A method and system for dynamic risk assessment of a storage tank

Abstract

This invention discloses a dynamic risk assessment method and system for storage tanks, comprising the following steps: S1, identifying potential risk sources of the storage tank, taking tank leakage as the assessment object; S2, inputting the potential risk sources and their possible consequences into a fuzzy Bow-Tie model to form a preliminary qualitative analysis of consequences and measures; S3, transforming the Bow-Tie model into a Bayesian network based on mapping relationships; S4, having experts score the corresponding risk sources and converting the scores into prior probabilities of the root nodes of the Bayesian network, determining the weights of each expert using the analytic hierarchy process (AHP), and converting the expert evaluation results into fuzzy probabilities to obtain the prior probabilities of the root nodes and determine the conditional probability tables of other nodes; S5, calculating the posterior probabilities and analyzing and determining the priority relationships of failure modes; S6, providing corresponding risk control measures based on the priority judgment results. This method realizes management from risk source identification to control measure output, overcoming the problem that traditional methods cannot cope with complex dynamic environments.

Description

Technical Field

[0001] This invention belongs to the field of safety management of storage tanks, and in particular relates to a method and system for dynamic risk assessment of storage tanks. Background Technology

[0002] Liquefied natural gas (LNG) is a clean energy source primarily composed of methane. It is produced by liquefying natural gas through cryogenic processing, significantly reducing its volume for easier transportation and storage. Due to its high combustion efficiency and low pollution characteristics, LNG has been widely used in the energy, power, industrial, and transportation sectors. However, the storage and transportation of LNG presents several safety hazards, such as leaks, overpressure, fires, and explosions. Therefore, effective risk assessment and management of LNG storage tanks are crucial. Consequently, research on how to achieve dynamic risk assessment of LNG storage tanks to ensure their safe and stable operation has become a key focus of current academic and industrial research. Traditional risk assessment methods often employ qualitative analysis or static models, which are insufficient to effectively address the complex dynamic environment of LNG storage tanks in actual operation. These methods typically cannot monitor and provide real-time warnings of risk events, nor can they comprehensively analyze the interactions between different risk sources.

[0003] Therefore, there is an urgent need to design a dynamic risk assessment method for storage tanks to solve the problems mentioned above. Summary of the Invention

[0004] To address the technical problem that traditional risk assessment methods mentioned in the background technology mostly employ qualitative analysis or static models, which are difficult to effectively cope with the complex dynamic environment of LNG storage tanks in actual operation, a dynamic risk assessment method and system for storage tanks is provided.

[0005] To achieve the above objectives, the specific technical solution of the dynamic risk assessment method and system for storage tanks of the present invention is as follows: A method for dynamic risk assessment of storage tanks includes the following steps: S1. Identify potential risk sources for storage tanks, using tank leakage as the assessment object; S2. Input the potential risk sources and possible consequences into the fuzzy Bow-Tie model to form a preliminary qualitative analysis of consequences and measures; S3. Transform the bow-tie model into a Bayesian model network based on the mapping relationship; S4. Experts score the corresponding risk sources and convert them into the prior probabilities of the root nodes of the Bayesian network. The weights of each expert are determined by the analytic hierarchy process, and the expert evaluation results are converted into fuzzy probabilities to obtain the prior probabilities of the root nodes and determine the conditional probability tables of other nodes. S5. Calculate the posterior probability and analyze and determine the priority relationship of failure modes; S6. Based on the priority judgment results, provide corresponding risk prevention and control measures.

[0006] Furthermore, prior to step S2, the following steps are also included: Compile preventive measures for tank failures before and after failures, and input the preventive measures data into a fuzzy Bow-Tie model to correct the parameters of the fuzzy Bow-Tie model.

[0007] Furthermore, the data on preventative measures for tank failures includes dynamic risk control data, tank structural integrity assurance data, tank shell maintenance status, and tank safety protection data.

[0008] Furthermore, the data on preventative measures following tank failures includes emergency response data, closed-loop management data, tank shell repair data, and natural environment data.

[0009] Furthermore, after inputting preventive measures data into the Bow-Tie model, fuzzy logic is used to fuzzify the risk factors in the Bow-Tie model in order to correct the parameters of the fuzzy Bow-Tie model.

[0010] Furthermore, fuzzy probability can determine the conditional probabilities of other nodes through forward reasoning and obtain a conditional probability table; Other nodes include intermediate nodes and child nodes.

[0011] Further, in step S5, By calculating posterior probabilities and performing sensitivity analysis based on Bayesian networks, the failure modes with the greatest impact on overall risk are identified and their priorities are determined in order to optimize risk control measures.

[0012] Further, step S4 includes the following steps: S41. Utilize the knowledge and experience of experts to estimate the probability of each event, and divide the probability of the event occurring into seven different levels from low to high, namely: extremely low, low, slightly low, medium, slightly high, high and extremely high. S42. Use the fuzzy membership function to fuzzify the expert scoring results, as shown in the following formula: S43. Convert FPS to fuzzy failure probability FFR. The conversion formula is: In the formula, .

[0013] Furthermore, between steps S41 and S42, the following steps are also included: Use AHP to determine the weights of factors related to an expert's overall capabilities; Factors contributing to an expert's overall competence include their professional knowledge, the fairness of their scoring and evaluation, and their relevant personal experience.

[0014] A dynamic risk assessment system for storage tanks uses the aforementioned dynamic risk assessment method for storage tanks. The dynamic risk assessment system for storage tanks includes a risk source identification module, a bow-tie model, an expert group scoring module, a Bayesian network model, and a risk prevention and control module.

[0015] The dynamic risk assessment method for storage tanks of this invention has the following advantages: By constructing a dynamic risk assessment process with "tank leakage" as the assessment object, the method achieves closed-loop management from risk source identification to the output of prevention and control measures. It overcomes the problem that traditional qualitative or static risk assessment methods cannot cope with complex dynamic environments by combining a fuzzy Bow-Tie model with a Bayesian network. Especially in transforming qualitative analysis into quantitative reasoning, it improves the scientific rigor and operability of the assessment results. This method can reflect changes in the operating status of storage tanks in real time, supports dynamic updates of prior probabilities, and significantly enhances risk warning capabilities and decision support levels.

[0016] The dynamic risk assessment system for LNG storage tanks of this invention has the following advantages: It provides an integrated dynamic risk assessment system for LNG storage tanks, comprising five modules: risk source identification, Bow-Tie modeling, expert scoring, Bayesian networks, and risk prevention and control, forming an integrated hardware and software solution. Each module has a clear division of labor and well-defined interfaces, making it easy to embed into existing SCADA or safety management platforms for online assessment and real-time early warning. The modular design of the system also facilitates future functional expansion and model iteration, possessing good engineering applicability and industrialization prospects. This system truly realizes closed-loop management of "assessment—analysis—decision—feedback," providing strong technical support for the safe operation of LNG storage tanks. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the dynamic risk assessment system for storage tanks according to the present invention; Figure 2 This is the Bow-tie model network of the present invention; Figure 3 This is the Bayesian model network of the present invention.

[0018] Explanation of markings in the diagram: Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0021] The following is a reference to the appendix. Figures 1 to 3 This invention describes a method and system for dynamic risk assessment of storage tanks.

[0022] This embodiment provides a method for dynamic risk assessment of storage tanks, such as... Figures 1-3 As shown, the method includes the following steps: S1. Identify potential risk sources for storage tanks, using tank leakage as the assessment object; S2. Input the potential risk sources and possible consequences into the fuzzy Bow-Tie model to form a preliminary qualitative analysis of consequences and measures; S3. Transform the Bow-tie model into a Bayesian model network based on the mapping relationship; S4. Experts score the corresponding risk sources and convert them into the prior probabilities of the root nodes of the Bayesian network. The weights of each expert are determined by the analytic hierarchy process, and the expert evaluation results are converted into fuzzy probabilities to obtain the prior probabilities of the root nodes and determine the conditional probability tables of other nodes. S5. Calculate the posterior probability and analyze and determine the priority relationship of failure modes; S6. Based on the priority judgment results, provide corresponding risk prevention and control measures.

[0023] Understandably, the dynamic risk assessment method for storage tanks achieves closed-loop management from risk source identification to the output of control measures by constructing a dynamic risk assessment process with "tank leakage" as the assessment object. The method combines a fuzzy Bow-Tie model with a Bayesian network, overcoming the limitations of traditional qualitative or static risk assessment methods in handling complex dynamic environments. Particularly in transforming qualitative analysis into quantitative reasoning, it enhances the scientific rigor and operability of the assessment results. This method can reflect changes in the operating status of storage tanks in real time, supports dynamic updates of prior probabilities, and significantly enhances risk warning capabilities and decision support levels.

[0024] Specifically, Table 1 shows the causes of LNG storage tank malfunctions; please refer to Table 1. Furthermore, before step S2, it also includes sorting out the preventive measures before and after the tank failure, and inputting the preventive measures data into the fuzzy Bow-Tie model to correct the parameters of the fuzzy Bow-Tie model.

[0025] Understandably, introducing a data input mechanism for pre- and post-fault preventative measures before fuzzy Bow-Tie modeling allows model parameters to be corrected based on actual operation and maintenance data, improving the model's adaptability and accuracy. Compared to traditional fixed-parameter models, this method more realistically reflects the safety status of storage tanks under different operating conditions, avoiding misjudgments of risk due to neglecting preventative control measures. This step enhances the model's practicality, making it suitable not only for theoretical analysis but also for on-site safety management optimization.

[0026] Furthermore, the data on preventative measures for tank failures includes dynamic risk control data, tank structural integrity assurance data, tank shell maintenance status, and tank safety protection data.

[0027] Understandably, refining pre-failure prevention measures into four categories of data—dynamic risk control, structural integrity assurance, shell maintenance, and safety protection—achieves a systematic classification and quantification of risk factors in the early stages of the storage tank's entire lifecycle. In particular, dynamic risk control data (such as over-rush testing, pressure balancing, and pre-cooling) is directly linked to operating procedures, helping to identify human error; while structural integrity data focuses on the health status of the equipment itself, enhancing inherent safety. This classification and modeling approach enhances the comprehensiveness and relevance of risk identification.

[0028] Specifically, Table 2 shows the preventive measures taken before the LNG storage tank fails. Referring to Table 2, the preventive measures data before the storage tank fails also include safety valve inspection data. Understandably, further refining the specific indicators of key data such as dynamic risk control and structural integrity allows for a precise mapping from macro-level categories to micro-level parameters. For example, parameters such as "whether pre-cooling treatment is performed," "vacuum degree of the interlayer," and "density of insulation material filling" are all monitorable and quantifiable operational parameters, facilitating integration into automated monitoring systems. The introduction of natural disaster coefficients and construction error prevention coefficients incorporates external environmental factors and human error, enhancing the model's comprehensive judgment. This fine-grained indicator design is conducive to promoting the development of risk assessment towards intelligence and digitalization.

[0029] Furthermore, the data on preventative measures following tank failures includes emergency response data, closed-loop management data, tank shell repair data, and natural environment data.

[0030] Understandably, dividing post-failure preventative measures into four data categories—emergency response, closed-loop management, shell repair, and natural environment—constructs a comprehensive accident response and recovery capability assessment system. Emergency response data focuses on the efficiency of personnel evacuation and leak source control, directly impacting the severity of the accident's consequences; closed-loop management data reflects the system's self-regulation capabilities; and natural environment data considers external influencing factors, enhancing the assessment's external adaptability. This categorization structure provides a more reliable basis for subsequent emergency resource allocation and contingency plan development, strengthening overall risk resilience.

[0031] Specifically, Table 3 shows the response measures after an LNG storage tank malfunction. Referring to Table 3, the preventive measures data after a storage tank malfunction also include safety valve inspection data. Understandably, clearly defining core parameters in emergency response and closed-loop management, such as leak source control coefficients, personnel evacuation coefficients, and closed-loop leak data, makes the previously difficult-to-quantify emergency management process assessable and comparable. For example, by setting a "personnel evacuation coefficient," the efficiency of emergency plan execution can be evaluated; the "closed-loop vacuum coefficient" can be used to measure the self-stabilizing ability of storage tanks after an anomaly. These quantitative indicators provide a reliable input basis for subsequent probability calculations of Bayesian networks, significantly improving the accuracy and practicality of posterior analysis.

[0032] Furthermore, after inputting preventive measures data into the Bow-Tie model, fuzzy logic is used to fuzzify the risk factors in the Bow-Tie model in order to correct the parameters of the fuzzy Bow-Tie model.

[0033] Understandably, fuzzy logic can handle linguistic descriptions (such as "higher risk" or "moderate probability"), converting them into mathematically meaningful membership values, thereby improving the integration between subjective experience and objective models. This method not only preserves the value of expert knowledge but also enhances the model's robustness to uncertain information, improving the overall credibility of risk assessment.

[0034] Specifically, the consequences of a malfunction in the LNG storage tank in step S2 can be found in Table 4. Furthermore, fuzzy probability can determine the conditional probabilities of other nodes through forward reasoning and obtain a conditional probability table; other nodes include intermediate nodes and child nodes.

[0035] Understandably, the proposed Conditional Probability Table (CPT) for determining intermediate and child nodes using fuzzy probability through forward reasoning solves the technical challenge of obtaining conditional probabilities in Bayesian networks. Traditional methods often rely on large amounts of historical statistical data, which is scarce in the storage tank field. This method generates the CPT through expert scoring, fuzzy transformation, and logical reasoning, reducing data dependence while ensuring the effectiveness of network reasoning. This mechanism is particularly suitable for risk modeling of low-frequency, high-risk events and has strong engineering application value.

[0036] Furthermore, in step S5, the posterior probability is calculated based on the Bayesian network and a sensitivity analysis is performed. Through the sensitivity analysis of the Bayesian network, the failure modes that have the greatest impact on the overall risk are identified, their priorities are determined, and risk prevention and control measures are optimized.

[0037] Understandably, introducing sensitivity analysis based on posterior probability calculations can identify the failure modes (i.e., critical paths) that contribute the most to overall risk. This function allows managers to prioritize the most impactful aspects, concentrating resources on rectification or monitoring to achieve "precise prevention and control." For example, if analysis shows that "vacuum failure in the interlayer" has the greatest impact on the overall risk, then vacuum monitoring and maintenance should be strengthened as a priority. This feature greatly improves the targeting and cost-effectiveness of risk prevention and control measures, providing crucial support for proactive safety management.

[0038] Further, step S4 includes the following steps: S41. Utilize the knowledge and experience of experts to estimate the probability of each event, and divide the probability of the event occurring into seven different levels from low to high, namely: extremely low, low, slightly low, medium, slightly high, high and extremely high. S42. Use the fuzzy membership function to fuzzify the expert scoring results, as shown in the following formula: S43. Convert FPS to fuzzy failure probability FFR. The conversion formula is: In the formula, .

[0039] Understandably, by using a seven-level probability grading system (extremely low to extremely high) and fuzzy membership functions to process expert scores, a standardized and mathematical expression of subjective judgments is achieved. This grading system covers a wide range of possibilities, avoiding the information loss caused by binary (yes / no) or three-stage (high / medium / low) evaluations. Combined with the conversion formula from FPS to FFR, linguistic evaluation is further transformed into fuzzy failure probabilities that can be used for Bayesian inference, bridging the gap between "human judgment" and "machine computation." This method significantly improves the efficiency of expert knowledge utilization and the scientific rigor of model input.

[0040] Specifically, Table 5 shows the calculated expert weight scoring table for this embodiment. Furthermore, between steps S41 and S42, the following steps are also included: Use AHP to determine the weights of factors related to an expert's overall capabilities; Factors contributing to an expert's overall competence include their professional knowledge, the fairness of their scoring and evaluation, and their relevant personal experience.

[0041] Understandably, introducing AHP (Analytic Hierarchy Process) to determine expert weights in the expert scoring process changes the previous crude approach of "average weighting." By comprehensively evaluating experts across three dimensions—professional knowledge, impartiality, and relevant experience—differentiated weights are assigned to different experts, ensuring more objective and authoritative evaluation results. Especially in scenarios involving multiple experts, it effectively suppresses the interference of individual non-professional opinions on the overall results, improves the accuracy of the root node's prior probability, and thus guarantees the reliability of the entire Bayesian network inference chain.

[0042] Specifically, after step S4, the weights of each expert are determined using the Analytic Hierarchy Process (AHP), and the expert evaluation results are converted into fuzzy probabilities to obtain the prior probability of the root node. Based on the logical relationships, forward reasoning using Batch Normalization (BN) is applied to determine the conditional probability tables (CPTs) of other nodes (including intermediate nodes and child nodes). The specific calculation results are shown in Table 6. Specifically, after step S5, the failure modes with the greatest impact on the overall risk are identified through sensitivity analysis of the Bayesian network, and their priorities are determined in order to optimize risk prevention and control measures. Table 7 is a comparison table of the probabilities of LNG storage tank failure events, and Table 8 is the sensitivity analysis calculation results of each basic event.

[0043] Table 7 Comparison of the prior and subsequent probabilities of events affecting LNG storage tank malfunctions Table 8. Sensitivity Analysis Results for Each Basic Event This embodiment also provides a dynamic risk assessment system for storage tanks, which uses the above-mentioned dynamic risk assessment method for storage tanks. The dynamic risk assessment system for storage tanks includes a risk source identification module, a Bow-tie model, an expert group scoring module, a Bayesian network model, and a risk prevention and control module.

[0044] This dynamic risk assessment system for LNG storage tanks provides an integrated solution encompassing five modules: risk source identification, Bow-Tie modeling, expert scoring, Bayesian networks, and risk prevention and control. These modules form a unified hardware and software solution. Each module has a clear division of labor and well-defined interfaces, making it easy to integrate into existing SCADA or safety management platforms for online assessment and real-time early warning. The system's modular design also facilitates future functional expansion and model iteration, demonstrating good engineering applicability and industrialization prospects. This system truly achieves closed-loop management of "assessment—analysis—decision—feedback," providing strong technical support for the safe operation of LNG storage tanks.

[0045] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for dynamic risk assessment of storage tanks, characterized in that, Includes the following steps: S1. Identify potential risk sources for storage tanks, using tank leakage as the assessment object; S2. Input the potential risk sources and possible consequences into the fuzzy Bow-Tie model to form a preliminary qualitative analysis of consequences and measures; S3. Transform the Bow-tie model into a Bayesian model network based on the mapping relationship; S4. Experts score the corresponding risk sources and convert them into the prior probabilities of the root nodes of the Bayesian network. The weights of each expert are determined by the analytic hierarchy process, and the expert evaluation results are converted into fuzzy probabilities to obtain the prior probabilities of the root nodes and determine the conditional probability tables of other nodes. S5. Calculate the posterior probability and analyze and determine the priority relationship of failure modes; S6. Based on the priority judgment results, provide corresponding risk prevention and control measures.

2. The dynamic risk assessment method for storage tanks according to claim 1, characterized in that, Before step S2, the following is also included: Compile preventive measures for tank failures before and after failures, and input the preventive measures data into a fuzzy Bow-Tie model to correct the parameters of the fuzzy Bow-Tie model.

3. The dynamic risk assessment method for storage tanks according to claim 2, characterized in that, Data on preventative measures for storage tank failures includes dynamic risk control data, data on ensuring the structural integrity of the storage tank, data on the maintenance status of the storage tank shell, and data on the safety protection of the storage tank.

4. The dynamic risk assessment method for storage tanks according to claim 2, characterized in that, Data on preventative measures following tank malfunctions includes emergency response data, closed-loop management data, tank shell repair data, and natural environment data.

5. The dynamic risk assessment method for storage tanks according to any one of claims 2-4, characterized in that, After inputting preventive measures data into the Bow-Tie model, the risk factors in the Bow-Tie model are fuzzified using fuzzy logic methods to correct the parameters of the fuzzy Bow-Tie model.

6. The dynamic risk assessment method for storage tanks according to claim 5, characterized in that, Fuzzy probability can determine the conditional probabilities of other nodes through forward reasoning and obtain a conditional probability table; Other nodes include intermediate nodes and child nodes.

7. The dynamic risk assessment method for storage tanks according to claim 1, characterized in that, In step S5, By calculating posterior probabilities and performing sensitivity analysis based on Bayesian networks, the failure modes with the greatest impact on overall risk are identified and their priorities are determined in order to optimize risk control measures.

8. The dynamic risk assessment method for storage tanks according to claim 1, characterized in that, Step S4 includes the following steps: S41. Utilize the knowledge and experience of experts to estimate the probability of each event, and divide the probability of the event occurring into seven different levels from low to high, namely: extremely low, low, slightly low, medium, slightly high, high and extremely high. S42. Use the fuzzy membership function to fuzzify the expert scoring results, as shown in the following formula: S43. Convert FPS to fuzzy failure probability FFR. The conversion formula is: In the formula, .

9. The dynamic risk assessment method for storage tanks according to claim 1, characterized in that, Between steps S41 and S42, the following steps are also included: Use AHP to determine the weights of factors related to an expert's overall capabilities; Factors contributing to an expert's overall competence include their professional knowledge, the fairness of their scoring and evaluation, and their relevant personal experience.

10. A dynamic risk assessment system for storage tanks, characterized in that, The dynamic risk assessment method for storage tanks as described in any one of claims 1-9 is used. The dynamic risk assessment system for storage tanks includes a risk source identification module, a Bow-tie model, an expert group scoring module, a Bayesian network model, and a risk prevention and control module.

Images