A method, system, device and medium for NCES condition signal reconstruction based on reachable graph difference analysis

Abstract

A NCES condition signal reconstruction method, system, device and medium based on reachable graph difference analysis, the method comprising: modeling a system under test as NCES, obtaining discrete event observation data of the system under test, generating an observed reachable graph and a theoretical reachable graph; topologically comparing the observed reachable graph with the theoretical reachable graph of the unconditional signal, extracting a difference reachable graph, and dividing a state set; based on the condition signal preliminary screening of the state count matrix, a set of feasible solutions of the condition signal is solved; an ILP model based on Big-M is constructed, and all feasible solutions of the condition signal after preliminary screening are solved; a relaxation variable is introduced into the ILP model, a minimum error objective function is constructed, and an optimal solution of the condition signal is solved; the system, device and medium are used to realize the method; the application adopts the condition signal characteristic equivalence representation module between the NCES; by analyzing the observation data, the unknown condition signal and the multi-condition signal processing mode are inversely inferred, and the accurate inference of the inter-module control logic is realized.

Description

Technical Field

[0001] This invention belongs to the field of NCES conditional signal reconstruction technology, and particularly relates to a method, system, device and medium for NCES conditional signal reconstruction based on reachability graph difference analysis. Background Technology

[0002] In the field of modern industrial automation and discrete event systems, with the increasing complexity and modularity of production systems, engineering maintenance, fault diagnosis, and legacy system upgrades face severe challenges. In actual engineering sites, due to long-term equipment operation, personnel turnover, or lack of supplier documentation, engineers often find it difficult to obtain the original, detailed control logic design documents of the system. Typically, they can only acquire the system's state change sequence during operation through sensors, i.e., the system's reachability graph (RG) and the internal structural information of each independent functional module.

[0003] To reproduce and understand the control behavior of a system, it is necessary to reverse-engineer the control logic between modules. In recent years, Network Conditional / Event Systems (NCES) have been widely used in the modeling of industrial automation systems due to their unique signaling mechanism. Furthermore, in such reverse-engineering problems, the NCES model provides a mathematical framework that can accurately describe the control relationships between modules, equating the physical control logic between modules in actual industrial settings (such as electrical interlocks, enable signals, and status allow bits in a PLC) to conditional signals in the NCES model. Unlike traditional Petri nets, NCES introduces "conditional signals" and "event signals." Conditional signals allow or suppress target transitions by detecting whether the source state is satisfied, and this process does not consume the source state's token (resource). This "non-consumable" and "read-only" logical characteristic perfectly matches the physical essence of industrial control signals—"reading only the state and not consuming resources"—but also makes conditional signals highly concealed in token-based flow observation data, making them difficult to directly identify using traditional flow-based reconstruction methods.

[0004] In summary, the existing technology has the following defects and shortcomings: 1. The computational complexity grows exponentially, making it difficult to handle large-scale systems; Typically, a global state space search strategy is used, which involves substituting all possible signal connection methods into the system model for verification. As the number of system modules and the scale of states increase, the search space explodes exponentially. This high computational cost limits existing methods to small-scale experimental systems, failing to meet the real-time or near-real-time diagnostic needs of large-scale systems in industrial settings.

[0005] 2. Weak ability to identify complex logical patterns; In practical industrial control, transitions are often enabled by a combination of multiple conditions (e.g., module C can only start when module A is ready and module B is idle). However, in actual industrial applications, simplification assumptions are often made, such as assuming that all signals are single-control or only support AND gate logic. This makes it difficult to distinguish and identify the complex control modes of OR gate logic, leading to inconsistencies between the reconstructed model and the actual system behavior.

[0006] 3. Highly sensitive to the integrity of observation data, lacking robustness; Existing algebraic or finite difference methods typically assume that the observed reachability graph is complete and accurate. However, in practical engineering, due to limitations in sensor layout or sampling frequency, observation data often contains blind spots or sensor noise. When blind spots exist, existing methods are prone to inference interruptions or incorrect attributions due to the lack of key state features, and lack robust decision-making strategies in the case of feature degradation.

[0007] 4. Lack of a mechanism for quickly removing irrelevant modules; In multi-module systems, only a few modules are tightly coupled in terms of control, while a large number of modules are independent of each other, i.e., unrelated modules. Existing methods lack an effective pre-screening mechanism, making it impossible to quickly identify and filter out these unrelated modules without applied control signals before fine-grained reconstruction. This results in the algorithm wasting most of its computational power on a large number of invalid candidate relationships, severely slowing down the overall reconstruction efficiency. Summary of the Invention

[0008] To address the problems of high computational complexity, sensitivity to observation blind spots, and difficulty in identifying complex logic gates (AND / OR gates) in existing discrete event system logic reconstruction, this invention aims to provide a method, system, device, and medium for NCES conditional signal reconstruction based on reachability graph difference analysis. It uses NCES as the modeling basis and leverages its unique conditional signal characteristics to equivalently represent the control logic between modules. By analyzing the system's observation data (such as reachability graphs), it inversely infers the lost conditional signals and their multi-conditional signal processing modes, thereby achieving accurate inference of the control logic between modules.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for reconstructing NCES conditional signals based on reachability graph difference analysis includes the following steps: Step 1: Model the system under investigation as an NCES, obtain the discrete event observation data of the system under investigation, and generate the observed reachability graph RG and the theoretical reachability graph RG0; Step 2: Perform a topological comparison between the observed reachability graph RG and the theoretical reachability graph RG0 of the unconditional signal, extract the difference reachability graph, and divide the state set, including the existence state set. Directly missing state set Indirect missing state set ; Step 3: Generate a state counting matrix based on the state set. By comparing the differences in the elements of the state counting matrix, perform preliminary screening of the condition signals and solve for a set of feasible solutions to the condition signals. Step 4: Construct an ILP model based on Big-M and solve for all feasible solutions to the conditional signals after initial screening; Step 5: Introduce slack variables into the ILP model from Step 4. We construct a minimum error objective function and solve for the optimal solution of the conditional signal.

[0010] The set of existence states : Includes all data in the observation data that cause the change The set of source states for normal emission; the set of states that are directly missing. : Includes all source states reachable from the observed data, but transitions The set of source states that are not enabled; the set of indirectly missing states. : Includes all changes in the observation data caused by the unreachability of the source state. The set of source states that are not enabled.

[0011] The process of generating the state counting matrix is ​​as follows: Each state According to the individual modules of the system to be investigated Decomposed into module state combinations Generate a state count matrix, including the theoretical triggering baseline matrix. Direct missing edge counting matrix and indirect missing edge counting matrix The theoretical triggering reference matrix This indicates the transition under unconditional signal control. The theoretical maximum number of triggers in each module state, the direct missing edge counting matrix. Represents the set of directly missing states. The number of times each module state appears, the indirect missing edge counting matrix Indicates the set of indirectly missing states. The number of times each module state occurs; the expression for the theoretical triggering benchmark matrix is ​​as follows: in, The number of reachable states for each module. For module The A reachable state.

[0012] The initial screening process for conditional signals based on the state counting matrix is ​​as follows: Definition The effective trigger baseline matrix represents the baseline of the total number of effective triggers after removing indirect missing triggers; the state count matrix is ​​compared for differences in matrix elements to initially screen condition signals: If there exists a module state that satisfies This indicates that all triggers are currently blocked, and the multi-condition signal processing mode is then determined. It is an AND gate, and if there are still transitions in the states of other modules in the current module. If it can be enabled normally, the module's state holds the condition library location of the token; If there exists a module state that satisfies And there exist other modules in the current module whose states satisfy the condition. When, it indicates the module state. The enable function remains unaffected; the module state is as follows. The holding of the treasury in China is a conditional treasury; If all module states of a certain module are... ,in If the value is a constant, it indicates that the current module is an irrelevant module. The connection weights of irrelevant modules are then reset to zero for dimensionality reduction.

[0013] The construction process of the ILP model based on Big-M is as follows: For different sets of states in the observation reachability graph RG, combined with mode decision variables... With logical decision variables Establish three types of hard constraints: For the set of existing states Forced change Conditional enable: , ; For the set of directly missing states Forced change Unconditional enable: , ; For the set of indirectly missing states Introducing logical decision variables To infer the potential enabling conditions of the transition: in, This is the system state row vector after binarization. Let be a row vector consisting entirely of 1s, where , For collection of warehouses, For the conditional signal matrix Change and Transformation Related column vectors, For larger numbers, usually take .

[0014] The introduction of slack variables The ILP model is as follows: For the set of existing states The elements in the text are: , ; For the set of directly missing states The elements in the text are: , ; For the set of indirectly missing states The elements in the text are: .

[0015] The objective function for minimizing the error is: in, Used to minimize the degree of deviation from observed data, i.e., error correction; Used to find the simplest structure for connecting lines according to Occam's razor. For slack variables The weighting coefficients, For change Hekushu The corresponding conditional signal.

[0016] This invention also provides an NCES conditional signal reconstruction system based on reachability graph difference analysis, comprising: Data acquisition and reachability graph generation module: Models the system under investigation as NCES, acquires discrete event observation data of the system under investigation, and generates the observation reachability graph RG and the theoretical reachability graph RG0; State set partitioning module: Performs topological comparison between the observed reachability graph RG and the theoretical reachability graph RG0 of the unconditional signal, extracts the difference reachability graph, and partitions the state set, including the existential state set. The set of directly missing states Indirect missing state set ; The state counting initial screening module includes a matrix generation unit and a difference analysis unit. Based on the state set, the matrix generation unit generates a state counting matrix; the difference analysis unit compares the element differences of the state counting matrix to perform initial screening of conditional signals and solves a set of feasible solutions for the conditional signals. LIP Model Building Module: Constructs an ILP model based on Big-M and solves all feasible solutions for the conditional signals after initial screening; LIP Model Optimization Module: Introducing Slack Variables into the ILP Model We construct a minimum error objective function and solve for the optimal solution of the conditional signal.

[0017] The present invention also provides an NCES conditional signal reconstruction device based on reachability graph difference analysis, comprising: Memory: A computer program that stores the above-described NCES conditional signal reconstruction method based on reachability map difference analysis, and is a computer-readable device; Processor: Used to implement the above-described NCES conditional signal reconstruction method based on reachability graph difference analysis when executing the computer program.

[0018] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described NCES conditional signal reconstruction method based on reachability graph difference analysis.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. It greatly reduces computational complexity and enables rapid reconfiguration of large-scale systems; This invention innovatively introduces a preliminary screening mechanism based on a state counting matrix. Unlike exhaustive searches in a vast global state space, this invention decomposes the global state into module-level local sub-states and performs algebraic operations by constructing an effective triggering benchmark matrix and a direct missing edge counting matrix. This approach reduces the algorithm's time complexity from exponential to polynomial. This enables the invention to process large-scale NCES containing numerous modules at extremely low computational cost, meeting the real-time diagnostic needs of industrial sites.

[0020] 2. It has the ability to quickly remove irrelevant modules, which significantly improves the convergence speed of the algorithm; This invention, by calculating the element differences between the effective triggering baseline matrix and the direct missing edge counting matrix, can intuitively identify which modules' state distributions show no significant characteristics before and after the transition. Based on this characteristic, the algorithm can automatically identify and filter out irrelevant modules without applied control signals before fine reconstruction. This mechanism is equivalent to significantly reducing the dimensionality of the problem before solving, avoiding invalid calculations, and thus significantly improving the overall reconstruction efficiency.

[0021] 3. It breaks through the limitations of a single logical assumption and achieves accurate identification of complex "AND / OR" logic patterns; In step 4, this invention introduces an integer linear programming (ILP) model and specifically designs logical pattern decision variables within the model. By constructing a hybrid linear constraint system covering both AND and OR gate scenarios, the algorithm no longer relies on manually preset single logical assumptions. Instead, it automatically searches for the logical combination that best matches the observed data through a mathematical programming solver. This enables the invention to perform simultaneous searches under multiple states and multiple transitions (events), ensuring a high degree of consistency between the reconstructed model and the behavior of the real physical system.

[0022] 4. Enhanced robustness to observation blind spots and data noise, making it suitable for non-ideal observation environments; To address the issues of observation blind spots (indirect missing data) and sensor noise, the ILP model of this invention introduces logical decision variables and slack variables, transforming the solution objective from a "hard constraint satisfaction problem" to a "global optimization problem of minimizing error." For observation blind spots, the model can automatically infer the most reasonable potential enabling conditions to fill in the missing data; for spurious states caused by sensor noise, the model can automatically tolerate and filter data conflicts by minimizing the error objective function. This enables the invention to output the optimal solution with the highest confidence even under harsh conditions where observation data is incomplete or contains errors, demonstrating strong engineering applicability.

[0023] In summary, compared with existing technologies, this invention adopts a hierarchical reconstruction strategy. It first utilizes state counting for low-cost, rapid dimensionality reduction and initial screening, and then employs the ILP method for high-precision logical optimization. This architecture avoids the computational bottleneck caused by directly using ILP to process the entire dataset, and also compensates for the algorithm failure caused by interference when using algebraic methods alone, achieving a balance between efficiency and accuracy. Attached Figure Description

[0024] Figure 1 This is a flowchart of the NCES conditional signal reconstruction method of the present invention.

[0025] Figure 2 This is a schematic diagram of the unconditional signal system in Example 1.

[0026] Figure 3 The observation reach diagram is for the unconditional signal system of Example 1.

[0027] Figure 4 This is the theoretical attainability diagram of the unconditional signal system in Example 1.

[0028] Figure 5 The difference can be obtained for the unconditional signal system of Example 1.

[0029] Figure 6 This is a schematic diagram of the conditional signal system in Example 1.

[0030] Figure 7 This is the reachability diagram of the conditional signal system in Example 1.

[0031] Figure 8 This is a schematic diagram of the unconditional signal system in Example 2.

[0032] Figure 9 This is the observation reach diagram for the unconditional signal system in Example 2.

[0033] Figure 10 This is the theoretical attainability diagram of the unconditional signal system in Example 2.

[0034] Figure 11 The difference can be obtained for the unconditional signal system of Example 2. Detailed Implementation

[0035] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0036] A method for reconstructing NCES conditional signals based on reachability graph difference analysis includes the following steps: Step 1: Model the system under investigation as an NCES, use the operating data of the system under investigation collected by the sensor as the observation data of the discrete events of the system, generate the observation reachability graph RG, and generate the theoretical reachability graph RG0 of the unconditional signal system by combining the NCES structure with the emission rules. Step 2: Perform a topological comparison between the observed reachability graph RG and the theoretical reachability graph RG0 of the unconditional signal, extract the difference reachability graph, and divide the state set; Traversing the differences leads to every transition in the graph. and status Determine the set of missing association pairs Based on the missing properties, they are divided into sets of directly missing pairs. and indirect missing pairs of sets According to the changes Whether the emission is directly missing, the set of transitions controlled by conditional signals is obtained. Further establish the following set of states: Existence of a set of states : Includes all data in the observation data that cause the change The set of source states for normal transmission; Directly missing state set : Includes all source states reachable from the observed data, but transitions The set of source states that are not enabled; these states represent transitions. It is subject to the limitations imposed by external condition signals; Indirect missing state set : Includes all changes in the observation data caused by the unreachability of the source state. The set of source states that are not enabled constitutes the identification blind zone; Step 3: Initial screening of conditional signals based on the state counting matrix; Each state According to the individual modules of the system to be investigated Decomposed into module state combinations The state counting matrix, including the theoretical triggering reference matrix, is generated through a matrix generation unit. Direct missing edge counting matrix and indirect missing edge counting matrix The theoretical triggering reference matrix This indicates the transition under unconditional signal control. The theoretical maximum number of triggers in each module state, the direct missing edge counting matrix. Represents the set of directly missing states. The number of times each module state appears, the indirect missing edge counting matrix Indicates the set of indirectly missing states. The number of times each module's state appears.

[0037] The theoretical triggering benchmark matrix expression is as follows: in, The number of reachable states for each module. For module The One reachable state; definition The effective trigger baseline matrix represents the baseline of the total number of effective triggers after removing indirect missing triggers. The state count matrix is ​​compared using a difference analysis unit, and the condition signals are generated only between modules. Initial screening of condition signals follows: If there exists a module state that satisfies This indicates that all triggers are currently blocked, and the multi-condition signal processing mode is then determined. It is an AND gate, and if there are still transitions in the states of other modules in the current module. If it can be enabled normally, then the module's state holds the token's location as a conditional location; if there exists a module state that satisfies... And there exist other modules in the current module whose states satisfy the condition. When, it indicates the module state. The enable function remains unaffected; the module state is as follows. The repository holding the token is the condition repository; if all module states of a certain module have... ,in If it is a constant, it indicates that the current module is an irrelevant module, and the transition is irrelevant. Without any control function, the connection weights of irrelevant modules are reset to zero, and irrelevant modules are removed, thus achieving dimensionality reduction.

[0038] Based on the above judgment, a set of feasible solutions for the conditional signal can be obtained. If it is necessary to obtain all feasible solutions for the current system, proceed to the next step.

[0039] Step 4: Construct an integer linear programming (ILP) model based on Big-M, solve all feasible solutions for the conditional signals after initial screening, and transform the logical relationship after initial screening into a mathematical programming problem; For different sets of states in the observation reachability graph RG, combined with mode decision variables With logical decision variables Three different types of hard constraints are established using a constraint generator: For the set of existing states Forced change Conditional enable: , ; For the set of directly missing states Forced change Unconditional enable: , ; The above two types of constraints are based on pattern decision variables. Selecting the automatic matching multi-condition signal processing mode ; For the set of indirectly missing states Further introduce logical decision variables To infer changes Potential enabling scenarios: in, This is the system state row vector after binarization. Let be a row vector consisting entirely of 1s, where , For collection of warehouses, For the conditional signal matrix Change and Transformation Related column vectors, For larger numbers, usually take ; Step 5: Introduce slack variables into the ILP model from Step 4. Construct an objective function that minimizes the error; To handle potential sensor noise or transmission errors (leading to spurious states) in industrial field observation data, a slack variable is set through a fault-tolerant configuration unit. The weighting coefficients are determined, and non-negative relaxation variables are introduced into the hard constraints in step 4. hard constraints ( Transformation into soft constraints To adapt to industrial environments with different signal-to-noise ratios; The introduction of slack variables The ILP model is as follows: For the set of existing states Forced change Conditional enable: , For the set of directly missing states Forced change Unconditional enable: , ; For the set of indirectly missing states ,change Potential enabling scenarios: The objective function for minimizing the error is: in, Used to minimize the degree of deviation from observed data, i.e., error correction; Used to find the simplest structure for connecting lines according to Occam's razor. For slack variables The weighting coefficients, For change Hekushu The corresponding conditional signal.

[0040] Step 6: Call an external mathematical programming solver to solve the ILP model from Step 5, and analyze the optimal solution vector of the conditional signals, i.e., output the conditional signal matrix. and multi-condition signal processing mode It automatically generates a data file containing complete conditional signal logic and visualizes the reconstructed system topology; The system outputs conditional signals and all feasible or optimal solutions of the multi-conditional signal processing mode, determining the connection source and target of the conditional signals; adding corresponding conditional arcs to the original NCES structure, generating a restored reachability graph according to the NCES emission rules, and comparing it with the observed reachability graph RG. If they match, it indicates that the correct conditional signal solution has been obtained, that is, the control logic between system modules has been restored.

[0041] The system first reads the observation reachability map RG of the system under investigation through the data acquisition interface. It then enters the first-level processing stage, using a constructed state counting algorithm to calculate the difference between the baseline matrix and the counting matrix. At this stage, the system automatically identifies statistically significant explicit conditional signals and irrelevant modules that have no impact on the control logic. If the initial screening results can completely determine all logical relationships, i.e., there are no fuzzy solutions, the system will directly output the reconstructed model, thus avoiding high computational costs. If undetermined candidate relationships still exist after the initial screening or an observation blind zone is detected, the system will automatically enter the second-level processing stage. At this point, the fuzzy logic mapping after removing irrelevant modules is transformed into a mixed-integer linear programming model. Logical pattern decision variables and slack variables are introduced, and a mathematical programming solver searches globally for the optimal solution that satisfies the physical constraints and minimizes the error. Finally, the system integrates the results of the two-level processing to generate an NCES network model containing complete conditional signals on the original modular structure.

[0042] To fully verify the effectiveness of the above-mentioned hierarchical processing procedure under different operating conditions, the following two representative differentiated implementation scenarios are set up: Example 1: For a typical scenario with a small number of observation blind spots and a large number of modules, this example verifies how the present invention can quickly eliminate irrelevant modules and locate signals, perform state counting and filtering, and verify the efficiency of the algorithm.

[0043] Taking a reactor cooling safety control system in a continuous chemical production scenario as the system under investigation, the conditional signal screening algorithm based on the state counting matrix proposed in step 3 is used to verify its effectiveness in a multi-module complex logic system. This system aims to ensure the production safety of the reactor through coordinated control of cooling pump start / stop, safety interlock switching, and load adaptation. The system consists of three independent functional subsystems: Cooling pump circulation subsystem : Responsible for starting and resetting the daily / emergency cooling pump; Interlocking state subsystem Controls the opening (disable operation) and disengagement (allow operation) of safety interlocks; Load Status Subsystem Monitor the migration of heat load levels (low / medium / high).

[0044] In practical engineering, the observed behavior trajectory of the system has been acquired through sensor networks. This system is then modeled as an NCES (Non-Computer Integrated System) to be investigated. ,in For a collection of modules, This represents the system's initial state, corresponding to the physical meanings of cooling ready, interlock activated, and low load. The object to be reconfigured is the condition signal matrix. and multi-condition signal processing mode The structure of an unconditional signaling system is as follows: Figure 2 As shown, the physical meaning of each location and transition is detailed in Table 1, and the codes of all possible reachable states involved in the system are detailed in Table 2.

[0045] Table 1. Physical meaning of each storage location and its changes Table 2. All possible reachable state codes involved in the system The observation reachability diagram of the unconditional signal system obtained based on long-term observations. ( Figure 3 The theoretical attainability diagram of unconditional signal systems ( Figure 4 Perform topological comparison to extract the difference reachability graph of the unconditional signal system, such as... Figure 5 As shown in red in the middle.

[0046] Analysis revealed a set of missing association pairs in the system. It contains 19 elements. Based on the missing properties, it is divided into: sets of directly missing pairs. There are 18 pairs. Such omissions indicate change. In a specific state The conditional signal directly suppresses the indirect loss of the set. .state Unreachable in the observation reachability map, leading to changes The behavior in this state cannot be directly observed. Based on this, the set of transitions controlled by conditional signals is located as follows: .

[0047] According to the algorithm flow, the set of directly missing states for each transition is first processed. and indirect missing state set The statistics were compiled, and the results are as follows: Subsequently, for all states Perform module-level decomposition and construct a direct missing edge counting matrix. (Table 3) Indirect missing edge counting matrix (Table 4) and the effective benchmark matrix (Table 5).

[0048] Table 3 Directly Missing Edge Count Matrix Table 4 Indirect Missing Edge Count Matrix Table 5 Effective Triggering Baseline Matrix By comparison and Based on the statistical characteristics, the conditional signals of each transition are inferred as follows: change (Daily Startup): In the module status ( Under the condition of holding a token, the direct missing count is 3, which perfectly matches the baseline value of 3, indicating that the transition is completely suppressed in this state; while in the state... ( If the missing count under (holding tokens) is not equal to the baseline value, it indicates that the transition can still be enabled normally. Irrelevantness determination: in module In all states, the missing count distribution is uniformly smaller than the effective baseline. Judgment module This is an irrelevant module. Conclusion: Transition Subject to module Single-signal control, its condition library is (Interlock Release), Multi-condition Signal Processing Mode For AND (a special case of single signal).

[0049] change (Emergency Startup): In the module status Completely missing below; in module status and The following is completely missing. Only when the module is... In state and In state Time, Change Complete loss did not occur. Conclusion: Change Multiple conditions must be met simultaneously to determine its multi-condition signal processing mode. For AND, the set of condition libraries is (Interlock released and medium load).

[0050] change (Interlock Release): In the module In state Below, the missing counts are all equal to the effective baseline value. This indicates that during the cooling process, the interlock release operation is completely prohibited, and in the module... In any module state, the condition for determining irrelevant modules is satisfied, therefore the transition can be inferred. It is only permitted when the cooling system is ready, as defined by its condition library. Multi-condition signal processing mode For AND.

[0051] change (Interlock Activation): In the module In state and Below, the missing counts are all equal to the effective baseline value. This indicates that under these two low / medium load conditions, the interlocking activation operation is completely prohibited. Conclusion: Inferring the transition. Only under high load conditions The following is allowed, and its condition library does so. Multi-condition signal processing mode For AND.

[0052] change and (Load fallback migration): For transitions Direct missing edge counting matrix The module states have non-zero values, but none of them are equal to the valid baseline value. This indicates that a single condition is insufficient to completely determine the enabling of the transition, conforming to the partial inhibition characteristic of an OR gate. By backtracking the enabling state, it is found that as long as the module... In state ( Holding a token or module In state ( Holding Token), Changes This can happen. Conclusion: Determine the transition. Multicondition signal processing mode OR, as defined by the condition library Similarly, determine the changes. Multicondition signal processing mode This is the OR pattern, as defined by the condition library. .

[0053] Based on the above inferences, the final conditional signal connection matrix is ​​obtained. With multi-condition signal processing mode : The reconstructed conditional signal connection matrix With multi-condition signal processing mode Substitute into an unconditional signal system Generate a conditional signal system ,like Figure 6 As shown. Formal verification shows that the generated reachability graph... ( Figure 7 ) and the original observation reach map ( Figure 3 Completely isomorphic.

[0054] The reconstruction results accurately reproduced the system's security control logic: This reveals a dual confirmation mechanism for emergency pump startup, which requires intervention only when the interlock is released and the load is medium, thus preventing misoperation. The circuit breaker mechanism of "high load triggering forced interlock" has been verified. , (OR gate logic): This reflects the flexibility of system load adjustment. Regardless of the subsystem state, as long as any safety condition is met, the system allows load degradation, embodying a fault-oriented safety design philosophy. The above results demonstrate that the state-count-based reconstruction algorithm can accurately extract implicit coupling logic between modules from limited observation data, providing a reliable tool for the digital auditing of legacy industrial systems.

[0055] Example 2: For complex scenarios with a large number of observation blind spots and sensor noise, the ILP model is verified to perform logic completion and noise-resistant inference, thus verifying the robustness of the algorithm.

[0056] Using an industrial pipeline valve interlocking system as the test system, the effectiveness of the conditional signal reconstruction algorithm based on integer linear programming proposed in step 4 of this invention is verified. The core function of this system is to achieve stable operation of the industrial fluid transport process through graded pressure control of the main pipeline, diversion control of branch pipelines, and pressure over-limit monitoring interlock. The system mainly consists of three functional units: Main pipeline valve unit : Responsible for graded regulation of fluid pressure (off / low pressure / medium pressure / high pressure); Branch pipe valve unit : To enable the diversion and switching of fluid to the production area or storage area; Pressure monitoring unit Monitor pipeline pressure status and trigger over-limit alarms.

[0057] The system is formally modeled as NCES. ,in For a set of modules, the initial state The corresponding physical meanings are main valve closed, branch valve closed, and pressure normal. The object to be reconstructed is the condition signal matrix. and multi-condition signal processing mode The structure of the system to be investigated is as follows: Figure 8 As shown in Table 6, the physical meaning of each node is shown in Table 7, and all the state codes that may be involved in the system operation are shown in Table 7.

[0058] Table 6 Physical meaning of each node in the system to be investigated Table 7 All state codes that may be involved in system operation The unconditional signal system observation map obtained through on-site data collection is available. ( Figure 9 The theoretical attainability diagram of unconditional signal systems ( Figure 10 Perform topological comparison to extract the difference reachability graph of the unconditional signal system, such as... Figure 11 As shown in red in the middle.

[0059] Analysis revealed that the system contains a large number of unreachable states (such as...). (etc.), making traditional algebraic reconstruction methods difficult to apply directly. First, extract the set of missing association pairs in the system. Determine the set of transitions based on the definition of direct missing. There must exist conditional signals. For these three key transitions, we further divide their state sets: Based on the above set partitioning, the Big-M based ILP model proposed in this invention is used. For example, for transitions... The model will simultaneously establish guarantees Enable constraints and guarantees Suppression constraints, and measures for blind spots Logical selection constraints.

[0060] After inputting the model into the solver (CPLEX / Gurobi), the solution set of all feasible condition signals for the system is shown in Table 8.

[0061] Table 8. Set of all feasible condition signals for the system Table 8 reveals the system control logic hidden behind the observation data. (Based on changes) Taking (branch 1 open) as an example, the solution set shows that all feasible solutions contain the core signal set. This indicates that a "low pressure" or "medium pressure" state is a necessary prerequisite for opening the branch valves in the production area, reflecting the protection logic to prevent high-pressure fluids from directly impacting production equipment. Regarding the changes... (Pressure alarm), the results show (Branch 2 opening) is the key factor that triggers the alarm, indicating that there may be a risk of pressure fluctuation in the pipeline in the storage area.

[0062] The ILP method provides a complete list of all logically equivalent feasible solutions, which not only verifies the effectiveness of the algorithm but also provides system maintainers with a means to identify redundant signals (such as those in the solution set). The theoretical basis for non-essential signals (such as signals that are not necessary) helps to further optimize the control logic design.

Claims

1. A method for reconstructing NCES conditional signals based on reachability graph difference analysis, characterized in that, Includes the following steps: Step 1: Model the system under investigation as an NCES, obtain the discrete event observation data of the system under investigation, and generate the observed reachability graph RG and the theoretical reachability graph RG0; Step 2: Perform a topological comparison between the observed reachability graph RG and the theoretical reachability graph RG0 of the unconditional signal, extract the difference reachability graph, and divide the state set, including the set of existential states. The set of directly missing states Indirect missing state set ; Step 3: Generate a state counting matrix. By comparing the differences in the elements of the state counting matrix, perform preliminary screening of the condition signals and solve for a set of feasible solutions to the condition signals. Step 4: Construct an ILP model based on Big-M and solve for all feasible solutions to the conditional signals after initial screening; Step 5: Introduce slack variables into the ILP model from Step 4. We construct a minimum error objective function and solve for the optimal solution of the conditional signal.

2. The NCES conditional signal reconstruction method according to claim 1, characterized in that, The set of existence states : Includes all data in the observation data that cause the change The set of source states for normal emission; the set of states that are directly missing. : Includes all source states reachable from the observed data, but transitions The set of source states that are not enabled; the set of indirectly missing states. : Includes all changes in the observation data caused by the unreachability of the source state. The set of source states that are not enabled.

3. The NCES conditional signal reconstruction method according to claim 1, characterized in that, The process of generating the state counting matrix is ​​as follows: Each state According to the individual modules of the system to be investigated Decomposed into module state combinations Generate a state count matrix, including the theoretical triggering baseline matrix. Direct missing edge counting matrix and indirect missing edge counting matrix The theoretical triggering reference matrix This indicates the transition under unconditional signal control. The theoretical maximum number of triggers in each module state, the direct missing edge counting matrix. Represents the set of directly missing states. The number of times each module state appears, the indirect missing edge counting matrix Indicates the set of indirectly missing states. The number of times each module state occurs; the expression for the theoretical triggering benchmark matrix is ​​as follows: in, The number of reachable states for each module. For module The A reachable state.

4. The NCES conditional signal reconstruction method according to claim 3, characterized in that, The initial screening process for conditional signals based on the state counting matrix is ​​as follows: definition The effective trigger baseline matrix represents the baseline of the total number of effective triggers after removing indirect missing triggers; the state count matrix is ​​compared for differences in matrix elements to initially screen condition signals: If there exists a module state that satisfies This indicates that all triggers are currently blocked, and the multi-condition signal processing mode is then determined. It is an AND gate, and if there are still transitions in the states of other modules in the current module. If it can be enabled normally, the module's state holds the condition library location of the token; If there exists a module state that satisfies And there exist other modules in the current module whose states satisfy the condition. When, it indicates the module state. The enable function remains unaffected; the module state is as follows. The holding of the treasury in China is a conditional treasury; If all module states of a certain module are... ,in If the value is a constant, it indicates that the current module is an irrelevant module. The connection weights of irrelevant modules are then reset to zero for dimensionality reduction.

5. The NCES conditional signal reconstruction method according to claim 1, characterized in that, The process of building an ILP model based on Big-M is as follows: For different sets of states in the observation reachability graph RG, combined with mode decision variables With logical decision variables Establish three types of hard constraints: For the set of existing states Forced change Conditional enable: ， ； For the set of directly missing states Forced change Unconditional enable: ， ； For the set of indirectly missing states Introducing logical decision variables To infer the potential enabling conditions of the transition: in, This is the system state row vector after binarization. Let be a row vector consisting entirely of 1s, where , For collection of warehouses, For the conditional signal matrix Change and Transformation Related column vectors, For larger numbers, usually take .

6. The NCES conditional signal reconstruction method according to claim 1, characterized in that, The introduction of slack variables The ILP model is as follows: For the set of existing states The elements in the text are: ， ； For the set of directly missing states The elements in the text are: ， ； For the set of indirectly missing states The elements in the text are: 。 7. The NCES conditional signal reconstruction method according to claim 1, characterized in that, The objective function for minimizing the error is: in, Used to minimize the degree of deviation from observed data, i.e., error correction; Used to find the simplest structure for connecting lines according to Occam's razor. For slack variables The weighting coefficients, For change Hekushu The corresponding conditional signal.

8. A conditional signal reconstruction system for NCES based on reachability graph difference analysis, characterized in that, include: Data acquisition and reachability graph generation module: Models the system under investigation as NCES, acquires discrete event observation data of the system under investigation, and generates the observation reachability graph RG and the theoretical reachability graph RG0; State set partitioning module: Performs topological comparison between the observed reachability graph RG and the theoretical reachability graph RG0 of the unconditional signal, extracts the difference reachability graph, and partitions the state set, including the existential state set. The set of directly missing states Indirect missing state set ; The state counting initial screening module includes a matrix generation unit and a difference analysis unit. Based on the state set, the matrix generation unit generates a state counting matrix; the difference analysis unit compares the element differences of the state counting matrix to perform initial screening of conditional signals and solves a set of feasible solutions for the conditional signals. LIP Model Building Module: Constructs an ILP model based on Big-M and solves all feasible solutions for the conditional signals after initial screening; LIP Model Optimization Module: Introducing Slack Variables into the ILP Model We construct a minimum error objective function and solve for the optimal solution of the conditional signal.

9. A conditional signal reconstruction device for NCES based on reachability graph difference analysis, characterized in that, include: Memory: A computer program for reconstructing NCES conditional signals based on reachability map difference analysis as described in any one of claims 1-7, and is a computer-readable device; Processor: Used to implement the NCES conditional signal reconstruction method based on reachability graph difference analysis as described in any one of claims 1-7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the NCES conditional signal reconstruction method based on reachability graph difference analysis as described in any one of claims 1-7.

Images