A petrochemical RBI risk management system

Abstract

The present application relates to the technical field of petrochemical industry equipment risk management, in particular to a petrochemical RBI risk management system, comprising an industrial internet of things collection terminal, a risk calculation server, a blockchain storage node and a collaborative terminal. The industrial internet of things collection terminal is used for collecting field equipment parameters; the risk calculation server runs a risk management program, realizes standardized access and fusion of multi-source equipment data through a data collection layer, constructs a quantitative model of failure probability and consequence and matches a failure mechanism through a risk modeling layer, dynamically generates and optimizes an inspection strategy through a strategy optimization layer, monitors risks and triggers multi-level early warning through a monitoring and early warning layer, and realizes task closed loop and data sharing through a collaborative management layer; the blockchain storage node is used for data credible storage. The present application realizes the digitization and intelligentization of petrochemical equipment risk management, solves the problems of data island, extensive evaluation and response lag, and significantly improves the level of equipment safety management and economy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of petrochemical industry equipment risk management, and more particularly to a petrochemical RBI risk management system. The RBI refers to risk-based inspection, a method for risk assessment and management of plant equipment (such as pressure vessels, pipelines, safety valves, etc.). It optimizes inspection plans and resource allocation by analyzing the likelihood of equipment failure and the severity of consequences, ensuring safety while reducing costs. BACKGROUND

[0002] In the prior art, the RBI (Risk-Based Inspection) risk management of petrochemical industry for equipment mainly relies on traditional decentralized data processing, empirical risk assessment, static inspection strategy and lagging monitoring and coordination scheme, and has not yet formed a digital, dynamic and intelligent whole-process management system. Specifically, the core implementation mode of the existing RBI risk management scheme includes: integrating equipment data scattered in the ERP (Enterprise Resource Planning) system, the DCS (Distributed Control System) and various offline detection reports by artificial means, relying on expert subjective experience to assess the likelihood of equipment failure (POF) and the consequences of failure (COF), adopting a fixed cycle (such as uniform 3 years) inspection strategy, and updating the equipment risk state through artificial analysis once every quarter or half a year, and the inspection task tracking is disconnected between departments (equipment department, process department, safety department) due to data barriers.

[0003] The existing petrochemical RBI risk management scheme has four core technical bottlenecks, which are difficult to meet the dual needs of high safety guarantee and low cost management of petrochemical industry for equipment, and the specific technical problems are as follows:

[0004] First, the data integration efficiency is low and the data quality is poor. The data required by the existing RBI management is stored in multiple independent systems such as ERP system, DCS system and offline detection report, and the data format is not unified, which requires artificial integration. The artificial integration process occupies more than 60% of the risk analysis cycle, seriously affecting the management efficiency; at the same time, unstructured data such as detection reports accounts for more than 50% of all data, and the error rate of manual extraction of such data is ≥20%, the data credibility is insufficient, which directly restricts the effectiveness of the subsequent risk assessment results.

[0005] Second, risk assessment relies on experience and judgment, and the assessment results are rough. The existing RBI management system (such as early RBI tools) lacks standardized quantitative model support and mainly relies on expert subjective judgment to carry out failure probability (POF) and failure consequence (COF) assessment, which can only divide the risk into three fuzzy levels of "high / medium / low", and cannot achieve accurate quantification. The measured data of a certain refinery shows that the error of using this type of assessment method can reach ±40%, which is easy to lead to missed detection of high-risk equipment and hidden safety hazards.

[0006] Third, the inspection strategy is static and fixed, and the adaptability is poor. The existing scheme adopts a "one-size-fits-all" fixed cycle inspection strategy, which does not consider the real-time state differences such as corrosion rate changes and process parameter fluctuations during equipment operation, and cannot dynamically adjust the inspection frequency and detection method. Industry statistics show that due to the rigidity of such strategy, the unnecessary cost caused by over-inspection in the petrochemical industry exceeds 3 billion yuan per year, and the leakage accidents of high-risk equipment caused by insufficient inspection frequency account for 65% of the total number of industry equipment accidents, making it difficult to balance safety and economy.

[0007] Fourth, the risk monitoring response is lagging and the multi-department cooperation is insufficient. The existing RBI management relies on periodic manual analysis for updating the risk state of equipment, and the update cycle is usually quarterly or semi-annually, which cannot timely capture the risk changes caused by dynamic corrosion, process fluctuations and other factors, and the risk dynamic missing judgment rate is high; in addition, there are serious data barriers between equipment, process, safety and other departments, and data cannot be shared, leading to disconnection of inspection task tracking, and the average response period after abnormal situation occurs is ≥7 days, which cannot meet the rapid disposal demand.

[0008] The above defects of the existing technology, combined with industry standard requirements, actual pain points and technical support conditions, promote the research and development of the technical scheme of the present application, and the specific research and development background is as follows:

[0009] From the perspective of standard driving, API581 version 4 (2025) has compulsorily required that the petrochemical RBI management needs to realize "dynamic risk assessment" and "multi-source data fusion", and the domestic "RBI Work Specification for Pressure Equipment" also clearly proposes that the RBI management system needs to complete "digital and intelligent upgrading", and the existing static and experiential RBI scheme cannot meet the latest industry standard requirements, and urgent technical upgrading is needed.

[0010] From the perspective of industry pain points, the annual equipment maintenance cost of petrochemical enterprises exceeds 200 million yuan, of which the accident loss caused by inaccurate RBI risk assessment accounts for 30%; the daily loss caused by unplanned shutdown of large-scale refining and chemical devices due to equipment failure exceeds 8 million yuan, and the industry has an urgent need for RBI technology that can realize accurate risk control and reduce cost and loss.

[0011] From the perspective of technical support, the industrial Internet of Things (5G communication, edge computing) technology can realize real-time collection of device data, machine learning (reinforcement learning, LSTM neural network) technology provides algorithm support for intelligent decision-making, and blockchain technology can guarantee the reliable storage of data. The maturity of the above technologies provides the feasibility for the digitalization and intelligentization upgrade of petrochemical RBI management, and creates conditions for solving the defects of existing technologies. SUMMARY

[0012] Therefore, the present application provides a petrochemical RBI risk management system, which aims to solve the problems of low integration efficiency caused by data silos, insufficient evaluation accuracy caused by experience dependence, test resource mismatch caused by static strategy (specifically, over-testing of low-risk equipment increases costs or insufficient testing of high-risk equipment), and delayed response caused by lagging monitoring in traditional petrochemical RBI risk management, thereby improving the digitalization, intelligentization and dynamization level of petrochemical equipment risk management, and achieving the comprehensive goal of improving the safety and optimizing the economy of petrochemical equipment risk management.

[0013] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0014] A petrochemical RBI risk management system, comprising: an industrial Internet of Things collection terminal, a risk calculation server, a blockchain storage node, and a collaborative terminal;

[0015] The industrial Internet of Things collection terminal is deployed in the petrochemical device field and is used for collecting device operation and environmental parameters;

[0016] The risk calculation server is in communication connection with the industrial Internet of Things collection terminal and is used for running a risk management program, which, when executed by a processor, is used to realize:

[0017] Through the data collection layer, the standardized access, cleaning and fusion of petrochemical equipment multi-source data are realized;

[0018] Through the risk modeling layer, a quantitative model of failure probability and consequence is constructed, and the device failure mechanism is matched;

[0019] Through the strategy optimization layer, an inspection strategy is generated and dynamically optimized based on the risk level, and the inspection resources are allocated;

[0020] Through the monitoring and early warning layer, the device risk changes are monitored in real time, the abnormalities are identified and the multi-level early warning is triggered;

[0021] Through the collaborative management layer, multi-department data sharing, inspection task closed-loop management and performance iterative optimization are realized;

[0022] The blockchain storage node is in communication connection with the risk calculation server and is used to realize the reliable storage and traceability of detection data and strategy scheme.

[0023] The cooperative terminal is in communication connection with the risk calculation server, and is used for providing a man-machine interactive interface for a user.

[0024] In a specific implementable scheme, the industrial Internet of Things collection terminal comprises an intelligent sensor and an edge gateway.

[0025] The intelligent sensor at least comprises a temperature sensor for measuring temperature, a pressure sensor for measuring pressure and a corrosion probe for measuring corrosion rate, wherein the temperature sensor adopts a PT100 type, the measurement range is -50~300℃, the pressure sensor adopts a strain gauge type, the measurement range is 0~10MPa, and the corrosion probe measurement range is 0~1mm / a.

[0026] The edge gateway is used for protocol conversion and data preprocessing, is configured with a 4-core ARM CPU and 2GB memory, supports data transmission through LoRa, 5G or Ethernet, has an IP65 protection level and is adapted to an industrial environment of -40~85℃.

[0027] In a specific implementable scheme, the implementation of the data collection layer comprises:

[0028] The enterprise resource planning system is interfaced to obtain equipment basic account information, and the account information at least covers equipment material, specification, design pressure, design temperature and service life;

[0029] Process operation parameters are collected through the Internet of Things gateway, the process parameters include temperature, pressure, flow, medium composition, the sampling frequency is 1 time / minute, and the transmission delay is ≤10 seconds;

[0030] Offline detection results of ultrasonic thickness measurement and ray detection are integrated, the results include wall thickness, defect size and corrosion rate, and are stored in association according to equipment position number;

[0031] Online corrosion probes and corrosion inhibitor injection records are accessed to establish a corrosion environment feature library, the feature library contains concentration, content, pH value and other parameters.

[0032] In a specific implementable scheme, the data collection layer is used for executing data cleaning and standardization, comprising:

[0033] 3σ criterion is adopted to identify outlier data, and whether to correct is determined in combination with equipment operation conditions (such as start-up and shutdown stages), and the abnormal treatment accuracy is ≥95%;

[0034] Random forest prediction model is adopted to fill in missing values of continuous data, and the error is ≤5%;

[0035] Mode filling discrete data missing values, to ensure data integrity;

[0036] Logical consistency verification is performed through device parameter correlation rules (such as wall thickness and design pressure matching relationship), and the verification pass rate is ≥97%.

[0037] In a specific implementation, the failure possibility quantification model in the risk modeling layer is a multi-factor coupling model based on corrosion damage, fatigue damage, stress level and detection effectiveness;

[0038] The failure possibility model is divided into five levels according to the calculation results, which are level 1 (<0.01), level 2 (0.01-0.05), level 3 (0.05-0.1), level 4 (0.1-0.2), and level 5 (>0.2);

[0039] The failure consequence assessment model is a three-dimensional consequence quantification model based on safety impact, environmental impact and economic loss, and is divided into five levels according to the assessment results, which are A (<2), B (2-4), C (4-6), D (6-8), and E (>8).

[0040] In a specific implementation, the matching device failure mechanism in the risk modeling layer is a decision tree algorithm, which matches with a preset petrochemical equipment damage mechanism library according to 15 input characteristics including temperature, medium composition and material;

[0041] The damage mechanism library covers at least 12 types of corrosion mechanisms (such as high-temperature sulfur corrosion and wet hydrogen sulfide corrosion) and 8 types of mechanical damage mechanisms (such as fatigue cracks and creep), and the mechanism matching accuracy is ≥90%.

[0042] In a specific implementation, the dynamic optimization of inspection strategy in the strategy optimization layer is to use a reinforcement learning algorithm to adaptively adjust the inspection period according to the device risk change rate ΔR, and the adjustment formula logic is new period = original period × (1-adjustment coefficient k × risk change rate ΔR), wherein the adjustment coefficient k is in the range of 0.2-0.5;

[0043] The strategy optimization layer also allocates inspection resources through an integer programming model to minimize the total detection cost, including equipment detection cost and equipment downtime cost, and the cost saving rate is ≥20%.

[0044] In a specific implementation, real-time monitoring in the monitoring and early warning layer is to use a sliding window algorithm to calculate the risk value in real time, and the sliding window size is set to 30 minutes, and the risk value update frequency is ≥1 times / hour;

[0045] At the same time, long short-term memory neural network is used to predict future risk trends, and the prediction period is 3 months in the future.

[0046] The multi-level early warning is triggered by combining static threshold and dynamic threshold, the static threshold is based on API standard setting, the deviation > 10% triggers early warning, the dynamic threshold is prediction value + 3 times error standard deviation, the early warning level is divided into general (yellow), important (orange) and urgent (red), and the response time of red early warning is ≤1 hour;

[0047] The monitoring and early warning layer also performs abnormality tracing based on the knowledge graph containing 5000+ petrochemical accident cases, the tracing time is ≤2 hours, and the accuracy is ≥90%.

[0048] In a specific implementable scheme, the multi-department data sharing of the collaborative management layer is based on a role access control model to assign different data permissions to the equipment department, the process department and the safety department, wherein the equipment department can access all data, the process department only accesses operation parameters, and the safety department only accesses risk reports, and the permission accuracy is 100%;

[0049] The inspection task closed-loop management is to automatically generate a work order, which contains equipment position number, detection method, responsible person and time limit information, and track the execution progress through a Gantt chart, and automatically remind if overdue, and the task completion rate is ≥98%;

[0050] The collaborative management layer updates the risk model based on the inspection data every quarter, calculates the risk management benefit, and the benefit logic is inspection cost saving + accident loss reduction - system investment, and the effectiveness of risk management and control is improved by ≥10% every quarter.

[0051] In a specific implementable scheme, the system realizes data interaction between each component through industrial Ethernet and RESTful API interface, and adopts Kafka message queue to process asynchronous communication, and the peak throughput is ≥500TPS;

[0052] The system adopts a hybrid data storage architecture, including a MySQL database (for storing structured data), an InfluxDB database (for storing time series data) and a blockchain (for storing certification data), and the data query response time is ≤1 second;

[0053] The system security conforms to the IEC 62443 standard, a firewall is configured, and an AES-256 encryption algorithm is used to encrypt data.

[0054] Compared with the prior art, the petrochemical RBI risk management system can realize the digital management of the whole process of the risk-based inspection of petrochemical equipment, and through the technical architecture of multi-source equipment data acquisition and cleaning, risk quantization modeling and failure mechanism matching, intelligent generation and optimization of inspection strategy, dynamic risk monitoring and early warning, and whole-process collaborative closed-loop management, the petrochemical equipment risk management is realized dynamically and intelligently by combining the industrial Internet of Things, machine learning prediction and blockchain storage technology, and the equipment safety management level and the economic efficiency of resource allocation are effectively improved, and the petrochemical RBI risk management system has the following beneficial effects:

[0055] First, the data integration efficiency and quality are significantly improved.

[0056] The petrochemical RBI risk management system realizes the automatic integration and cleaning of equipment basic data, operation data, detection data and corrosion data by multi-source equipment data standardized acquisition and fusion processing and a multi-source data fusion algorithm based on Dempster-Shafer evidence theory (in line with the API581 fourth edition (2025) multi-source data reliability evaluation recommended method), solves the problems of low efficiency and high error rate of traditional manual integration, and makes the data integration efficiency improved by 90% and the data accuracy rate improved to more than 96%.

[0057] Data support statement:

[0058] The calculation basis of the 90% efficiency improvement is that, taking 200 key equipment (including 30 towers, 20 reactors and 150 pipe sections) of a refinery as test samples, the traditional manual integration needs to extract data from ERP, DCS and detection reports one by one and manually associate, and the average time consumption is 72 hours; the system can complete the integration of the same samples only by 7.2 hours through protocol adaptation (Modbus / OPC UA) and automatic association algorithm, (Symbol explanation: T 传统 is the time consumption of traditional manual integration, and T 系统 is the time consumption of automatic integration of the system).

[0059] The verification basis of the accuracy rate of more than 96% is that 5000 multi-source data (including 1000 equipment basic data, 2000 operation data, 1200 detection data and 800 corrosion data) are input during the test, 4802 of which are "real correct data" marked by manual, and 198 of which are error data identified after system fusion,

[0060]

[0061] (Symbol explanation: is the total amount of multi-source data for testing, is the amount of error data identified after system fusion), which meets the quantitative index of "more than 96%".

[0062] Algorithm reliability basis: the fusion logic of Dempster-Shafer evidence theory has been verified by API581 Appendix B "Multi-source data reliability assessment case" in the fourth edition, and the accuracy rate of the system in the test for processing conflicting data (such as wall thickness deviation of the same device by different detection methods) reaches 97%, which is consistent with the standard case.

[0063] Second, greatly improve the accuracy and reliability of risk assessment.

[0064] The present application realizes the accurate division of risk grade and the accurate identification of failure mechanism by constructing a multi-factor coupled failure probability (POF) quantification model and a safety-environment-economy three-dimensional consequence (COF) assessment model, and matching the typical failure mechanism of petrochemical equipment based on the decision tree algorithm (training set containing 5000+ petrochemical equipment failure cases), the risk assessment error is controlled within 10%, the mechanism matching accuracy is increased to 90%, and the missed judgment of high-risk equipment is effectively avoided.

[0065] Data support statement:

[0066] Verification basis of POF calculation error ≤10%: 100 devices with known risk grade are selected (from API581 Appendix C "POF verification case library" in the fourth edition, the reference POF value range is 0.02-0.18), the average deviation of the POF value calculated by the system and the reference value is 4.17%. For example: the reference POF of a hydrogenation reactor is 0.12, the calculated value of the system is 0.115,

[0067]

[0068] (Symbol interpretation: POF 基准 is the reference failure probability value in the API581 verification case library, POF 系统 is the failure probability value calculated by the system), which meets the "error ≤10%".

[0069] Calculation basis of mechanism matching accuracy 90%: the failure mechanism library clearly contains 20 typical mechanisms (12 types of corrosion: high-temperature sulfur corrosion, wet hydrogen sulfide corrosion, naphthenic acid corrosion, etc.; 8 types of mechanical damage: fatigue crack, creep, erosion, hydrogen-induced cracking, etc., according to the classification in "Petrochemical Equipment Corrosion and Protection Handbook" (2022 edition)); 180 device failure cases (covering 20 types of mechanisms) are tested, and the system correctly matches 162, ;

[0070] Among them, is the total number of device failure cases tested, is the number of cases correctly matched by the system.

[0071] Third, realize the balance optimization of economic and safety of inspection strategy.

[0072] The application generates a dynamic inspection scheme based on the risk level of the equipment, and realizes adaptive optimization of the inspection cycle and resources through a reinforcement learning algorithm (reward function aiming at "safety improvement + cost saving") and an integer programming model, while ensuring that the defect detection rate of high-risk equipment is not less than 98%, the excessive inspection cost of low-risk equipment is significantly reduced.

[0073] Verification basis for high-risk equipment defect detection rate ≥98%: select 50 high-risk equipment (POF ≥4 level, 50 known hidden defects, from third-party detection agency preset defect samples), traditional fixed strategy detects 45, the system optimization strategy (100% coverage + ultrasonic + X-ray detection) detects 49, ;

[0074] Among them, is the number of known hidden defects, is the number of detected defects.

[0075] Fourth, greatly improve the risk response speed and early warning ability.

[0076] The application realizes real-time calculation and trend prediction of risk value based on sliding window algorithm (window size 30 minutes) and LSTM neural network (training set containing 3-year equipment risk time series data) through real-time risk monitoring and multi-level early warning mechanism, and uses knowledge graph (containing 5000+ petrochemical accident cases) for abnormal source analysis, risk update frequency is increased to once per hour, red early warning response time is shortened to within one hour, which fundamentally solves the problem of lagging response of traditional risk management.

[0077] Data support statement:

[0078] Test conditions for risk update frequency once per hour: the test environment is industrial Ethernet (gigabit bandwidth), deploy Kafka message queue (peak throughput 500TPS), sliding window algorithm calculates risk value every 30 minutes, 1 hour completes the whole process of "data collection-calculation-storage-visualization update", without manual intervention, meeting the real-time requirement of "once per hour".

[0079] Statistical range of red early warning response time ≤1 hour: the response time is defined as the whole process of "system triggering early warning → work order sending to inspection personnel terminal", including automatic early warning push (within 10 seconds, based on RESTful API), work order generation (within 5 minutes, based on preset template), personnel receiving confirmation (within 30 minutes, mobile terminal push); test 20 times of simulated red early warning scene (such as corrosion rate sudden increase of 0.2mm / a), average response time 42 minutes, maximum 55 minutes, all ≤1 hour.

[0080] Verification of traceability time ≤2 hours: 30 abnormal early warning cases (such as sudden pressure rise, corrosion exceeding standard) are selected, and the system locates the root cause based on the knowledge graph (including "abnormal index-direct cause-root cause" causal chain), with an average time of 1.2 hours and a maximum time of 1.8 hours, meeting the requirement of "≤2 hours".

[0081] Fifth, form multi-department cooperation and management closed loop.

[0082] The application realizes data sharing and task cooperation tracking of multiple departments such as equipment, process and safety through the permission management and work order closed loop management mechanism based on the RBAC model (role permission conforms to ISO31000 "Risk Management Role and Responsibility Specification"), so that the inspection task completion rate is improved to more than 98%, the annual equipment risk reduction rate reaches 15%, and the overall management efficiency and continuous optimization capability are significantly improved.

[0083] Data support statement:

[0084] Statistical basis for task completion rate ≥98%: The statistical period is 1 quarter (90 days), the system generates 100 inspection work orders (covering high / medium / low risk equipment), 2 are overdue (due to temporary maintenance conflict of equipment), 98 are completed, the completion rate is , the statistical range includes "work order generation-execution-acceptance-archiving" full closed loop, and does not include special work orders exempted by humans.

[0085] Annual risk reduction rate ≥15% comparison benchmark: Take the average risk value of equipment in 2023 before the system goes online as the benchmark (65 points, 100-point system, based on 5x5 risk matrix quantification), and the average risk value in 2024 after the system goes online is reduced to 55.25 points, with a reduction rate , the risk reduction is due to the reduction of high-risk equipment from 20 to 17 (defect repair in advance) and the optimization of inspection strategy for medium-risk equipment from 50 to 42.

[0086] RBAC permission accuracy rate 100% verification: Test 100 times of cross-department data access requests (equipment department accesses all data, process department accesses running data, safety department accesses risk report), and there is no permission boundary or access failure, meeting the requirement of "permission accuracy rate 100%", verified according to ISO27001 "Information Security Permission Management Specification". BRIEF DESCRIPTION OF DRAWINGS

[0087] In order to more clearly illustrate the technical solutions of the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only embodiments of the application, and those skilled in the art can obtain other drawings according to the provided drawings without creating any inventive labor.

[0088] Figure 1 The overall architecture and data flow diagram of the petrochemical RBI risk management system. DETAILED DESCRIPTION

[0089] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0090] The present application takes petrochemical equipment full life cycle data fusion as the core, realizes the fusion management and real-time mapping of multi-source heterogeneous data by constructing a dynamic twin body covering the full life cycle data of the equipment. The system performs the full-process operation of "multi-source equipment data acquisition and cleaning → risk quantification modeling and failure mechanism matching → inspection strategy intelligent generation and optimization → dynamic risk monitoring and early warning → full-process collaborative closed-loop management" through a computer program, and realizes dynamic identification, causal tracing and quantitative evaluation of risks relying on intelligent algorithms.

[0091] The present application integrates risk matrix theory, corrosion mechanism analysis, machine learning prediction and other technologies, aims to solve the problems of data island, extensive risk assessment, static strategy and lagging response in traditional RBI management, and automatically generates risk inspection reports and rectification schemes that meet industry standards, and finally realizes the full-process digitalization, dynamic and intelligent management of petrochemical equipment based on risk-based inspection (RBI), so as to comprehensively improve the accuracy, real-time and compliance of petrochemical enterprise equipment risk management, and effectively reduce the probability of unplanned shutdown and safety accidents.

[0092] As shown in Figure 1 The petrochemical RBI risk management system of the present application is applied to the field of petrochemical industry equipment integrity management and risk-based inspection, and the hardware carrier composition of the management system includes: an industrial Internet of Things collection terminal, a risk calculation server, a block chain storage node and a collaborative terminal.

[0093] ① Industrial Internet of Things collection terminal: deployed on the site of petrochemical devices, containing intelligent sensors (temperature: PT100, -50~300℃; pressure: strain gauge, 0~10MPa; corrosion: resistance probe, 0~1mm / a), edge gateway (4-core ARM CPU, 2GB memory, supporting LoRa / 5G / Ethernet), sampling frequency 1Hz-1kHz adjustable, protection level IP65, suitable for -40~85℃ industrial environment.

[0094] ②Risk calculation server: industrial server (16-core CPU + 32 GB memory + 2 TB SSD), deploy risk quantification model and failure mechanism library, support parallel computing (≥100 devices evaluated simultaneously), single device risk calculation time ≤10 seconds.

[0095] ③Blockchain storage node: 3 redundant servers (8-core CPU + 16 GB memory), adopt consortium chain architecture, realize detection data and strategy scheme on-chain storage, block generation time ≤30 seconds, data cannot be tampered with and can be traced.

[0096] ④Collaborative terminal: engineer workstation (i7 CPU + 32 GB memory), mobile inspection terminal (Android system, supports offline data entry), communicates with the system through industrial Ethernet, response time ≤2 seconds.

[0097] In this embodiment, the industrial Internet of Things collection terminal is deployed in the petrochemical device field to collect device operation and environmental parameters such as temperature, pressure, and corrosion rate;

[0098] The risk calculation server is in communication connection with the industrial Internet of Things collection terminal, and the risk management program running inside is configured to perform the following functions: through the data acquisition layer, realize standardized access, cleaning and fusion of multi-source device data; through the risk modeling layer, build quantitative models of failure probability and consequences and match device failure mechanisms; through the strategy optimization layer, dynamically generate and optimize inspection strategies based on risk levels; through the monitoring and early warning layer, real-time monitor risk changes and trigger multi-level early warning; through the collaborative management layer, realize multi-department data sharing and inspection task closed-loop management;

[0099] The blockchain storage node is in communication connection with the risk calculation server, and is used for trusted storage and traceability of key detection data and strategy scheme;

[0100] The collaborative terminal is in communication connection with the risk calculation server, and provides a human-computer interaction interface for device engineers and inspectors, for receiving early warnings, processing work orders and entering data.

[0101] Further, the software architecture of the management system adopts a layered design, including:

[0102] ①Data acquisition layer

[0103] Function: Standardized access, cleaning and fusion of multi-source data of petrochemical equipment

[0104] Core components:

[0105] Protocol adaptation module (supports 8+ industrial protocols such as Modbus / OPC UA / HTTP)

[0106] Data cleaning engine (abnormal value detection, missing value filling algorithm library)

[0107] Multi-source fusion module (credibility fusion tool based on Dempster-Shafer theory)

[0108] Technical features: support 50,000 data access per second, fusion accuracy ≥ 96%.

[0109] This layer serves as the foundation of the system, mainly responsible for the standardized access, cleaning and fusion of multi-source data of petrochemical equipment. Its core function is to support multiple industrial communication protocols through the protocol adaptation module to achieve comprehensive collection of equipment basic data, running data, detection data and corrosion data; through the built-in data cleaning engine, the collected data is subjected to outlier detection and missing value filling; finally, through the multi-source fusion module based on evidence theory, the fusion calculation of the credibility of heterogeneous data is completed, providing a high-quality and consistent data basis for the upper risk modeling.

[0110] ②Risk modeling layer

[0111] Function: Failure probability (POF) and consequence quantification (COF), failure mechanism matching

[0112] Core components:

[0113] POF calculation engine (multi-factor coupling model of API581)

[0114] COF evaluation module (three-dimensional consequence quantification tool for safety / environment / economy)

[0115] Mechanism matching system (decision tree classification model + 20 class damage feature library)

[0116] Technical features: risk level division accuracy ≥ 92%, mechanism matching accuracy ≥ 90%.

[0117] This layer is the core calculation engine of the system, mainly responsible for the quantitative evaluation of failure probability and consequence, and the intelligent matching of equipment failure mechanism. Through the failure probability calculation engine, it calculates the precise failure probability value according to the multi-factor coupling model; through the failure consequence evaluation module, it quantifies the severity of failure consequences from the safety, environmental and economic dimensions; through the mechanism matching system integrated with the decision tree algorithm and the damage feature library, it automatically identifies the potential failure mode of the equipment, and finally outputs the quantitative risk level and specific failure mechanism.

[0118] ③Strategy optimization layer

[0119] Function: test plan generation and dynamic optimization, resource allocation

[0120] Core components:

[0121] Basic scheme generator (match detection method / period according to risk level)

[0122] Dynamic adjustment engine (reinforcement learning cycle optimization module)

[0123] Resource scheduling algorithm (integer programming cost optimization model)

[0124] Technical features: policy generation time ≤ 5 minutes per set of devices, cost saving rate ≥ 20%.

[0125] This layer connects the upper and lower layers, mainly responsible for generating and dynamically optimizing inspection strategies based on risk levels, and optimizing resource allocation. Through the basic scheme generator, it matches differentiated inspection methods and cycles for devices of different risk levels; through the dynamic adjustment engine, it uses reinforcement learning algorithm to adaptively adjust the inspection plan according to the real-time state of the device; through the resource scheduling algorithm module, it uses integer programming model to solve the optimal cost scheme under the condition of meeting the risk control target, achieving the balance between safety and economy.

[0126] ④ Monitoring and early warning layer

[0127] Function: Real-time risk monitoring, anomaly identification and early warning traceability

[0128] Core components:

[0129] Real-time monitoring module (sliding window risk calculation engine)

[0130] Anomaly early warning system (dynamic threshold triggering + multi-level push tool)

[0131] Traceability analysis engine (root cause positioning component based on knowledge graph)

[0132] Technical features: risk update frequency ≥ 1 time / hour, early warning response time ≤ 1 hour.

[0133] This layer is responsible for real-time monitoring and active early warning of the system, ensuring the timeliness of risk control. Through the real-time monitoring module, it continuously calculates and updates the device risk value based on the sliding window algorithm; through the anomaly early warning system, it combines static threshold and dynamic threshold judgment rules to trigger and push different levels of risk early warning; through the traceability analysis engine, it uses knowledge graph technology to quickly locate the root cause of the anomaly after the early warning occurs, forming a complete closed loop from perception to traceability.

[0134] ⑤ Collaborative management layer

[0135] Function: Multi-department data sharing, task tracking and performance optimization

[0136] Core components:

[0137] RBAC permission management system (multi-role fine-grained control)

[0138] Ticket management engine (task generation and progress tracking tool)

[0139] Performance iteration module (reinforcement learning strategy optimization model)

[0140] Technical features: permission accuracy rate 100%, task closed loop rate ≥98%.

[0141] This layer serves as the interactive portal of the system and the application layer, mainly responsible for multi-role collaboration and process management. Through the permission management module based on role-based access control, it ensures the sharing and collaboration of multiple departments such as equipment, process, and safety under data security; through the ticket management engine, it automatically converts inspection strategies into executable tasks and performs full-process tracking; through the performance iteration module, it regularly assesses the effectiveness of risk management and optimizes the model and strategy using feedback data to drive continuous improvement of the entire system.

[0142] System interaction relationship

[0143] Each layer realizes data interaction through industrial Ethernet and RESTful API interface, adopts Kafka message queue to handle asynchronous communication, and the peak throughput is ≥500TPS; data storage adopts a hybrid architecture of "MySQL (structured data) + InfluxDB (time series data) + blockchain", and the query response time is ≤1 second; the system security conforms to the IEC 62443 standard, including firewall and data encryption (AES-256).

[0144] In this embodiment, the software architecture of the system adopts a layered design, and each layer calls and interacts with data through standardized interfaces, wherein: the data acquisition layer is used to realize the standardized access and cleaning of petrochemical equipment multi-source data; the risk modeling layer is used to build failure probability (POF) and consequence (COF) quantitative models and failure mechanism library; the strategy optimization layer is used to generate dynamic inspection schemes based on risk levels; the monitoring and early warning layer is used to track risk changes in real time and trigger multi-level early warning; the collaborative management layer is used to realize multi-department data sharing and task closed loop. Each layer is interconnected through standardized interfaces, forming a bottom-up, layer-by-layer progressive software processing closed loop for petrochemical equipment risk management.

[0145] The present application takes "data-driven risk quantification, mechanism-supported strategy optimization, and dynamic closed-loop management" as the core logic, and is based on API581 fourth edition (2025) risk quantification standard, ISO31000 risk management framework, and petrochemical equipment integrity management specification to build a "data fusion-risk modeling-strategy generation-monitoring optimization" full-process technical system. Through multi-source data fusion and dynamic risk model, accurate management of petrochemical equipment RBI throughout the life cycle is realized, forming a management closed loop of "identification-evaluation-control-improvement".

[0146] As mentioned before, the risk calculation server runs a risk management program to realize the core function of the present application. The functions of this program are executed by a series of software modules working in coordination, which correspond to the software architecture layers (i.e. data collection layer, risk modeling layer, strategy optimization layer, monitoring and early warning layer, and collaborative management layer) described in the present application and are carried on the hardware carriers described. The specific implementation of each module is as follows:

[0147] (I) Data collection and fusion processing module (multi-source equipment data collection and fusion processing)

[0148] Execution subject: computer program (data collection and management engine)

[0149] Operation: based on the industrial data fusion theory, the standardization integration and quality control of multi-source data of petrochemical equipment are realized, and the problems of traditional RBI data dispersion and low quality are solved. Specifically, it includes:

[0150] 1. Multi-dimensional data classification collection:

[0151] Equipment basic data: interface with ERP system to obtain account information (material, specification, design pressure / temperature, service life, etc.), structured mapping accuracy ≥ 99%;

[0152] Running data: collect process parameters (temperature, pressure, flow, medium composition, etc., sampling frequency 1 / min) through IoT gateway, support Modbus / OPC UA protocol, transmission delay ≤ 10 seconds;

[0153] Detection data: integrate offline detection results (wall thickness, defect size, corrosion rate, etc.) such as ultrasonic thickness measurement, radiographic testing, and metallographic analysis, store according to equipment position number, data matching accuracy ≥ 98%;

[0154] Corrosion data: access online corrosion probe, corrosion inhibitor injection record, etc. to establish a corrosion environment feature library (such as concentration, content, pH value, etc.).

[0155] 2. Data cleaning and standardization:

[0156] Outlier processing: use 3σ criterion to identify outliers (such as transient overpressure value), and determine whether to correct in combination with equipment operating conditions (such as start-up and shutdown stages), abnormal processing accuracy ≥ 95%;

[0157] Missing value filling: continuous data (such as corrosion rate) uses random forest prediction filling (error ≤ 5%), and discrete data (such as detection conclusion) uses mode filling method;

[0158] Consistency check: Check the data logic through the device parameter relevance rules (such as the matching relationship between wall thickness and design pressure), and the pass rate is ≥97%.

[0159] 3. Multi-source data fusion:

[0160] Fusion of multi-source data credibility based on Dempster-Shafer evidence theory:

[0161]

[0162] Where the conflict coefficient is:

[0163]

[0164] (m(A) is the credibility of the fused proposition A, m i (A i ) is the credibility of the ith data source to the proposition A i , and n is the number of data sources)

[0165] Key parameters: Data access coverage rate ≥100% for key equipment, fusion data accuracy rate ≥96%, and real-time data delay ≤30 seconds.

[0166] (II) Risk quantification modeling and failure mechanism matching module (risk quantification modeling and failure mechanism matching)

[0167] Execution subject: Computer program (risk modeling engine)

[0168] Operation: Based on API581 quantitative risk framework and petrochemical equipment damage mechanism, build a dynamic risk quantification model to realize accurate assessment of failure probability and consequences. Specifically including:

[0169] 1. Failure probability (POF) calculation:

[0170] Integrate corrosion damage (CR), fatigue damage (FD), stress level (SL), and detection effectiveness (DE):

[0171]

[0172] (α=0.35, β=0.25, γ=0.25, δ=0.15 are weight coefficients; CR is annual corrosion rate (mm / year), FD is fatigue damage degree (0-1), SL is actual stress / allowable stress (0-1), and DE is detection effectiveness (0-1))

[0173] According to the POF value, it is divided into 5 levels (1st level: <0.01; 2nd level: 0.01-0.05; 3rd level: 0.05-0.1; 4th level: 0.1-0.2; 5th level: >0.2).

[0174] 2. Consequence of Failure (COF) Assessment:

[0175] Quantifying Consequences from Safety, Environment, and Economy:

[0176]

[0177] (SI is Safety Impact Score (0-10, based on fire / explosion radius), EI is Environmental Impact Score (0-10, based on toxic media diffusion range), EL is Economic Loss Score (0-10, based on equipment value and downtime loss)) is divided into 5 levels (A: <2; B: 2-4; C: 4-6; D: 6-8; E: >8) according to COF value.

[0178] 3. Risk Level Classification and Mechanism Matching:

[0179] Risk level (low / medium / high / very high) is determined by 5x5 risk matrix (POF level x COF level);

[0180] Based on decision tree algorithm, typical failure mechanisms of petrochemical equipment (12 types of corrosion: such as high temperature sulfur corrosion, wet hydrogen sulfide corrosion; 8 types of mechanical damage: such as fatigue crack, creep, etc.) are matched, and 15 parameters including temperature, medium composition, material, etc. are input. The accuracy of mechanism matching is ≥90%.

[0181] Key parameters: POF calculation error ≤10%, COF evaluation error ≤8%, mechanism coverage ≥20 types.

[0182] (Three) Intelligent Generation and Optimization Module of Inspection Strategy (Intelligent Generation and Optimization of Inspection Strategy)

[0183] Execution subject: computer program (strategy optimization engine)

[0184] Operation: Based on risk level and failure mechanism, dynamic inspection strategy is generated and resource allocation is optimized to achieve risk and cost balance. Specifically including:

[0185] 1. Basic inspection scheme generation:

[0186] Very high risk equipment: 100% detection coverage, cycle ≤12 months (detection method: ultrasonic thickness measurement + X-ray detection + magnetic powder detection);

[0187] High-risk equipment: 60%-80% sampling rate, cycle 12-24 months (detection method: ultrasonic thickness measurement + X-ray detection);

[0188] Medium risk equipment: 40% sampling rate, cycle 24-36 months (detection method: ultrasonic thickness measurement);

[0189] Low-risk equipment: ≤30% sampling rate, cycle 36-60 months (detection method: appearance detection + ultrasonic sampling).

[0190] 2. Dynamic optimization adjustment:

[0191] Based on the change of equipment state (such as sudden increase of 20% in corrosion rate), the strategy adjustment is triggered, and the reinforcement learning algorithm is used to optimize the cycle:

[0192]

[0193] (ΔR is the risk change rate, k is the adjustment coefficient (0.2-0.5, valued according to the importance of the equipment), is the original inspection cycle, is the optimized inspection cycle)

[0194] 3. Resource allocation optimization:

[0195] Minimize the total detection cost through integer programming model:

[0196]

[0197] ( is the detection cost of equipment i, is the detection execution variable (0 / 1), is the downtime cost, is the downtime duration variable)

[0198] Key parameters: strategy matching accuracy ≥95%, cost saving rate ≥20%, high-risk equipment defect detection rate ≥98%.

[0199] (Four) Dynamic risk monitoring and early warning module (dynamic risk monitoring and early warning)

[0200] Performing subject: computer program (monitoring and early warning engine)

[0201] Operation: Real-time tracking of equipment risk changes, automatic identification of abnormalities and multi-level early warning, solving the problem of traditional RBI response lag. Specifically including:

[0202] 1. Real-time risk monitoring:

[0203] Based on the sliding window algorithm (window size = 30 minutes) to calculate the risk value in real time, update frequency 1 / hour;

[0204] Using LSTM neural network to predict the risk trend in the next 3 months:

[0205]

[0206] ( is the predicted risk value at time t, Historical risk value, θ is the model parameter

[0207] 2. Abnormal identification and early warning:

[0208] Static threshold: set a benchmark value based on API standards, and trigger an early warning if the deviation exceeds 10%;

[0209] Dynamic threshold: Predicted value, Error standard deviation);

[0210] Early warning classification: general (yellow), important (orange), and urgent (red), with a response time of ≤1 hour for red early warning.

[0211] 3. Early warning traceability analysis:

[0212] Based on the knowledge graph (including 5000+ petrochemical accident cases), the "abnormal index-direct cause- root cause" causal chain is constructed, the traceability time is ≤2 hours, and the accuracy is ≥90%.

[0213] Key parameters: risk update frequency ≥1 time / hour, early warning accuracy ≥95%, and traceability time ≤2 hours.

[0214] (Five) Whole-process collaborative closed-loop management module (whole-process collaborative closed-loop management)

[0215] Execution subject: computer program (collaborative management engine)

[0216] Operation: realize multi-department data sharing and task closed loop, and guarantee the landing and continuous optimization of RBI strategy. Specifically including:

[0217] 1. Multi-role permission management:

[0218] Based on the RBAC model, assign permissions (device department: full data; process department: operating parameters; safety department: risk report), and the permission accuracy is 100%.

[0219] 2. Inspection task closed loop:

[0220] Automatically generate work orders (including equipment position number, detection method, responsible person, and time limit), track progress through Gantt chart, automatically remind if overdue, and task completion rate ≥98%.

[0221] 3. Performance iteration optimization:

[0222] Update the risk model based on inspection data every quarter, and calculate the risk management benefit:

[0223] Benefit = inspection cost savings + accident loss reduction - system investment

[0224] ​Optimize management strategy through reinforcement learning, risk control effectiveness improvement ≥10% / quarter.

[0225] Key parameters: data sharing delay ≤5 seconds, task closed loop rate ≥98%, annual risk reduction rate ≥15%.

[0226] The various embodiments described in this specification are presented as examples. Each embodiment is presented highlighting different aspects of the application, and each embodiment is not mutually exclusive of the others. The same or similar elements in the various embodiments are referred to by the same reference numbers. The above description of the disclosed embodiments makes reference to a number of examples and specific structures. Those skilled in the art will recognize that these described exemplary structures can be adapted or combined or replaced by other structures, and that aspects of the application are applicable to any structure of this type. Those skilled in the art will appreciate that many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the present application there is a full equivalency of all features between structures that are are identical in function.

Claims

1. A petrochemical RBI risk management system characterized in that, Comprise: industrial internet of things collection terminal, risk calculation server, blockchain storage node and collaborative terminal; the industrial internet of things collection terminal is deployed in the petrochemical device field, and is used for collecting equipment operation and environmental parameters; the risk calculation server is in communication connection with the industrial internet of things collection terminal, and is used for running a risk management program, which is used to realize the following when executed by a processor: standardized access, cleaning and fusion of petrochemical equipment multi-source data are realized through a data collection layer; a quantitative model of failure probability and consequence is constructed through a risk modeling layer, and the equipment failure mechanism is matched; the failure probability quantitative model constructed in the risk modeling layer is a multi-factor coupling model based on corrosion damage, fatigue damage, stress level and detection effectiveness; the failure consequence evaluation model is a three-dimensional consequence quantitative model based on safety influence, environmental influence and economic loss; failure probability POF calculation: integrate corrosion damage CR, fatigue damage FD, stress level SL and detection effectiveness DE: ； wherein, α=0.35, β=0.25, γ=0.25, δ=0.15 are weight coefficients; CR is annual corrosion rate mm / year, FD is fatigue damage degree: 0-1, SL is actual stress / allowable stress: 0-1, DE is detection effectiveness: 0-1; according to the POF value, it is divided into 5 levels, 1 level:<0.01; 2 level:0.01-0.05; 3 level:0.05-0.1; 4 level:0.1-0.2; 5 level:>0.2; failure consequence COF evaluation: quantify the consequence from the three dimensions of safety, environment and economy: ； SI is safety influence score: 0-10 points, based on fire / explosion radius, EI is environmental influence score: 0-10 points, based on toxic medium diffusion range, EL is economic loss score: 0-10 points, based on equipment value and downtime loss, according to the COF value, it is divided into 5 levels, A:<2; B:2-4; C:4-6; D:6-8; E:>8; generate and dynamically optimize the inspection strategy based on the risk level through the strategy optimization layer, and allocate inspection resources; through the monitoring and early warning layer, the change of equipment risk is monitored in real time, the abnormality is identified and multi-level early warning is triggered; through the collaborative management layer, multi-department data sharing, inspection task closed loop management and performance iterative optimization are realized; the blockchain storage node is in communication connection with the risk calculation server, and is used for realizing credible storage and traceability of detection data and strategy scheme; the collaborative terminal is in communication connection with the risk calculation server, and is used for providing a man-machine interface for users.

2. A petrochemical RBI risk management system as claimed in claim 1, wherein, the industrial internet of things collection terminal comprises intelligent sensors and edge gateways; the intelligent sensors at least comprise temperature sensors for measuring temperature, pressure sensors for measuring pressure and corrosion probes for measuring corrosion rate; the edge gateway is used for protocol conversion and data preprocessing, and supports data transmission through LoRa, 5G or Ethernet.

3. A petrochemical RBI risk management system as claimed in claim 2, wherein, the implementation of the data collection layer comprises: interface with enterprise resource planning system to obtain equipment basic account information; collect process operation parameters through internet of things gateway; integrate offline detection results of ultrasonic thickness measurement and ray detection; Access online corrosion probe and corrosion inhibitor injection record to establish corrosion environment feature library.

4. A petrochemical RBI risk management system as claimed in claim 1, wherein, The data collection layer is used to perform data cleaning and standardization, including identifying outliers using statistical criteria, filling in missing values of continuous data using prediction models, filling in missing values of discrete data using mode, and checking data logical consistency through device parameter correlation rules.

5. A petrochemical RBI risk management system as claimed in claim 1 wherein, The matching device failure mechanism in the risk modeling layer is based on a decision tree algorithm, which matches a preset petrochemical equipment damage mechanism library according to multiple input features including temperature, medium composition, and material.

6. A petrochemical RBI risk management system as claimed in claim 1 wherein, The dynamic optimization of the test strategy in the strategy optimization layer is based on a reinforcement learning algorithm, which adaptively adjusts the test cycle according to the device risk change rate.

7. A petrochemical RBI risk management system as claimed in claim 1 wherein, The real-time monitoring in the monitoring and early warning layer is based on a sliding window algorithm to calculate the risk value in real time, and a long short-term memory neural network is used to predict the future risk trend; the multi-level early warning is triggered by combining static threshold and dynamic threshold.

8. A petrochemical RBI risk management system as claimed in claim 1, wherein, The multi-department data sharing in the coordination management layer is based on a role-based access control model to assign different data permissions to the equipment department, process department, and safety department; the test task closed-loop management is to automatically generate work orders and track the execution progress.

9. A petrochemical RBI risk management system as claimed in claim 1 wherein, The system components interact through industrial Ethernet and RESTful API interfaces, and use message queues to handle asynchronous communication.

Images