Thermal power plant steam turbine installation VR auxiliary teaching system and method based on logic constraint atlas

Abstract

The invention discloses a thermal power plant steam turbine installation VR auxiliary teaching system and method based on a logic constraint map. In order to solve the problem that existing VR training lacks process logic constraints and is difficult to cultivate standardized operation habits, the method comprises the following steps: abstracting a turbine installation process into a directed acyclic graph (DAG) to construct a logic constraint graph; the system dynamically generates an executable task list through a sequence planning engine, and performs real-time precondition verification and interception on user operation by using a process rule verifier; through semantic modeling of parts, a dynamic assembly sequence verification algorithm and a virtual-real mapping interaction mechanism, operation guidance of strong logic driving is realized. Hidden construction specifications are converted into dominant and executable logic rules, the process error rate and the key component omission rate of students are effectively reduced, and teaching transformation from skill simulation to engineering thinking development is achieved.

Description

Technical Field

[0001] This invention relates to the field of virtual reality (VR) technology and industrial education and training, and in particular to a VR-assisted teaching system and method for the installation of steam turbines in thermal power plants based on logic constraint graphs. Background Technology

[0002] Steam turbines in thermal power plants are the core power equipment that converts the thermal energy of steam into mechanical energy. Their installation quality directly determines the safe and stable operation of the power system and the thermal economy of the unit. Modern large steam turbines (such as 600MW and 1000MW ultra-supercritical units) have precise structures, and their installation involves thousands of parts. There are strict time priority relationships and state dependencies between the processes. For example, the flow clearance must be measured before the main cover is installed, and the bearing housing must be ground and leveled before the bearing plate is in place.

[0003] Traditional steam turbine installation training faces challenges such as difficulty in hands-on training, lack of visibility into internal structures, and difficulty in assessing process specifications. Virtual reality (VR) technology, with its immersive and interactive advantages, offers a new solution for this field. Existing technologies include VR-based steam turbine maintenance or disassembly / assembly assistance systems, but these systems have significant drawbacks: they emphasize roaming displays or simple disassembly / assembly animations, focusing on visual realism while neglecting the core logical constraints of the installation process; they lack mandatory constraints on the order of operations, and free disassembly / assembly can easily lead to trainees developing improper operating habits; implicit process specifications are not made explicit, making it difficult for trainees to understand the causal logic behind the operations; and incorrect operations are only indicated as failed, failing to achieve intelligent attribution and precise guidance.

[0004] Therefore, there is an urgent need to develop a logic-driven VR-assisted teaching system that transforms process logic into executable constraint rules, thereby enabling the mandatory development of operational standards and the in-depth cultivation of engineering thinking. Summary of the Invention

[0005] The technical problems this invention aims to solve are: existing VR teaching systems lack effective process logic constraint mechanisms, cannot intercept out-of-order operations, and are difficult to cultivate standardized operating habits; implicit process specifications (such as dependencies and micro-sequence of operations) in turbine installation are difficult to make explicit, and students have weak causal logic cognition; feedback on erroneous operations is vague, lacks intelligent traceability capabilities, and teaching guidance is not targeted enough; the lack of tactile feedback in the virtual environment makes it difficult to align parts, affecting the teaching experience and operational accuracy.

[0006] To address the aforementioned problems, this invention provides a VR-assisted teaching system for the installation of steam turbines in thermal power plants based on logic constraint graphs, comprising: The data layer is used to store basic data including a process logic knowledge graph, which is a directed acyclic graph (DAG) that abstracts the turbine installation process. Nodes represent installation operations or components, and directed edges represent strong dependencies between nodes. The support layer provides underlying engine support and hardware interfaces, including the Unity3D engine, XR interaction toolkit, and SteamVR SDK; The logic layer includes a sequence planning engine and a process rule validator. The sequence planning engine performs real-time topological sorting on the directed acyclic graph to generate a list of currently executable tasks. The process rule validator responds to user interaction requests for target components and determines whether the interaction is allowed based on the list of executable tasks and the status of the predecessor nodes of the target node in the directed acyclic graph. If not allowed, an error message containing information about incomplete predecessor nodes is generated. The presentation layer is used to render the visual state of components in the VR scene based on the determination result of the logic layer, and to display operation feedback and error prompts to the user.

[0007] Furthermore, the sequence planning engine is also used to dynamically prune subsequent branches of a user operation path and issue a warning when it detects that the user operation path may lead to a logical deadlock.

[0008] Furthermore, the logic layer also includes a component state manager, which manages multiple micro-process states of a single component based on a finite state machine (FSM). When making a judgment, the process rule verifier needs to simultaneously verify whether the macro-installation sequence and micro-state transition of the target component are compliant.

[0009] Furthermore, the system also includes a virtual-real mapping interaction module, used for: Render the user's handheld object into a freely movable holographic projection; Set up invisible adsorption areas at the correct installation positions of the components; When the holographic projection enters the adsorption area and the process rule verifier determines that the operation is allowed, the adsorption logic is triggered, and the corresponding three-dimensional model is controlled to automatically move to the pose defined by the adsorption area.

[0010] Furthermore, the logic layer also includes a teaching evaluation module, which records user operation sequences, compares them with a standard process knowledge graph, calculates the logic deviation degree, and generates a comprehensive report that includes error attribution and skills assessment.

[0011] Furthermore, the 3D models of the components stored in the data layer are associated with semantic logical data objects, which at least include the component's unique semantic ID, a list of preceding node IDs, and an installation status identifier; the process logic knowledge graph is constructed from these logical data objects.

[0012] Furthermore, the logical data object defines the logical vector of the component according to formula (1). : (1) Among them, ParentID is the parent component, Preconditions is the set of component IDs that must be installed first; Actions is the sequence of atomic operations required for this component, and Type is the component type.

[0013] Furthermore, the logic for the process rule verifier to perform dynamic assembly sequence verification is as follows: The installation process is modeled as a directed acyclic graph. V is the set of vertices, representing installation steps or components. E is the set of directed edges. This indicates that process u must be completed before process v; When the user attempts to install the target component X At that time, execute: a) Obtain the set of preceding nodes ; b) Traverse and verify the state of the preceding nodes, and calculate the decision function. : ; Where Status(u) indicates whether the installation status of node u is complete; c) If If true, unlock and allow interaction; if false, lock and generate a set of missing nodes. As a notification message.

[0014] On the other hand, the present invention also provides a VR-assisted teaching method for the installation of steam turbines in thermal power plants based on logic constraint graphs, which is applied to the above-mentioned system and includes the following steps: S1: Construct a process logic knowledge graph that transforms the steam turbine installation process into a directed acyclic graph (DAG); S2: Load the 3D model of the component in the VR environment and associate it with the corresponding logical node in the directed acyclic graph; S3: Respond to the user's interactive operation on the target component and trigger dynamic assembly sequence verification: query all the preceding nodes of the target component in the directed acyclic graph and their completion status; S4: Execute the branch operation based on the verification result: If all preceding nodes are in a completed state, the current operation is allowed, and the relevant states are updated. If there are any incomplete preceding nodes, the current operation is intercepted and feedback information is generated, which indicates the specific incomplete preceding node; S5: Based on the user's entire operation record, generate a teaching evaluation report that reflects their logical compliance.

[0015] Furthermore, after the operation is permitted in step S4, the virtual-real mapping adsorption step is further performed: the holographic projection held by the user is matched with the preset installation position adsorption area, and the three-dimensional model is driven to perform smooth and precise automatic positioning; and / or, in step S4, the feedback information is displayed in the form of a tree diagram to show the unfinished task chain traced back from the current operation node.

[0016] The beneficial effects of this invention are: Logic-driven and standardized training: By using a directed acyclic graph (DAG) to transform textual specifications into computable logical constraints, trainees are forced to follow standard operating procedures. This fundamentally solves the problems of arbitrary operation and reversed procedures in VR training, and effectively cultivates trainees' engineering logic thinking and standardized operating habits.

[0017] Real-time verification and intelligent guidance: The system can perform real-time logical verification and precondition review for each operation, which not only prevents erroneous operations, but also clearly tells the reason for the error through intelligent source tracing, realizing in-depth teaching from informing about the error to explaining the reason.

[0018] Combining the virtual and real for a realistic experience: Through an interactive mechanism that maps the virtual and real worlds, the system retains the freedom of VR operation while solving the problem of precise installation and alignment. Furthermore, through features such as accident simulation, it enhances the understanding of the consequences of improper operation.

[0019] The assessment is scientific and based on evidence: the system can record the entire operation flow and compare it with the standard graph, generating a quantitative logical deviation degree and skills assessment report, providing an objective and accurate basis for teaching assessment. Attached Figure Description

[0020] Figure 1 This is the architecture diagram of the turbine installation auxiliary teaching system based on logic constraint graph of the present invention.

[0021] Figure 2 An illustration of error messages in this invention. Detailed Implementation

[0022] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] 1. System Framework Addressing the complex process logic and strong process dependencies inherent in the installation of steam turbines in thermal power plants, this system is designed with a focus on logic-driven interaction and visual scene reconstruction. The system adopts a classic four-layer architecture, from bottom to top: data layer, support layer, logic layer, and presentation layer. These layers communicate through standardized API interfaces. The logic layer, as the core of the system, is responsible for all tasks related to installation sequence, precondition verification, and process specification determination. The system architecture diagram is shown below. Figure 1 As shown.

[0024] (1) Data layer The data layer is the cornerstone of the system, responsible for storing static asset data and dynamic process rules. This layer consists of a 3D component model library, a process logic knowledge graph (DAG), a process parameter database, and a user record operation library. The main components of this layer are: ① 3D Model Library for Components: Stores high-precision models of the turbine body (such as rotor, cylinder, diaphragm) and tools; all models are accompanied by a unique semantic ID, which serves as a key index for the logical layer to identify objects.

[0025] ② Process Logic Knowledge Graph: This is the core data structure of this system. The system abstracts the installation process into a directed acyclic graph (DAG) data structure. The graph contains the following elements: Node: Represents a specific component or atomic operation (such as "platform in place" or "bolts tightened hot").

[0026] Edge: Represents a strong dependency. For example, if node A points to node B, it means "B can only proceed after A is completed".

[0027] Attributes: Each node contains a status flag (not started / in progress / completed) and a mutual exclusion condition (such as "this operation is irreversible").

[0028] ③ Process parameter database: records the process parameters for each installation stage; ④ User operation log: Saves historical data of user operations.

[0029] (2) Support layer The support layer provides the underlying engine support and hardware interfaces required for system operation. The main modules within this layer are: ① Unity3D Engine: Responsible for scene management, graphics rendering, physics simulation (PhysX), and audio processing.

[0030] ② Interaction SDK: Integrates SteamVR SDK or XRInteractionToolkit, responsible for processing tracking data and controller input signals from head-mounted displays such as HTC Vive / Oculus.

[0031] (3) Logic Layer This layer is a crucial part of the system, responsible for transforming static knowledge into dynamic teaching guidance. The functions of each component within this layer are as follows: ① Sequence Planning Engine: Topology sorting module: When the system starts up or its state changes, it performs real-time topology sorting on the DAG graph to generate the current "Executable Task List (ActiveTaskList)".

[0032] Dynamic pruning: When a student selects an incorrect installation path, the engine will calculate the subsequent branches of that path in real time to determine whether it will lead to a deadlock, and thus decide whether to issue a warning or forcibly block it.

[0033] ② Component State Manager (FSMManager): Manages the lifecycle states of individual components. For example, a bearing housing not only has a "position state" (ground / base) but also a "clean state" (dirty / clean). The system uses a finite state machine to force the trainer to perform the "cleaning" action before performing the "installation" action.

[0034] ③ ComplianceValidator: When a user attempts to perform an operation, the validator immediately compares the current operation ID with the "List of Executable Tasks". The response is either Pass or Fail. Pass means the operation is allowed and the DAG graph state is updated. Fail means the operation is blocked, and the incomplete preceding nodes are retrieved from the knowledge graph, generating an error message.

[0035] ④ Teaching Assessment Module: This module is responsible for monitoring the trainees' action stream in real time, comparing it with the Standard Process Graph, calculating the logical deviation, and generating a comprehensive assessment report that includes error attribution, skill radar chart, and operation replay.

[0036] (4) Presentation layer The presentation layer directly faces the user and is responsible for presenting the results of the logic layer's operations in an intuitive visual and auditory format. The functions of each component in this layer are as follows: ① Visual guidance system Provide the following functions Highlight: For components that are logically allowed to be installed, render a green outline shader.

[0037] Ghostly Guidance: When the trainee picks up the part, a semi-transparent ghostly image appears at the correct target installation position, indicating the placement point.

[0038] State Lock: For parts that do not meet the prerequisites, render them in gray or red and disable interaction.

[0039] ② Intelligent UI Feedback: Error blocking pop-up: When a trainee attempts to perform an operation that violates regulations (such as "lifting the rotor without installing the lower bearing shell"), the UI interface will not only display "Operation failed" but will also dynamically generate a reason explanation: "Logical error: The lower bearing shell of bearing #3 is missing. Please install the lower bearing shell first." 2. Construction and Implementation of the Sequence Logic Model

[0040] (1) Semantic modeling of components Based on traditional 3D models, for each component object P Add a logical vector : (1) Among them, ParentID is the parent component (e.g., blades belong to rotors), Preconditions is the set of prerequisite dependencies (a list of component IDs that must be installed first); Actions is the sequence of atomic operations required for this part (e.g., cleaning → hoisting → alignment → fastening).

[0041] (2) Assembly sequence planning algorithm This is the core algorithm of this system. The turbine installation process is modeled as a directed acyclic graph (DAG). (2) V is the set of vertices, representing installation steps or components. E is the set of directed edges. This indicates that process u must be completed before process v.

[0042] Its core algorithm logic is that when a user attempts to grab a component... X During installation, the system performs the following checks: ① Obtain the set of predecessor nodes Query Image G Find all pointersX Node set (3) ②State traversal verification (4) ③ Logical feedback like If true, part X is unlocked, the system allows interaction, and a green highlight guide is displayed; like If the result is false, then part X is locked, interaction is not allowed, and the set of incomplete preceding nodes is calculated. Missing : List on the UI Missing The content serves as a prompt.

[0043] (3) State machine management of process steps The installation of a single component (such as tightening a bolt) often involves microscopic sequences. A finite state machine (FSM) is used for control in this process. S0 (initial state) → Cleaning with detergent → S1 (clean state) S1 → Insert into the mounting hole → S2 (Pre-installed state) S2 → Torque wrench tightening → S3 (Tightening state / Complete) Constraint: If the tightening operation is performed directly in the S0 state, the system will determine it as invalid and prompt "Assembly with oil stains is strictly prohibited".

[0044] 3. System Implementation Framework and Key Functions (1) Development environment and software architecture pattern This system is developed based on the Windows 10 operating system. The core logic is written in C# and implemented using the Unity 2022 LTS engine. The software architecture strictly follows the MVC (Model-View-Controller) design pattern. The Model (data model layer) is responsible for parsing JSON-formatted process diagram data and maintaining the real-time status of components (e.g., Is Installed, IsCleaned). The View (view layer) is responsible for 3D model rendering, highlight shaders, UI panels, and sound feedback in the VR scene. The core of the Controller (controller layer) is the installation manager, which is responsible for responding to user input, calling algorithms to verify permissions, and updating the Model status.

[0045] (2) In-depth implementation of core functions ① Data structure definition of process logic diagram In order to transform the "Construction Specifications" into computer-processable data, the system defines a set of JSON-based serialized data structures, in which each component is not only a 3D model, but also a data object carrying logical attributes.

[0046] Data structure definition (C# pseudocode): public class PartData{ public string partID; / / Unique semantic ID (e.g., "LP_Rotor") public string partName; / / Display name (e.g., "Low-pressure rotor") Public List <string>preNodes; / / List of preceding node IDs (core of logical dependencies) public bool isInstalled; / / Current installation status public string targetZoneID; / / The collision box ID of the target installation area Public string errorMsg; / / Specific message displayed when this dependency is violated } When the system starts, the Graph Loader module reads the external JSON configuration file, deserializes it to generate a logically directed graph (Logical DAG) in memory, which is convenient for real-time querying.

[0047] ② Dynamic Sequence Validation Algorithm This is the core algorithm of the system. When a user attempts to manipulate a component in the VR scene, the system does not perform a physical simulation, but instead triggers a recursive logical check. Algorithm flowchart: Algorithm implementation logic: A. Raycasting: Use Physics.Raycast to obtain the component ID pointed to by the user.

[0048] B. Dependency traversal: The system retrieves the PartData corresponding to this ID.

[0049] Iterate through its list of preNodes.

[0050] C. Status determination: IF: The isInstalled property of all preceding nodes is true: →ReturnTRUE (Allow interaction, activate the adsorption logic).

[0051] ELSE: →ReturnFALSE (Reject interaction).

[0052] → Collect the names of all preceding nodes whose state is false and pass them to the UI system.

[0053] ③ Implementation of an interaction mechanism based on virtual-real mapping To address the alignment difficulties caused by the lack of tactile feedback in the virtual environment for steam turbine installation training, the system implements a virtual-real separation interaction mechanism, defined as follows: Hand: The user holds a holographic projection of the component (GhostObject), which moves 1:1 with the hand and is semi-transparent.

[0054] Target: Set up an invisible SnapZone in the correct installation location (such as above the bearing housing).

[0055] Adsorption Logic: When Ghost enters the Target's range and passes the logic check, Ghost turns green and is highlighted. After releasing the handle trigger, the component automatically interpolates and smoothly moves (Lerp) to the precise target coordinates and rotation angle. The system broadcasts an OnInstallComplete event to update the map status.

[0056] ④ Intelligent Error Tracing and Feedback System Traditional installation simulation systems typically only display an "operation error" message. This system, however, uses a reverse tracing algorithm to pinpoint the source of the error. When verification fails, the system not only intercepts the operation but also generates a "missing task chain." For example: Scenario: The trainee wants to directly attach the cylinder head; Logic: The system detected that the rotor installation process, a prerequisite for rotor installation, was not completed. → Tracing back, it was found that the bearing installation process, a prerequisite for rotor installation, was also not completed. UI Output: A holographic panel pops up in front of the trainee's field of vision, listing the following in a tree diagram: Operation Rejected! Prerequisites Incomplete: Low-pressure rotor hoisting root cause incomplete: #3 bearing lower bearing installation ⑤ Progress persistence and replay system To support instructional assessment, the system utilizes C#'s BinaryFormatter to implement serialized storage of the installation progress. Archiving: The current Graph state vector (i.e., the isInstalled boolean values ​​of all components) is saved as a binary file. Replay: The system records the timestamp and ID of each student's operation, forming an operation log. Teachers can load the logs to recreate the entire installation process from a top-down perspective, allowing analysis of where students frequently make mistakes.

[0057] 4. Typical Examples Taking the installation process of the low-pressure cylinder and rotor of a domestically produced 600MW supercritical steam turbine unit as an example, this process is the most logically complex, involves the most components, and is the riskiest part of the steam turbine installation.

[0058] (1) Construction of logical topology for the example The system first constructs the installation logic of this stage into a directed acyclic graph (DAG) containing 24 key nodes, based on the "Technical Specifications for Construction and Acceptance of Power Projects".

[0059] The key nodes are defined as follows: N_1: Low-pressure cylinder platform positioning and leveling N_2: Lower cylinder assembly and positioning N_3: Diaphragm and ring centering N_4: #3 and #4 bearing housings are in place. N_5: Lower Bearing Shell Installation and Scraping N_6: Rotor journal cleanliness inspection N_7: Rotor Positioning – [Key Milestone] N_8: Flow Passage Clearance Measurement Logical constraint set C: The system's sequence planning engine primarily monitors the following three types of strong constraints: ① Support constraints: N_1→N_2→N_4 (the foundation must precede the upper structure).

[0060] ② Protective constraints: N_5→N_7 (The bearing must be installed first before the rotor can fall; otherwise, the journal will rub directly against the bearing housing, causing it to be scrapped).

[0061] ③ Closed constraint: N_3→N_7 (The internal partition must be installed first before the rotor can be inserted; otherwise, the partition cannot be installed).

[0062] (2) Detailed operation process and system logic response The following records the specific operation steps of the trainees in the VR environment, as well as the real-time judgment logic of the system's "sequence planning engine".

[0063] Phase 1: Installation of foundation and stator components (preparation of prerequisites) Table 1. Preparation of Prerequisites

[0064] Phase Two: Critical Path for Rotor Lifting (Core Conflict Demonstration) This phase demonstrates how the system prevents catastrophic equipment damage.

[0065] Scenario description: The trainee has completed the installation of the cylinder and partition, and the crane has lifted the rotor and is hovering above the cylinder.

[0066] Incorrect operation attempt: The trainee forgot to install the lower bearing (N_5) and directly clicked the "Lower" button to try to put the rotor into the bearing housing.

[0067] System logical operation process: 1. Event Trigger: "Rotor Descending" event detected.

[0068] 2. Logical retrieval: Query the Support_Surface attribute of the rotor object.

[0069] 3. Status verification: Target location: Bearing_Pedestal_3 (Bearing housing #3) Current contents: Null Expected contents: Lower Bearing Shell 4. Judgment result: Current!=Expected→CRITICAL_FAILURE (fatal error).

[0070] System performance: A. Operation Freeze: The rotor is forcibly stopped by the system at a distance of 10cm from the bearing housing and cannot continue to descend, giving an error message, such as... Figure 2 As shown.

[0071] B. Accident Simulation: If the system is set to "Assessment Mode" (no protection mode), descent is allowed, but a metal impact effect is triggered at the moment of contact, the screen shatters, and an "Accident Report: Severe scratches on the journal, equipment scrapped, assessment failed" pops up.

[0072] C. Teaching guidance: In the "teaching mode", the "lower half bearing" and "lubricating oil can" on the tool rack are highlighted, and the prompt "Please install the lower bearing and apply turbine oil first" is displayed.

[0073] Phase 3: Gap Adjustment and Cover Closure (Post-Logic Verification) A. Scenario: The rotor has been safely positioned.

[0074] B. Operation: The trainee picks up the "upper cylinder head" and prepares to close the cover.

[0075] C. Logical verification: The system checks the data records of node N_8 (flow gap measurement).

[0076] If the Clearance_Data in the database is empty (i.e., the student has not performed the measurement), the system will prompt: "Violation: Hidden works have not been inspected. Please use the feeler gauge / plumb line method to measure the blade tip clearance and enter the data." (3) Analysis of example data To verify the effectiveness of this logical constraint mechanism, we conducted a comparative test on two classes (60 students in total): Control group (Group A): Using a VR system that can be freely assembled and disassembled without logical constraints.

[0077] Experimental group (Group B): Using this system (based on DAG logical constraints).

[0078] In the subsequent unannounced assessment, the performance data of the two groups of trainees on the task of "low-voltage rotor installation" are as follows: Table 2 Comparison of Assessment Data

[0079] As the above examples demonstrate, this system goes beyond simply showing how to tighten screws; more importantly, it shows "why you can't do it that way." By transforming the clauses in the construction specifications into logical nodes in a DAG diagram, the system successfully makes implicit process discipline explicit, ensuring that the operational habits developed by trainees in the virtual environment can be directly transferred to high-risk real-world operations. The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.< / string>

Claims

1. A VR-assisted teaching system for the installation of steam turbines in thermal power plants based on logic constraint graphs, characterized in that, include: The data layer is used to store basic data including a process logic knowledge graph, which is a directed acyclic graph that abstracts the turbine installation process. Nodes represent installation operations or components, and directed edges represent strong dependencies between nodes. The support layer provides underlying engine support and hardware interfaces, including the Unity3D engine, XR interaction toolkit, and SteamVRSDK; The logic layer includes a sequence planning engine and a process rule validator. The sequence planning engine is used to perform real-time topological sorting on the directed acyclic graph and generate a list of currently executable tasks. The process rule validator is used to respond to user interaction requests for target parts and to determine whether to allow the interaction operation based on the list of executable tasks and the state of the predecessor nodes of the target node in the directed acyclic graph. If not allowed, an error message containing information about incomplete predecessor nodes will be generated; The presentation layer is used to render the visual state of components in the VR scene based on the determination result of the logic layer, and to display operation feedback and error prompts to the user.

2. The system according to claim 1, characterized in that, The sequence planning engine is also used to dynamically prune subsequent branches of a user operation path and issue a warning when it detects that the path may lead to a logical deadlock.

3. The system according to claim 1, characterized in that, The logic layer also includes a component state manager, which manages multiple micro-process states of a single component based on a finite state machine. When making a judgment, the process rule verifier needs to verify whether the macro-installation sequence and micro-state transition of the target component are compliant.

4. The system according to claim 1, characterized in that, The system also includes a virtual-real mapping interaction module, used for: Render the user's handheld object into a freely movable holographic projection; Set up invisible adsorption areas at the correct installation positions of the components; When the holographic projection enters the adsorption area and the process rule verifier determines that the operation is allowed, the adsorption logic is triggered, and the corresponding three-dimensional model is controlled to automatically move to the pose defined by the adsorption area.

5. The system according to claim 1, characterized in that, The logic layer also includes a teaching assessment module, which records user operation sequences, compares them with a standard process knowledge graph, calculates the logic deviation degree, and generates a comprehensive report that includes error attribution and skills assessment.

6. The system according to claim 1, characterized in that, The 3D models of the components stored in the data layer are associated with semantic logical data objects. The logical data objects contain at least the unique semantic ID of the component, a list of preceding node IDs, and an installation status identifier. The process logic knowledge graph is constructed from the logical data objects.

7. The system according to claim 6, characterized in that, The logical data object defines the logical vector of the component according to formula (1). : （1） Here, ParentID is the parent component, Preconditions is the set of component IDs that must be installed first, Actions is the sequence of atomic operations required for this component, and Type is the component type.

8. The system according to any one of claims 1-7, characterized in that, The logic of the process rule verifier performing dynamic assembly sequence verification is as follows: The installation process is modeled as a directed acyclic graph. V is the set of vertices, representing installation steps or components. E is the set of directed edges. This indicates that process u must be completed before process v; When the user attempts to install the target component X At that time, execute: a) Obtain the set of predecessor nodes ; b) Traverse and verify the states of the preceding nodes, and calculate the decision function. : , where Status(u) indicates whether the installation status of node u is complete; c) If If true, unlock and allow interaction; if false, lock and generate a set of missing nodes. As a notification message.

9. A VR-assisted teaching method for the installation of steam turbines in thermal power plants based on logic constraint graphs, characterized in that, The system applied to any one of claims 1-8 includes the following steps: S1: Construct a process logic knowledge graph that transforms the steam turbine installation process into a directed acyclic graph; S2: Load the 3D model of the component in the VR environment and associate it with the corresponding logical node in the directed acyclic graph; S3: Respond to the user's interactive operation on the target component and trigger dynamic assembly sequence verification: query all the preceding nodes of the target component in the directed acyclic graph and their completion status; S4: Execute the branch operation based on the verification result: If all preceding nodes are in a completed state, the current operation is allowed, and the relevant states are updated. If there are any incomplete preceding nodes, the current operation is intercepted and feedback information is generated, which indicates the specific incomplete preceding node; S5: Based on the user's entire operation record, generate a teaching evaluation report that reflects their logical compliance.

10. The method according to claim 9, characterized in that, After the operation is allowed in step S4, the virtual-real mapping adsorption step is further performed: the holographic projection held by the user is matched with the preset installation position adsorption area, and the three-dimensional model is driven to perform smooth and accurate automatic positioning; and / or, in step S4, the feedback information is displayed in the form of a tree diagram to show the unfinished task chain traced back from the current operation node.

Images