A Physics Engine-Based Method for Automatic Generation and Verification of Device Disassembly Sequence

Abstract

This invention discloses an automatic generation and verification method for equipment disassembly sequence based on a physics engine, belonging to the field of equipment disassembly technology. This method acquires a 3D model of the equipment and initializes a physics engine simulation environment, loading component physical properties and dynamic connection constraints. Within the physics engine simulation environment, the feasibility of disassembly operations is iteratively evaluated, accurately capturing physical behaviors such as gravity and collisions, and identifying disassembly obstacles and preset operations to generate candidate operations. A path search algorithm is used to intelligently explore and generate creative disassembly sequences, overcoming the limitations of traditional manual experience and static methods. Through a multi-dimensional nonlinear total cost function, factors such as the number of disassembly steps, collision risk, tool replacement, and accessibility are comprehensively quantified to verify and select the optimal solution that balances efficiency, safety, and operational convenience.

Description

Technical Field

[0001] This invention relates to the field of equipment disassembly, and in particular to a method for automatically generating and verifying equipment disassembly sequence based on a physics engine. Background Technology

[0002] Equipment disassembly plays a crucial role in various stages of product design, manufacturing, maintenance, recycling, end-of-life disposal, reverse engineering, and quality analysis. Accurately and efficiently determining the disassembly sequence is of profound significance for improving operational efficiency, reducing safety risks, minimizing component damage, optimizing resource recycling processes, and guiding product design. Traditionally, determining the disassembly sequence relies primarily on human experience, engineers' expertise, or trial-and-error during actual operation. However, this experience-based approach has significant limitations. For new equipment with complex structures, numerous components, diverse connections, or that has never been disassembled before, relying solely on human experience makes it difficult to quickly and accurately identify the optimal disassembly path. The entire process is time-consuming and labor-intensive, easily limited by the operator's individual judgment, leading to low disassembly efficiency, high potential operational risks, and even potential component damage or safety accidents.

[0003] While some existing technologies employ geometric interference checking tools based on computer-aided design models or simple graph theory algorithms to assist in planning disassembly sequences, these methods typically focus on detecting static geometric conflicts and fail to adequately consider the dynamic physical behavior during disassembly. For example, they struggle to accurately simulate gravity, friction, inertia, and the actual trajectories of components under stress; they also cannot effectively assess the severity of potential collisions during disassembly or their impact on subsequent operations, as these physical effects are not simply geometric superpositions. Furthermore, these methods often lack the ability to comprehensively evaluate multiple dimensions such as disassembly efficiency, tool change frequency, and operational accessibility. They also struggle to generate creative, non-preset disassembly sequences using intelligent search algorithms under complex physical constraints and to select the truly optimal disassembly solution from among them. Therefore, existing technologies fall short in meeting the demands for high-precision, high-efficiency, and low-risk automated disassembly sequence planning, and cannot satisfy the higher requirements of Industry 4.0 for intelligent, efficient, and safe production. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, the present invention aims to provide a method for automatically generating and verifying device disassembly sequence based on a physics engine, comprising the following steps: Step S1 involves preparing the model and initializing the physics engine environment for the equipment to be disassembled, including acquiring the 3D model of the equipment and identifying the geometric information, material information, and initial connection constraint information of the components; constructing the physics engine simulation environment, importing the components as physical objects, setting physical properties, and loading releasable physical constraints.

[0005] Step S2 involves performing an iterative feasibility assessment and candidate disassembly operation generation within the physics engine simulation environment. This includes evaluating the accessibility of each disassembly operation through the physics engine simulation, identifying disassembly obstacles and pre-set operation components, and determining constraint removal operations to generate a series of currently feasible candidate disassembly operations.

[0006] Step S3: Based on the dynamic simulation of the physics engine, explore and generate a disassembly sequence, including using a path exploration algorithm to iteratively select and execute candidate disassembly operations from the initial disassembly state, update the physics engine state after each operation and generate new candidate operations in real time until the disassembly is completed, so as to gradually generate one or more disassembly sequences.

[0007] Step S4: Perform multi-dimensional optimality evaluation and verification on the generated disassembly sequence, including calculating the multi-dimensional total cost value of the disassembly sequence based on the physics engine simulation data, and selecting the disassembly sequence with the lowest total cost value as the optimal disassembly sequence; output the optimized optimal disassembly sequence.

[0008] Furthermore, the step of preparing the model and initializing the physics engine environment for the device to be disassembled includes: Step S101: Obtain a complete 3D model of the device to be disassembled. The 3D model includes the geometric information, material information, and initial connection constraint information of each component of the device. Based on the 3D model, each component is independently identified, and a hierarchical relationship between the components is constructed.

[0009] Step S102: Generate collision information corresponding to the component based on the geometric information of the component. The collision information is used for collision detection in the physics engine. And set corresponding physical properties for the component based on the material information and geometric information of the component. The physical properties are parameters and characteristics that can describe the physical object required for simulation in the physics engine.

[0010] Step S103: Initialize the physics engine simulation environment, import each component in the device's 3D model as an independent physical object into the physics engine; and based on the initial connection constraint information, identify and extract resolvable connection constraints, convert them into constraint types recognizable by the physics engine, and load the resolvable connection constraints as initial physical constraints between the physical objects.

[0011] Furthermore, the feasibility assessment and candidate disassembly operation generation for iterative disassembly operations include: Step S201: Analyze the spatial distribution and connection constraint status of the device components in the current physical engine simulation state, and identify all components that are currently not occluded and can be accessed by tools or humans, and can be used as potential disassembly targets.

[0012] Step S202: For each potential disassembly target component, a trial disassembly operation is simulated by applying virtual forces and displacements to it through the physics engine to detect whether there are collisions or interferences with other undisassembled components on its disassembly path, thereby assessing its disassembly accessibility and identifying disassembly obstacles. If disassembly obstacles exist, the component causing the disassembly obstacle is marked as a preset operation component for the current potential disassembly target component. The preset operation component must be disassembled before the current potential disassembly target component. If the potential disassembly target component has the initial physical constraint, the preset operation required to release the initial physical constraint is identified and marked as the constraint release operation for the current potential disassembly target component. The constraint release operation is a physical operation that invalidates the initial physical constraint, and its type is selected according to the nature of the connection constraint.

[0013] Step S203: Based on the disassembly accessibility assessment results, preset operation components, and constraint release operations, generate a series of currently feasible candidate disassembly operations.

[0014] Furthermore, the dynamic simulation based on the physics engine, executing the path exploration algorithm, exploring and generating the disassembly sequence includes: Step S301: Define an initial disassembly state in which the device is in a fully assembled state as the starting point for the path exploration.

[0015] Step S302: Starting from the root dismantling state node corresponding to the initial dismantling state, the search tree is traversed through a node selection strategy that balances the exploration degree of the dismantling path with the utilization degree of the current optimal dismantling path, and a dismantling state node to be expanded is selected.

[0016] Step S303: Expand the disassembly state node to be expanded. Based on the disassembly state in the current physics engine simulation environment and the feasibility assessment of the disassembly operation, generate one or more feasible disassembly operations that have not yet been explored under the node to be expanded, and create a corresponding new sub-disassembly state node for each feasible disassembly operation.

[0017] Step S304: Starting from the new sub-disassembly state node, in the physical engine simulation environment, based on the dynamic physical characteristics of the physical engine, a Monte Carlo simulation is performed to randomly select subsequent feasible disassembly operations to complete a complete disassembly sequence until all components of the device are successfully disassembled or the preset disassembly target is achieved; and for the simulated disassembly sequence, the corresponding total cost value is calculated as the negative cost evaluation value of this simulation through a preset cost function and weighting factor.

[0018] Step S305: Based on the simulated total cost value, update the statistical information of all dismantling state nodes on the path from the new sub-dismantling state node to the root node in the search tree in reverse. The statistical information includes the number of visits to the node and the cumulative cost assessment value.

[0019] Step S306: Repeat the selection, expansion, simulation execution, and backpropagation process until the preset search depth or computational resource limit is reached; finally, based on the statistical information of all sub-disassembly state nodes under the root node of the search tree, identify the disassembly sequence with the lowest cumulative cost evaluation value as the optimal disassembly sequence.

[0020] Furthermore, the multi-dimensional optimality evaluation and verification of the generated decomposition sequence includes: Step S401: For each of the generated disassembly sequences, calculate its corresponding total cost value based on the simulation data of the physics engine.

[0021] Step S402: Compare the total cost values ​​of all generated disassembly sequences, and select the disassembly sequence with the smallest total cost value as the optimal disassembly sequence.

[0022] Furthermore, the formula for calculating the total cost value is as follows: ,in, This represents the total cost value corresponding to the disassembly sequence; Indicates a specific disassembly sequence; Represents the disassembly sequence The corresponding number The original values ​​of the evaluation indicators include the total number of disassembly operations, the number of potential collisions, the cumulative value of collision severity, the number of tool replacements, and the cumulative value of accessibility assessment. Indicates the first The cost function of each evaluation indicator, the cost function will convert the original value The mapping is a non-linear cost branch, which applies heavier penalties and specific mappings to different evaluation metrics. Indicates the first The preset weighting factors corresponding to each evaluation indicator are used to adjust the relative importance of each indicator in the total cost; This indicates the total number of the evaluation indicators.

[0023] Furthermore, the optimized disassembly sequence is presented in a readable format, which provides textual descriptions of the disassembly steps, required tools, estimated time consumption, and dependencies between steps. It also features 3D visualization, dynamic simulation, multi-angle viewing, anomaly alerts, user interaction control, component highlighting, and operation feedback.

[0024] Furthermore, the initial connection constraint information in the three-dimensional model of the device indicates the structural connection relationship between the components. The structural connection relationship includes bolted connections achieved by removing fasteners; welding and bonding achieved by local destruction; snap-fit ​​connections and pin connections separated by specific operations; the initial connection constraint information includes sliding fits and rotational fits that allow relative movement of the components.

[0025] Furthermore, the physics engine can recognize constraint types used to accurately simulate the physical connection relationships and relative motion degrees of freedom between the components. The constraint types include fixed joints that restrict the physical object to a completely fixed relative position and posture; sliding joints that allow the physical object to translate along a single axis; and hinge joints that allow the physical object to rotate around a single axis. The constraint types are derived from releasable connection constraints.

[0026] Furthermore, during the process of iteratively assessing the feasibility of disassembly operations and generating candidate disassembly operations, after each disassembly operation is executed, the spatial relationships and connection status between components are updated in real time based on the dynamic simulation of the physics engine, and the disassembly accessibility and disassembly obstacles are recalculated.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention introduces a physics engine to dynamically simulate the 3D model of equipment, accurately capturing the real physical behavior of components during disassembly, such as gravity, collision, friction, and inertia. Based on the current physical state, iteratively evaluates the feasibility of the disassembly operation and identifies disassembly obstacles and pre-defined operational components. This overcomes the limitations of existing technologies that rely solely on human experience or static geometric checks, making the generated disassembly sequence more practical and safer, and significantly improving the disassembly efficiency and planning accuracy of complex equipment.

[0028] This invention utilizes a path exploration algorithm combined with dynamic simulation from a physics engine to intelligently explore and discover creative, non-preset optimal dismantling paths. Simultaneously, by introducing a multi-dimensional, non-linear total cost function to comprehensively quantify and evaluate the dismantling sequence, this invention can comprehensively consider factors such as the total number of dismantling operations, the cumulative value of potential collisions, the number of tool changes, and the cumulative value of reachability assessments. This allows for the selection of a truly optimal dismantling solution that balances efficiency, safety, and operational convenience, providing more reliable and efficient decision support for intelligent manufacturing, maintenance, and recycling. Attached Figure Description

[0029] Figure 1 This is an exemplary flowchart of the disassembly sequence generation and verification method of the present invention.

[0030] Figure 2 This is an exemplary flowchart of the steps for initializing the engine environment according to the present invention.

[0031] Figure 3 This is an exemplary flowchart of a step-by-step feasibility assessment of the disassembly operation of the present invention.

[0032] Figure 4 This is an exemplary flowchart of the path exploration algorithm of the present invention.

[0033] Figure 5 This is an exemplary flowchart of the steps for optimal evaluation and verification of the present invention. Detailed Implementation

[0034] The present invention will be further described below with reference to specific embodiments.

[0035] This application discloses a method for automatically generating and verifying the disassembly sequence of equipment based on a physics engine. It aims to solve the problems of low disassembly efficiency, high risk, and complex operation caused by inter-component dependencies and spatial interference during the disassembly of complex equipment. This method, through 3D model preparation, physics engine simulation, feasibility assessment of disassembly operations, intelligent path exploration and sequence generation, and multi-dimensional cost optimization, can automatically generate and verify the optimal disassembly sequence, significantly improving the safety, efficiency, and automation level of disassembly operations.

[0036] like Figure 1 The diagram shown is an exemplary flowchart of the automatic generation and verification method based on disassembly sequence in this embodiment, including the following steps: Step S1 involves preparing the model and initializing the physics engine environment for the equipment to be disassembled, including acquiring the 3D model of the equipment and identifying the geometric information, material information, and initial connection constraint information of the components; constructing the physics engine simulation environment, importing the components as physical objects, setting physical properties, and loading releasable physical constraints.

[0037] like Figure 2 The diagram shown is an exemplary flowchart of the steps involved in preparing the model and initializing the physics engine environment for the device to be disassembled in this embodiment, including: Step S101: Obtain a complete 3D model of the equipment to be disassembled. The 3D model includes the geometric information, material information, and initial connection constraint information of each component of the equipment. Based on the 3D model, each component is independently identified, and the hierarchical relationship between the components is constructed.

[0038] In one embodiment, the complete 3D model of the device to be disassembled typically originates from its CAD files, or from point cloud data obtained through reverse engineering of the physical object via 3D scanning, which is then converted into a usable 3D mesh model. The 3D model of the device should contain precise 3D shape, dimensions, surface normals, and other geometric information of each component, which is crucial for subsequent collision detection and spatial analysis. Simultaneously, the material information of the components, such as density, elastic modulus, coefficient of friction, and strength, will be used by the physics engine to calculate the component's mass, inertia, collision response, and other physical characteristics. Most critically, the initial connection constraint information describes the interconnection methods between the components in the assembled state. This includes structural connections, such as bolted connections achieved by removing fasteners, welded or bonded connections achieved through localized destruction, and snap-fit ​​or pin connections separated by specific operations. Furthermore, the initial connection constraint information also covers the relative motion fits allowed between components, such as sliding fits and rotational fits. This information is a prerequisite for constructing the physics engine simulation environment and accurately simulating the disassembly process.

[0039] Step S102: Generate collider information corresponding to the component based on the component's geometric information. The collider information is used for collision detection in the physics engine. Based on the component's material information and geometric information, set corresponding physical properties for the component. The physical properties are parameters and characteristics that can describe the physical object required for simulation in the physics engine.

[0040] Step S103: Initialize the physics engine simulation environment, import each component in the device's 3D model as an independent physical object into the physics engine; and based on the initial connection constraint information, identify and extract removable connection constraints, convert them into constraint types that the physics engine can recognize, and load the removable connection constraints as the initial physical constraints between physical objects.

[0041] In one embodiment, initializing the physics engine simulation environment includes selecting and configuring a suitable physics engine, and setting global physical parameters such as simulation step size, gravity vector, and friction coefficient. Subsequently, the system imports each component model processed in S101 and S102, along with its collision information and physical properties, as an independent "rigid body" or "physical object" into the physics engine. Each physical object will have an independent motion state. Based on the initial connection constraint information obtained in S101, the system intelligently identifies and extracts connection constraints that can be released during disassembly, such as detachable structures like bolted connections, snap-fit ​​connections, and pin connections. These real-world connections are transformed into "joints" or "constraints" that the physics engine can understand and simulate. These constraint types can include fixed joints that restrict the physical object to a completely fixed relative position and orientation, sliding joints that allow the physical object to translate along a single axis, and hinged joints that allow the physical object to rotate around a single axis. The constraint types that these physics engines can recognize are derived from the properties of removable connection constraints. For example, a bolted connection might appear as a fixed joint in the physics engine, which would be released when the "unscrew the bolt" operation is performed. By loading these removable connection constraints as initial physical constraints, it ensures that the device is in a fully assembled state at the start of the simulation, with components connected by these constraints, and that these constraints can be selectively released during subsequent disassembly operations.

[0042] Step S2 involves conducting an iterative feasibility assessment and candidate disassembly operation generation within a physics engine simulation environment. This includes evaluating the accessibility of each disassembly operation through physics engine simulation, identifying disassembly obstacles and pre-set operation components, and determining constraint removal operations to generate a series of currently feasible candidate disassembly operations.

[0043] like Figure 3 The diagram illustrates the feasibility assessment and candidate disassembly operation generation for iterative disassembly operations in this embodiment, including: Step S201 involves analyzing the spatial distribution and connection constraints of the device components under the current physics engine simulation state, identifying all currently unobstructed components that can be accessed by tools or personnel and serve as potential disassembly targets. In one embodiment, step S201 is an "environmental scan" step before each round of disassembly decision-making. The system acquires the position, orientation, and connection relationships of all components in the physics engine in real time, such as which components have been removed, which bolts have been removed, and which components are in a free state. Subsequently, the system iterates through all components that have not yet been disassembled and filters them based on their current state. Only components that meet the two conditions of being "unobstructed" and "accessible by tools or personnel" are identified as current "potential disassembly target components" and become the objects of the next evaluation step.

[0044] Step S202: For each potential disassembly target component, a physics engine is used to simulate a trial disassembly operation applying virtual forces and displacements to it. This is to detect whether there are collisions or interferences with other undisassembled components on its disassembly path, thereby assessing its disassembly accessibility and identifying disassembly obstacles. If disassembly obstacles exist, the component causing the disassembly obstacle is marked as a preset operation component for the current potential disassembly target component. The preset operation component must be disassembled before the current potential disassembly target component. If the potential disassembly target component has initial physical constraints, the preset operation required to remove the initial physical constraints is identified and marked as the constraint removal operation for the current potential disassembly target component. The constraint removal operation is a physical operation that invalidates the initial physical constraints, and its type is selected according to the nature of the connection constraints. Step S203 generates a series of currently feasible candidate disassembly operations based on the disassembly reachability assessment results, pre-defined operating components, and constraint removal operations. In one embodiment, step S203 summarizes the results of the preceding analysis to form a list of executable actions. The system uses the disassembly operations of components assessed as "reachable" in S202, as well as the disassembly operations of all identified "pre-defined operating components" and "constraint removal operations," as the list of "feasible candidate disassembly operations" in the current state. Each candidate operation contains explicit instructions, such as: "remove component X," "unscrew bolt Z on component Y," or "press the clip on component A." These operations constitute all possible branches of the path exploration algorithm in the current disassembly state, providing action space for subsequent path search.

[0045] Step S3: Based on the dynamic simulation of the physics engine, explore and generate a disassembly sequence. This includes using a path exploration algorithm to iteratively select and execute candidate disassembly operations from the initial disassembly state, updating the physics engine state and generating new candidate operations in real time after each operation, until the disassembly is completed, so as to gradually generate one or more disassembly sequences.

[0046] like Figure 4 The diagram shows the implementation of this embodiment, which uses a physics engine-based dynamic simulation to execute a path exploration algorithm, explore and generate a dismantling sequence, including: Step S301 defines an initial disassembly state where the device is in a fully assembled state as the starting point for path exploration. In one embodiment, the initial disassembly state refers to a scenario where all components of the device are assembled according to the design, and all initial physical constraints are active. This state is the starting point of the physics engine simulation and the root node of the path search algorithm. All subsequent disassembly operations will start from this fully assembled state and gradually move towards the disassembly target.

[0047] Step S302: Starting from the root dismantling state node corresponding to the initial dismantling state, traverse the search tree and select a dismantling state node to be expanded by a node selection strategy that balances the exploration degree of the dismantling path with the utilization degree of the current optimal dismantling path. Step S303 involves expanding the disassembly state node to be expanded. Based on the disassembly state in the current physics engine simulation environment and the feasibility assessment of the disassembly operation, one or more feasible disassembly operations that have not yet been explored under the node to be expanded are generated, and a corresponding new sub-disassembly state node is created for each feasible disassembly operation. In one embodiment, when the system selects a "node to be expanded," it starts from the state represented by that node and tries all the "currently feasible candidate disassembly operations" generated in S203. For each attempted disassembly operation, the system performs a virtual simulation in the physics engine to verify the actual effect and potential problems of the operation. During the simulation, the system re-evaluates whether the operation is truly feasible, such as whether it will cause component collisions, whether constraints can be successfully released, and whether it conforms to physical laws, ensuring that only truly feasible path branches are generated. Only operations evaluated as feasible are adopted, and new "sub-disassembly state nodes" are created in the search tree. These newly generated child nodes represent the new state of the device after performing the operation and become the new starting point for subsequent iterations.

[0048] Step S304: Starting from the new sub-disassembly state node, in the physics engine simulation environment, based on the dynamic physical characteristics of the physics engine, a Monte Carlo simulation is performed to randomly select subsequent feasible disassembly operations to complete a complete disassembly sequence until all components of the device are successfully disassembled or the preset disassembly target is achieved; and for the simulated disassembly sequence, the corresponding total cost value is calculated as the negative cost evaluation value of this simulation through the preset cost function and weighting factor.

[0049] Step S305: Based on the simulated total cost value, update the statistical information of all dismantling state nodes on the path from the new sub-dismantling state node to the root node in the search tree in reverse. The statistical information includes the number of times the node is accessed and the cumulative cost assessment value.

[0050] Step S306: Repeat the selection, expansion, simulation execution, and backpropagation process until the preset search depth or computational resource limit is reached; finally, based on the statistical information of all sub-disassembly state nodes under the root node of the search tree, identify the disassembly sequence with the lowest cumulative cost evaluation value as the optimal disassembly sequence.

[0051] Step S4: Perform multi-dimensional optimality evaluation and verification on the generated disassembly sequence, including calculating the multi-dimensional total cost value of the disassembly sequence based on the physics engine simulation data, and selecting the disassembly sequence with the lowest total cost value as the optimal disassembly sequence; output the optimized optimal disassembly sequence.

[0052] like Figure 5 The following is an example of multi-dimensional optimality evaluation and verification of the generated disassembly sequence in this embodiment, including: In step S401, for each generated disassembly sequence, the system calculates its corresponding total cost value based on the simulation data from the physics engine. The system then performs a high-precision simulation of each candidate disassembly sequence in the physics engine again. This simulation is typically more detailed than the Monte Carlo simulation in S304, recording more comprehensive data, such as the actual time of each operation, the precise location and force of collisions, the degree of tool wear, and refined values ​​of energy consumption. Based on this detailed simulation data, and using preset multi-dimensional cost functions and weighting factors, the system calculates the "total cost value" for each disassembly sequence.

[0053] Step S402: Compare the total cost values ​​of all generated disassembly sequences and select the disassembly sequence with the smallest total cost value as the optimal disassembly sequence.

[0054] The formula for calculating the total cost is: ,in, This represents the total cost value corresponding to the disassembly sequence; Indicates a specific disassembly sequence; Represents the disassembly sequence The corresponding number The raw values ​​of the evaluation indicators include the total number of disassembly steps, the number of potential collisions, the cumulative value of collision severity, the number of tool replacements, and the cumulative value of accessibility assessment. Indicates the first The cost function of each evaluation indicator will be mapped to a non-linear cost branch, which will be used to impose heavier penalties or specific mappings on different evaluation indicators. Indicates the first The preset weighting factors corresponding to each evaluation indicator are used to adjust the relative importance of each indicator in the total cost; This indicates the total number of evaluation indicators.

[0055] In one embodiment, this formula quantifies the overall cost of a single decomposition sequence and serves as the basis for the final decision. Wherein, Represents a specific disassembly sequence The overall total cost value is calculated; the smaller this value, the better the decomposition sequence, and it serves as the basis for ultimately selecting the optimal sequence. The formula is derived by... The costs of different evaluation indicators are weighted and summed, where Indicates the first Each evaluation metric has a preset weighting factor, which is configurable and used to adjust the relative importance of each cost within the total cost. For example, in scenarios prioritizing ultimate safety, collision-related metrics will have extremely high weights; in scenarios prioritizing efficiency, the total number of disassembly steps and the number of tool changes will have higher weights. These weights can be configured based on the specific needs of the actual disassembly task, equipment value, operational risks, and other factors, or optimized through expert experience, historical data analysis, or even machine learning methods, to reflect the relative importance of different considerations. Indicates the first The cost function of each evaluation indicator is a non-linear function that converts the original evaluation indicator values... This is mapped to a cost branch. Nonlinear mapping allows for "aggravated penalties" or "specific mappings" for certain metrics. For example, a minor collision might have a low cost, but the cost of multiple minor collisions or a single serious collision can increase exponentially to reflect its significant impact on equipment damage or personnel safety; for tool replacement frequency, each replacement might introduce fixed time or labor costs, therefore... It can be a step function or a linear function; for the cumulative value of the accessibility assessment, the penalty may increase non-linearly for regions that are more difficult to operate on. Represents the disassembly sequence The corresponding number The raw values ​​of each evaluation indicator, which are derived from the decomposed sequence in S401. Data obtained directly or through cumulative calculation during high-precision physics engine simulations includes the total number of disassembly steps, the number of potential collisions, the cumulative value of collision severity, the number of tool changes, and the cumulative value of accessibility assessment. This represents the total number of evaluation indicators. This formula provides a comprehensive, flexible, and highly customizable total cost evaluation mechanism for decomposition sequences by weighted nonlinearly synthesizing multiple key evaluation indicators. This allows for precise quantification of the advantages and disadvantages of different decomposition sequences, guiding the selection of the optimal solution that best meets actual needs.

[0056] The output is the optimized disassembly sequence, presented in a readable format. This readable format provides textual descriptions of the disassembly steps, required tools, estimated time, and dependencies between steps. It also features 3D visualization, dynamic simulation, multi-angle viewing, anomaly alerts, user interaction control, component highlighting, and operation feedback.

[0057] The initial connection constraint information in the equipment's 3D model indicates the structural connection relationships between components. These structural connection relationships include bolted connections achieved by removing fasteners; welding and bonding achieved through localized destruction; and snap-fit ​​and pin connections separated by specific operations. The initial connection constraint information also includes sliding and rotational fits that allow relative movement between components.

[0058] The physics engine can recognize constraint types used to accurately simulate the physical connections and relative degrees of freedom of motion between components. Constraint types include fixed joints that restrict the physical object to a completely fixed relative position and posture; sliding joints that allow the physical object to translate along a single axis; and hinge joints that allow the physical object to rotate around a single axis. Constraint types are derived from releasable connection constraints.

[0059] During the process of conducting iterative disassembly feasibility assessment and generating candidate disassembly operations, after each disassembly operation is executed, the spatial relationships and connection status between components are updated in real time based on the dynamic simulation of the physics engine, and the disassembly accessibility and disassembly obstacles are recalculated.

[0060] Those skilled in the art will understand that the above embodiments are merely exemplary, and various modifications and equivalent substitutions can be made without departing from the spirit and scope of the invention. For example, specific feature point algorithms, optimizer selection, distortion model details, etc., can be adjusted according to actual needs. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for automatically generating and verifying the device disassembly sequence based on a physics engine, characterized in that, Includes the following steps: Step S1: Prepare the model and initialize the physics engine environment for the equipment to be disassembled, including acquiring the 3D model of the equipment and identifying the geometric information, material information, and initial connection constraint information of the components; build the physics engine simulation environment, import the components as physical objects, set physical properties, and load releasable physical constraints; Step S2: In the physics engine simulation environment, perform iterative feasibility assessment of disassembly operations and generate candidate disassembly operations. This includes evaluating the accessibility of each disassembly operation through the physics engine simulation, identifying disassembly obstacles and preset operation components, and determining constraint removal operations to generate a series of currently feasible candidate disassembly operations. Step S3: Based on the dynamic simulation of the physics engine, explore and generate a disassembly sequence, including using a path exploration algorithm to iteratively select and execute candidate disassembly operations from the initial disassembly state, update the physics engine state after each operation and generate new candidate operations in real time until the disassembly is completed, so as to gradually generate one or more disassembly sequences. Step S4: Perform multi-dimensional optimality evaluation and verification on the generated disassembly sequence, including calculating the multi-dimensional total cost value of the disassembly sequence based on the physical engine simulation data, and selecting the disassembly sequence with the lowest total cost value as the optimal disassembly sequence. Output the optimized decomposition sequence; The step of preparing the model and initializing the physics engine environment for the device to be disassembled includes step S103: initializing the physics engine simulation environment, importing each component in the 3D model of the device as an independent physical object into the physics engine; and based on the initial connection constraint information, identifying and extracting removable connection constraints, converting them into constraint types recognizable by the physics engine, and loading the removable connection constraints as initial physical constraints between the physical objects. The feasibility assessment and candidate disassembly operation generation for iterative disassembly operations include: Step S201: Analyze the spatial distribution and connection constraint status of the device components in the current physical engine simulation state, and identify all components that are not currently obscured and can be touched by tools or humans, and can be used as potential disassembly targets. Step S202: For each potential disassembly target component, a trial disassembly operation is simulated by applying virtual forces and displacements to it through the physics engine to detect whether there are collisions or interferences with other undisassembled components on its disassembly path, thereby assessing its disassembly accessibility and identifying disassembly obstacles; if disassembly obstacles exist, the component causing the disassembly obstacle is marked as a preset operation component for the current potential disassembly target component, and the preset operation component must be disassembled before the current potential disassembly target component; if the potential disassembly target component has the initial physical constraint, the preset operation required to release the initial physical constraint is identified and marked as the constraint release operation for the current potential disassembly target component, the constraint release operation is a physical operation that invalidates the initial physical constraint, and its type is selected according to the nature of the connection constraint; Step S203: Based on the disassembly accessibility assessment results, preset operation components, and constraint release operations, generate a series of currently feasible candidate disassembly operations; The dynamic simulation based on the physics engine, executing the path exploration algorithm, exploring and generating the disassembly sequence includes: Step S301: Define an initial disassembly state in which the device is in a fully assembled state as the starting point for the path exploration; Step S302: Starting from the root dismantling state node corresponding to the initial dismantling state, the search tree is traversed through a node selection strategy that balances the exploration degree of the dismantling path with the utilization degree of the current optimal dismantling path, and a dismantling state node to be expanded is selected. Step S303: Expand the disassembly state node to be expanded. Based on the disassembly state in the current physics engine simulation environment and the feasibility assessment of the disassembly operation, generate one or more feasible disassembly operations that have not yet been explored under the state node to be expanded, and create a corresponding new sub-disassembly state node for each feasible disassembly operation. Step S304: Starting from the new sub-disassembly state node, in the physical engine simulation environment, based on the dynamic physical characteristics of the physical engine, a Monte Carlo simulation is performed to randomly select subsequent feasible disassembly operations to complete a complete disassembly sequence until all components of the device are successfully disassembled or the preset disassembly target is achieved; and for the simulated disassembly sequence, the corresponding total cost value is calculated as the negative cost evaluation value of this simulation through a preset cost function and weighting factor. Step S305: Based on the simulated total cost value, update the statistical information of all dismantling state nodes on the path from the new sub-dismantling state node to the root node in the search tree in reverse. The statistical information includes the number of visits to the node and the cumulative cost assessment value. Step S306: Repeat the selection, expansion, simulation execution, and backpropagation process until the preset search depth or computational resource limit is reached; finally, based on the statistical information of all sub-disassembly state nodes under the root node of the search tree, identify the disassembly sequence with the lowest cumulative cost evaluation value as the optimal disassembly sequence. The multi-dimensional optimality evaluation and verification of the generated decomposition sequence includes: Step S401: For each of the generated disassembly sequences, calculate its corresponding total cost value based on the simulation data of the physics engine; Step S402: Compare the total cost values ​​of all generated disassembly sequences, and select the disassembly sequence with the smallest total cost value as the optimal disassembly sequence; The formula for calculating the total cost is as follows: ,in, Represents the disassembly sequence The corresponding total cost value; Indicates a specific disassembly sequence; Indicates the first decomposition sequence corresponding to the The original values ​​of the evaluation indicators include the total number of disassembly operations, the number of potential collisions, the cumulative value of collision severity, the number of tool replacements, and the cumulative value of accessibility assessment. Indicates the first The cost function of each evaluation indicator, the cost function will convert the original value The mapping is a non-linear cost branch, which applies heavier penalties and specific mappings to different evaluation metrics. Indicates the first The preset weighting factors corresponding to each evaluation indicator are used to adjust the relative importance of each indicator in the total cost; This indicates the total number of the evaluation indicators.

2. The method for automatically generating and verifying the device disassembly sequence based on a physics engine according to claim 1, characterized in that: The step of preparing the model and initializing the physics engine environment for the device to be disassembled includes: Step S101: Obtain a complete 3D model of the device to be disassembled. The 3D model includes the geometric information, material information, and initial connection constraint information of each component of the device. Based on the 3D model, each component is independently identified, and a hierarchical relationship between the components is constructed. Step S102: Generate collision information corresponding to the component based on the geometric information of the component. The collision information is used for collision detection in the physics engine. And set corresponding physical properties for the component based on the material information and geometric information of the component. The physical properties are parameters and characteristics that can describe the physical object required for simulation in the physics engine.

3. The method for automatically generating and verifying the device disassembly sequence based on a physics engine according to claim 1, characterized in that: The optimized disassembly sequence is presented in a readable format, which provides textual descriptions of the disassembly steps, required tools, estimated time, and dependencies between steps. It also features 3D visualization, dynamic simulation, multi-angle viewing, anomaly alerts, user interaction control, component highlighting, and operation feedback.

4. The method for automatically generating and verifying the device disassembly sequence based on a physics engine according to claim 1, characterized in that: The initial connection constraint information in the three-dimensional model of the equipment indicates the structural connection relationship between the components. The structural connection relationship includes bolted connections achieved by removing fasteners; welding and bonding achieved by local destruction; snap-fit ​​connections and pin connections separated by specific operations; the initial connection constraint information includes sliding fits and rotational fits that allow relative movement of the components.

5. The method for automatically generating and verifying the device disassembly sequence based on a physics engine according to claim 1, characterized in that: The physics engine can recognize constraint types used to accurately simulate the physical connection relationships and relative motion degrees of freedom between the components. The constraint types include fixed joints that restrict the physical object to a completely fixed relative position and posture; sliding joints that allow the physical object to translate along a single axis; and hinge joints that allow the physical object to rotate around a single axis. The constraint types are derived from releasable connection constraints.

6. The method for automatically generating and verifying the device disassembly sequence based on a physics engine according to claim 1, characterized in that: During the process of iterative disassembly operation feasibility assessment and candidate disassembly operation generation, after each disassembly operation is executed, the spatial relationship and connection status between components are updated in real time based on the dynamic simulation of the physics engine, and the disassembly accessibility and disassembly obstacles are recalculated.

Images