Method, system, equipment and medium for detecting defects in CNC machine tool machining trajectory planning
By combining traditional trajectory planning algorithms, physical simulation and multimodal large models, efficient and accurate detection and optimization of CNC machine tool trajectory planning defects are achieved, solving the problems of errors and unevenness in traditional methods and improving production efficiency and product quality.
Patent Information
- Application Number
- CN202411636553.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-09-30
- Estimated Expiration
- 2044-11-15
AI Technical Summary
Traditional CNC machine tool trajectory planning and detection methods are prone to errors and non-smoothness problems in complex machining environments, affecting machining accuracy and efficiency.
Combining traditional trajectory planning algorithms, physical simulation environments, and multimodal large models, efficient and accurate trajectory planning defect detection is performed by collecting multimodal data. The Dijkstra algorithm and A* algorithm are used to plan the trajectory, combined with issac sim software for simulation, and cameras and laser scanners are used to collect data. Frame-by-frame detection and optimization are performed through a multimodal CNC machine tool trajectory planning defect detection large model.
It realizes efficient and accurate detection and optimization of CNC machine tool trajectory planning defects, reduces the trial and error costs in the actual processing process, improves detection accuracy and production efficiency, and is suitable for different types of CNC machine tools and processing technologies.
Smart Images
Figure CN119511951B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of numerical control machining, and in particular to a method, system, equipment and medium for detecting defects in machining trajectory planning of numerical control machine tools. Background Art
[0002] CNC machine tools play a vital role in the manufacturing industry. The accuracy and optimization of their machining trajectories directly impact product quality and production efficiency. Traditional trajectory planning and detection methods are prone to errors and defects such as unevenness in complex machining environments, making it difficult to ensure optimal, accurate, and executable trajectories.
[0003] CNC machine tools, through their high-precision control and transmission systems, enable rapid production of complex parts while ensuring consistent product quality. Their automated processing significantly reduces manual labor, improves production efficiency, and reduces costs. Their flexibility and programmability enable them to adapt to changing production needs, satisfying personalized and small-batch production requirements, providing strong support for the modernization and intelligentization of the manufacturing industry.
[0004] The accuracy of the machining trajectory directly impacts part precision and surface quality. Deviations in the trajectory can lead to dimensional issues, increased surface roughness, and other issues, impacting overall product performance. The degree of machining trajectory optimization directly impacts production efficiency. An optimized trajectory reduces tool idle time, improves cutting efficiency, and thus reduces production costs. Furthermore, an optimized trajectory reduces shock and vibration on the machine tool, thereby extending its service life.
[0005] Traditional trajectory planning and detection methods are prone to errors in complex machining environments due to factors such as the precision limitations of the machine tool itself, tool wear, and workpiece deformation. These errors directly affect the accuracy and optimization of the machining trajectory. Traditional trajectory planning and detection methods may also cause uneven trajectories, which are usually caused by imperfect planning algorithms and insufficient precision of detection equipment. Unsmooth trajectories increase the impact and vibration of the machine tool, affecting machining quality and production efficiency. To overcome the limitations of traditional methods, modern CNC machine tools need to adopt more advanced trajectory planning and detection methods to overcome the limitations of traditional methods and improve machining accuracy and production efficiency. Summary of the Invention
[0006] The purpose of the present invention is to address the problems in the above-mentioned prior art and provide a method, system, equipment and medium for detecting defects in CNC machine tool machining trajectory planning. By combining traditional trajectory planning algorithms, physical simulation environments and multimodal large models, efficient and accurate detection and optimization of CNC machine tool machining trajectory planning defects can be achieved.
[0007] In order to achieve the above object, the present invention has the following technical solutions:
[0008] In a first aspect, an embodiment of the present invention provides a method for detecting defects in CNC machine tool machining trajectory planning, comprising:
[0009] Collect CNC machine tool trajectory planning defect data and train a multi-modal CNC machine tool trajectory planning defect detection model;
[0010] Randomly plan the machining trajectory of CNC machine tools and simulate the actual machining process of CNC machine tools in a physical simulation environment;
[0011] Collect multimodal data during the actual machining process of the simulated CNC machine tool, and fuse the CNC machine tool machining trajectory with the collected multimodal data;
[0012] The fused multimodal data is input into the multimodal CNC machine tool trajectory planning defect detection model for frame-by-frame detection to identify CNC machine tool machining trajectory planning defects.
[0013] As a preferred solution, the step of randomly planning the CNC machine tool processing trajectory is implemented using the Dijkstra algorithm and the A* algorithm.
[0014] As a preferred solution, the physical simulation environment is simulated using issac sim software to simulate different processing conditions and physical properties.
[0015] As a preferred solution, after the step of identifying the defects in the CNC machine tool processing trajectory planning, it also includes feedback on the specific location information and defect category of the processing trajectory planning defects, optimizing and adjusting the trajectory based on the feedback, and repeating the above steps until the trajectory meets the optimization requirements.
[0016] As a preferred solution, the step of collecting multimodal data during the actual processing of the simulated CNC machine tool includes, during the simulation process, using multiple sensors to respectively collect simulated visual images and point cloud data, the types of sensors including cameras and laser scanners, and synchronizing and preprocessing the collected visual images and point cloud data to form multimodal data.
[0017] As a preferred solution, the step of randomly planning the CNC machine tool processing trajectory generates a trajectory planning result by a trajectory generator, and sends it to the actuator in the simulation environment for physical simulation; when running to each position of the trajectory, the image and point cloud data of the workpiece are collected, and are encoded by the image encoder and the 3D point cloud encoder respectively to obtain image features and point cloud features; the generated trajectory is first tokenized and mapped into a subword sequence; then the trajectory token is fused with the image features and point cloud features; the fused features are input into a multi-modal CNC machine tool trajectory planning defect detection large model, and the token output by the multi-modal CNC machine tool trajectory planning defect detection large model is output through a trajectory defect classification head to diagnose the trajectory defect: whether there is a defect and the specific type of the defect.
[0018] As a preferred solution, the step of collecting multimodal data during the actual machining process of the simulated CNC machine tool and fusing the CNC machine tool machining trajectory with the collected multimodal data includes: when running to each position or moment of the trajectory, collecting the corresponding workpiece image and workpiece point cloud through a camera and a laser scanner; mapping the discrete pose sequence of the trajectory into a token sequence, and encoding the workpiece image and workpiece point cloud by using an image encoder and a point cloud encoder to obtain respective features;
[0019] The trajectory token and the output of the image encoder are passed through the Q-Former layer to obtain the image features most relevant to the trajectory token; the trajectory token and the output of the point cloud encoder are passed through the 3D Q-Former layer to obtain the point cloud features most relevant to the trajectory token; the image features most relevant to the trajectory token and the point cloud features most relevant to the trajectory token are respectively passed through the fully connected projection layer and spliced together, and provided as input to the multimodal CNC machine tool trajectory planning and defect detection large model.
[0020] As a preferred solution, when the multi-modal CNC machine tool trajectory planning defect detection large model detects the occurrence of a defect, it relies on the trajectory generator to regenerate the trajectory planning result, and avoids the position where the defect has been detected when generating the path; the detection process is repeated until the multi-modal CNC machine tool trajectory planning defect detection large model does not detect the occurrence of a defect after the trajectory execution is completed.
[0021] In a second aspect, an embodiment of the present invention provides a CNC machine tool machining trajectory planning defect detection system, comprising:
[0022] The defect detection large model training module is used to collect CNC machine tool trajectory planning defect data and train to obtain a multi-modal CNC machine tool trajectory planning defect detection large model;
[0023] The physical simulation module is used to randomly plan the machining trajectory of CNC machine tools and simulate the actual machining process of CNC machine tools in a physical simulation environment;
[0024] The multimodal data fusion module is used to collect multimodal data during the actual processing of the simulated CNC machine tool and fuse the CNC machine tool processing trajectory with the collected multimodal data;
[0025] The detection and recognition module is used to input the fused multimodal data into the multimodal CNC machine tool trajectory planning defect detection model for frame-by-frame detection to identify CNC machine tool processing trajectory planning defects.
[0026] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the method for detecting defects in machining trajectory planning of CNC machine tools is implemented.
[0027] Compared with the prior art, the first aspect of the present invention has at least the following beneficial effects:
[0028] By fine-tuning the existing pre-trained multimodal large model, a multimodal CNC machine tool trajectory planning defect detection large model can be obtained. Combined with a physical simulation environment and an automatic feedback mechanism, efficient trajectory optimization is achieved, reducing the trial-and-error cost during the actual processing process. By simulating the actual processing process of the CNC machine tool in a physical simulation environment, collecting multimodal data during the simulated actual processing process of the CNC machine tool, fusing the CNC machine tool processing trajectory with the collected multimodal data, and inputting the fused multimodal data into the multimodal CNC machine tool trajectory planning defect detection large model for frame-by-frame detection, CNC machine tool trajectory planning defects can be identified. The present invention fully utilizes the pre-trained multimodal large model's ability to understand the physical world. The multimodal large model can accurately identify trajectory defects in complex environments, improving detection accuracy. The CNC machine tool trajectory planning defect detection method based on the multimodal large model and physical simulation environment is applicable to different types of CNC machine tools and processing processes, and has good adaptability and scalability. At the same time, the CNC machine tool trajectory planning defect detection method of the present invention has a high degree of automation, reduces manual intervention, and can effectively improve production efficiency and product quality.
[0029] It can be understood that the beneficial effects of the second to third aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0031] Figure 1 Schematic diagram of the architecture of a CNC machine tool machining trajectory planning and defect detection system according to an embodiment of the present invention;
[0032] Figure 2 Schematic diagram of the trajectory planning and physical simulation process of the detection method according to an embodiment of the present invention;
[0033] Figure 3 Schematic diagram of multimodal data acquisition and fusion processing of the detection method according to an embodiment of the present invention;
[0034] Figure 4 Schematic diagram of defect detection results and feedback optimization process of the detection method according to an embodiment of the present invention;
[0035] Figure 5 Schematic diagram of the network structure of a large model for multi-modal CNC machine tool trajectory planning and defect detection according to an embodiment of the present invention. DETAILED DESCRIPTION
[0036] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.
[0037] With the development of computer vision and deep learning technologies, in order to ensure the optimality, accuracy and executability of CNC machine tool trajectories, an embodiment of the present invention proposes a method for detecting defects in CNC machine tool machining trajectory planning. By combining traditional trajectory planning algorithms, physical simulation environments and multimodal large models, efficient and accurate detection and optimization of trajectory planning defects are achieved.
[0038] See also Figures 2 to 4 The method for detecting defects in CNC machine tool machining trajectory planning according to an embodiment of the present invention comprises the following steps:
[0039] S1. Collect CNC machine tool trajectory planning defect data and train a multi-modal CNC machine tool trajectory planning defect detection model;
[0040] S2. Randomly plan CNC machine tool processing trajectories and simulate the actual processing process of CNC machine tools in a physical simulation environment;
[0041] S3. Collecting multimodal data during the actual machining process of the simulated CNC machine tool, and fusing the CNC machine tool machining trajectory with the collected multimodal data;
[0042] S4. Input the fused multimodal data into the multimodal CNC machine tool trajectory planning defect detection model for frame-by-frame detection to identify CNC machine tool machining trajectory planning defects.
[0043] In a possible implementation, step S1 performs LORA (Low-Rank Adaptation) fine-tuning based on the existing pre-trained multimodal large model to obtain a multimodal CNC machine tool trajectory planning defect detection large model.
[0044] This step fully leverages the pre-trained multimodal large model's ability to understand the physical world, thereby ensuring the effectiveness of the trained multimodal CNC machine tool trajectory planning and defect detection large model. LORA (Low-Rank Adaptation) fine-tuning is an efficient fine-tuning method for large pre-trained models. The core concept of LORA fine-tuning is to fine-tune the model's internal parameters using low-rank decomposition techniques to reduce training parameters and GPU memory usage while maintaining high model performance. Specifically, LORA adds a small model branch alongside the original pre-trained model and introduces two low-rank matrices (A and B) to simulate the effect of full parameter fine-tuning. Matrix A is used to reduce the dimensionality of the input data, while matrix B is used to increase the dimensionality of the reduced data. The increased dimensionality is then superimposed with the output of the original pre-trained model to achieve fine-tuning of the model. LORA significantly improves fine-tuning efficiency by reducing training parameters and computational complexity. In resource-constrained environments, LORA can complete the model fine-tuning process more quickly. Because LORA only trains two low-rank matrices, A and B, it requires much less storage space than full parameter fine-tuning. This makes LoRa more suitable for use in edge devices or scenarios with limited resources. Although LoRa reduces training parameters, it retains key information in the pre-trained model through low-rank decomposition technology, thereby maintaining the high performance of the model. Experimental results show that LoRa performs better than or close to full parameter fine-tuning on multiple tasks. LoRa allows users to flexibly adjust the rank (r value) of the low-rank matrices A and B according to different task requirements. By adjusting the r value, users can find the best balance between model performance and computing resources. With the continuous improvement of computing resources, the LoRa fine-tuning method is expected to achieve better performance improvements on larger-scale data sets. In addition, the LoRa fine-tuning method can also be combined with other technologies, such as knowledge distillation, transfer learning, etc., to jointly improve the performance of deep learning models.
[0045] In a possible implementation, the step of randomly planning the machining trajectory of the CNC machine tool in step S2 is implemented using a traditional trajectory planning algorithm such as the Dijkstra algorithm and the A* algorithm to plan the machining trajectory of the CNC machine tool.
[0046] Dijkstra algorithm and A* algorithm are two commonly used path search algorithms. They both find the shortest path in a graph structure, but have different characteristics and application scenarios.
[0047] The Dijkstra algorithm, also known as the Dijkstra algorithm, is a classic graph search algorithm used to solve the shortest path problem in weighted graphs. Starting from a vertex (source), the algorithm uses a greedy strategy, selecting the nearest unvisited vertex to the source for expansion each time, updating the distance values of the adjacent vertices until the path reaches the endpoint or all vertices are traversed. The Dijkstra algorithm is suitable for graphs without negative-weight edges. It uses a distance array (or vector) to record the shortest distance from the source to each vertex. By continuously selecting the vertex with the shortest distance to expand, it gradually determines the shortest path from the source to other vertices.
[0048] The A* algorithm is a heuristic search algorithm that considers both the actual movement cost and the estimated distance cost. The priority of a node is evaluated using the evaluation function f(n) = g(n) + h(n), where g(n) represents the actual movement cost from the starting point to the current node, and h(n) represents the estimated cost from the current node to the target node. The A* algorithm uses a priority queue (usually a minimum heap) to manage the nodes to be processed, selecting the node with the smallest f(n) value for expansion each time. The A* algorithm is applicable to various path search problems, especially when there are obstacles or complex terrain. The evaluation function guides the search direction, achieving high efficiency and accuracy. Under optimal conditions (i.e., when the h(n) function is accurate), the shortest path is guaranteed to be found.
[0049] The Dijkstra algorithm is suitable for scenarios where the graph lacks negatively weighted edges, while the A* algorithm is more suitable for path search problems involving obstacles or complex terrain. The A* algorithm uses an evaluation function to guide the search direction and is generally more efficient and accurate than the Dijkstra algorithm. However, the performance of the A* algorithm also depends on the accuracy of the evaluation function h(n). The Dijkstra algorithm is relatively simple and easy to implement. In contrast, the A* algorithm requires the implementation of an evaluation function and data structures such as a priority queue, making its implementation more complex.
[0050] Furthermore, step S2 simulates the actual machining process of the CNC machine tool in a physical simulation environment. The physical simulation environment is simulated using issac sim software to simulate different machining conditions and physical properties, such as material rigidity, friction, etc.
[0051] Isaac Sim is a robotics simulation toolkit for the NVIDIA Omniverse platform. It provides essential features for building virtual robotic worlds and experiments, providing researchers and practitioners with the tools and workflows necessary to create robust, physically accurate simulations and synthetic datasets. Isaac Sim supports the simulation of diverse sensor data, which can be used with various computer vision techniques. Isaac Sim integrates the latest PhysX and RTX engines for physics and graphics simulation, providing users with a realistic simulation environment. Isaac Sim incorporates the USD (Universal Scene Description) format, developed by Pixar, as a description format for robots and complex scenes. This extensible and open-source format facilitates content creation and exchange between different tools. Isaac Sim communicates freely with ROS and ROS2 systems, significantly facilitating robotics software development. The Isaac Gym feature in Isaac Sim allows users to simultaneously simulate large numbers of identical robots and obtain operational data from these robots for training reinforcement learning controllers.
[0052] In one possible embodiment, after the step of identifying the defects in the CNC machine tool processing trajectory planning described in step S4, it also includes the steps of result feedback and optimization. Specifically, the specific location information and defect category of the processing trajectory planning defects are fed back, and the trajectory is optimized and adjusted based on the feedback. The above steps are repeated until the trajectory meets the optimization requirements.
[0053] In one possible embodiment, the step S3 of collecting multimodal data during the actual processing of the simulated CNC machine tool includes, during the simulation process, using multiple sensors to respectively collect simulated visual images and point cloud data, where the types of sensors include cameras and laser scanners, and synchronizing and preprocessing the collected visual images and point cloud data to form multimodal data; thereafter, fusing the CNC machine tool processing trajectory with the collected multimodal data.
[0054] Multimodal data includes various types of information from different sensors, such as color, texture, and shape features in visual images, and geometric information such as spatial position, shape, and size in point cloud data. These complementary pieces of information together form a comprehensive description of the simulation environment, enhancing both the richness and completeness of the information.
[0055] Preprocessing the collected visual images and point cloud data, such as denoising, filtering, and registration, can eliminate redundancy and erroneous information in the data, improving its quality and accuracy. Simultaneously processing data from different sensors ensures consistency and coordination between the data, further improving its reliability and availability.
[0056] Multimodal data provides important data support for the optimization and improvement of algorithms and models. Through in-depth analysis and mining of multimodal data, we can discover the underlying laws and patterns in the data, thereby guiding the optimization and improvement of models.
[0057] like Figure 3 As shown, in a possible implementation manner, the step S3 of collecting multimodal data during the actual machining process of the simulated CNC machine tool and fusing the CNC machine tool machining trajectory with the collected multimodal data includes:
[0058] When running to each position or moment of the trajectory, the corresponding workpiece image and workpiece point cloud are collected by the camera and laser scanner; the discrete pose sequence of the trajectory is mapped to a token sequence, and the workpiece image and workpiece point cloud are encoded by the image encoder and point cloud encoder to obtain their respective features; the trajectory token and the output of the image encoder are passed through the Q-Former layer to obtain the image features most relevant to the trajectory token; the trajectory token and the output of the point cloud encoder are passed through the 3D Q-Former layer to obtain the point cloud features most relevant to the trajectory token; the image features most relevant to the trajectory token and the point cloud features most relevant to the trajectory token are respectively passed through the fully connected projection layer and spliced together, and provided as input to the multimodal CNC machine tool trajectory planning and defect detection large model.
[0059] In the multimodal data processing flow of the embodiment of the present invention, the discrete pose sequence of the trajectory is mapped to a token sequence, and the features from the image and point cloud encoders are fused using Q-Former (or a similar Transformer structure).
[0060] First, the discrete pose sequence of a trajectory (typically including position and orientation information) is mapped into a series of tokens through quantization, embedding, and other methods. These tokens can be viewed as discrete representations of the trajectory, with each token containing information about the trajectory at a specific point in time or location. Next, an image encoder (such as a convolutional neural network (CNN)) is used to process the input image to obtain an image feature representation. These image features, along with the trajectory tokens, are then fed into a Q-Former layer (or a similar attention mechanism layer). The Q-Former layer calculates attention weights between the trajectory tokens and the image features to identify the image features most relevant to the trajectory tokens. Similarly, a point cloud encoder (such as PointNet or PointConv) is used to process the input point cloud data to obtain a point cloud feature representation. These point cloud features, along with the trajectory tokens, are then fed into a 3D Q-Former layer (or a similar 3D attention mechanism layer). The 3D Q-Former layer calculates 3D attention weights between the trajectory tokens and the point cloud features to identify the point cloud features most relevant to the trajectory tokens. Finally, the image features and point cloud features most relevant to the trajectory token are passed through a fully connected projection layer (also known as a fully connected layer or linear layer) for dimensionality adjustment and feature fusion. These projected features are then concatenated to form a multimodal feature representation that combines image and point cloud information. This multimodal data processing pipeline integrates information from different modalities, improving the system's perception capabilities and decision-making accuracy. Specifically, image features provide rich color and texture information, aiding in object and scene recognition, while point cloud features provide precise spatial location information, helping to understand an object's shape and size. Combining this information allows for the construction of a more accurate and reliable perception model.
[0061] In addition, using Transformer structures such as Q-Former to process multimodal data has the following potential advantages:
[0062] The Transformer structure can process input data in parallel to improve computational efficiency.
[0063] Through the attention mechanism, Transformer can capture the global contextual relationship between input data.
[0064] The Transformer structure is easy to extend and modify and can adapt to different tasks and datasets.
[0065] It is important to note that the specific implementation details of this multimodal data processing process may vary depending on the application scenario and dataset. Therefore, in actual applications, adjustments and optimizations need to be made based on specific circumstances.
[0066] like Figure 4 As shown, in a possible implementation manner, when the multimodal CNC machine tool trajectory planning defect detection large model described in step S4 of the embodiment of the present invention detects the occurrence of a defect, it relies on the trajectory generator to regenerate the trajectory planning result, and avoids the position where the defect has been detected when generating the path; the detection process is repeated until the multimodal CNC machine tool trajectory planning defect detection large model does not detect the occurrence of a defect after the trajectory execution is completed.
[0067] See also Figure 1 Another embodiment of the present invention further provides a CNC machine tool machining trajectory planning defect detection system, comprising:
[0068] The defect detection large model training module is used to collect CNC machine tool trajectory planning defect data and train to obtain a multi-modal CNC machine tool trajectory planning defect detection large model;
[0069] The physical simulation module is used to randomly plan the machining trajectory of CNC machine tools and simulate the actual machining process of CNC machine tools in a physical simulation environment;
[0070] The multimodal data fusion module is used to collect multimodal data during the actual processing of the simulated CNC machine tool and fuse the CNC machine tool processing trajectory with the collected multimodal data;
[0071] The detection and recognition module is used to input the fused multimodal data into the multimodal CNC machine tool trajectory planning defect detection model for frame-by-frame detection to identify CNC machine tool processing trajectory planning defects.
[0072] The network structure of the multi-modal CNC machine tool trajectory planning defect detection model in the embodiment of the present invention is as follows: Figure 5 As shown, the steps of randomly planning the CNC machine tool processing trajectory are as follows: the trajectory generator generates the trajectory planning result, which is sent to the actuator in the simulation environment for physical simulation; when running to each position of the trajectory, the image and point cloud data of the workpiece are collected, and the image encoder and 3D point cloud encoder are used to encode the data to obtain image features and point cloud features respectively; the generated trajectory is first tokenized and mapped into a subword sequence; then the trajectory token is fused with the image features and point cloud features; the fused features are input into the multimodal CNC machine tool trajectory planning defect detection large model, and the token output by the multimodal CNC machine tool trajectory planning defect detection large model is output through the trajectory defect classification head to diagnose the trajectory defect: whether there is a defect and the specific type of the defect.
[0073] The CNC machine tool machining trajectory planning defect detection system according to the embodiment of the present invention mainly performs the following operations:
[0074] 1) Training a large model for multimodal CNC machine tool trajectory planning and defect detection:
[0075] Collect a dataset of CNC machine tool trajectory planning defects in simulation and real environments;
[0076] Based on the existing pre-trained multimodal large model, fine-tuning is performed to obtain a multimodal CNC machine tool trajectory planning defect detection large model;
[0077] 2) Using a multi-modal CNC machine tool trajectory planning defect detection model:
[0078] Trajectory planning: Use traditional trajectory planning algorithms to randomly plan the machining trajectory of CNC machine tools;
[0079] Physical simulation: Import the planned trajectory into the physical simulation environment to simulate the actual machining process of the CNC machine tool;
[0080] Data acquisition: During the simulation process, a variety of sensors are used to collect simulated visual images and point cloud data;
[0081] Multimodal data fusion: The collected images and point cloud data are synchronized and preprocessed to form multimodal data, and the processing trajectory is fused with the collected multimodal data set in the form of images and point clouds at the same time.
[0082] Defect detection: Input the multimodal dataset into the multimodal large model for frame-by-frame detection to identify defects in the trajectory;
[0083] Result feedback and optimization: The detected trajectory defect information is fed back to the trajectory planning module to optimize the trajectory. The above steps are repeated until the trajectory meets the optimization requirements.
[0084] Furthermore, the result feedback and optimization achieves gradual optimization of the trajectory by automatically adjusting the parameters of the trajectory planning algorithm.
[0085] By combining traditional trajectory planning algorithms, physical simulation and multimodal data analysis, the present invention achieves efficient and accurate detection and optimization of processing trajectories, and has significant technical advantages and broad application prospects.
[0086] Another embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the method for detecting defects in machining trajectory planning of a CNC machine tool.
[0087] Another embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the method for detecting defects in machining trajectory planning of CNC machine tools is implemented.
[0088] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device, medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal and software distribution medium that can carry the computer program code. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals. For ease of explanation, the above content only shows the part related to the embodiment of the present invention. For specific technical details not disclosed, please refer to the method part of the embodiment of the present invention. The computer-readable storage medium is non-transitory and can be stored in a storage device formed by various electronic devices, and can implement the execution process recorded in the method of the embodiment of the present invention.
[0089] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.
[0090] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0091] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.
[0092] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.
Claims
1. A method for detecting defects in CNC machine tool machining trajectory planning, characterized in that: include: Collect CNC machine tool trajectory planning defect data and train a multi-modal CNC machine tool trajectory planning defect detection model; Randomly plan the machining trajectory of CNC machine tools and simulate the actual machining process of CNC machine tools in a physical simulation environment; Collect multimodal data during the actual machining process of the simulated CNC machine tool, and fuse the CNC machine tool machining trajectory with the collected multimodal data; The fused multimodal data is input into the multimodal CNC machine tool trajectory planning defect detection model for frame-by-frame detection to identify CNC machine tool machining trajectory planning defects.
2. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 1, characterized in that: The step of randomly planning the machining trajectory of the CNC machine tool is implemented using the Dijkstra algorithm and the A* algorithm.
3. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 1, wherein: The physical simulation environment is simulated using issac sim software to simulate different processing conditions and physical properties.
4. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 1, wherein: After the step of identifying the defects in the CNC machine tool machining trajectory planning, the specific location information and defect category of the machining trajectory planning defects are also fed back, the trajectory is optimized and adjusted based on the feedback, and the above steps are repeated until the trajectory meets the optimization requirements.
5. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 1, wherein: The step of collecting multimodal data during the actual processing of the simulated CNC machine tool includes, during the simulation process, using multiple sensors to respectively collect simulated visual images and point cloud data, the types of sensors including cameras and laser scanners, and synchronizing and preprocessing the collected visual images and point cloud data to form multimodal data.
6. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 5, characterized in that: In the step of randomly planning the machining trajectory of the CNC machine tool, the trajectory generator generates a trajectory planning result, which is sent to the actuator in the simulation environment for physical simulation; when running to each position of the trajectory, the image and point cloud data of the workpiece are collected, and the image encoder and 3D point cloud encoder are used to encode the image features and point cloud features respectively; The generated trajectory is first tokenized and mapped into a subword sequence; then the trajectory token is fused with image features and point cloud features; the fused features are input into the multimodal CNC machine tool trajectory planning defect detection large model. The token output by the multimodal CNC machine tool trajectory planning defect detection large model is output through the trajectory defect classification head to diagnose the trajectory defect: whether there is a defect and the specific type of defect.
7. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 6, wherein: The steps of collecting multimodal data during the actual machining process of the simulated CNC machine tool and fusing the CNC machine tool machining trajectory with the collected multimodal data include: when running to each position or moment of the trajectory, collecting corresponding workpiece images and workpiece point clouds through a camera and a laser scanner; mapping the discrete pose sequence of the trajectory into a token sequence, and encoding the workpiece images and workpiece point clouds by using an image encoder and a point cloud encoder to obtain respective features; The trajectory token and the output of the image encoder are passed through the Q-Former layer to obtain the image features most relevant to the trajectory token; the trajectory token and the output of the point cloud encoder are passed through the 3D Q-Former layer to obtain the point cloud features most relevant to the trajectory token; the image features most relevant to the trajectory token and the point cloud features most relevant to the trajectory token are respectively passed through the fully connected projection layer and spliced together, and provided as input to the multimodal CNC machine tool trajectory planning and defect detection large model.
8. The method for detecting defects in CNC machine tool machining trajectory planning according to claim 6, wherein: When the multi-modal CNC machine tool trajectory planning defect detection large model detects the occurrence of a defect, it relies on the trajectory generator to regenerate the trajectory planning result, and avoids the position where the defect has been detected when generating the path; The detection process is repeated until the multi-modal CNC machine tool trajectory planning defect detection model does not detect the occurrence of defects after the trajectory is executed.
9. A CNC machine tool machining trajectory planning defect detection system, characterized in that: include: The defect detection large model training module is used to collect CNC machine tool trajectory planning defect data and train to obtain a multi-modal CNC machine tool trajectory planning defect detection large model; The physical simulation module is used to randomly plan the machining trajectory of CNC machine tools and simulate the actual machining process of CNC machine tools in a physical simulation environment; The multimodal data fusion module is used to collect multimodal data during the actual processing of the simulated CNC machine tool and fuse the CNC machine tool processing trajectory with the collected multimodal data; The detection and recognition module is used to input the fused multimodal data into the multimodal CNC machine tool trajectory planning defect detection model for frame-by-frame detection to identify CNC machine tool processing trajectory planning defects.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, the method for detecting defects in machining trajectory planning of a CNC machine tool according to any one of claims 1 to 8 is implemented.