A method and system for generating and traceable output of an algorithm-driven process safety decision tree for standard gas production
By generating a structured process safety decision tree, combined with local knowledge and safety rules, the inconsistencies caused by human experience in standard gas production are resolved, enabling traceable and verifiable process decisions and improving production safety and efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 重庆朝阳气体有限公司
- Filing Date
- 2026-04-15
- Publication Date
- 2026-06-09
Smart Images
Figure CN122173659A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the interdisciplinary field of artificial intelligence and industrial control, specifically relating to an algorithm-driven process safety decision tree generation and traceable output method and system for standard gas production. Background Technology
[0002] Standard gases, as key substances in metrology, environmental monitoring, and instrument calibration, are characterized by high precision, high risk, diverse varieties, and small batches in their research and production. The production process typically involves multiple complex steps, including filling, mixing, valve volume matching, dilution and homogenization, and reaction risk control. In the traditional model, the formulation of these process decisions and safety operation documents relies heavily on expert experience and manual writing, leading to a series of inherent problems: First, decision-making standards are inconsistent and difficult to trace. Operating documents written by different technicians based on their personal experience often differ in terms of filling sequence, mixing strategy, and safety boundary setting. Especially when facing production tasks with multiple varieties, small batches, and frequent changes, the differences in personnel experience can amplify systemic risks, resulting in poor product quality consistency. Moreover, once deviations occur, it is difficult to pinpoint the root cause of the problem due to the lack of quantitative decision-making basis.
[0003] Secondly, there is a risk of omissions in safety risk assessments. When manual operation documents are written, the focus is often on routine procedures, which may overlook potential risks under specific working conditions (such as runaway reaction, sudden pressure changes, etc.) or fail to clearly define safe operating boundaries, thus creating potential safety hazards in production.
[0004] With the development of artificial intelligence technology, some existing technologies have begun to explore the use of AI models to assist in process decision-making. For example, Chinese patent CN114647741A discloses a process reasoning method based on temporal knowledge graphs and partially observable Markov decision processes (POMDPs). This method can train models using historical data to automatically generate or recommend process paths. However, the decision-making process of such models based on pure data-driven or reinforcement learning is similar to a "black box," and the generated process solutions often lack clear numerical support and a traceable chain of evidence. When the process parameters or operating steps suggested by the model pose risks, operators find it difficult to judge their rationality, let alone conduct effective verification. The results generated are difficult to directly apply to standard gas production sites with extremely high safety and reliability requirements.
[0005] Furthermore, Chinese patent CN118296165A proposes a method for generating process routes by combining case-based reasoning and rule-based reasoning. Although this scheme introduces a rule base, its decision-making process is more based on the similarity matching of existing cases and the logical deduction of preset rules. It still fails to strongly bind dynamic calculation results, such as quality ratio, dew point pressure, and volume matching value, to each node of the decision tree. Similarly, it lacks an integrated closed-loop mechanism that can deeply integrate evidence retrieval, accurate calculation, and safety rules to generate traceable and verifiable operation documents.
[0006] For example, Chinese patent CN114647741A discloses an automated process decision-making method that uses temporal knowledge graphs and reinforcement learning models for process reasoning. However, this method encapsulates the decision-making process in a black-box model, making the decision-making basis difficult for operators to understand and verify, and lacking traceability assurance for the decision-making process. Similarly, Chinese patent CN118296165A discloses a process scheme generation method based on case and rule reasoning. Its decision-making relies on a historical case library and a preset rule library. The generated process scheme lacks dynamic correlation with the precise calculation of specific physical quantities in the current production task, resulting in insufficient accuracy and adaptability when production parameters change. None of the above methods address the urgent need for complete evidence chains and verifiable decisions in standard gas production.
[0007] More importantly, existing technologies lack a collaborative mechanism that can ensure the accuracy of numerical calculations while preserving the advantages of large models in natural language understanding and document generation. How to construct a method that can integrate local knowledge, precise calculations, and safety rules to automatically generate structured, traceable, and verifiable process decision documents is a technical challenge that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0008] In view of this, the present invention aims to solve the technical problems in existing standard gas production process decision-making, such as reliance on human experience leading to inconsistent standards, difficulty in traceability of results, and easy omission of safety boundaries, as well as the lack of verifiable evidence chains in existing AI decision-making models.
[0009] To achieve the above objectives, the present invention provides a method and system that can integrate local knowledge, accurate calculations and safety rules to automatically generate structured, traceable and verifiable process decision documents.
[0010] An algorithm-driven method for generating and traceable outputting process safety decision trees for standard gas production includes: Data Input: Acquire input data related to the standard gas production task. The input data includes at least the standard terms retrieved from the local knowledge base, gas property parameters obtained from the locally deployed property database, and order requirements and equipment parameters. Calculation verification: Based on the acquired input data, the locally deployed algorithm function is called to accurately calculate the key physical quantities in the production process, and the calculation results containing the calculation formula, the input parameters, and the intermediate values are output. Decision generation: The calculation results are mapped to a preset rule base to dynamically generate a process safety decision tree; wherein, the decision tree contains multiple decision nodes, and the generation of each decision node is driven by the corresponding calculation results and rules, and each node is bound to a traceable chain of evidence. The chain of evidence clearly records: (1) the specific calculation result values from the calculation verification steps that support the decision of the node and their corresponding calculation formulas and input parameters; (2) the original text or identifier of the standard clauses from the data input steps that serve as the legal or authoritative basis for the decision of the node; Document output: Based on the decision tree and the preset document template, a structured operation list and emergency plan are generated, and traceable metadata is embedded in the generated document. The traceable metadata includes at least parameter source, calculation version and evidence fragment location information, so that each operation instruction in the final operation list and emergency plan can be traced back to the calculation basis and standard clauses behind it.
[0011] Preferably, the data input step further includes real-time access to sensor data from the production site via an IoT interface. When the deviation between the sensor data and the predicted value of the calculation result exceeds a preset threshold, a dynamic correction event is triggered, and the decision generation step is re-executed based on the dynamic correction event to adjust the subsequent process strategy in order to achieve closed-loop, adaptive process control.
[0012] Preferably, the calculation verification step further includes a dual verification mechanism, namely, checking the range and dimension consistency of the input parameters and cross-validating the calculation results, thereby ensuring the accuracy and reliability of the calculation results.
[0013] Preferably, the method further includes a conflict detection step, that is, during the decision generation step, conflicts between different rules or constraints are detected in parallel; when a conflict is detected, a prompt message containing at least two alternative solutions and their risk levels is generated for the user to select and confirm. Further, when a resource shortage conflict is detected, an optimization engine is activated. This optimization engine, based on a preset weighted scoring model, assigns weights to factors such as order urgency, substitutability, procurement cycle, and production cost, calculates the comprehensive cost and risk scores of different decision solutions, and uses the Pareto optimality method to weigh multiple objectives and recommend the optimal solution.
[0014] Preferably, the method further includes an interactive verification step, which presents the decision tree through a web-based operating platform and responds to the user's modification of the decision tree nodes. The modified decision is simulated and executed in a digital twin environment, and the changes in key process parameters (such as final product concentration, production time, energy consumption, and safety risk level) before and after the modification are displayed for user verification and optimization.
[0015] This invention also provides an algorithm-driven process safety decision tree generation and traceable output system for standard gas production. The system includes a data input module, a calculation and verification module, a decision generation module, a conflict detection module, a document output module, and a web-based interactive platform, used to implement the methods described in any of the above-mentioned embodiments.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) Realize the transformation of decision-making from experience-driven to evidence-driven. By strongly binding the calculation verification results, standard clause references and each node of the decision tree, each process decision has clear numerical support and a traceable evidence chain, which fundamentally solves the problems of inconsistent standards and difficulty in tracing results in manual decision-making, and significantly improves the consistency and verifiability of decision-making.
[0017] (2) Improve production safety and compliance. The conflict detection mechanism exposes the contradictions between different rules or resource constraints in advance and provides risk classification and alternative solutions. By automatically generating emergency plans with binding evidence, human error and systemic risks can be effectively prevented, the probability of production accidents can be reduced, and high standards of safety and compliance requirements can be met.
[0018] (3) Enhance the scalability and adaptability of the system. The rule base supports automatic generation and optimization based on machine learning, enabling the system to continuously learn and evolve from historical successful cases. Real-time data is accessed through the Internet of Things interface, realizing dynamic correction capability, enabling decisions to adapt to real-time changes in the production site. The flexibility of the output format enables it to adapt to production lines with different levels of automation.
[0019] (4) The digital accumulation and reuse of process knowledge are realized. The entire decision-making process, intermediate calculations, evidence chains and final output operation documents are saved in a structured and versioned manner, forming reusable enterprise knowledge assets. This not only accelerates the rapid switching of multi-variety and small-batch production tasks, but also provides a data foundation for the continuous improvement of future processes.
[0020] (5) Optimize the efficiency of human-machine collaborative decision-making. Through the Web-based visual interaction platform and digital twin simulation verification, technicians can intuitively understand the decision-making logic, conduct trial-and-error optimization, and then distribute it to the production environment. While ensuring the security of decision-making, it greatly improves decision-making efficiency and user participation.
[0021] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description
[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the architecture of an algorithm-driven process safety decision tree generation and traceable output system for standard gas production according to the present invention. Figure 2 This is a schematic diagram of the process safety decision tree in an embodiment of the present invention. Detailed Implementation
[0023] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0024] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.
[0025] Example 1 This embodiment provides an algorithm-driven process safety decision tree generation and traceable output system for standard gas production. This system aims to solve the technical problems of inconsistent standards, difficulty in traceable results, and easy omission of safety boundaries in traditional manual decision-making.
[0026] The system architecture is as follows Figure 1 As shown, it mainly includes: a data input module, a calculation and verification module, a decision generation module, a conflict detection module, a document output module, and an optional web-based interactive platform.
[0027] 1. Data Input Module The data input module is the system's data entry point, providing structured and standardized input for subsequent calculations and decision-making. Its data sources include, but are not limited to: (1) Local knowledge base / standard clause search results: These results can be provided by a separate hybrid search system, such as the "A Method and System for Fully Localized Hybrid Search and Knowledge Graph Construction for Standard Gas R&D and Production" submitted on the same day as this invention. The input data includes clauses, rules, and expert experience related to the current production task retrieved from national / industry standards, internal enterprise technical specifications, and historical process documents. For example, for the task of "preparing carbon monoxide standard gas in nitrogen with a concentration of 10 ppm", this module may retrieve the provisions on filling pressure in "GB / T 5274.1-2018" and the enterprise's internal "Safety Operating Procedures for Carbon Monoxide Gas Handling".
[0028] (2) Physical property database: Stores the basic physicochemical properties of key gas components, such as gas molar mass, critical temperature, critical pressure, dew point coefficient, explosion limits, etc. The data is stored in a structured form. For example, for "carbon monoxide", the database can return its molar mass as 28.01 g / mol and its explosion limits in air as 12.5% to 74.2%, etc.
[0029] (3) Order requirements and equipment parameters: Order requirements include the type, concentration, filling volume, purity requirements, and delivery cycle of the target gas. Equipment parameters include the currently available cylinder volume, valve model, pipe diameter, maximum operating pressure of the equipment, and safety valve setting. These data can be obtained in real time through API interfaces with Enterprise Resource Planning (ERP) and Manufacturing Execution System (MES).
[0030] 2. Calculation and Verification Module The calculation and verification module receives data from the data input module and calls the locally deployed algorithm function library to accurately calculate key physical quantities in the production process, and outputs calculation results with intermediate processes.
[0031] In one example, for a gas cylinder filling task, the calculation verification module performs the following steps: 1. Parameter extraction: Parse structured parameters from order requirements and physical property databases. For example, the target gas is "carbon monoxide (CO)", the target concentration is "10ppm" (volume fraction), the balance gas is "nitrogen (N2)", the cylinder volume is "40L", and the filling pressure is "12.5MPa".
[0032] 2. Function calculation: 1) Mass ratio calculation: Calculate the required mass of carbon monoxide using the ideal gas law PV=nRT. First, calculate the amount of substance of the total gas at 40L and 12.5MPa. Then, based on the target concentration of 10 ppm, calculate the amount of carbon monoxide. Finally, considering the molar mass of carbon monoxide... Calculate the required mass of carbon monoxide. = The calculation process will output the numerical values for each step, such as "The total amount of substance is approximately 20.12 mol, and the amount of carbon monoxide is approximately 2.012 × 10⁻⁶". -4 "The required mass of carbon monoxide is approximately 5.63 mg."
[0033] 2) Dew point calculation: Based on the composition and pressure of the mixed gas, the dew point temperature of the mixed gas is estimated by looking up a table or calling the dew point calculation function.
[0034] 3) Volume matching calculation: Based on the gas cylinder volume and the total amount of gas to be prepared, determine whether multiple gas cylinders are needed or select a suitable gas mixing cylinder.
[0035] 4) Pressure calculation: Calculate the pressure changes inside the gas cylinder at different filling stages.
[0036] The calculation verification module also has a dual verification mechanism: first, it checks the range and dimension consistency of the input parameters, such as whether the input concentration is within the safety threshold and whether the pressure exceeds the rated pressure of the equipment; second, it performs cross-verification of the calculation results, such as comparing the calculated mass ratio with the experience value of similar historical tasks. If the deviation exceeds the preset threshold (such as ±5%), a warning is triggered and the user is prompted to perform manual verification.
[0037] 3. Decision Generation Module The decision generation module is the core of the system. It maps the calculation results to a preset rule base and dynamically generates a structured process safety decision tree.
[0038] Rule base construction: Rule bases can be constructed in various ways. One approach is manual construction based on expert experience, where senior process engineers explicitly translate long-accumulated tacit knowledge, such as valve selection principles, filling step priority, and risk assessment experience, into clear rule statements. Another approach is automatic generation and continuous optimization based on machine learning from historical successful process cases. For example, the system can perform association rule mining on hundreds of historical process documents and operation records, automatically discovering association patterns such as "when the target gas is hydrogen and its concentration is higher than 1%, 80% of successful cases used the multi-step dilution method," and providing this as a new rule suggestion for expert review and inclusion in the rule base, thus achieving self-evolution of the rule base.
[0039] Figure 2 An exemplary structure of the process safety decision tree generated in this embodiment is shown. Based on the above calculations and rules, the decision tree constructed by the decision generation module can be as follows: Root node: Start Sub-node 1: Valve Selection Evidence binding: The calculation result is "Gas type: Carbon monoxide (toxic)", and the rule base rule is "Toxic gases must be handled using a special valve made of stainless steel with a self-locking function".
[0040] Decision output: Select "Model A-XXX Stainless Steel Self-Locking Valve".
[0041] Sub-node 2: Filling steps Evidence binding: The calculation result is "Target concentration: 10ppm (low concentration)", and the rule base rule is "For easily adsorbed gases with concentrations below 100ppm, a multi-step dilution method should be used".
[0042] Decision output: The filling scheme is determined as follows: "Step 1: Prepare 1000ppm intermediate concentration gas; Step 2: Dilute 1000ppm gas to 10ppm".
[0043] This decision node is bound to the following chain of evidence: Computational evidence: derived from the computational verification module. Calculation formula: Substitute the parameter: target concentration Final gas cylinder volume intermediate concentration Calculation results: Required intermediate concentration gas volume The calculation results show that the amount of source gas required to directly prepare 10 ppm of gas is extremely small and difficult to control precisely. Therefore, according to the engineering experience rule, "for low-concentration easily adsorbed gases, a multi-step dilution method should be used."
[0044] Reference to standard clause: According to GB / T 5274.1-2018 "Preparation of mixed gases for gas analysis calibration - Part 1: Preparation by weighing method", Clause 5.3.2 states that "For components with concentrations below 100 ppm, it is recommended to use a stepwise dilution method to improve the preparation accuracy".
[0045] This chain of evidence makes the decision to "adopt the multi-step dilution method" completely transparent, and any operator can verify the rationality and compliance of the decision based on this chain.
[0046] Child node 3: Mixing strategy Evidence binding: The calculation result is "the equilibrium gas is N2, the target gas is CO, and the two have similar densities", and the rule base rule is "gases with similar densities can use the static diffusion method, and the static time must be ≥24 hours".
[0047] Decision output: Adopt the "natural settling method, settling for 24 hours".
[0048] Sub-node 4: Risk Assessment and Emergency Response Evidence binding: The calculation result states that "the explosion limit of carbon monoxide in air is 12.5% to 74.2%, and the concentration of carbon monoxide in the currently prepared mixed gas is far below this value." The rule base rule states that "when the gas concentration is far below the lower explosion limit, it is defined as low risk."
[0049] Decision output: Risk level is "low risk", emergency response path is "If a leak occurs, immediately cut off the gas source, start local ventilation, no evacuation is required".
[0050] It is worth noting that each node in the decision tree is bound to corresponding computational evidence and a reference to a standard clause. For example, the decision output of child node 2 will be accompanied by an "evidence chain" label, which includes the calculation formula, key parameters, calculation results, and the referenced standard clause number and original text excerpt. This makes the entire decision-making process transparent and traceable.
[0051] Decision trees can be stored and represented in various data formats. One approach is to store them as structured data formats such as JSON or XML, facilitating parsing and retrieval by other modules. Another approach, for scenarios requiring user understanding and review, is to use a graphical engine to render the decision tree as a visual flowchart. Users can then intuitively view the complete reasoning path from the root node to the leaf nodes, the decision basis for each node, and its risk level on the web.
[0052] 4. Conflict Detection Module The conflict detection module works in parallel with the decision generation module. It is responsible for checking for contradictions between different rules or constraints and handling them according to preset strategies.
[0053] For example: A. Rule Conflict: Rule A requires that "to achieve high purity, the gas cylinder must be vacuumed three times before filling," while Rule B from the equipment parameters limits "the number of vacuuming operations to a maximum of two to avoid damaging the gas cylinder valve." Upon recognizing this conflict, the conflict detection module will automatically trigger a prompt and generate a message containing two options and their risk levels: "Option 1: Use two vacuum purging operations, but there is a low risk of not meeting purity standards; Option 2: Replace with a special valve that meets the three-purging requirement (increased cost). Please confirm your selection." B. Rule and Calculation Conflict: The calculation results show that completing this order requires 15kg of Class A gas, but the inventory management database indicates that there is only 10kg of Class A gas remaining. The conflict detection module will immediately output a "Resources Insufficient" warning and provide the following suggestion: "Order resources are insufficient; it is recommended to initiate emergency procurement or adjust the production plan." 5. Document Output Module The document output module receives the complete decision tree from the decision generation module and automatically generates a structured operation list and emergency plan based on a preset template.
[0054] Operation list generation: The decision nodes in the decision tree are converted into step-by-step operation instructions according to the execution order. For example: Operation List (Task ID: CO-10ppm-001) (1) Preparation: Take stainless steel self-locking valves of model A-XXX from the warehouse.
[0055] (2) First step filling (preparation of intermediate concentration): 0.563g (calculated value) of carbon monoxide is filled into a 40L gas cylinder, and nitrogen is added until the pressure reaches 2.5MPa to obtain an intermediate concentration gas of 1000ppm.
[0056] (3) Mix well: Let the gas cylinder stand for 24 hours.
[0057] (4) Second step filling (final dilution): Use the gas in the above gas cylinder as "source gas" to fill another vacuum 40L empty cylinder, and add nitrogen to the pressure of 12.5MPa.
[0058] (5) Finally mix well: let stand for 24 hours again.
[0059] (6) Safety confirmation: Verify that the risk level is low and confirm the emergency plan.
[0060] (7) Emergency plan generation: Generate corresponding emergency plans based on the output of the risk assessment node. For example, for a "low-risk" scenario, the emergency plan may be simplified to "wear personal protective equipment and keep the room well-ventilated".
[0061] (8) Embedding Traceable Metadata: The system embeds traceable metadata in XML or JSON format within the generated document. This metadata can be parsed by other systems for auditing and review. Metadata includes, but is not limited to: ParameterSource: {"Gas Mass": "Based on calculations, according to the formula PV=nRT, molar mass 28.01 g / mol"} CalculationVersion: {"Proportion Calculation": "V1.2"} EvidenceFragment: {“Standard Clause”:“GB / T 5274.1-2018, Clause 5.3.2”} ApprovalRecord: {"Generator": "System", "Reviewer": "Pending Human Signature"} It is worth noting that the output format of the operation list is flexible. In addition to generating natural language text for human operation, the system can also generate instructions in other formats depending on the downstream execution equipment. For example, when the production line is highly automated, the system can convert the above operation steps into a machine-readable instruction set, such as directly converting it into code that can be recognized by a programmable logic controller (PLC), to drive the automated equipment to execute, thereby achieving full-process automation from decision-making to execution.
[0062] 6. Web-based interactive platform The system interacts with users through a web-based operating platform. This platform can be further integrated with a digital twin model to form a more powerful decision support capability. Specifically, after a user modifies a node in the decision tree on the web, that modification can be immediately simulated in the digital twin environment. Based on the modified decision, the system calculates its impact on the final product concentration, production time, energy consumption, and safety risks in virtual space. By observing the simulation results, users can "trial and error" and optimize without actually consuming materials or interrupting production, confirming the optimal solution before issuing the final decision to the real production environment.
[0063] Example 2 This embodiment expands upon Embodiment 1 by adding a data input module. In addition to a local knowledge base and database, the data input module also accesses sensor data from the production site in real time via an Internet of Things (IoT) interface.
[0064] For example, during the "dilution and mixing" step, the system can read pressure and temperature sensor data from the gas cylinder in real time. If the real-time pressure deviates from the pressure predicted by the calculation module beyond a preset threshold, the system triggers a "real-time anomaly detection" event and feeds this event back to the decision generation module as new input. The "dynamic correction rule" in the decision generation module is activated, recalculating and adjusting subsequent filling or mixing strategies (e.g., extending the settling time or starting auxiliary stirring), thereby achieving closed-loop, adaptive process control.
[0065] Example 3 This embodiment optimizes the conflict detection module based on Embodiment 1. The conflict detection module not only handles conflicts between rules but also introduces a risk assessment and cost analysis model.
[0066] For example, when a "resource shortage" conflict is detected, the system doesn't just issue a simple warning; instead, it activates a lightweight optimization engine. This engine calculates the overall cost and risk score of different decision-making options based on the urgency of the order, the availability of substitutes for the out-of-stock gas, procurement cycle time, and production costs, and recommends the optimal solution. This elevates decision support from a compliance perspective to an economic one.
[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for algorithm-driven process safety decision tree generation and traceable output for standard gas production, characterized in that, include: Data Input: Acquire input data related to the standard gas production task. The input data includes at least the standard terms retrieved from the local knowledge base, gas property parameters obtained from the locally deployed property database, and order requirements and equipment parameters. Calculation verification: Based on the acquired input data, the locally deployed algorithm function is called to accurately calculate the key physical quantities in the production process, and the calculation results containing the calculation formula, the input parameters, and the intermediate values are output. Decision generation: The calculation results are mapped to a preset rule base to dynamically generate a process safety decision tree; wherein, the decision tree contains multiple decision nodes, and each decision node is bound to a traceable chain of evidence, the chain of evidence including at least: the specific calculation result value from the calculation verification step and its corresponding calculation formula used to support the decision of the node, and the original text or identifier of the standard clause from the data input step as the basis for the decision; Document output: Based on the decision tree and the preset document template, a structured operation list and emergency plan are generated, and traceable metadata is embedded in the generated document. The traceable metadata includes at least parameter source, calculation version and evidence fragment location information, so that each operation instruction in the final operation list and emergency plan can be traced back to the calculation basis and standard clauses behind it.
2. The method according to claim 1, characterized in that, The data input also includes: real-time access to sensor data from the production site via an IoT interface, wherein the sensor data includes at least pressure data and temperature data; the method also includes: when the deviation between the sensor data and the predicted value of the calculation result exceeds a preset threshold, a dynamic correction event is triggered, and the decision generation step is re-executed based on the dynamic correction event to adjust the subsequent process strategy.
3. The method according to claim 1, characterized in that, The key physical quantities include one or more of mass ratio, dew point, pressure, and volume matching; the calculation verification step also includes a dual verification mechanism: checking the range and dimension consistency of the input parameters, and cross-validating the calculation results.
4. The method according to claim 1, characterized in that, The decision tree is stored in JSON or XML structured data formats, or presented graphically.
5. The method according to claim 1, characterized in that, It also includes a conflict detection step: during the execution of the decision generation step, conflicts between different rules or constraints are detected in parallel; when a conflict is detected, a prompt message containing at least two alternative solutions and their risk levels is generated, and the user's confirmation of the selection of the alternative solutions is received.
6. The method according to claim 5, characterized in that, The conflict detection also includes: when a conflict of insufficient resources is detected, an optimization engine is launched to calculate the comprehensive cost and risk score of different decision-making schemes based on at least one of the following factors: order urgency, substitutability, procurement cycle and production cost, and recommend the optimal scheme.
7. The method according to claim 1, characterized in that, It also includes an interactive verification step: presenting the decision tree through a web-based operating platform; responding to the user's modification operation on the nodes in the decision tree, simulating the execution of the modified decision in a digital twin environment, and displaying a comparison of the changes in key process parameters before and after the modification for user verification.
8. An algorithm-driven process safety decision tree generation and traceable output system for standard gas production, characterized in that, The system for implementing the method as described in any one of claims 1 to 7 includes: A data input module is used to perform the data input step as described in claim 1; The calculation verification module is used to perform the calculation verification steps as described in claim 1; The decision generation module is used to execute the decision generation steps as described in claim 1; The document output module is used to perform the document output step as described in claim 1; A conflict detection module, when the method includes the conflict detection step as described in claim 5, is used to execute the conflict detection step; When the method includes the interaction verification step described in claim 7, the web-based interactive platform is used to execute the interaction verification step.