MOC Intelligent Identification and Process Analysis Management System Based on Digital Twin

Abstract

This invention relates to the fields of digital twin, chemical process control, and industrial intelligent management, specifically to a digital twin-based MOC intelligent identification and process analysis management system. The system includes a data acquisition module, an intent parsing and injection module, a topology map construction module, a spatiotemporal transmission simulation module, and a health assessment and feedback module. It acquires change intent data, digital twin base data, and real-time operational data; extracts the change content as parameterized features and generates micro-perturbation vectors, injects them into target topology nodes, constructs a cross-domain process topology map, iteratively generates predicted values ​​for each process parameter according to a preset time step, compares these values ​​with safety critical thresholds, and outputs a global health assessment result. When there is a risk of exceeding limits, it calculates the adjustment amount in reverse and generates a process compensation scheme; when safe, it outputs normal operation confirmation information, achieving early identification and compensation guidance for future risks of local changes.

Description

Technical Field

[0001] This invention relates to the fields of digital twin, chemical process control and industrial intelligent management technology, specifically to a digital twin-based MOC intelligent identification and process analysis management system. Background Technology

[0002] In change management in the fine chemical and continuous process industries, the approval and risk analysis of equipment replacement, control parameter adjustment, process route modification and material boundary changes are gradually evolving from static ledger verification and manual experience judgment to intelligent assessment methods that combine digital twin models, real-time operating data and process correlations. Currently, the analysis methods for production unit change management generally support single-point identification of the changed object, changed location, or changed parameter. If it is necessary to further determine the impact of the change on the entire process chain, control chain, and equipment health status, it is usually necessary to rely on manual review of drawings, operating procedures, interlock logic, and historical operation records to obtain the corresponding risk conclusions. However, when approving changes using the above methods, judgments are often based solely on local equipment information or static design information. It is difficult to accurately translate the change intent into calculable disturbance quantities and inject them into the overall device model. It is also impossible to continuously track the cross-domain transmission process of disturbances along materials, heat, pressure, and control logic, resulting in difficulties in identifying delayed, cumulative, and coupled risks. Furthermore, this approach relies heavily on prior human knowledge, leading to low consistency in assessments when faced with missing equipment tag numbers, ambiguous descriptions, abnormal real-time data, or incomplete material boundaries. Moreover, it lacks the ability to perform reverse compensation analysis based on global topology relationships, making it difficult to simultaneously provide engineering-executable process compensation solutions after risk detection, resulting in low closed-loop handling capabilities. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a digital twin-based intelligent identification and process analysis management system for MOCs. Specifically, the technical solution of this invention includes: The data acquisition module is used to acquire change intent data, digital twin base data, and real-time operational data; The intent parsing and injection module communicates with the data acquisition module and is used to extract features from the change intent data to generate parameterized features. Based on the parameterized features, it generates a micro-perturbation vector containing node identifiers and perturbation values, identifies one or more topological nodes corresponding to the micro-perturbation vector as target topological nodes, and injects the micro-perturbation vector into the target topological nodes contained in the digital twin base data. The topology map construction module communicates with the data acquisition module and is used to construct cross-domain process topology maps based on digital twin base data. The spatiotemporal transmission simulation module is connected to the intent parsing and injection module, the topology map construction module, and the data acquisition module, respectively. It is used to take real-time running data as the initial operating condition baseline, and based on the cross-domain process topology map with injected micro-perturbation vectors, iteratively deduce according to a preset time step to generate spatiotemporal ripple effect simulation data containing the predicted values ​​of each process parameter. The health assessment and feedback module, which communicates with the spatiotemporal conduction simulation module, compares the predicted values ​​of each process parameter in the spatiotemporal ripple effect simulation data with the preset safe critical threshold range of the corresponding process parameter to generate a global health assessment result. The configuration is as follows: if the global health assessment result indicates a risk of exceeding limits, it calculates the parameter deviation between the predicted process parameter value and the safe critical threshold, calculates the adjustment amount required to offset the parameter deviation based on the cross-domain process topology map, and generates a process compensation scheme based on the adjustment amount; if the global health assessment result indicates a safe state, it outputs confirmation information that the system is operating normally.

[0004] Optionally, the digital twin base data includes physical space topology data, process logic topology data, and material property library data; Among them, physical space topology data is used to characterize the spatial location relationships of devices; Among them, process logic topology data is used to characterize the logical relationships of the control system.

[0005] Optionally, the intent parsing and injection module includes: The feature extraction unit is used to perform natural language processing on the change intent data to extract parameterized features that represent the change object, change type, and change value. The vector generation unit, connected to the feature extraction unit, is used to generate micro-perturbation vectors based on parameterized features; The node matching unit, connected to the vector generation unit, is used to identify the target topology node corresponding to the micro-perturbation vector and inject the micro-perturbation vector into the target topology node contained in the digital twin base data.

[0006] Optionally, the topology map building module is used for: Extract equipment node data and process parameter node data from the digital twin base data; Based on equipment node data and process parameter node data, a directed graph containing physical connection edges and logical connection edges is constructed. The directed graph is output as a cross-domain process topology graph.

[0007] Optionally, the spatiotemporal transmission simulation module includes: The reduced-order multiphysics coupling unit is used to perform physical field coupling calculations on the cross-domain process topology map using a reduced-order model that extracts low-order dominant features through dimensionality reduction mapping, and generate coupled state data. Specifically, the dimensionality reduction mapping can use principal component analysis or intrinsic orthogonal decomposition algorithm. By performing eigenvalue decomposition on the full physical field state data corresponding to the cross-domain process topology map, the eigenvectors corresponding to the top N largest eigenvalues ​​are extracted to construct a projection matrix, thereby extracting a few dominant features that are most sensitive to safety consequences from the high-dimensional full physical state. The time acceleration simulation unit, connected to the reduced-order multiphysics coupling unit, is used to perform time-step iterative simulation based on coupled state data, micro-perturbation vectors, and real-time running state data, using a time-series prediction algorithm based on the state transition matrix, to generate simulation data of the spatiotemporal ripple effect.

[0008] Optionally, the time acceleration simulation unit is specifically used for: Using real-time operating data as the initial operating baseline, along the connecting edges of the cross-domain process topology map, the parameter propagation trend of the micro-disturbance vector within a preset time period is calculated, where the preset time period is greater than or equal to the preset process response delay time. Based on the parameter propagation trend, the simulation data of the spatiotemporal ripple effect is output, where the predicted values ​​of each process parameter are represented as the attenuation or amplification values ​​of the parameter relative to the initial operating condition baseline.

[0009] Optionally, the health assessment and feedback module includes: The threshold comparison unit is used to compare the predicted values ​​of each process parameter contained in the simulation data of the spatiotemporal ripple effect with the preset safety critical threshold of the corresponding process parameter. The status determination unit, connected to the threshold comparison unit, is used to determine that if the predicted value of the process parameter exceeds the safety critical threshold range, the topology node corresponding to the predicted value of the process parameter has a risk of exceeding the limit, and generate a global health assessment result containing risk warning information. The status determination unit is also used to determine that the topology node corresponding to the predicted process parameter value is in a safe state if the predicted process parameter value is within the safe critical threshold range, and to generate a global health assessment result containing safe passage information.

[0010] Optionally, the system also includes: The 3D visualization module communicates and connects with the health assessment and feedback module to generate a dynamic risk heat map in the 3D digital twin model built on the digital twin base data based on the global health assessment results. Among them, the dynamic risk heat map is used to characterize the transmission path and scope of impact of risk over time.

[0011] Optionally, the health assessment and feedback module also includes: The compensation scheme generation unit, connected to the status determination unit, is used to respond to risk warning information and calculate the adjustment amount required to offset parameter deviations based on the cross-domain process topology map. The compensation scheme generation unit is also used to encapsulate the adjustment amount into a process compensation scheme and output it.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. This system uses a data acquisition module to jointly acquire change intent data, digital twin base data, and real-time operational data. The intent parsing and injection module performs natural language processing, parameter extraction, micro-disturbance vector generation, and target topology node injection on the change object, change type, and change value. This enables textual and form-based change requests to be transcribed into calculable engineering disturbances and accurately entered into the digital twin model. This effectively solves the problems in existing technologies that make it difficult to accurately convert change intents into calculable disturbance quantities and difficult to inject them into the overall device model. At the same time, through field integrity verification, alias dictionary matching, conflict semantic recognition, and fuzzy description interval evaluation, the system improves the consistency and usability of evaluation in scenarios such as missing device tag numbers and fuzzy descriptions. 2. This system subdivides the digital twin base data into physical space topology data, process logic topology data, and material property library data. The topology map construction module extracts equipment node data and process parameter node data to construct a cross-domain process topology map that includes both physical and logical connection edges. This allows material flow paths, heat exchange paths, control loop action paths, and material tolerance boundaries to be tracked and interpreted in a unified map. This effectively solves the problem that existing technologies can only make judgments based on local equipment information or static design information and are unable to continuously track the cross-domain transmission process of disturbances along materials, heat, pressure, and control logic. Furthermore, when material boundaries are missing or logical relationships are incomplete, a conservative boundary processing or template supplementation mechanism can be used to avoid the risk of misjudgment caused by default safety. 3. This system employs a reduced-order multiphysics coupling and time-accelerated extrapolation based on the state transition matrix through a spatiotemporal conduction simulation module. Using real-time operating data as the initial baseline, it iterates and extrapolates over a preset time period not less than the process response delay time. The output is spatiotemporal ripple effect simulation data, which manifests as parameter attenuation or amplification values. This enables the system not only to identify disturbance propagation paths but also to reveal the speed, intensity, duration, and buffering or amplification trends of the impact propagation. This effectively solves the problem that existing technologies struggle to identify delayed, cumulative, and coupled risks. Furthermore, it can switch to a reliable baseline or conservative mode under abnormal operating conditions such as key sensor offline, timestamp drift, and start-up / shutdown, ensuring the robustness of the evaluation results. 4. This system compares the predicted values ​​of each process parameter with the preset safety thresholds item by item through the health determination and feedback module, and generates a global health assessment result containing risk warning information or safe passage information based on a unified status determination logic. This enables complex multi-domain spatiotemporal simulation results to be converged into objective, standardized judgment conclusions that can be directly used for approval decisions. This effectively reduces the degree to which existing technologies rely on manual review of drawings, interlocking logic and historical records, and on human experience to extract risk conclusions. At the same time, in cases of threshold conflicts, short-term overruns, and insufficient data integrity at key nodes, a more conservative judgment benchmark is preferred and the entire sequence log is retained, which improves the rigor and traceability of the approval process. 5. This system, upon detecting a risk exceeding limits, calculates the adjustment amount required to offset parameter deviations based on a cross-domain process topology map. This adjustment amount is then encapsulated into a process compensation scheme that includes the target object, execution conditions, and duration. Simultaneously, a 3D visualization module generates a dynamic risk heatmap in the digital twin model, representing the risk's propagation path and impact range over time. This allows the system not only to determine the existence of a risk but also to provide closed-loop mitigation suggestions on how to offset the risk through measures such as improving cooling loop capacity, adjusting preheating targets, and shortening inspection cycles. The system spatializes and temporally displays the risk propagation process, effectively addressing the shortcomings of existing technologies, such as a lack of reverse compensation analysis capabilities based on global topology relationships, difficulty in simultaneously providing engineering-executable process compensation schemes after risk detection, and low risk communication efficiency. Attached Figure Description

[0013] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0015] like Figure 1 As shown, a digital twin-based intelligent identification and process analysis management system for MOCs is implemented. The system includes: a data acquisition module, used to acquire change intent data, digital twin base data, and real-time operational data; The intent parsing and injection module communicates with the data acquisition module and is used to extract features from the change intent data to generate parameterized features. Based on the parameterized features, it generates a micro-perturbation vector containing node identifiers and perturbation values, identifies one or more topological nodes corresponding to the micro-perturbation vector as target topological nodes, and injects the micro-perturbation vector into the target topological nodes contained in the digital twin base data. The topology map construction module communicates with the data acquisition module and is used to construct cross-domain process topology maps based on digital twin base data. The spatiotemporal transmission simulation module is connected to the intent parsing and injection module, the topology map construction module, and the data acquisition module, respectively. It is used to take real-time running data as the initial operating condition baseline, and based on the cross-domain process topology map with injected micro-perturbation vectors, iteratively deduce according to a preset time step to generate spatiotemporal ripple effect simulation data containing the predicted values ​​of each process parameter. The health assessment and feedback module, which communicates with the spatiotemporal conduction simulation module, compares the predicted values ​​of each process parameter in the spatiotemporal ripple effect simulation data with the preset safe critical threshold range of the corresponding process parameter to generate a global health assessment result. The configuration is as follows: if the global health assessment result indicates a risk of exceeding limits, it calculates the parameter deviation between the predicted process parameter value and the safe critical threshold, calculates the adjustment amount required to offset the parameter deviation based on the cross-domain process topology map, and generates a process compensation scheme based on the adjustment amount; if the global health assessment result indicates a safe state, it outputs confirmation information that the system is operating normally.

[0016] This embodiment provides an implementation mechanism for a digital twin-based MOC intelligent identification and process analysis management system. Specifically, the system is deployed in a continuous fine chemical production line scenario for change approval and operation assurance. The production line includes raw material storage tanks, feed pumps, regulating valves, preheating heat exchangers, main reactors, condensers, and subsequent distillation units. The system does not interpret change management merely as identifying which equipment has been modified, but rather as a continuous judgment process of whether local modifications will be gradually transmitted along heat, flow rate, pressure, control logic, and material tolerance, and will be transformed into global risks at some point in the future. The details are as follows: The data acquisition module receives three types of inputs: First, change intent data, which can come from change management (MOC) application forms, maintenance work orders, parameter adjustment records, or piping and instrumentation diagram modification instructions filled out by process engineers; second, digital twin base data, reflecting the original design status of the unit, equipment layout relationships, control logic relationships, and material boundaries; and third, real-time operational data, from the field distributed control system, safety instrumented system, online analyzer, and equipment status monitoring terminal. After receiving the change intent, the system does not directly perform static approval, but first, the intent parsing and injection module converts the textual or form-based change content into a parameterized expression with engineering meaning. For example, replacing the B-type feed pump and reducing the inlet regulating valve opening from 50% to 45% is no longer just textual information, but will be converted into disturbance information that can be injected into the twin model, such as changes in flow capacity, head, pulsation characteristics, and corresponding valve resistance. The topology map construction module organizes the equipment nodes, process parameter nodes and control relationships in the production line into a cross-domain process topology map, so that material flow paths, heat exchange paths and control loop action paths can all be tracked in the same map; the spatiotemporal conduction simulation module uses the current on-site conditions as a baseline and gradually deduces how disturbances spread, are buffered or amplified in the map at a preset time step. For example, an increase in feed resistance within a preset deviation range may only manifest as pump outlet fluctuations in the first preset time period, but in subsequent periods it will cause changes in the residence time of the preheating section, causing deviations in the reactor feed temperature and ratio, ultimately affecting the heat release and condensation load. The health assessment and feedback module compares the predicted values ​​of each parameter obtained from the simulation with the safety critical threshold item by item, and outputs the global health assessment result. When there is a risk of future exceedances, it does not simply prohibit changes, but further searches backward along the graph to find compensation locations that can be used to offset the deviations and generates corresponding process compensation schemes. When the prediction results show that the system is within the safe range, it outputs a confirmation message. To illustrate how data flows within the system, a simplified simulation structure can be used. Consider equipment nodes N1 as the feed pump, N2 as the regulating valve, N3 as the preheater, and N4 as the main reactor. Corresponding parameter nodes P1 represent flow rate, P2 as pressure, P3 as inlet temperature, and P4 as reaction temperature rise. If a change in intent results in a disturbance—namely, a change in N1's head within a preset range or an increase in N2's resistance—this disturbance is injected into N1 and N2. The simulation then proceeds along the physical path N1→N2→N3→N4 and the logical path P1→P2→P3→P4. The system is not concerned with a single changed value, but rather with whether these changes, over time, will lead to heat accumulation or insufficient control margin near N4. Regarding the anomaly handling mechanism, if the change intent data contains missing equipment codes, missing change values, or conflicting descriptions, the system can first perform field integrity verification. If the verification passes, it proceeds to parameterization processing; if the verification fails, the change is marked as pending supplementation, and the injection of disturbances into the simulation link is suspended to avoid erroneous approval. If the real-time operational data has key sensors offline, timestamp drift, or sudden changes in operating conditions, the system prioritizes using stable data from the most recent reliable time period as a temporary operating condition baseline. If the degree of missing data exceeds the preset tolerance range, a prompt is output indicating that only the structural-level assessment has been completed and the operational-level assessment has not been completed, to prevent incomplete data from being mistakenly used as a safety basis. If a node in the simulation link lacks material boundary or control strategy information, the system can adopt conservative boundary processing for that node, that is, increase the risk level in the approval result instead of defaulting to safety. For example, in a fine chemical plant, process engineers plan to replace feed pump P-201 with a spare pump of another specification, and simultaneously reduce the opening of regulating valve V-203 from 50% to 45% to suppress upstream flow fluctuations. The data acquisition module reads this request from the MOC work order, retrieves the relationships between P-201, V-203, preheater E-301, and reactor R-401 from the 3D model and control database, and then reads the feed temperature, reactor jacket temperature, cooling water flow rate, and tower inlet pressure for the current shift from the field. After intent parsing, the system identifies this change as a combined disturbance of increased upstream resistance and altered pump characteristics, and injects it into the corresponding node. Simulation results show that in the early period after the change is implemented, the DCS control loop can still maintain the surface temperature stability of the reactor; however, in the second preset time period, which is longer than the first preset time period, due to the amplified fluctuation of the preheater outlet temperature, the local heat release peak of the reactor tends to approach the safety boundary; the system then gives a conditional approval result: if the cooling loop flow rate is increased simultaneously and the heat exchanger inspection cycle is shortened, the change can be implemented; if no process compensation is added, there is a risk of exceeding the limit in the future. The purpose of this step is to extend MOC approval from static object identification to dynamic deduction based on future operational consequences, so as to establish a traceable and explainable causal link between local changes and global security, thereby enabling early identification and compensation guidance for hidden risks. In this embodiment, the digital twin base data includes physical space topology data, process logic topology data, and material property library data; Among them, physical space topology data is used to characterize the spatial location relationships of devices; Among them, process logic topology data is used to characterize the logical relationships of the control system.

[0017] This embodiment provides a mechanism for organizing digital twin base data. Specifically, in the aforementioned continuous fine chemical production line scenario, simply recording equipment names and parameter limits is usually insufficient to support real MOC analysis, because many risks do not come from single-point values ​​themselves, but from the spatial proximity of equipment, the logical coupling of control loops, and the tolerance boundaries of the material itself. Therefore, this embodiment subdivides the base data into three categories: physical space topology data, process logic topology data, and material property library data, and makes the three together constitute the basic constraints for subsequent simulations. The following is a detailed description: Physical spatial topology data is used to describe the actual location relationships, connectivity methods, and relative elevations of equipment in the plant, pipe racks, and skids; for example, the feed pump is located on the ground floor, the preheater is located on the second-floor platform, and there is a high-temperature short pipe section between the reactor and the condenser; the significance of this type of information is that certain pressure pulsations, heat accumulation, or leakage consequences can be affected by spatial distance, layout direction, and adjacent equipment; Process logic topology data reflects which parameter is affected by which control loop, and which interlock depends on the logical association of which measuring point. For example, when the reactor outlet temperature is higher than the set value, the opening of the cooling water regulating valve should be increased; when the pressure in front of the tower increases, the reflux regulating loop may be activated; such logical relationships determine that the same change will have different consequences under different control modes; material property library data is used to describe the temperature resistance, pressure resistance, corrosion resistance and thermal conductivity of pipes, seals, heat exchange surfaces and reactor linings; the reason is that when local parameters have not yet exceeded the process control threshold, long-term damage to the material level may have already begun to accumulate, such as the effect of being in an over-temperature state that exceeds the preset range of the set threshold on the aging rate of gaskets; For ease of understanding, a simplified structural description can be used. Let equipment E1, E2, and E3 be the feed pump, preheater, and reactor, respectively. Spatially, E1 and E2 are connected by a riser pipe, and E2 and E3 are connected by an insulated pipe of a preset length. Logically, T1 is the preheater outlet temperature measuring point, T2 is the reactor jacket temperature measuring point, and L1 is the cooling valve control signal. Changes in T1 affect T2, and T2 further drives the control chain of L1. In terms of materials, the lifespan of the heat exchanger tube bundle will decrease rapidly when the long-term operating temperature is higher than a certain value. Thus, when upstream changes increase flow pulsation, the system not only knows that the flow rate has changed, but also knows that this change will be quickly transmitted to the front end of the reactor via the riser pipe and the insulated short pipe, causing additional impact on the heat exchange surface before the control loop intervenes. Regarding anomaly handling, if only one or two of the three types of base data are present, the system can still perform limited assessments, but the capability boundaries need to be marked in the results. For example, if only spatial topology is available but control logic is lacking, the system can determine the transmission paths of materials and energy, but it is difficult to estimate the buffering effect of the control loop. If only logical relationships are available but material boundaries are lacking, the system can determine the interlocking trigger sequence, but it cannot accurately assess the long-term material loss risk. For new model components missing from the material library, the system can call the conservative reference range of the same type of material, or automatically raise the approval conclusion by one level of caution until the material data is completed. For example, in the aforementioned device, the change in flow characteristics after the replacement of P-201 initially occurs at the pump outlet. However, due to the short-distance high-temperature pipe section arrangement between the preheater E-301 and the reactor R-401, the thermal inertia is small, and the disturbance is more easily and quickly transmitted to the reactor inlet. At the same time, there is a certain delay in the action of the jacket cooling valve in the reactor temperature control loop. In addition, the E-301 tube bundle material is more prone to fatigue under long-term temperature fluctuations. Based on this, the system judges that even if the main reactor temperature has not immediately exceeded the limit on the surface, the heat exchanger and the near-end pipe section may still become the early damage location. The purpose of this step is to provide a low-level reference with engineering constraints for subsequent intent injection, graph construction and spatiotemporal extrapolation, thereby realizing the transformation from data existence to risk interpretability; In this embodiment, the intent parsing and injection module includes: The feature extraction unit is used to perform natural language processing on the change intent data to extract parameterized features that represent the change object, change type, and change value. The vector generation unit, connected to the feature extraction unit, is used to generate micro-perturbation vectors based on parameterized features; The node matching unit, connected to the vector generation unit, is used to identify the target topology node corresponding to the micro-perturbation vector and inject the micro-perturbation vector into the target topology node contained in the digital twin base data.

[0018] This embodiment provides a change intent parsing and injection mechanism. Specifically, simply archiving MOC applications as documents makes it difficult to enter the simulation chain. However, directly sending the original natural language text into the simulation lacks stable engineering semantic support. Therefore, this embodiment breaks down the process into three consecutive steps: feature extraction, vector generation, and node matching, so that unstructured change descriptions can be transcribed into accurate micro-perturbation expressions that can be accepted by the twin model. The details are as follows: The feature extraction unit performs natural language processing on the change intent data, including entity association scanning and syntactic mapping, to ensure accurate extraction of parameterized features representing the change object, change type, and change value. For example, replacing the feed pump with a B model and reducing the valve opening by at least 5% instead of relying on fuzzy condition extraction, the changes are refined into specific change objects, such as the feed pump and regulating valve, change types, such as pump model replacement and flow component opening reduction, and highly specific change values, such as the rated flow characteristic function difference coefficient and equivalent control amount of -5%. In contrast, the scheme of directly extracting non-quantitative operating conditions instead of converting and calculating the actual change value will hinder the simulation of the lower plate due to the lack of numerical basis. The vector generation unit connected to the feature extraction unit generates the corresponding micro-perturbation vector based on the parameterized features. The micro-perturbation vector must encapsulate two types of associated verification quantities in its structure: one is the node identifier, which is used to index the root cause of the impact; the other is to take over the aforementioned change value and use it as the actual perturbation value assignment measure. Following that, at the end of the processing flow, the node matching unit actively connects to the underlying ledger and identifies the target topology node corresponding to the micro-perturbation vector. After confirming that the link is unambiguous, the micro-perturbation vector is injected into the target topology node contained in the digital twin base data. For ease of understanding, a simplified system model can be used. If the application text is to replace P-201 with a type B pump and adjust the opening of V-203 from 50% to 45%, then after feature extraction, two sets of strictly matched parameterized features can be formed: the first set of features directly corresponds to the changed object P-201, its change type is equipment-level replacement operation, and the change value is mapped to a function variable of the head and displacement difference; the second set aligns with the changed object V-203, the change type is labeled as opening reduction, and the change value is directly parsed as -5%; when generating vectors, based on these attributes, two-body micro-perturbation vectors M1 and M2 carrying specific node identifiers and actual disturbance values ​​are directly derived; when matching nodes, the system searches for P-201 in the equipment ledger and V-203 in the control chart. After completing the precise correspondence, the node matching unit injects the incremental vector with the disturbance value into the above-mentioned pump node and valve node; Regarding anomaly handling, if the text contains abbreviations, colloquial names, or historical reference numbers, such as the old No. 2 valve of the front main pump, the node matching unit can call the alias dictionary and historical mapping table for secondary matching. If multiple candidate objects still appear after secondary matching, the system will not directly inject the data but will instead generate a list for manual confirmation. If the text only contains vague descriptions such as adjusting the flow rate within a preset range to reduce the preset temperature, the system can first extract the change direction and mark the change value as incomplete. During the deduction, a range perturbation method will be used for conservative evaluation. If even the change type or direction cannot be determined, the automatic approval will be suspended, and only the parsing traces will be retained for subsequent supplementary entry. If an application contains contradictory change descriptions, such as the same valve being both closed and fully open, the system should identify it as conflicting semantics and refuse to enter the injection stage. For example, in the aforementioned production line, the MOC submitted by the process engineer states: Considering the recent fluctuations in raw material viscosity, it is proposed to replace P-201 with a spare type B pump, and reduce the opening of V-203 from 50% to 45%, while keeping the reactor load unchanged during implementation; After system analysis, two core change objects and their derived deterministic change values ​​are identified; Since keeping the reactor load unchanged means that the downstream control loop will actively compensate for the upstream disturbance, during injection, not only are disturbances with clear disturbance values ​​delivered to the P-201 and V-203 nodes to trigger propagation, but also pre-adjusted instruction action tags are attached to the reactor load control related logic nodes to restore the deduction and amplification on the coupling chain; The purpose of this step is to transform the MOC description, which is originally only suitable for human reading, into an engineering perturbation unit with a unique mathematical quantification standard that can be propagated in a digital twin environment, thereby achieving a one-to-one correspondence between the semantic changes and the simulation object and strengthening the constraints. In this embodiment, the topology map construction module is used for: Extract equipment node data and process parameter node data from the digital twin base data; Based on equipment node data and process parameter node data, a directed graph containing physical connection edges and logical connection edges is constructed. The directed graph is output as a cross-domain process topology graph.

[0019] This embodiment provides a mechanism for constructing a cross-domain process topology graph. Specifically, if the device is only viewed as an equipment list, the system can only answer which devices exist; if the device is only viewed as a parameter table, the system can only answer which data exceeds the limit; while MOC risk really needs to answer in what direction a change at a certain node will be transmitted to which objects; therefore, this embodiment integrates equipment nodes and process parameter nodes into a single directed graph and explicitly distinguishes between physical connection edges and logical connection edges. The details are as follows: Equipment node data is used to characterize physical objects such as pumps, valves, heat exchangers, reactors, towers, sensors, and controllers; process parameter node data is used to characterize state objects such as flow rate, pressure, temperature, liquid level, component concentration, vibration amplitude, and valve position feedback; physical connection edges mainly correspond to pipes, cables, heat exchange surface contact paths, rotating machinery coupling paths, etc., reflecting the actual direction in which materials, energy, or mechanical forces can be transmitted; logical connection edges reflect the control or causal influence between parameters, such as flow rate changes affecting residence time, residence time affecting outlet temperature, outlet temperature affecting reaction rate, and reaction rate affecting heat release intensity; by retaining both types of edges simultaneously, the system can not only know which pipeline the disturbance travels through, but also when it crosses the physical domain, transforming from a flow problem into a thermal problem or from a thermal problem into a control problem; To illustrate the diagram construction method, a simplified example can be used. Assume the equipment nodes include D1 as the feed pump, D2 as the regulating valve, D3 as the preheater, and D4 as the main reactor; the parameter nodes include Q1 as the feed flow rate, P1 as the pump outlet pressure, T1 as the preheater outlet temperature, and T2 as the reactor temperature rise; the physical connection edges can be set as D1→D2, D2→D3, and D3→D4; the logical connection edges can be set as Q1→T1, T1→T2, and P1→Q1. Thus, when a slight disturbance occurs at the pump node, the system can trace the equipment chain along D1→D2→D3→D4 and the parameter chain along P1→Q1→T1→T2. If the controller node C1 also issues a regulation command to the cooling valve node D5, the logical edges T2→C1 and C1→D5 can be added, thereby forming a closed-loop influence chain. Compared with the basic scheme, if the deduction is based solely on a single physical connection, the delay or amplification effect of control actions on risk propagation may be underestimated under certain extreme conditions; if the deduction is based solely on control logic, the thermal inertia, pressure drop and local stress caused by the actual pipeline and equipment layout may be ignored; therefore, this embodiment strengthens the previous scheme by using a bilateral type diagram to make the risk transmission path complete. Regarding anomaly handling, if a device node exists but has no online parameter nodes (e.g., a standby pump that has not been put into use for a long time), the system can still retain it as a static device node and activate the relevant parameter edges when implementing changes. If the logical relationship cannot be automatically extracted from the control system, it can be supplemented by the typical impact rules preset in the process knowledge base. If the automatic extraction result conflicts with the manually maintained interlocking table, the interlocking table that has been reviewed and published takes priority. For isolated nodes, i.e., nodes that have neither physical edges nor logical edges, the system can mark them as objects to be checked and not include them in the main link health calculation to avoid meaningless propagation. For example, in the aforementioned production line, P-201, V-203, E-301, and R-401 constitute a continuous process chain. After the graph is constructed, the system can simultaneously express the physical path of the material from the pump outlet entering the preheater and then the reactor through the valve, as well as the parameter path of the change in heat exchange efficiency of the preheater caused by flow fluctuations, which affects the peak heat release of the reactor. In this way, when P-201 is replaced, the subsequent simulation no longer stops at the static understanding of equipment replacement, but can track how the replacement propagates across equipment and parameter levels. The purpose of this step is to establish a computable and traceable cross-domain correlation network, thereby achieving a unified description of the propagation direction, propagation level, and potential amplification links of MOC disturbances. In this embodiment, the spatiotemporal transmission simulation module includes: The reduced-order multiphysics coupling unit is used to perform physical field coupling calculations on the cross-domain process topology map using a reduced-order model that extracts low-order dominant features through dimensionality reduction mapping, generating coupled state data. Specifically, the dimensionality reduction mapping can employ principal component analysis or intrinsic orthogonal decomposition algorithms. By performing eigenvalue decomposition on the full physical field state data corresponding to the cross-domain process topology map, the eigenvectors corresponding to the top N largest eigenvalues ​​are extracted to construct a projection matrix, where the value of N is determined based on a preset cumulative variance contribution rate threshold. This extracts a few dominant features that are most sensitive to safety consequences from the high-dimensional full physical state. The time acceleration simulation unit, connected to the reduced-order multiphysics coupling unit, is used to perform time-step iterative simulation based on coupled state data, micro-perturbation vectors, and real-time running state data, using a time-series prediction algorithm based on the state transition matrix, to generate simulation data of the spatiotemporal ripple effect.

[0020] This embodiment provides a mechanism for implementing spatiotemporal conduction simulation. Specifically, while performing full-scale fluid, heat transfer, structural, and control joint simulations on the entire plant is theoretically more refined, it often fails to meet timeliness requirements during MOC approval and is limited by model size and computational resources. Conversely, completely abandoning physical coupling and relying solely on empirical rules makes it difficult to identify delayed and cumulative risks. Therefore, this embodiment combines reduced-order multiphysics coupling with time-accelerated extrapolation to achieve a feasible approval speed while preserving key process mechanisms. The following is a detailed explanation: The role of the reduced-order multiphysics coupling unit is to extract the original high-dimensional, continuous, and computationally complex full physical state into a few dominant states that are most sensitive to safety consequences. Taking a continuous reaction device as an example, what truly determines the development of risk is usually not all local details, but rather dominant characteristics such as main flow stability, key heat exchange efficiency, reaction heat release intensity, control valve adjustable margin, and key material heat load. This unit combines these dominant characteristics into coupled state data to express the overall response trend of the current device under the coupling of multiple domains of heat, flow, and control. The time acceleration extrapolation unit, based on this, uses the injected micro-perturbations and real-time on-site conditions as the starting point for future extrapolation, and extrapolates subsequent states step by step according to a preset time step to generate simulation data of the spatiotemporal ripple effect. The acceleration here does not skip the mechanism, but replaces the second-by-second full recalculation with a stable state evolution relationship, so that the risk profile can be seen in the approval window several hours, days, or even weeks later. A simplified system model is used to illustrate the data organization method; the coupled state data includes S1 as the stability of the feed, S2 as the heat exchange margin, S3 as the reaction exothermic load, and S4 as the control loop adjustment margin; when the micro-disturbance M1 acts on the feed pump node and M2 acts on the valve node, the system first obtains the initial coupled state based on the cross-domain spectrum, such as the decrease in the stability of the feed and the slight contraction of the heat exchange margin; The time acceleration simulation unit operates according to a preset time. , , The state is updated sequentially: In the early time slices, the control loop may absorb most of the fluctuations, making the change in S3 not obvious; in subsequent time slices, if S2 continues to decrease, S3 may begin to increase, further compressing S4; what is ultimately formed is not a single conclusion, but a state trajectory that evolves over time. Specifically, this time series prediction process can be implemented by defining discretized state deduction logic; Coupled state data vector at time step Combined with the injected micro-perturbation vector, i.e., the time-driven operation vector Ut, the state transition matrix reflects the inherent coupling evolution relationship between the dominant features of the device under the current operating conditions. And a mapping input matrix that reflects the immediate impact weights of external change actions on each state characteristic. Calculate after a preset time step After State vector at time step ,in, Represents matrix multiplication. The preset time step is expressed by the following formula:

[0021] Wherein, the state transition matrix and mapping input matrix It can be obtained by subspace system identification of the historical steady-state operation data of the device, or by off-line linearization expansion of the full-size process mechanism model at the current operating condition reference point; By initializing the real-time running data to the initial state By continuously executing the time series prediction algorithm with a preset time step, the transmission trajectory of each dominant state in the sequence time can be quickly deduced and obtained, and finally the simulation data of the spatiotemporal ripple effect is generated. Compared to the aforementioned schemes that only construct graphs, simply knowing the nodes and edges is insufficient to support approval, because the graph itself does not answer how fast, how strong, or how long the impact will last. Therefore, this embodiment introduces reduced-order coupling and time-accelerated extrapolation to strengthen the previous layer scheme and transform the structural relationships into predictable dynamic results. In terms of anomaly handling, if a subsystem has not yet completed high-fidelity modeling, a reduced-order unit can be replaced by a similar process template model, and the output should indicate that this part is the template evaluation result; if the real-time operating conditions deviate significantly from the applicable range of the model, such as the unit being started up or shut down or undergoing emergency load shedding, the system can pause long-cycle simulations and only provide short-term trend warnings; if micro-disturbances involve multiple coupled units, resulting in multiple solutions or instability in the state evolution, the system will prioritize outputting the most unfavorable operating condition path for conservative judgment approval; if the time step setting is greater than the preset step threshold, causing critical transition processes to be masked, the system can automatically refine the time resolution of the risk escalation phase. For example, in the aforementioned device, the replacement of P-201 and the reduction of the opening of V-203 together cause changes in the front-end feeding characteristics; the down-order unit maps this change to a coupled state of decreased feeding stability, increased fluctuations in the heat exchange efficiency of the preheating section, and a compression of the reactor cooling adjustment margin; after accelerated time simulation, the system found that the reactor main temperature was still within the control range in the first preset time period after implementation, but in the subsequent operation stage, due to the superposition of heat exchanger fouling coefficient and flow fluctuations, the cooling loop gradually approaches saturation, and the amplitude of reaction heat fluctuations shows signs of amplification in the future period; in this way, the approvers can see the full picture of the consequences developing over time before the change is implemented; The purpose of this step is to compress the multi-domain coupling consequences of complex devices into dynamic simulation results that can be completed within the approval timeframe, while ensuring the interpretability of the process mechanism, thereby enabling early exposure of long-term accumulated risks. In this embodiment, the time acceleration simulation unit is specifically used for: Using real-time operating data as the initial operating baseline, along the connecting edges of the cross-domain process topology map, the parameter propagation trend of the micro-disturbance vector within a preset time period is calculated, where the preset time period is greater than or equal to the preset process response delay time. Based on the parameter propagation trend, the simulation data of the spatiotemporal ripple effect is output, where the predicted values ​​of each process parameter are represented as the attenuation or amplification values ​​of the parameter relative to the initial operating condition baseline.

[0022] This embodiment provides a mechanism for extrapolating the trend of parameter propagation. Specifically, although the previous solution can provide the overall state of evolution over time, in some process scenarios, the approver also needs to know whether the disturbance is gradually controlled to shrink during propagation or is amplified by certain abnormal boundaries. Especially for continuous production devices with huge thermal inertia, hysteresis regulation and long cascade control, if the selected observation and extrapolation time period is too short, that is, the preset time period is shorter than the preset process response delay time, the transient hysteresis effect will mistakenly determine the non-limited state of the system during the response delay period as a safe state. Therefore, this embodiment imposes a strict requirement on the time axis coverage depth of the simulation: the simulation must use real-time operating data as the initial operating condition baseline, and the preset time period used for evaluation must be greater than or equal to the preset process response delay time, so as to ensure that it can run through the entire dynamic and static transformation period and accurately distinguish the parameter propagation trend. The following is a detailed explanation: Using real-time operating data as the initial operating baseline means that before the time-accelerated simulation unit begins simulation and prediction, it determines the physical baseline boundary conditions of the current device, such as the load rate, stability margin, and pipeline back pressure of the current device. After determining the physical simulation baseline, the simulation process begins to calculate the parameter propagation trend of the micro-disturbance vector within a long preset time period under controlled conditions, step by step, along the connecting edges of the cross-domain process topology map. If a node along the simulation chain, such as an intermediate tank with a large capacity buffer or a pressure stabilizing point that relies on the strong integral effect of the controller to offset the pressure, has an effective buffering function, it can significantly dissipate and absorb excessive fluctuation energy potential during the process. Conversely, if it encounters a sensitive node at a design bottleneck, such as a phase change heat exchange surface in a near-full-load state or even a reactor in the extreme of the sensitive zone, it is very easy to cause the fluctuation amplitude to diverge or even increase exponentially. Therefore, based on this, in the subsequent spatiotemporal ripple effect simulation data output outward based on the parameter propagation trend, all the monitoring metrics, i.e., the predicted values ​​of each process parameter, are no longer static numerical values, but are structured as attenuation or amplification values ​​of the parameters. By directly representing them as attenuation or amplification values, it is possible to accurately characterize whether a certain risk fluctuation is being gradually digested and weakened by the intrinsic tolerance or is continuously accumulating within the system. For ease of understanding, a simplified chain can be used; let the node chain be N1 feed pump → N2 valve → N3 preheater → N4 reactor, and the parameter chain be Q flow rate → T inlet temperature → H exothermic intensity; if the baseline judgment indicates that the current cooling water allocation reserve is higher than the preset allocation margin threshold, then after a sufficiently long time delay, N3 will inevitably have a significant suppression and smoothing effect on the Q anisotropic fluctuation transmitted from upstream, causing the relative disturbance generated by the amplitude of T relative to Q to be attenuated according to a preset ratio. At this time, the predicted values ​​of each process parameter generated in the simulation output are exactly the attenuation values ​​of the parameters for the fluctuation. Conversely, if the system is already approaching the critical polarity zone of deterioration, then if T experiences a small positive fluctuation, it will cause H to rise exponentially or exceed the safety threshold. This will be clearly shown as an amplified value of the parameter in the generated predictive data stream array. Therefore, the final feedback of the system points directly to the core: it not only outlines where the route leads, but also clearly indicates in which segment the attenuation value was digested and in which segment the amplified value was deteriorated. Regarding the anomaly handling mechanism, if real-time operating data indicates that the device is in the start-up, shutdown, flushing, or fault switching phase, the normal response time mechanism under normal production conditions will no longer hold constant. In this case, rigidly applying the normal hysteresis time domain would be unreliable. The system should automatically switch to a strong conservative mode and dynamically extend the upper limit of the preset process response delay time tolerance window to support calculations for a longer preset time period. If the increase or decrease of the feedback prediction data at a certain local node is lower than the preset fluctuation threshold compared to the inherent signal-to-noise ratio low-frequency jitter of the instrument and is judged as a failure stray value, it will be automatically removed as a steady-state filter item in the graph output to prevent excessive noise alarms from diluting the true amplified evolutionary risks. For example, in the aforementioned device, the site is currently operating at 85% of its designed full load, and the reactor cooling loop has been over-tuned. The simulation module starts its judgment based on the baseline of this period, and calculates the propagation process covering the entire link through a preset time period. The results show that although the initial impact of the front-end feeding initially reflects a mild attenuation value characteristic in the thermal vessel through short-range hysteresis, it eventually spreads to the end and the reserved time limit has been made up. At the end of the parameter response delay period, it inversely amplifies the sensitive area amplification effect, causing the exothermic peak to rise with an extremely steep amplification value at a certain delayed point in the future, approaching the safety line. This clearly shows that only by setting a simulation span longer than the fixed process hysteresis period can the hidden extension and aggravation of the safety crisis become traceable and visible. The purpose of this step is to deepen the analysis of pure disturbance propagation path by locking the threshold of the extrapolation time slice covered by the long-wave delay compensation mechanism, so as to give a positive and negative impact feedback array judgment sequence with trend polarity, thereby intercepting any cumulative hidden risk conditions with delay effect and apparent stability. In this embodiment, the health status determination and feedback module includes: The threshold comparison unit is used to compare the predicted values ​​of each process parameter contained in the simulation data of the spatiotemporal ripple effect with the preset safety critical threshold of the corresponding process parameter. The status determination unit, connected to the threshold comparison unit, is used to determine that if the predicted value of the process parameter exceeds the safety critical threshold range, the topology node corresponding to the predicted value of the process parameter has a risk of exceeding the limit, and generate a global health assessment result containing risk warning information. The status determination unit is also used to determine that the topology node corresponding to the predicted process parameter value is in a safe state if the predicted process parameter value is within the safe critical threshold range, and to generate a global health assessment result containing safe passage information.

[0023] This embodiment provides a global health assessment and feedback mechanism. Specifically, although the modules of the aforementioned scheme have achieved multi-dimensional calculation and can completely output spatiotemporal propagation simulation trajectories with directionality and time correlation, for engineering approval and execution, if the parameters exceeding the limit are extracted by human subjective consciousness through the lack of rigid standardized unified standard alignment process, the review threshold and workload will be high, and the scale control will be more prone to dangerous and uncertain operational risks such as individual differences, inconsistent strictness, and even omissions. Therefore, this embodiment sets up a strict threshold comparison unit and a state judgment logic mapping unit based on binary rules, namely the state judgment unit, which is responsible for purifying and converging the generalized spatiotemporal ripple effect simulation values ​​into an array of accurate evaluation conclusions with unique judgment standards that can be intuitively processed and can be used for manual review before downstream delivery. The details are as follows: When the judgment process is initiated, the threshold comparison unit serves as the entry point. Its core operational responsibility is to sequentially capture the predicted values ​​of each process parameter contained in the simulation data of the spatiotemporal ripple effect generated by the upstream calculation through the mapping traversal action, and perform a hard numerical comparison action that is highly aligned with the safety critical threshold of the corresponding process parameter called by the storage and distribution in real time. These numerous safety critical thresholds, which serve as the cornerstone of the measurement scale, are not only derived from the single conventional operation control upper limit value, but also include and even deeply integrate and encompass a group of stringent red line rules from the entire domain, such as the safety limit of equipment design, the boundary of material fatigue resistance long-cycle tolerance value, and even the lower limit of redundant interlocking control feedback adjustment capacity dissipation. Following the completion of the basic value surface verification, the status determination unit connected by the control flow receives the status sequence response and forms a system-level report. The specific configuration determination process is divided into two branches with absolute orientation. For one branch: if it is confirmed in the column-by-column verification that the predicted value of the process parameter is greater than or equal to the safety critical threshold, then the fault alarm lock-up mechanism is immediately triggered in strict accordance with the downgrade red line rule and it is determined that the single point state has been affected by the system imbalance. The risk judgment is immediately issued and determined that the topology node corresponding to the predicted value of the process parameter has the risk of exceeding the limit. Then, this risk element state is extracted and injected into the information flow to generate a global health assessment result reporting matrix containing risk warning information. In the corresponding parallel verification branch process: if the state calibration module backtests and confirms that the predicted value of the process parameter is still completely under control and strictly below the safety critical threshold, then the risk of breaking through the safety defense line is clearly ruled out and the topology node corresponding to the predicted value of the process parameter is determined to be in a safe state. At the same time, a complete global health assessment result containing safe passage information is generated to complete a comprehensive map report. The global result emphasized here is not, in essence, a purely computational summary of the out-of-limit points. It is a set of related judgments formed by back-pushing back from each end parameter and on the macro topology coordinate flow line. It assigns which risk element becomes the initial risk node that triggers the crisis, and ensures that safety endorsement is issued to the unaffected parts. To facilitate understanding, a simplified judgment process can be adopted; assume that the spatiotemporal model prediction results after long-span evolution are sent to the outflow end: the T2 parameter sequence is assigned as the array of predicted values ​​for reactor temperature rise trend parameters; after the flow tube is connected, it is pressed into the detection: if the calculation and identification determine that the T2 curve trend has gradually deteriorated and is approaching the upper limit extreme value of its inherent matching temperature allowable tolerance zone, that is, its exclusive preset safety critical threshold upper limit point, then the judgment module forcibly marks and anchors this point and marks and confirms that the reactor node itself and its corresponding process parameters have a high risk of exceeding the limit, and at the same time packages the results with a direct risk warning label as an early warning feedback report output; Conversely, if the inspection and comparison confirm that the predicted values ​​in the operation sequence of all associated nodes, such as the pressure extreme value and the tolerance of the operating valve orifice threshold, are all less than or equal to their respective safety critical thresholds, then the sub-process will release the confirmation beacon according to the order to mark and verify that all nodes in this series are in a safe state. The resulting integrated output will be a list of global health assessment results containing safety pass information, which will be presented to the decision-makers in the approval chain with the characteristics of execution confirmation identification. Through this design, the two-way investigation has complete interpretability, thereby achieving the status of two-way clearing reports. Regarding the anomaly handling mechanism, since the multi-dimensional fusion inevitably leads to redundant and exclusionary judgments, if the thresholds extracted from different reference dimension sources expose numerical collisions and rule contradictions, for example, when the permissible fault tolerance extreme value in the process operation standard manual is abnormally higher than the bottom limit parameter of the long-term damage tolerance assessment of a certain special aging material, the strong regulatory logic is given priority and must call the most conservative, strict, and numerically restricted set of red line scales as the current unified and only comparison benchmark to avoid false expectation errors. If a node in the approved batch exhibits parameter over-limit behavior within the first preset time period, but the duration window shrinks drastically and then quickly recovers, the system still does not allow this slight overload data to be directly filtered out. Instead, the entire sequence log is continuously exported and written to the record result pool matrix for review to determine its acceptability, in order to avoid automatic filtering by the algorithm. If the integrity and reliability verification of the data stream of a critical high-risk node fails to pass the check, the default pass-through behavior will also be terminated, and the node status will be locked for manual strong review and verification without issuing an unconditional pass. For example, in the aforementioned system linkage instance application, spatiotemporal simulation, after extending the detection delay, reveals signs of deteriorating feedback: the local extreme exothermic peak generated in the core section inside reactor R-401 shows an increasing trend of approaching the outer edge of the safety isolation control package and exhibiting a tendency to lose connection within the extended time domain; when this crisis prediction is reviewed at the downlink of the health monitoring feedback link, a hard judgment is made after threshold comparison and strict comparison by the internal judgment logic chain's penetrating state determination unit: R-401 is undoubtedly clearly identified as having high density The key risk points that exceed the limits are marked as the main core risk points. The warning feedback is directly extracted and sealed together with the status of other surrounding auxiliary edge weak warning devices. After compilation, a global health assessment result warning message containing full-dimensional analysis and warning level of the defense zone is broadcast to the control screen and the final review flow terminal. This plan and related actions have triggered a deep-level chain thermal imbalance and related component runaway risk warning in the coupled scenario. Under the premise of not having effective process compensation measures, this change will be prohibited. The purpose of this step is to forcibly use a consistent hard index comparison scale to reduce and standardize all the complex and difficult high-dimensional results of the calculation fluctuations in each domain to one dimension and output them as the final execution status and action notification approval order set. In this way, intelligent review and adjudication can replace experience and visual inspection, and generate objective, quantifiable and unambiguous standardized evaluation conclusions. In this embodiment, the system further includes: The 3D visualization module communicates and connects with the health assessment and feedback module to generate a dynamic risk heat map in the 3D digital twin model built on the digital twin base data based on the global health assessment results. Among them, the dynamic risk heat map is used to characterize the transmission path and scope of impact of risk over time.

[0024] This embodiment provides a three-dimensional visualization display mechanism. Specifically, the previous layer solution can output structured health assessment results. However, in large continuous equipment, relying solely on textual conclusions or two-dimensional lists often makes it difficult for approvers and operators to quickly understand where the risk originates, along what path it takes, and which areas it ultimately affects. Especially in cross-unit equipment, the risk path may simultaneously pass through the equipment chain, control chain, and adjacent spatial areas. Therefore, this embodiment adds a three-dimensional visualization display module to map the global health assessment results onto a digital twin model, forming a dynamic risk heat map. The following is a detailed description: The 3D visualization module reads the equipment's geometric information, spatial coordinates, pipeline layout, and layered structure from the base, and then combines this with the risk nodes, propagation direction, and time segments from the assessment results to complete the 3D projection of the risk path; The so-called dynamic risk heat map is not a static way of coloring dangerous equipment red, but rather a way of showing the expansion process of risk intensity over time; For example, initially only a mild warning color appears near the feed pump outlet, which deepens along the valve to the preheater pipe section, and a high-risk hot zone forms near the reactor inlet and jacket; In this way, the operator can intuitively see the risk migrating from the front-end feed chain to the core reaction section, rather than just seeing a few discrete parameter exceedance points; A simplified scenario can be used: In the 3D model, region A is the feeding area, region B is the heat exchange area, and region C is the reaction area. If the assessment results show that the risk propagation chain is A→B→C, with a brief buffer at B and a significant amplification at C, then the heat map can show A slightly bright at time slice t1, B at medium brightness at t2, and C at high brightness with path connections at t3. In this way, approvers not only know the final risk but also see the intermediate transmission chain. Compared to solutions that only output evaluation text, the lack of three-dimensional dynamic representation leads to lower risk communication efficiency in complex environments with multiple personnel involved, and it is easy to overlook spatial proximity effects, such as the close proximity between high-risk equipment and flammable material corridors. Therefore, this embodiment uses visualization to fill the gaps in understanding in engineering decision-making scenarios. In terms of anomaly handling, if a device does not yet have a high-precision 3D model, the system can use a simplified geometry for placeholder display, but still retain its risk color and propagation direction; if the risk path in the assessment results crosses multiple floors or areas, the module can switch between layered views, section views, or path-focused views to avoid occlusion; if the risk credibility of individual nodes is insufficient, the heatmap can use different textures or borders to indicate that it needs to be reviewed, without confusing it with confirmed high-risk nodes; if the system detects that the current display terminal performance is insufficient, it can prioritize displaying key devices and the main propagation chain, and temporarily not render secondary details; For example, in the aforementioned device, after the digital twin 3D interface is brought up during the approval meeting, the system displays the risk propagation path between P-201, V-203, E-301, and R-401 in the form of a dynamic heat map. Meeting participants can see that after the change is implemented, a slight risk coloring first appears in the front-end pump outlet area; the color of the short pipe sections at the preheater inlet and outlet is enhanced; then, a significant high-temperature risk hot zone is formed near the reactor inlet and jacket, accompanied by time stamps showing that the risk increases after a delay of several process cycles; this performance enables the equipment, process, and safety parties to quickly reach a consensus. The purpose of this step is to transform the abstract simulation evaluation results into a spatial and temporal expression that can be intuitively understood by on-site engineers, thereby achieving more efficient approval communication and risk identification. In this embodiment, the health determination and feedback module further includes: The compensation scheme generation unit, connected to the status determination unit, is used to respond to risk warning information and calculate the adjustment amount required to offset parameter deviations based on the cross-domain process topology map. The compensation scheme generation unit is also used to encapsulate the adjustment amount into a process compensation scheme and output it.

[0025] This embodiment provides a process compensation scheme generation mechanism. Specifically, if the system can only prohibit implementation after a risk is detected, although some accidents are avoided, a large number of changes that could be safely implemented with additional conditions may still be rejected, affecting production flexibility. Especially in continuous production units, many local changes are not absolutely impossible to implement, but require simultaneous adjustment of cooling, reflux, inspection frequency or control parameters. Therefore, this embodiment further generates a process compensation scheme after a risk warning. The details are as follows: The compensation scheme generation unit reads the risk nodes, parameter deviations, and propagation paths output by the status determination unit, and then performs a reverse search along the cross-domain process topology map to find adjustable points that can offset the deviation. The so-called reverse search does not mean mechanically backtracking the equipment list, but rather tracing back along the influence chain to find which upstream or bypass parameters have adjustment capabilities and whose direction of action can offset the current risk. For example, if the reactor heat load prediction is too high, nodes with intervention capabilities such as cooling loop flow rate, preheater outlet temperature setting, front-end feeding cycle time, reflux ratio, or interlocking early warning threshold can be searched first. The system generates adjustment amounts for these nodes and encapsulates them into process compensation schemes with execution conditions, target objects, and durations. These compensation schemes can be single adjustments or combinations of multiple adjustments to avoid concentrating pressure on a single control point. More specifically, the reverse calculation of adjustment amounts is based on the system graph and its inherent variable sensitivity linkages. When the state determination unit identifies that future spatiotemporal parameter predictions will face exceeding safety thresholds, it calculates the parameter deviation vector that needs to be eliminated. ; Simultaneously, during the reverse search, a set of compensation nodes containing multiple adjustment parameters is located; the system then extracts a sensitivity mapping matrix that matches the operating conditions, recording the influence of the local partial derivative of the compensation amount of each unit node on the target constraint exceeding the limit under the current cross-domain network. The local partial derivative influence value within this matrix can be obtained by simulating and testing each relevant node of the cross-domain process topology map by injecting a unit step disturbance, and recording the calculated ratio of the steady-state changes generated at the target constraint node; an adjustment vector is established to solve for the adjustment required to offset the aforementioned parameter deviations. Guiding model:

[0026] Among them, the right side of the formula Represents the deviation vector from the parameters Zero vectors of the same dimension; This represents an element-by-element comparison of vectors, where each element of the left-hand vector is less than or equal to... Furthermore, the derivation of this formula is based on the scenario where the predicted parameter value exceeds the upper limit of the preset safety critical threshold, i.e. This represents a positive deviation; if the evaluation scenario involves a parameter drop below a preset safety threshold, the inequality sign is reversed accordingly. ; The system uses the remaining control margin of real equipment as a boundary constraint to solve matrix equations for optimization. The optimization objective is to make the adjustment vector... The norm is the smallest, which means the control cost is the lowest; thus, it is possible to accurately quantify and obtain operational guidance data such as how much the opening of the compensation water valve should be increased by to offset the upstream heat load, and finally encapsulate it into the output process compensation scheme. A simplified example can be used to illustrate this: If the evaluation results show that the temperature rise of reactor T2 exceeds the safety boundary, and its main propagation chain comes from the flow fluctuation of Q1 and the high inlet temperature of T1, then the system can find two types of compensation points during the reverse search: one is to reduce the preheating degree of T1, and the other is to increase the heat absorption capacity of the cooling loop for T2; when encapsulating the output, the system can generate a combined recommendation to implement the change, lower the preheating target and increase the cooling flow rate, rather than just indicating that there is a risk; Compared to the previous layer's solution, which can only provide risk assessment, the lack of a compensation solution generation mechanism often leads to a rejection or manual discussion in the approval process, making it difficult to form a stable and rapid closed loop. Especially in high-frequency MOC scenarios, relying entirely on expert meetings to formulate compensation measures item by item is inefficient and inconsistent. Therefore, this embodiment strengthens the previous layer's solution through automated reverse intervention suggestions. Regarding anomaly handling, if the adjustable point obtained from the reverse search is already close to its operational limit, such as a cooling valve that is almost fully open, the system will no longer recommend this node as the primary compensation method to avoid creating a new bottleneck. If multiple compensation actions conflict with each other, such as reducing the preheating target affecting the downstream separation efficiency, the system can output multiple alternative solutions and indicate their respective impact ranges for approvers to choose from. If sufficient compensation capabilities cannot be found, the system should clearly state that there is no feasible compensation solution, rather than outputting formalized suggestions. If the compensation solution involves changes to the safety instrumented system settings, the system must upgrade it to a mandatory manual review item and not allow automatic issuance. For example, in the aforementioned device, the system determines that replacing P-201 and lowering the opening of V-203 will lead to a risk of R-401 exceeding the heat release peak limit during future operation. After reverse analysis along the spectrum, the compensation scheme generation unit finds that this risk can be offset by two main measures: first, increasing the jacket cooling water flow rate and intervening in control in advance; second, appropriately reducing the preheating target of E-301 to reduce the heat load at the reaction inlet. The system finally outputs a conditional approval recommendation: pump replacement and valve position adjustment are allowed, but the cooling circuit capacity must be increased simultaneously, the preheating setting must be lowered, and the E-301 inspection cycle must be shortened from the original plan. If it is confirmed on-site that the cooling system currently has no additional capacity, the system automatically switches the conclusion to not recommending implementation. The purpose of this step is to provide an engineering-feasible mitigation path after a risk is identified, thereby extending MOC management from problem identification to closed-loop handling.

[0027] 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 technical solutions of the present invention.

Claims

1. A digital twin-based intelligent identification and process analysis management system for MOCs, characterized in that, The system includes: The data acquisition module is used to acquire change intent data, digital twin base data, and real-time operational data; The intent parsing and injection module communicates with the data acquisition module and is used to extract features from the change intent data to generate parameterized features. Based on the parameterized features, it generates a micro-perturbation vector containing node identifiers and perturbation values, identifies one or more topological nodes corresponding to the micro-perturbation vector as target topological nodes, and injects the micro-perturbation vector into the target topological nodes contained in the digital twin base data. The topology map construction module communicates with the data acquisition module and is used to construct cross-domain process topology maps based on digital twin base data. The spatiotemporal transmission simulation module is connected to the intent parsing and injection module, the topology map construction module, and the data acquisition module, respectively. It is used to take real-time running data as the initial operating condition baseline, and based on the cross-domain process topology map with injected micro-perturbation vectors, iteratively deduce according to a preset time step to generate spatiotemporal ripple effect simulation data containing the predicted values ​​of each process parameter. The health assessment and feedback module communicates with the spatiotemporal conduction simulation module. It is used to compare the predicted values ​​of each process parameter contained in the spatiotemporal ripple effect simulation data with the preset safety critical threshold range of the corresponding process parameter to generate a global health assessment result. The configuration is as follows: if the global health assessment result indicates that there is a risk of exceeding the limit, the parameter deviation between the predicted value of the process parameter and the safety critical threshold is calculated, the adjustment amount required to offset the parameter deviation is calculated in reverse based on the cross-domain process topology map, and a process compensation scheme is generated based on the adjustment amount. If the global health assessment result indicates that the system is in a safe state, a confirmation message indicating that the system is operating normally will be output.

2. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 1, characterized in that, The digital twin foundation data includes physical space topology data, process logic topology data, and material property library data; Among them, physical space topology data is used to characterize the spatial location relationships of devices; Among them, process logic topology data is used to characterize the logical relationships of the control system.

3. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 1, characterized in that, The intent parsing and injection module includes: The feature extraction unit is used to perform natural language processing on the change intent data to extract parameterized features that represent the change object, change type, and change value. The vector generation unit, connected to the feature extraction unit, is used to generate micro-perturbation vectors based on parameterized features; The node matching unit, connected to the vector generation unit, is used to identify the target topology node corresponding to the micro-perturbation vector and inject the micro-perturbation vector into the target topology node contained in the digital twin base data.

4. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 1, characterized in that, The topology graph building module is used for: Extract equipment node data and process parameter node data from the digital twin base data; Based on equipment node data and process parameter node data, a directed graph containing physical connection edges and logical connection edges is constructed. The directed graph is output as a cross-domain process topology graph.

5. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 2 or 4, characterized in that, The spatiotemporal transmission simulation module includes: The reduced-order multiphysics coupling unit is used to perform physical field coupling calculations on the cross-domain process topology map using a reduced-order model that extracts low-order dominant features through dimensionality reduction mapping, and generate coupled state data. Specifically, the dimensionality reduction mapping can use principal component analysis or intrinsic orthogonal decomposition algorithm. By performing eigenvalue decomposition on the full physical field state data corresponding to the cross-domain process topology map, the eigenvectors corresponding to the top N largest eigenvalues ​​are extracted to construct a projection matrix, thereby extracting a few dominant features that are most sensitive to safety consequences from the high-dimensional full physical state. The time acceleration simulation unit, connected to the reduced-order multiphysics coupling unit, is used to perform time-step iterative simulation based on coupled state data, micro-perturbation vectors, and real-time running state data, using a time-series prediction algorithm based on the state transition matrix, to generate simulation data of the spatiotemporal ripple effect.

6. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 5, characterized in that, The time acceleration simulation unit is specifically used for: Using real-time operating data as the initial operating baseline, along the connecting edges of the cross-domain process topology map, the parameter propagation trend of the micro-disturbance vector within a preset time period is calculated, where the preset time period is greater than or equal to the preset process response delay time. Based on the parameter propagation trend, the simulation data of the spatiotemporal ripple effect is output, where the predicted values ​​of each process parameter are represented as the attenuation or amplification values ​​of the parameter relative to the initial operating condition baseline.

7. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 1, characterized in that, The health assessment and feedback module includes: The threshold comparison unit is used to compare the predicted values ​​of each process parameter contained in the simulation data of the spatiotemporal ripple effect with the preset safety critical threshold of the corresponding process parameter. The status determination unit, connected to the threshold comparison unit, is used to determine that if the predicted value of the process parameter exceeds the safety critical threshold range, the topology node corresponding to the predicted value of the process parameter has a risk of exceeding the limit, and generate a global health assessment result containing risk warning information. The status determination unit is also used to determine that the topology node corresponding to the predicted process parameter value is in a safe state if the predicted process parameter value is within the safe critical threshold range, and to generate a global health assessment result containing safe passage information.

8. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 7, characterized in that, The system also includes: The 3D visualization module communicates and connects with the health assessment and feedback module to generate a dynamic risk heat map in the 3D digital twin model built on the digital twin base data based on the global health assessment results. Among them, the dynamic risk heat map is used to characterize the transmission path and scope of impact of risk over time.

9. The MOC intelligent identification and process analysis management system based on digital twin as described in claim 7, characterized in that, The health assessment and feedback module also includes: The compensation scheme generation unit, connected to the status determination unit, is used to respond to risk warning information and calculate the adjustment amount required to offset parameter deviations based on the cross-domain process topology map. The compensation scheme generation unit is also used to encapsulate the adjustment amount into a process compensation scheme and output it.

Images