Machine tool adjustment control method fusing digital twin and virtual reality

Abstract

The application discloses a machine tool installation and adjustment control method fusing digital twinning and virtual reality, which realizes the deep fusion of virtual and real in the machine tool installation and adjustment process by constructing virtual-real mapping relationship, establishing a data-driven two-way interaction channel, and introducing an adaptive adjustment and predictive feedback mechanism. A digital twinning model is constructed in a virtual environment, and dynamic synchronization of virtual and physical machine tools is realized in combination with real-time sensing data. Through a VR interaction device, an operator can not only observe the running state of each component of the machine tool in an immersive manner, but also can perform a pre-performance and error analysis of the assembly operation in the virtual environment, and meanwhile, key parameters in the operation process are fed back to the digital twinning model, so that the real-time state of the physical machine tool is compared, and accurate guidance on the installation and adjustment process is realized. Compared with existing installation and adjustment modes, the embodiment can complete virtual verification of the installation and adjustment process before actual operation, potential problems are found in advance, and a large amount of trial and error cost and material waste are avoided.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology, specifically a machine tool assembly and adjustment control method that integrates digital twins and virtual reality. Background Technology

[0002] With the rapid development of intelligent manufacturing and Industry 4.0, machine tools, as core equipment in modern manufacturing, play a decisive role in product quality and production efficiency through their assembly and debugging. Traditional machine tool assembly and debugging methods often rely on manual experience and on-site operation, which is not only time-consuming and labor-intensive, but also prone to problems such as low efficiency, insufficient accuracy, and even safety hazards when encountering complex structures or high-precision requirements. In recent years, with the gradual maturation of new-generation technologies such as information technology, virtual reality (VR), and digital twins, how to deeply apply these technologies to the machine tool assembly and debugging process has become a key research focus that urgently needs to be addressed in the field of intelligent manufacturing.

[0003] For example, Chinese patent CN116301390A discloses a machine tool assembly guidance method. Applying this method to AR glasses standardizes the assembly process, but it struggles to adapt to the rapid fusion of multi-source data under complex working conditions. When the system structure changes or the application scenario shifts, it often requires readjusting the sensor layout and data interfaces, resulting in insufficient versatility, high system maintenance costs, and limiting its widespread application in large-scale equipment manufacturing.

[0004] The emergence of digital twin technology offers a new approach to solving the aforementioned problems. Digital twins can not only construct highly realistic digital models of physical machine tools in virtual space, but also achieve real-time monitoring and dynamic mapping of real machine tools through sensors and data acquisition systems, thereby realizing two-way interaction and dynamic synchronization between the virtual and real worlds. However, the current application of digital twins in machine tool assembly and adjustment is still in the exploratory stage, focusing more on monitoring equipment operating status and predictive maintenance, lacking deep integration applications for assembly and adjustment processes. Furthermore, how to combine digital twins with immersive virtual reality environments to achieve intuitive operation guidance, error detection, and collaborative optimization during the assembly and adjustment process remains a challenge that has not yet been fully resolved. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a machine tool assembly and adjustment control method that integrates digital twins and virtual reality. Through data-driven, predictive feedback and adaptive control mechanisms, it realizes bidirectional linkage and adaptive interaction between different modules, and can achieve real-time synchronization between virtual machine tools and physical machine tools. Combined with AI intelligent algorithms for assisted decision-making and path optimization, it improves the intelligence, visualization and safety of the assembly and adjustment process.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A machine tool assembly and adjustment control method integrating digital twin and virtual reality includes the following steps: Step 1: Multi-source data acquisition and fusion: Multi-source sensors are deployed at key parts of the physical machine tool to collect signals including position, displacement, torque, vibration, and temperature; the collected signals are transmitted to the virtual machine tool digital twin model through a standardized communication protocol to build a real-time mapping between the physical machine tool and the virtual machine tool digital twin model; A cross-modal association matching model for multi-source heterogeneous data is constructed based on support vector machine; principal component analysis and clustering algorithms are used to perform scene division, cluster dimensionality reduction and feature compression on the data to construct a key data feature set; a joint Kalman filter algorithm is used to dynamically fuse and estimate the key data feature set to obtain the machine tool assembly and adjustment status with high precision dynamic perception. Step 2: Intelligent assembly and adjustment interaction based on virtual reality: An immersive virtual reality interactive environment is provided to the operator through a head-mounted display device; in the virtual reality interactive environment, the operator's control commands can be received through gesture recognition and voice recognition; an AI-assisted decision-making module is introduced, which is based on deep learning and hierarchical reinforcement learning algorithms to jointly analyze historical assembly and adjustment data and real-time sensor data to optimize the assembly scheme, including assembly sequence optimization and path planning, and provide assembly and adjustment suggestions to the operator; Step 3: Closed-loop bidirectional transmission and adaptive optimization: Construct a closed-loop control mechanism of "prediction-feedback-reassembly"; feed back the machine tool assembly and adjustment status obtained in Step 1 to the virtual reality interactive environment and the virtual machine tool digital twin model in real time, update the virtual machine tool digital twin model so that it can accurately map the operating status and assembly changes of the physical machine tool, and realize bidirectional transmission and dynamic consistency between the virtual and the real.

[0007] Furthermore, in step one, the multi-source assembly and adjustment data fusion sensing is achieved through a posterior probability estimation model within a probabilistic statistical framework, the expression of which is: in: This is the prior probability; For each data source, there is a likelihood function; It is a collection of multi-source observation data; Indicates the first Observational data from multiple data sources; This represents the actual assembly state.

[0008] Furthermore, in step two, the AI-assisted decision-making module adopts a collaborative framework that integrates multimodal knowledge graphs, domain-specific large language model semantic reasoning, and hierarchical reinforcement learning, and includes: The multimodal knowledge graph integrates 3D CAD structure, part attributes, assembly tolerance information and historical process text, and constructs an assembly semantic topology graph; The domain-specific large language model performs assembly relationship completion and dependency inference; The hierarchical reinforcement learning adopts a two-layer policy network structure: the upper layer network uses an attention-sequence hybrid encoder to predict several candidate assembly sub-blocks; the lower layer network searches for the optimal assembly order within the constrained feasible region through policy gradient and hierarchical adaptive reinforcement learning.

[0009] Furthermore, the constrained feasible region is jointly defined by the assembly feasibility heuristic rule and the fast interference determination algorithm.

[0010] Furthermore, the hierarchical reinforcement learning introduces a state embedding method based on contrastive learning and a dynamic failure case memory pool to penalize and reweight conflict-prone operation sequences.

[0011] Furthermore, during the assembly scheme optimization process, the assembly scheme is evaluated and the optimal scheme is selected based on multi-objective Pareto hierarchical ranking. The domain-wide language model is used to automatically generate an interpretive evaluation report and proof of key constraint satisfaction. The optimization objectives include working hours, energy consumption, number of fixture changes, and risk score.

[0012] Furthermore, in step two, the gesture recognition includes at least a first gesture for grasping virtual objects and a second gesture for emergency stop operation.

[0013] Furthermore, in step three, before updating the virtual machine tool digital twin model, assembly tolerance analysis and simulation verification are performed to determine whether the current assembly state meets the process and accuracy requirements. If yes, the virtual machine tool digital twin model is directly updated; if not, parameter sensitivity identification is performed to identify key parameters affecting the error. The assembly process knowledge base and parts database are combined to generate optimization suggestions for assembly adjustment strategies and interaction parameters, and the assembly adjustment process is adaptively adjusted.

[0014] Furthermore, the adaptive adjustment and assembly process includes dynamically adjusting the interaction sensitivity, feedback rate, and assembly sequence of the virtual reality interactive environment based on real-time feedback.

[0015] The beneficial effects of this invention are as follows: This invention presents a machine tool assembly and adjustment control method integrating digital twins and virtual reality. By deploying multi-source sensors in the core areas of the machine tool, it acquires multi-dimensional signals such as displacement, torque, vibration, and temperature. Combined with a real-time communication mechanism, the physical operating state is seamlessly mapped to a virtual machine tool digital twin model. Unlike traditional methods relying on offline data and human experience, this method continuously senses and proactively analyzes potential deviations during assembly and adjustment. Coupled with artificial intelligence algorithms, the system can not only automatically diagnose tolerance errors but also quickly generate adjustment suggestions based on real-time feedback, thereby achieving adaptive optimization of the assembly and adjustment process. This data-driven real-time iterative mechanism effectively improves assembly and adjustment accuracy and stability, reducing human trial and error and repetitive operations.

[0016] By integrating virtual reality technology, operators can enter an immersive virtual reality interactive environment to observe and control the machine tool's operating status in real time. The system supports multimodal interaction, including gestures and voice, allowing users to issue commands naturally, such as executing emergency stops, invoking AI suggestions, or adjusting assembly schemes, significantly improving operational intuitiveness and response speed. In terms of system design, this invention adopts a four-layered hierarchical architecture, ensuring clear logic and functional decoupling between modules while maintaining high scalability. This architecture not only enhances the system's flexibility and maintainability but also provides technical support for future functional expansion and cross-platform applications.

[0017] In summary, the machine tool assembly and adjustment control method integrating digital twins and virtual reality of the present invention has the following technical effects: (1) Significantly improve assembly and adjustment accuracy and quality: The high-precision assembly and adjustment status obtained based on multi-source data fusion perception (joint Kalman filtering) is far superior to the accuracy of relying on a single sensor or manual judgment; automated tolerance analysis and parameter sensitivity identification can detect micro deviations in real time during the assembly process and accurately locate the root cause of the problem, thereby guiding precise adjustments and ensuring that the final assembly quality meets high standard process requirements.

[0018] (2) Significantly improve assembly efficiency and reduce costs: Immersive VR interaction and AI-assisted decision-making enable operators to quickly understand complex assembly tasks and obtain optimized assembly sequences and paths, avoiding repeated exploration and thinking, and significantly shortening assembly time; the closed-loop two-way transmission mechanism realizes "virtual trial and error, physical execution", and most potential problems and interference risks have been discovered and resolved in the virtual environment, greatly reducing repeated disassembly and assembly in the physical world, saving huge material and time costs.

[0019] (3) Enhanced operational safety and reliability: Operators mainly conduct drills and planning in a virtual environment, avoiding the safety risks that may be caused by directly operating complex heavy machine tools; the interactive functions such as gesture emergency stop provided by the system provide additional safety guarantees for actual operation; data-driven decision-making reduces the reliance on unreliable human experience, making the entire assembly and adjustment process more scientific, reliable, and the results more predictable.

[0020] (4) Realize the intelligence and adaptability of the system: The method of the present invention is no longer a static, predefined program, but a dynamic system that can learn from data and adapt in real time. Whether it is the adaptation to different machine tool models or the response to sudden situations (such as minor errors in parts or environmental changes) during the assembly and adjustment process, the system can flexibly generate coping strategies through its closed-loop feedback and adaptive adjustment mechanism, showing good versatility and scalability. Attached Figure Description

[0021] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a schematic diagram illustrating the principle framework of the machine tool assembly and adjustment control method integrating digital twins and virtual reality as described in this invention. Figure 2 This is a flowchart of the machine tool assembly and adjustment control method integrating digital twin and virtual reality in this embodiment; Figure 3 A flowchart for multi-source data acquisition and fusion; Figure 4 A flowchart for intelligent assembly and adjustment interaction based on virtual reality; Figure 5 The interaction flowchart for the first gesture control; Figure 6 The interaction flowchart for the second gesture control; Figure 7 This is a flowchart for closed-loop bidirectional transfer and adaptive optimization. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0023] This embodiment integrates digital twin and virtual reality in its machine tool assembly and adjustment control method. By constructing a virtual-real mapping relationship, establishing a data-driven two-way interactive channel, and introducing adaptive adjustment and predictive feedback mechanisms, it achieves deep integration of virtual and real worlds during machine tool assembly and adjustment. A digital twin model is constructed in the virtual environment and combined with real-time sensor data to achieve dynamic synchronization between the virtual and physical machine tools. Through VR interactive devices, operators can not only immerse themselves in observing the operating status of various machine tool components, but also perform pre-rehearsals and error analyses of assembly operations in the virtual environment. Simultaneously, key parameters during the operation are fed back to the digital twin model, comparing them with the real-time status of the physical machine tool, thus providing precise guidance for the assembly and adjustment process. Compared with existing assembly and adjustment methods, this embodiment can complete virtual verification of the assembly and adjustment process before actual operation, identifying potential problems in advance and avoiding significant trial-and-error costs and material waste.

[0024] Specifically, such as Figure 1 As shown, this embodiment integrates digital twins and virtual reality in its machine tool assembly and adjustment control method. It uses multi-source information as the foundation, employs a virtual-real fusion approach as the core, utilizes network communication as the central channel, and treats the user as the driver. This achieves mapping and feedback throughout the entire process from physical assembly and adjustment to virtual simulation assembly and adjustment interaction, providing reliable engineering support and structural assurance for the digital management and intelligent manufacturing of complex machine tools.

[0025] Specifically, such as Figure 2 As shown in the figure, the machine tool assembly and adjustment control method integrating digital twin and virtual reality in this embodiment includes the following steps.

[0026] Step 1: Multi-source data acquisition and fusion.

[0027] Multi-source sensors are deployed at key parts of the physical machine tool to collect signals including position, displacement, torque, vibration, and temperature. These signals are transmitted to a virtual machine tool digital twin model via a standardized communication protocol, establishing a real-time mapping between the physical and virtual machine tool digital twin models. A cross-modal correlation matching model for multi-source heterogeneous data is constructed based on support vector machines. Principal component analysis and clustering algorithms are used to perform scene segmentation, cluster dimensionality reduction, and feature compression on the data, constructing a key data feature set. A joint Kalman filter algorithm is employed to dynamically fuse and estimate the key data feature set, obtaining a high-precision dynamically sensed machine tool assembly and adjustment status.

[0028] Specifically, this embodiment addresses the problem of mismatched twin models caused by lag in state acquisition and missing sensors during existing complex assembly and adjustment processes by proposing a multi-source assembly and adjustment data fusion method, such as... Figure 3As shown, a fast communication method for multi-source data in the machine tool assembly and adjustment process is proposed by developing a standardized communication server for data interaction; based on support vector machine theory, a multi-source heterogeneous data association and matching mechanism is studied, combined with principal component analysis and clustering algorithms; a machine tool debugging data feature compression method based on scene division and cluster dimensionality reduction is studied to construct a key data feature set; and a joint Kalman filter algorithm is used to achieve fusion perception of machine tool assembly and adjustment data.

[0029] Establishing a unified semantics and timeline across multi-source data is crucial for achieving deep fusion. Based on Support Vector Machine (SVM) theory, an association matching model is constructed to learn the implicit relationships between different types of data through training samples, enabling automatic matching of cross-modal data. Simultaneously, addressing the redundancy and computational complexity of high-dimensional data, dimensionality reduction is performed using Principal Component Analysis (PCA) and clustering algorithms, retaining key feature components and compressing invalid information to ultimately form a key data feature set usable for modeling and fusion. This process not only improves data processing efficiency but also provides data assurance for real-time debugging and control. In the fusion sensing stage, a joint Kalman filter (UKF) algorithm is introduced to dynamically estimate and fuse multi-source data, resolving inconsistencies caused by measurement errors and time synchronization issues. Thus, by combining prior estimates of the machine tool assembly state with real-time sensor observations, high-precision dynamic sensing of the machine tool's state can be achieved.

[0030] Multi-source data fusion sensing can be described within the framework of probability statistics and optimization theory. In this embodiment, multi-source assembly and adjustment data fusion sensing is implemented through a posterior probability estimation model within the probability statistics framework, assuming the existence of a multi-source observation data set. The corresponding actual assembly state is The fusion process can then be modeled as a posterior probability estimation problem: in: This is the prior probability; For each data source, there is a likelihood function; It is a collection of multi-source observation data; Indicates the first Observational data from multiple data sources; This represents the actual assembly state.

[0031] The optimal state estimation result can be obtained by maximizing the posterior probability (MAP estimation). In practical applications, due to the complex noise characteristics and correlations of various data sources, covariance modeling and weighting strategies need to be introduced to construct a more accurate fusion estimate. Furthermore, when dealing with nonlinear and non-Gaussian noise problems, approximate solutions can be achieved through extended Kalman filtering, particle filtering, or variational inference. Further, within the deep learning framework, features from different modalities can be mapped to a unified representation through a shared latent space, and then attention mechanisms or graph neural networks can be used to capture cross-modal dependencies, achieving end-to-end perceptual modeling.

[0032] Step Two: Intelligent Assembly and Adjustment Interaction Based on Virtual Reality: An immersive virtual reality interactive environment is provided to the operator through a head-mounted display device; in the virtual reality interactive environment, the operator's control commands can be received through gesture recognition and voice recognition; an AI-assisted decision-making module is introduced, which is based on deep learning and hierarchical reinforcement learning algorithms to jointly analyze historical assembly and adjustment data and real-time sensor data to optimize the assembly scheme, including assembly sequence optimization and path planning, and provide assembly and adjustment suggestions to the operator.

[0033] The virtual reality-based intelligent interaction integrates an assembly interface, a head-mounted display device, gesture recognition, voice recognition, and an AI-assisted decision-making module, forming an immersive interactive solution for operators. AI-assisted decision-making is the core of the intelligence. This embodiment proposes a deep learning-based assembly scheme optimization method that jointly analyzes historical assembly data and real-time sensor data to provide suggestions on assembly schemes.

[0034] To address the shortcomings of existing assembly sequence generation methods, such as reliance on expert experience, weak comprehensive optimization capabilities under multiple constraints (process feasibility, assembly accessibility, and interference risk), and lack of interpretability, this paper constructs a collaborative framework of "multimodal knowledge graph + large-scale model semantic reasoning + hierarchical reinforcement learning guided by constraint feasible domain." This embodiment integrates 3D CAD structure, part attributes, assembly tolerance information, and historical process text to generate an assembly semantic topology graph. A domain-specific large-scale language model, fine-tuned by instructions, is used to complete assembly relationships and infer potential assembly sequence dependencies, outputting an interpretable logical chain. A two-layer policy network is employed: the upper layer uses an attention-sequence hybrid encoder to predict several candidate assembly sub-blocks, while the lower layer searches for the optimal assembly order in the action space trimmed from the constraint feasible domain (by rapid feasibility heuristics and rapid interference determination) through policy gradients and hierarchical adaptive reinforcement learning. To improve search efficiency, state embedding based on contrastive learning and a dynamic "failure case memory pool" are introduced to penalize and reweight easily conflicting operation sequences. To address the quality of assembly results, this embodiment introduces time-series cycle time and collision risk simulations with differentiable approximations, and jointly backpropagates simulation feedback and policy gradients to form an online closed-loop self-correction. Finally, candidate solutions are selected through multi-objective Pareto hierarchical ranking (time, energy consumption, number of fixture changes, and risk score), and the large model automatically generates interpretive evaluation reports and proofs of key constraint satisfaction, thereby achieving high-quality, interpretable, and rapidly iterative optimization of assembly solutions. Figure 4 As shown.

[0035] In summary, this embodiment employs a collaborative framework integrating multimodal knowledge graphs, domain-wide language model semantic reasoning, and hierarchical reinforcement learning. This framework includes: a multimodal knowledge graph that integrates 3D CAD structures, part attributes, assembly tolerance information, and historical process text to construct an assembly semantic topology graph; a domain-wide language model for assembly relationship completion and dependency inference; and a hierarchical reinforcement learning system with a two-layer policy network structure: the upper layer uses an attention-sequence hybrid encoder to predict several candidate assembly sub-blocks; and the lower layer searches for the optimal assembly order within the constrained feasible region using policy gradients and hierarchical adaptive reinforcement learning. Furthermore, the constrained feasible region is jointly defined by assembly feasibility heuristic rules and a fast intervention determination algorithm. The hierarchical reinforcement learning introduces a state embedding method based on contrastive learning and a dynamic failure case memory pool to penalize and weight conflict-prone operation sequences. Furthermore, during the assembly scheme optimization process, the assembly scheme is evaluated and the optimal scheme is selected based on multi-objective Pareto hierarchical ranking. The domain-wide language model is used to automatically generate an interpretive evaluation report and proof of key constraint satisfaction. The optimization objectives include working hours, energy consumption, number of fixture changes, and risk score.

[0036] Furthermore, this embodiment introduces immersive visual interaction technology, allowing operators to enter a virtual assembly and adjustment environment through a head-mounted display device and observe the machine tool's operating status in real time. The virtual assembly and adjustment environment also supports gesture and voice interaction, enabling operators to communicate more naturally with the virtual environment. For example... Figure 5 The diagram shows the process of gesture-based interaction. Figure 6 This is a gesture-based emergency stop process. In this embodiment, gesture recognition includes at least a first gesture for grasping virtual objects and a second gesture for emergency stop operation. This embodiment adjusts the viewing angle or invokes the AI ​​module to provide suggestions via voice commands. Immersive interaction not only improves operational efficiency but also enhances the understanding and control of complex assembly scenarios.

[0037] Step 3: Closed-Loop Bidirectional Transmission and Adaptive Optimization: Construct a closed-loop control mechanism of "prediction-feedback-reassembly"; the machine tool assembly status obtained in Step 2 is fed back in real time to the virtual reality interactive environment and the virtual machine tool digital twin model, updating the virtual machine tool digital twin model so that it can accurately map the operating status and assembly changes of the physical machine tool, realizing bidirectional transmission and dynamic consistency between the virtual and real systems. Specifically, in this embodiment, before updating the virtual machine tool digital twin model, assembly tolerance analysis and simulation verification are performed to determine whether the current assembly status meets the process and accuracy requirements: if so, the virtual machine tool digital twin model is directly updated; if not, parameter sensitivity identification is performed to identify key parameters affecting the error, and optimization suggestions for assembly and adjustment strategies and interaction parameters are generated by combining the assembly process knowledge base and the parts database, and the assembly and adjustment process is adaptively adjusted. Specifically, the adaptive adjustment of the assembly and adjustment process includes dynamically adjusting the interaction sensitivity, feedback rate, and assembly sequence of the virtual reality interactive environment based on real-time feedback.

[0038] In this embodiment, by deploying multi-source sensors during the assembly and adjustment process, key operating parameters such as position, displacement, torque, vibration, and temperature are collected in real time and transmitted to the virtual machine bed digital twin model. These parameters are then compared and updated with the dynamic parameters in the virtual machine bed digital twin model, thereby achieving synchronous mapping between the physical entity and the virtual model. Through this mechanism, the virtual reality interactive environment can dynamically reproduce the actual state of the physical assembly and adjustment process, allowing operators to intuitively grasp the assembly and adjustment progress and promptly identify potential deviations within the virtual reality interactive environment.

[0039] like Figure 7As shown, the bidirectional transmission method in the assembly and adjustment process involves timely feedback of the latest parameter information to the operator after dynamic parameters are collected and updated by sensors during the process. Based on the feedback, the operator re-executes the initial assembly plan, ensuring consistency between the assembly and adjustment process and the latest state. Subsequently, assembly tolerance analysis and simulation verification are performed on the updated key parameters to determine whether the current assembly state meets process requirements and accuracy standards. If the verification results show deviations, a sensitivity analysis module is triggered to further identify key parameters affecting errors and generate targeted optimization suggestions based on the assembly process knowledge base and parts database. Simultaneously, the virtual model is updated in real-time during the assembly and adjustment process, ensuring accurate mapping of the physical entity's operating state and assembly changes, thus achieving bidirectional transmission and dynamic consistency between the virtual and physical systems. This mechanism not only guarantees the real-time performance and accuracy of the assembly and adjustment process but also effectively reduces trial-and-error costs and improves operational intuitiveness and assembly reliability through a closed-loop predictive-feedback-reassembly model.

[0040] This embodiment integrates digital twins and virtual reality into a machine tool assembly and adjustment control method. Through the deep integration of digital twins and virtual reality, it achieves dynamic consistency, real-time feedback, and adaptive optimization between the virtual environment and physical objects during the machine tool assembly and adjustment process, and includes the following contents.

[0041] (1) In the data acquisition and virtual-real mapping stage, by embedding multi-source sensors in the physical machine tool, real-time acquisition of multi-physical field signals such as position, displacement, torque, vibration and temperature is realized, and these signals are compared and updated with the dynamic parameters of the virtual model to construct a real-time correspondence between the physical entity and the virtual model.

[0042] (2) Regarding assembly error handling, a mechanism of "automated tolerance analysis and parameter sensitivity identification" is proposed. Through assembly and adjustment tolerance analysis and simulation, deviations in the operation process are detected and judged in real time. When the virtual assembly and adjustment results fail to meet the accuracy or process requirements, the sensitivity analysis mechanism is automatically invoked to identify the key parameters that cause the tolerance, and parameter optimization is performed in combination with the parts database and assembly process knowledge base. At the same time, the AI-driven decision module generates assembly and adjustment suggestions based on the analysis results, including adjusting the assembly sequence of key parts or replanning the process path, thereby realizing dynamic correction of the assembly and adjustment process.

[0043] (3) An innovative mechanism was proposed for the interactive updating of virtual and real systems. When key parameters in the assembly process are optimized and corrected, the results are automatically fed back to the virtual machine tool model and dynamic rendering is completed in real time. Operators can intuitively obtain the latest status through a visual interface or head-mounted display device, realizing consistent updates between the virtual model and the physical machine tool. The isomorphic interaction between virtual and real systems not only improves the intuitiveness of operation, but also provides real-time assistance to operators in complex assembly tasks, enabling complex assembly tasks to be completed in a more intelligent and safer way, thereby significantly reducing human error.

[0044] (4) In the adaptive control phase, the system supports automatic optimization of the interaction method and assembly strategy. Through gesture recognition, voice recognition, and immersive head-mounted display devices, operators can interact with the virtual assembly environment in a natural and intuitive way and trigger operation commands. The system dynamically adjusts the interaction sensitivity, feedback rate, and assembly sequence based on feedback during the actual assembly process, thereby ensuring the continuity of operation and high-precision execution of assembly.

[0045] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A machine tool assembly and adjustment control method integrating digital twin and virtual reality, characterized in that: Includes the following steps: Step 1: Multi-source data acquisition and fusion: Multi-source sensors are deployed at key parts of the physical machine tool to collect signals including position, displacement, torque, vibration, and temperature; the collected signals are transmitted to the virtual machine tool digital twin model through a standardized communication protocol to build a real-time mapping between the physical machine tool and the virtual machine tool digital twin model; A cross-modal association matching model for multi-source heterogeneous data is constructed based on support vector machine; principal component analysis and clustering algorithms are used to perform scene division, cluster dimensionality reduction and feature compression on the data to construct a key data feature set; a joint Kalman filter algorithm is used to dynamically fuse and estimate the key data feature set to obtain the machine tool assembly and adjustment status with high precision dynamic perception. Step 2: Intelligent assembly and adjustment interaction based on virtual reality: An immersive virtual reality interactive environment is provided to the operator through a head-mounted display device; in the virtual reality interactive environment, the operator's control commands can be received through gesture recognition and voice recognition; an AI-assisted decision-making module is introduced, which is based on deep learning and hierarchical reinforcement learning algorithms to jointly analyze historical assembly and adjustment data and real-time sensor data to optimize the assembly scheme, including assembly sequence optimization and path planning, and provide assembly and adjustment suggestions to the operator; Step 3: Closed-loop bidirectional transmission and adaptive optimization: Construct a closed-loop control mechanism of "prediction-feedback-reassembly"; feed back the machine tool assembly and adjustment status obtained in Step 1 to the virtual reality interactive environment and the virtual machine tool digital twin model in real time, update the virtual machine tool digital twin model so that it can accurately map the operating status and assembly changes of the physical machine tool, and realize bidirectional transmission and dynamic consistency between the virtual and the real.

2. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 1, characterized in that: In step one, the multi-source assembly and adjustment data fusion sensing is achieved through a posterior probability estimation model within a probabilistic statistical framework, the expression of which is: in: This is the prior probability; For each data source, there is a likelihood function; It is a collection of multi-source observation data; Indicates the first Observational data from multiple data sources; This represents the actual assembly state.

3. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 1, characterized in that: In step two, the AI-assisted decision-making module adopts a collaborative framework that integrates multimodal knowledge graphs, domain-wide large language model semantic reasoning, and hierarchical reinforcement learning, and includes: The multimodal knowledge graph integrates 3D CAD structure, part attributes, assembly tolerance information and historical process text, and constructs an assembly semantic topology graph; The domain-specific large language model performs assembly relationship completion and dependency inference; The hierarchical reinforcement learning adopts a two-layer policy network structure: the upper layer network uses an attention-sequence hybrid encoder to predict several candidate assembly sub-blocks; the lower layer network searches for the optimal assembly order within the constrained feasible region through policy gradient and hierarchical adaptive reinforcement learning.

4. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 3, characterized in that: The constrained feasible region is jointly defined by the assembly feasibility heuristic rule and the fast interference determination algorithm.

5. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 3, characterized in that: The hierarchical reinforcement learning introduces a state embedding method based on contrastive learning and a dynamic failure case memory pool to penalize and reweight conflict-prone operation sequences.

6. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 3, characterized in that: During the assembly scheme optimization process, the assembly scheme is evaluated and the optimal scheme is selected based on multi-objective Pareto hierarchical ranking. The domain large language model is used to automatically generate an interpretive evaluation report and proof of key constraint satisfaction. The optimization objectives include working hours, energy consumption, number of fixture changes and risk score.

7. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 1, characterized in that: In step two, the gesture recognition includes at least a first gesture for grasping virtual objects and a second gesture for emergency stop operation.

8. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 1, characterized in that: In step three, before updating the virtual machine tool digital twin model, assembly tolerance analysis and simulation verification are performed to determine whether the current assembly state meets the process and accuracy requirements. If yes, the virtual machine tool digital twin model is updated directly. If not, parameter sensitivity identification is performed to identify key parameters that affect the error. The assembly process knowledge base and parts database are combined to generate optimization suggestions for assembly adjustment strategies and interaction parameters, and the assembly adjustment process is adaptively adjusted.

9. The machine tool assembly and adjustment control method integrating digital twin and virtual reality according to claim 8, characterized in that: The adaptive adjustment and assembly process includes dynamically adjusting the interaction sensitivity, feedback rate, and assembly sequence of the virtual reality interactive environment based on real-time feedback.

Images