, thus forming the global evidence parameters in step S3.

[0062] In this embodiment, the component splitting operation specifically includes:

[0063] Read the risk assessment result vector generated in step S2, and perform an index positioning operation on the risk assessment result vector according to the preset position index table of each component in the risk assessment result vector. The position index table clearly records the fixed position intervals of the probability component, stability component, data quality component, similarity component, and time continuity component in the risk assessment result vector. According to the index intervals defined in the position index table, perform a sub-vector truncation operation on the risk assessment result vector to extract the values ​​belonging to the same risk component type from the risk assessment result vector in index order to form the corresponding component-level input vectors. Perform the above index truncation operation on the probability component, stability component, data quality component, similarity component, and time continuity component respectively to obtain a set of component-level input vectors that correspond one-to-one with the risk component type. Each component-level input vector is consistent with the risk assessment result vector in the time index dimension and is used for subsequent mapping calculation of component-level evidence parameters.

[0064] In this embodiment, the component-level evidence parameters for calculating the corresponding risk component specifically include:

[0065] According to the risk component type, each component-level input vector is read sequentially from the component-level input vector set, maintaining the order of the time index positions in the component-level input vector; the component-level input vector is input into the component-level evidence mapping matrix corresponding to the risk component type, and matrix multiplication is performed with the component-level evidence mapping matrix to obtain a linear mapping result vector; component-level evidence bias vectors are added to each component of the linear mapping result vector to generate an intermediate evidence mapping result vector; non-negative constraint mapping operation is performed on the intermediate evidence mapping result vector, mapping values ​​less than zero in the intermediate result vector to zero, and keeping values ​​greater than or equal to zero unchanged, to obtain a component-level evidence parameter vector; the component-level evidence parameter vector is stored in non-negative numerical form, and a one-to-one correspondence is established between it and the corresponding risk component type identifier in the parameter storage structure to form component-level evidence parameters used for subsequent hierarchical evidence modeling.

[0066] In this embodiment, the evidence concentration constraint control of S3 specifically includes:

[0067] S31. Read the set of component-level evidence parameters and the set of source-level evidence parameters. Read the set of mapping terms from the component-level evidence parameters to the source-level evidence parameters. The mapping term set records the component-level evidence parameter identifier, data source identifier, and initial value of the mapping coefficient corresponding to each mapping term. S32. Perform component summation on the component-level evidence parameters corresponding to each data source identifier to generate a component-level evidence concentration value. Perform component summation on the source-level evidence parameters corresponding to each data source identifier to generate a source-level evidence concentration value. S33. Read the preset concentration threshold parameter. Perform a threshold comparison operation on the component-level evidence concentration value and the source-level evidence concentration value to generate a concentration constraint marker value. S34. S35. For mapping items with concentration constraint markers as constraint trigger values, perform a scaling operation on the initial value of the mapping coefficient and the concentration threshold parameter to generate a corrected value of the mapping coefficient. For mapping items with concentration constraint markers as non-constraint trigger values, retain the initial value of the mapping coefficient. S36. For each mapping item in the mapping item set, perform a multiplication operation on the corrected value of the mapping coefficient and the corresponding component-level evidence parameter to obtain the mapping contribution value. Perform a summation operation on the mapping contribution values ​​of the same data source identifier to update the source-level evidence parameter set. S37. Read the mapping item set from the source-level evidence parameters to the global evidence parameters, perform concentration constraint calculation on the source-level evidence parameter set according to S32 to S35, and update the global evidence parameters.

[0068] In this embodiment, S4 specifically includes:

[0069] Read the component-level evidence parameters and source-level evidence parameters in the order of time index. Arrange the component-level evidence parameters in the order of risk component type at each time index position to form a component-level evidence vector. Then read the source-level evidence parameters at the same time index position to form a source-level evidence vector.

[0070] The component summation operation is performed on the component-level evidence vector and the source-level evidence vector respectively to obtain the component-level evidence concentration value and the source-level evidence concentration value. The division normalization operation is performed on the component-level evidence vector and the source-level evidence vector respectively to generate the normalized component-level evidence vector and the normalized source-level evidence vector. Then, the component absolute difference calculation and component summation operation are performed on the normalized component-level evidence vector and the normalized source-level evidence vector to obtain the evidence difference value.

[0071] An interval mapping operation is performed on the evidence difference values ​​to generate evidence consistency measure values ​​and evidence conflict measure values, which are then written into the supporting evidence time series and conflicting evidence time series respectively in time index order.

[0072] In this embodiment, S5 includes the following steps:

[0073] Read the supporting evidence time series as the observation sequence for the first sequential probability ratio statistic, and read the conflicting evidence time series as the observation sequence for the second sequential probability ratio statistic. The supporting evidence time series and the conflicting evidence time series constitute the classification input of the sequential statistic.

[0074] Read the support decision parameter set and the conflict suppression decision parameter set. The support decision parameter set includes the set of support null hypothesis probability mapping parameters, the set of support alternative hypothesis probability mapping parameters, the first upper boundary threshold and the first lower boundary threshold. The conflict suppression decision parameter set includes the set of conflict null hypothesis probability mapping parameters and the set of conflict alternative hypothesis probability mapping parameters, the second upper boundary threshold and the second lower boundary threshold. Perform initial value setting on the first sequential probability ratio statistic and the second sequential probability ratio statistic.

[0075] Traverse the supporting evidence time series in time index order, read the supporting evidence value at each time index position, input the supporting evidence value into the supporting null hypothesis probability mapping parameter set and the supporting alternative hypothesis probability mapping parameter set respectively to obtain the supporting null hypothesis probability value and the supporting alternative hypothesis probability value, perform a ratio operation on the supporting alternative hypothesis probability value and the supporting null hypothesis probability value and then perform a logarithmic mapping to obtain the first statistic increment, and perform an addition operation on the first statistic increment and the current value of the first sequential probability ratio statistic to obtain the updated value of the first sequential probability ratio statistic;

[0076] Traverse the time series of conflict evidence in chronological order by time index, read the conflict evidence value at each time index position, input the conflict evidence value into the conflict null hypothesis probability mapping parameter set and the conflict alternative hypothesis probability mapping parameter set respectively to obtain the conflict null hypothesis probability value and the conflict alternative hypothesis probability value, perform a ratio operation on the conflict alternative hypothesis probability value and the conflict null hypothesis probability value and then perform a logarithmic mapping to obtain the second statistic increment, and perform an addition operation on the second statistic increment and the current value of the second sequential probability ratio statistic to obtain the updated value of the second sequential probability ratio statistic;

[0077] Write the security state determination trigger flag value when the updated value of the first sequential probability ratio statistic and the first upper boundary threshold satisfy the support decision condition and the updated value of the second sequential probability ratio statistic and the second lower boundary threshold satisfy the conflict suppression decision condition.

[0078] In this embodiment, S6 specifically includes:

[0079] Read the updated values ​​of the first sequential probability ratio statistic, the second sequential probability ratio statistic, the support decision parameter set, and the conflict suppression decision parameter set; perform a threshold comparison operation on the updated value of the first sequential probability ratio statistic and the first upper boundary threshold in the support decision parameter set to generate a first decision condition marker value; perform a threshold comparison operation on the updated value of the second sequential probability ratio statistic and the second lower boundary threshold in the conflict suppression decision parameter set to generate a second decision condition marker value; when both the first and second decision condition marker values ​​are satisfied, write the security state identification result into the security state cache; when either the first or second decision condition marker value is not satisfied, maintain an undetermined state record in the security state cache.

[0080] Example 1:

[0081] To verify the feasibility and stability of this invention in industrial production safety identification, it was applied to a type of continuously operating industrial production scenario. In this scenario, the production process involves the coordinated operation of multiple devices, and their operating status is constantly changing. There are significant differences between different devices and environmental conditions, and the operating data comes from diverse sources and changes frequently. In this type of scenario, traditional safety identification methods based on a single indicator or fixed threshold are difficult to accurately reflect the true safety status. They are prone to misjudgment or frequent switching of judgment results when data fluctuations or local anomalies occur, thus affecting the reliability of the overall production safety assessment.

[0082] In this scenario, the system continuously collects operational data from multiple data sources, including data related to equipment operating status, environmental status, process parameters, and data quality. Due to differences in sampling rhythms and numerical ranges among the various data sources, the system first performs time alignment processing on the collected operational data, establishing correspondences among different data types under a unified time index. Further, it performs normalization processing on the numerical values, thereby constructing a consistent multi-source risk assessment input vector. This approach avoids risk assessment biases caused by inconsistent data scales or sampling misalignments.

[0083] After constructing the input vector, the system performs risk assessment calculations on the multi-source risk assessment input vector. During the risk assessment process, instead of relying on a single scoring result, it simultaneously generates probability components, stability components, data quality components, similarity components, and temporal continuity components. The probability component reflects the risk score interval corresponding to the current operating state; the stability component characterizes the degree of fluctuation in the risk score over a continuous time period; the data quality component reflects the completeness and anomalies of the input data; the similarity component describes the degree of closeness between the current operating state and historical reference states; and the temporal continuity component reflects the changing trend of the risk state over time. Through this multi-component structure, the risk assessment results can more comprehensively characterize the operating state features.

[0084] After the risk assessment results are generated, the system further performs hierarchical evidence modeling processing on the risk assessment result vector. First, at the risk component layer, component-level evidence parameters are generated for each risk component, ensuring that different types of risk information are expressed independently. Then, at the data source layer, component-level evidence parameters corresponding to the same data source are aggregated to generate source-level evidence parameters. Finally, at the global layer, source-level evidence parameters from each data source are fused to generate global evidence parameters. During the mapping process from component-level evidence parameters to source-level evidence parameters and from source-level evidence parameters to global evidence parameters, an evidence concentration constraint control mechanism is introduced to limit the proportion of evidence participation, thereby preventing the amplification of local abnormal evidence during the fusion process.

[0085] After evidence modeling is completed, the system calculates an evidence consistency metric based on component-level and source-level evidence parameters. By quantifying the differences between evidence parameters at different levels, it generates supporting evidence time series and conflicting evidence time series. The supporting evidence time series describes the trend of consistency among different pieces of evidence, while the conflicting evidence time series describes deviations between different pieces of evidence. Compared with traditional methods, this approach can explicitly distinguish between consistent and conflicting information, making the subsequent judgment process clearer.

[0086] During the security status determination phase, the system performs a cumulative calculation of the sequential probability ratio statistics on both the supporting evidence time series and the conflicting evidence time series. As the process continues, the statistics gradually accumulate over time. When the supporting statistics meet the support decision condition and the conflicting statistics meet the conflict suppression decision condition, the system outputs the security status identification result. If either decision condition is not met, the system maintains an undetermined state, thereby avoiding giving an unstable judgment when there is insufficient evidence.

[0087] Statistical analysis of continuous operation data in the above scenarios revealed that after introducing hierarchical evidence modeling and a dual-channel sequential judgment mechanism, the security status identification results showed stronger continuity in the time dimension, the frequency of judgment result switching was significantly reduced, and the unjudged state could effectively cover the stage of insufficient evidence, thereby reducing the occurrence of misjudgments.

[0088] To further verify the effectiveness of the present invention, the changes in risk assessment components, evidence consistency measurement results, and cumulative changes in sequential statistics of multi-source operational data at different stages were compared and statistically analyzed. The relevant data are summarized below.

[0089] Table 1: Comparison of the Implementation Effects of Industrial Production Safety Identification

[0090] Time Index Sequence Risk score Evidence consistency metric Supports cumulative statistics Cumulative values ​​of conflict statistics Safety status output flag T1 0.42 0.78 1.35 0.22 Undecided T2 0.45 0.81 2.10 0.30 Undecided T3 0.47 0.84 3.05 0.38 safe status T4 0.46 0.83 3.90 0.40 safe status T5 0.44 0.79 4.20 0.41 safe status

[0091] In Table 1, the risk score value changed from 0.42 to 0.47 under the continuous time index and then fell back to 0.44, with an overall variation range of 0.05. This indicates that after the multi-source operational data underwent time alignment and numerical normalization, the resulting risk assessment input vector did not experience abrupt changes during the scoring mapping process, and the scoring results maintained continuity. The risk score reached a local peak at the intermediate time index and then fell back, but the decline was limited and did not have a discontinuous impact on subsequent evidence modeling and judgment processes.

[0092] The consistency measure values ​​in Table 1 range from 0.78 to 0.84, with a maximum variation of 0.06 between adjacent time indices. This reflects the stable correspondence between component-level and source-level evidence parameters under the constraints of hierarchical mapping and evidence concentration. The consistency measure gradually increases in the early stages, indicating that the evidence structure is gradually converging. Although there is a slight decrease in the later stages, the value remains within a continuous range and no significant evidence split has occurred.

[0093] The cumulative value of the supporting statistic gradually increased from 1.35 to 4.20, exhibiting a monotonically increasing trend. The increment range for each time index was between 0.30 and 0.95, indicating that the supporting evidence time series consistently provided a positive contribution to the sequential probability ratio statistic accumulation calculation. A significant increment appeared in the cumulative value at an intermediate time index, corresponding to the trigger point of the safe state, suggesting that the supporting evidence had a crucial impact on the decision at that moment.

[0094] The cumulative value of the conflict statistic increased from 0.22 to 0.41, with a total increase of 0.19. The increment of adjacent time indices gradually decreased, and the later increment did not exceed 0.02, indicating that the conflict evidence was suppressed and controlled during the sequential accumulation process, and there was no rapid amplification. The difference between the support statistic and the conflict statistic at the same time index increased from 1.13 to 3.79, and the difference continued to increase, forming a stable judgment interval.

[0095] In Table 1, the safe state output remains in an undetermined state in the early time index. It turns into a safe state at the time index position where the supporting statistics meet the support decision condition and the conflict statistics meet the conflict suppression decision condition, and remains unchanged in subsequent time indices. This indicates that the decision result is dominated by the statistical accumulation process and is not affected by single scoring or local evidence fluctuations, reflecting the characteristics of non-windowed and continuous decision.

[0096] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A risk assessment-based identification system and method for industrial production safety, characterized in that, Includes the following steps: S1. Collect operational data from several data sources during the industrial production process, align the execution time of the operational data and normalize the values ​​to form a multi-source risk assessment input vector; S2. Perform risk assessment calculations on the multi-source risk assessment input vector to generate a risk assessment result vector containing probability components, stability components, data quality components, similarity components, and temporal continuity components. S3. Perform hierarchical evidence modeling on the risk assessment result vector, generating component-level evidence parameters, source-level evidence parameters, and global evidence parameters at the risk component layer, data source layer, and global layer, respectively. The evidence parameter mapping process is controlled by evidence concentration constraints. S4. Calculate evidence consistency metrics based on component-level evidence parameters and source-level evidence parameters to form supporting evidence time series and conflicting evidence time series. S5. Perform cumulative calculations of the first sequential probability ratio statistic and the second sequential probability ratio statistic on the supporting evidence time series and conflicting evidence time series, respectively. S6. Determine the security status identification result when the first sequential probability ratio statistic meets the support decision condition and the second sequential probability ratio statistic meets the conflict suppression decision condition; maintain an undetermined state when neither decision condition is met. S7. Output the safety status identification result or undetermined status as the industrial production safety identification result.

2. The industrial production safety identification system and method based on risk assessment according to claim 1, characterized in that, The operational data in S1 specifically includes: Collect equipment operating status data corresponding to industrial production equipment, including speed, current, voltage, temperature, pressure, and vibration amplitude; collect environmental status data corresponding to the industrial production environment, including ambient temperature and humidity; collect process parameter data corresponding to the industrial production process, including process setpoints and process feedback values; collect data quality status data corresponding to the industrial production system, including sampling time intervals, missing data markers, and anomaly markers; arrange all types of data in a unified time index order to form the operating data in step S1.

3. The industrial production safety identification system and method based on risk assessment according to claim 2, characterized in that, The execution risk assessment calculation for S2 specifically includes: Traverse the multi-source risk assessment input vector sequence in time index order, input the input vector corresponding to each time index into the scoring weight matrix and the scoring bias vector, and generate a risk score numerical sequence arranged by time index. Using the risk score numerical sequence as the calculation object, the probability component is obtained by performing interval mapping operation on the risk score numerical sequence, the stability component is obtained by performing variance calculation on the risk score numerical sequence within a preset window length, the data quality component is obtained by performing ratio calculation on the sampling time interval value, the data missing mark value, and the data anomaly mark value, the similarity component is obtained by performing similarity calculation on the input vector and the reference input vector set, and the temporal continuity component is obtained by performing adjacent difference calculation on the risk score numerical sequence. The probability component, stability component, data quality component, similarity component, and time continuity component are arranged in a preset order to form a risk assessment result vector.

4. The industrial production safety identification system and method based on risk assessment according to claim 3, characterized in that, S3 specifically includes: Read the risk assessment result vector generated in step S2, and perform component splitting operation on the risk assessment result vector according to the position index of probability component, stability component, data quality component, similarity component and time continuity component to obtain a set of component-level input vectors distinguished by risk component type; For each component-level input vector in the component-level input vector set, input the component-level evidence mapping matrix and the component-level evidence bias vector, calculate the component-level evidence parameters of the corresponding risk component, store the component-level evidence parameters in non-negative numerical form and establish a correspondence with the risk component type identifier; Grouping of the component-level evidence parameters according to the data source identifier, and arranging the component-level evidence parameters in order of risk component type within each data source group to form a sequence of component-level evidence parameters within the data source. Input the source-level evidence mapping matrix and the source-level evidence bias vector into the component-level evidence parameter sequence within the data source, calculate the source-level evidence parameters of the corresponding data source, and record the corresponding data source identifier while keeping the source-level evidence parameters in non-negative numerical form. An evidence concentration constraint is imposed on the mapping process from component-level evidence parameters to source-level evidence parameters. The evidence concentration constraint is achieved by limiting the proportion of numerical participation of component-level evidence parameters in the mapping calculation. Arrange the source-level evidence parameters in the order of the data source identifiers to form a set of source-level evidence parameters. Input the set of source-level evidence parameters into the global evidence mapping matrix and the global evidence bias vector to calculate the global evidence parameters. An evidence concentration constraint is imposed on the mapping process from source-level evidence parameters to global evidence parameters. The evidence concentration constraint is achieved by limiting the proportion of the numerical participation of source-level evidence parameters in the mapping calculation, thus forming the global evidence parameters in step S3.

5. The industrial production safety identification system and method based on risk assessment according to claim 4, characterized in that, The evidence concentration constraint control in S3 specifically includes: S31. Read the set of component-level evidence parameters and the set of source-level evidence parameters. Read the set of mapping terms from the component-level evidence parameters to the source-level evidence parameters. The mapping term set records the component-level evidence parameter identifier, data source identifier, and initial value of the mapping coefficient corresponding to each mapping term. S32. Perform component summation on the component-level evidence parameters corresponding to each data source identifier to generate a component-level evidence concentration value. Perform component summation on the source-level evidence parameters corresponding to each data source identifier to generate a source-level evidence concentration value. S33. Read the preset concentration threshold parameter. Perform a threshold comparison operation on the component-level evidence concentration value and the source-level evidence concentration value to generate a concentration constraint marker value. S34. S35. For mapping items with concentration constraint markers as constraint trigger values, perform a scaling operation on the initial value of the mapping coefficient and the concentration threshold parameter to generate a corrected value of the mapping coefficient. For mapping items with concentration constraint markers as non-constraint trigger values, retain the initial value of the mapping coefficient. S36. For each mapping item in the mapping item set, perform a multiplication operation on the corrected value of the mapping coefficient and the corresponding component-level evidence parameter to obtain the mapping contribution value. Perform a summation operation on the mapping contribution values ​​of the same data source identifier to update the source-level evidence parameter set. S37. Read the mapping item set from the source-level evidence parameters to the global evidence parameters, perform concentration constraint calculation on the source-level evidence parameter set according to S32 to S35, and update the global evidence parameters.

6. The industrial production safety identification system and method based on risk assessment according to claim 5, characterized in that, S4 specifically includes: Read the component-level evidence parameters and source-level evidence parameters in the order of time index. At each time index position, arrange the component-level evidence parameters in the order of risk component type to form a component-level evidence vector. Then read the source-level evidence parameters at the same time index position to form a source-level evidence vector. The component summation operation is performed on the component-level evidence vector and the source-level evidence vector respectively to obtain the component-level evidence concentration value and the source-level evidence concentration value. The division normalization operation is performed on the component-level evidence vector and the source-level evidence vector respectively to generate the normalized component-level evidence vector and the normalized source-level evidence vector. Then, the component absolute difference calculation and component summation operation are performed on the normalized component-level evidence vector and the normalized source-level evidence vector to obtain the evidence difference value. An interval mapping operation is performed on the evidence difference values ​​to generate evidence consistency measure values ​​and evidence conflict measure values, which are then written into the supporting evidence time series and conflicting evidence time series respectively in time index order.

7. The industrial production safety identification system and method based on risk assessment according to claim 6, characterized in that, S5 includes the following steps: Read the supporting evidence time series as the observation sequence for the first sequential probability ratio statistic, and read the conflicting evidence time series as the observation sequence for the second sequential probability ratio statistic. The supporting evidence time series and the conflicting evidence time series constitute the classification input of the sequential statistic. Read the support decision parameter set and the conflict suppression decision parameter set. The support decision parameter set includes the set of support null hypothesis probability mapping parameters, the set of support alternative hypothesis probability mapping parameters, the first upper boundary threshold and the first lower boundary threshold. The conflict suppression decision parameter set includes the set of conflict null hypothesis probability mapping parameters and the set of conflict alternative hypothesis probability mapping parameters, the second upper boundary threshold and the second lower boundary threshold. Perform initial value setting on the first sequential probability ratio statistic and the second sequential probability ratio statistic. Traverse the supporting evidence time series in time index order, read the supporting evidence value at each time index position, input the supporting evidence value into the supporting null hypothesis probability mapping parameter set and the supporting alternative hypothesis probability mapping parameter set respectively to obtain the supporting null hypothesis probability value and the supporting alternative hypothesis probability value, perform a ratio operation on the supporting alternative hypothesis probability value and the supporting null hypothesis probability value and then perform a logarithmic mapping to obtain the first statistic increment, and perform an addition operation on the first statistic increment and the current value of the first sequential probability ratio statistic to obtain the updated value of the first sequential probability ratio statistic; Traverse the time series of conflict evidence in chronological order by time index, read the conflict evidence value at each time index position, input the conflict evidence value into the conflict null hypothesis probability mapping parameter set and the conflict alternative hypothesis probability mapping parameter set respectively to obtain the conflict null hypothesis probability value and the conflict alternative hypothesis probability value, perform a ratio operation on the conflict alternative hypothesis probability value and the conflict null hypothesis probability value and then perform a logarithmic mapping to obtain the second statistic increment, and perform an addition operation on the second statistic increment and the current value of the second sequential probability ratio statistic to obtain the updated value of the second sequential probability ratio statistic; Write the security state determination trigger flag value when the updated value of the first sequential probability ratio statistic and the first upper boundary threshold satisfy the support decision condition and the updated value of the second sequential probability ratio statistic and the second lower boundary threshold satisfy the conflict suppression decision condition.

8. The industrial production safety identification system and method based on risk assessment according to claim 7, characterized in that, S6 specifically includes: Read the updated values ​​of the first sequential probability ratio statistic, the second sequential probability ratio statistic, the support decision parameter set, and the conflict suppression decision parameter set; perform a threshold comparison operation on the updated value of the first sequential probability ratio statistic and the first upper boundary threshold in the support decision parameter set to generate a first decision condition marker value; perform a threshold comparison operation on the updated value of the second sequential probability ratio statistic and the second lower boundary threshold in the conflict suppression decision parameter set to generate a second decision condition marker value; when both the first and second decision condition marker values ​​are satisfied, write the security state identification result into the security state cache; when either the first or second decision condition marker value is not satisfied, maintain an undetermined state record in the security state cache.

Images