Mechanical arm simulation system supporting industrial quality inspection

Abstract

The invention discloses a mechanical arm simulation system supporting industrial quality inspection, which relates to the technical field of industrial automation and comprises a multi-physics field simulation engine module, a digital twin interface module, a defect sample library module and a cooperative control module. According to the method, traditional entity mechanical arm debugging is replaced by rigid body dynamics simulation based on a physical engine, loss and production line shutdown caused by repeated trial and error of equipment are avoided, rapid construction of the digital twin environment of the production line is achieved by means of an efficient CAD model conversion process and a lightweight loading technology, and high-fidelity optical simulation is achieved through a parameterized defect model library and high-fidelity optical simulation. Various common industrial defects and complex optical effects are covered, the modeling bottleneck of a quality inspection scene is broken through, real-time closed-loop interaction of visual inspection and motion control is realized by using thread separation and an efficient communication technology, and the recognition robustness of a visual algorithm to tiny defects is guaranteed; the reliability of algorithm verification is improved through the quantitative output of algorithm performance indexes and the comprehensive test of extreme working conditions by the simulation system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial automation, in particular to a mechanical arm simulation system supporting industrial quality inspection. BACKGROUND

[0002] The mechanical arm simulation system is a comprehensive virtual platform integrating computer graphics, physical simulation, digital twinning and other multidisciplinary technologies. It constructs a virtual space consistent with the real industrial environment in a digital way, and can accurately reproduce the structural characteristics, motion law, dynamic characteristics of the mechanical arm, as well as the physical properties of the workpiece, environmental lighting, equipment layout and other key elements. In the virtual environment, the motion control logic of the mechanical arm, visual detection algorithm, production line operation process, etc. can be simulated, tested and optimized in all directions.

[0003] In the process of intelligent manufacturing transformation, the mechanical arm simulation system has irreplaceable significance. It breaks the dependence on physical equipment and production line for entity debugging, providing a safe, flexible and low-cost test environment for the research, verification and iteration of industrial quality inspection technology. Through virtual simulation, enterprises can complete mechanical arm trajectory planning, defect detection algorithm verification, production line layout optimization, etc. before the production line is formally put into operation, effectively shortening the technology landing cycle, reducing equipment wear and tear and trial and error risks, and is the core technical support for promoting the intelligent and efficient upgrade of industrial quality inspection.

[0004] In the current industrial quality inspection field, the existing technical system still has certain defects. The entity mechanical arm debugging is heavily dependent on physical equipment, and repeated trial and error not only leads to increased equipment wear and tear and shortened service life, but also causes production line downtime, resulting in high debugging costs and long cycle. The traditional simulation system has low modeling efficiency, and it is difficult to quickly convert CAD models into simulation scenes, and it is difficult to efficiently build a production line digital twinning environment. The quality inspection scene modeling capability is weak, and there is a lack of high-fidelity defect sample library, which cannot accurately simulate various industrial defects and complex optical effects, and cannot reproduce the reality of the quality inspection scene. The visual detection algorithm and the mechanical arm motion control lack effective coordination mechanisms, and the independent operation of the two leads to the inability to verify the robustness of the algorithm in the simulation stage. The entity system is prone to false detection due to motion jitter, and the generalization ability of the algorithm under extreme working conditions cannot be tested in advance, which seriously affects the quality inspection reliability. Therefore, it is of great significance to develop a mechanical arm simulation system supporting industrial quality inspection. SUMMARY

[0005] The mechanical arm simulation system supporting industrial quality inspection provided by the present application can realize high-fidelity virtual simulation of an industrial quality inspection scene by constructing a complete architecture of a multi-physics field simulation engine, a digital twin interface, a defect sample library and a collaborative control module, aiming to solve problems such as high cost and long cycle of entity mechanical arm debugging, insufficient modeling capability of traditional simulation systems in quality inspection scenes, and difficulty in verifying visual detection algorithms and mechanical arm movement coordination, and to reduce debugging costs, improve modeling efficiency and ensure algorithm reliability through technical means such as physical level simulation to replace entity debugging, efficient modeling and defect simulation, real-time collaborative control and the like.

[0006] To solve the above technical problems, the present application provides the following technical solutions: a mechanical arm simulation system supporting industrial quality inspection, comprising: a multi-physics field simulation engine module, a digital twin interface module, a defect sample library module and a collaborative control module. The multi-physics field simulation engine module is used to simulate mechanical arm movement, visual imaging and light propagation, and the multi-physics field simulation engine module comprises a rigid body dynamics unit, a visual sensor simulation unit and a ray tracing rendering unit. The digital twin interface module is used to import a CAD model and construct a simulation scene, and the digital twin interface module comprises a CAD model import unit and a scene construction unit. The defect sample library module is used to generate a parameterized defect model and realize defect distribution management, and the defect sample library module comprises a parameterized defect modeling unit and a defect distribution management unit. The collaborative control module is used to realize data interaction and real-time control between visual detection, trajectory planning and the physical engine, and the collaborative control module comprises a data interaction unit and a real-time control unit. The multi-physics field simulation engine module, the digital twin interface module, the defect sample library module and the collaborative control module sequentially interact with data to construct a complete virtual simulation environment for industrial quality inspection.

[0007] Further, the rigid body dynamics unit of the multi-physics field simulation engine module adopts the Bullet physical engine kernel of ammo.js, creates a btRigidBody object for each link of the mechanical arm, establishes a rotating joint constraint for adjacent links through btHingeConstraint, configures a rotating axis vector and a limiting angle parameter, processes collision interaction between a workpiece and a clamp through a btCollisionDispatcher component, and the joint rotation angle calculation satisfies the formula: wherein is the initial angle, is the angle at time t, is the initial angle, is the harmonic amplitude, and is the angular frequency. and phase offset determined by offline fitting of kinematic calibration data of the robot arm joints, is a harmonic order.

[0008] Further, the visual sensor simulation unit of the multi-physics simulation engine module creates a WebGLRenderTarget rendering target based on a THREE.js extended rendering pipeline, sets the type to THREE.FloatType to support HDR imaging, configures the focal length, field of view angle, and depth of field parameters of the industrial camera, and outputs image data containing color and depth information, and the depth of field rendering effect satisfies the formula: wherein is the pixel value after depth of field processing, is the original pixel value, is the pixel depth value, is the focus depth, is the blur coefficient, determined according to the lens aperture parameters of the industrial camera and the simulation scene scale conversion.

[0009] Further, the ray tracing rendering unit of the multi-physics simulation engine module is implemented based on the Raycaster tool of THREE.js, creates a raycaster object and sets the firstHitOnly attribute to true, emits a sampling ray from the center of the virtual camera pixel, performs intersection detection on the objects in the scene and records the intersection information.

[0010] Further, the CAD model import unit of the digital twin interface module performs a conversion process from a STEP / IGES file to a GLTF format, after completing the format conversion through the Blender tool, loads the model using the GLTFLoader loader of THREE.js, and enables the dracoCompression compression function to optimize the model loading efficiency.

[0011] Further, the scene construction unit of the digital twin interface module establishes a conversion matrix of the world coordinate system and the base coordinate system of the robot arm, simulates the environmental lighting using an HDR panoramic lighting map, and realizes the spatial position mapping of each object in the scene through the coordinate conversion matrix.

[0012] Further, the parameterized defect modeling unit of the defect sample library module simulates scratches by perturbing the surface normal using a noise texture, calculates the normal perturbation value using the vec3perturbNormal function, and the function parameters include the position vector pos and the noise scaling coefficient noiseScale, simulates concave defects using the SDF deformation method, and the normal perturbation satisfies the formula: wherein is the perturbed normal, is the original normal vector, is the perturbation coefficient, the perturbation coefficient According to the preset threshold range corresponding to the defect level (slight / moderate / severe), selection, is the noise function gradient.

[0013] Further, the data interaction unit of the cooperative control module constructs a closed-loop data link of visual detection, trajectory planning and physical engine, the visual detection module outputs defect coordinates to the trajectory planning module, the trajectory planning module generates joint angle instructions and sends them to the physical engine, and the physical engine returns collision feedback information to the visual detection module.

[0014] Compared with the prior art, the mechanical arm simulation system supporting industrial quality inspection has the following beneficial effects: The present application replaces the traditional entity mechanical arm debugging by rigid body dynamics simulation based on the physical engine, avoids the loss and production line downtime caused by repeated trial and error of the equipment, reduces the debugging cost, realizes the rapid construction of the production line digital twin environment by means of efficient CAD model conversion process and lightweight loading technology, solves the problem of low modeling efficiency of the traditional simulation system, covers various common industrial defects and complex optical effects through the parameterized defect model library and high-fidelity optical simulation, breaks through the bottleneck of quality inspection scene modeling, realizes the real-time closed-loop interaction of visual detection and motion control by using thread separation and efficient communication technology, guarantees the robustness of the visual algorithm for identifying small defects, eliminates the collaborative barriers of control and detection, quantifies the output of algorithm performance indicators and comprehensively tests the extreme working conditions through the simulation system, provides data support for algorithm iteration, avoids on-site sudden failures, and improves the reliability of algorithm verification.

[0015] Other advantages, objects, and features of the application will be set forth in part in the following specification taken in conjunction with the accompanying drawings, and in part will become apparent to those skilled in the art from a consideration of the following specification and drawings, or can be learned from the practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0017] Figure 1 is a structural schematic diagram of the mechanical arm simulation system supporting industrial quality inspection; Figure 2 is a work flow diagram of the mechanical arm simulation system supporting industrial quality inspection; Figure 3 This is a flowchart of the collaborative control module. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0019] This invention discloses a robotic arm simulation system that supports industrial quality inspection. The system achieves high-fidelity virtual simulation of industrial quality inspection scenarios by constructing a complete architecture including a multiphysics simulation engine, a digital twin interface, a defect sample library, and a collaborative control module. It aims to solve problems such as high debugging costs and long cycles of physical robotic arms, insufficient modeling capabilities of traditional simulation systems for quality inspection scenarios, and difficulties in verifying the collaborative motion of visual inspection algorithms and robotic arms.

[0020] See Figure 1 The system specifically includes a multiphysics simulation engine module, a digital twin interface module, a defect sample library module, and a collaborative control module. Each module interacts with data in sequence to build a complete virtual simulation environment for industrial quality inspection.

[0021] The multiphysics simulation engine module is mainly used to simulate robotic arm motion, visual imaging, and light propagation. It includes a rigid body dynamics unit, a vision sensor simulation unit, and a ray tracing rendering unit. The rigid body dynamics unit uses the Bullet physics engine kernel of ammo.js to create corresponding objects for each link of the robotic arm, establish rotational joint constraints between adjacent links through specific components, and handle collision interactions between the workpiece and the fixture. The vision sensor simulation unit is based on THREE.js to extend the rendering pipeline, create rendering targets to support HDR imaging, configure various parameters of the industrial camera, and output image data containing color and depth information. The ray tracing rendering unit is implemented based on the Raycaster tool of THREE.js. After creating relevant objects, it emits sampling rays from the pixel center of the virtual camera, performs intersection detection on objects in the scene, and records the intersection point information.

[0022] The digital twin interface module is used to import CAD models and build simulation scenes. It includes a CAD model import unit and a scene building unit. The CAD model import unit performs the conversion process from STEP / IGES files to GLTF format. After the format conversion is completed with the help of Blender, the GLTFLoader loader of THREE.js is used to load the model and the compression function is enabled to optimize the model loading efficiency. The scene building unit establishes the transformation matrix between the world coordinate system and the robot arm base coordinate system, uses HDR panoramic light map to simulate ambient lighting, and realizes the spatial position mapping of each object in the scene through the coordinate transformation matrix.

[0023] The defect sample library module is used to generate parametric defect models and implement defect distribution management. It includes a parametric defect modeling unit and a defect distribution management unit. The parametric defect modeling unit simulates scratches by disturbing the surface normal with noise texture, calculates the normal disturbance value using a specific function, and also simulates dent-type defects using the SDF deformation method.

[0024] The collaborative control module is used to realize data interaction and real-time control between visual inspection, trajectory planning and physics engine. It includes a data interaction unit and a real-time control unit. The data interaction unit builds a closed-loop data link between visual inspection, trajectory planning and physics engine. The visual inspection module outputs defect coordinates to the trajectory planning module. The trajectory planning module generates joint angle commands and sends them to the physics engine. The physics engine sends collision feedback information back to the visual inspection module.

[0025] This invention reduces debugging costs, improves modeling efficiency, and ensures algorithm reliability by using physical-level simulation to replace physical debugging, efficient modeling and defect simulation, and real-time collaborative control.

[0026] Example 1

[0027] This embodiment addresses the scenario of detecting surface defects in automotive engine cylinder blocks. It utilizes a robotic arm simulation system supporting industrial quality inspection to construct a virtual inspection environment, enabling automated simulation detection of typical defects such as scratches and dents on the cylinder block surface. In this scenario, traditional physical inspection requires debugging the robotic arm's motion trajectory and optimizing visual inspection algorithm parameters while the production line is shut down. This presents problems such as high equipment wear and tear, long debugging cycles, and high risks associated with testing under extreme conditions. (See [link to previous document]). Figure 1 , Figure 2 and Figure 3 This simulation system can reproduce the physical characteristics, optical environment and defect features of the cylinder production line in a virtual environment, and complete the collaborative verification of the robotic arm motion control and visual inspection algorithms, significantly reducing the cost and technical risks of physical debugging.

[0028] During the system initialization phase, the digital twin interface module first executes the CAD model import process. Operators upload the STEP format 3D model of the car engine cylinder block and the IGES format model of the production line equipment through the CAD model import unit. The system calls the Blender tool to batch convert the above models to GLTF format. During the conversion process, the geometric accuracy and material properties of the model are automatically preserved. After the conversion is completed, the GLTFLoader of THREE.js is used to import the model data. At the same time, the dracoCompression function is enabled to compress the model vertex data to reduce memory usage. The scene construction unit builds a virtual production line scene based on the imported model data. By establishing a transformation matrix between the world coordinate system and the robot arm base coordinate system, the spatial position mapping of objects such as the cylinder block, conveyor belt, and robot arm is realized. To simulate the lighting conditions of the real production environment, the system loads HDR panoramic light maps and simulates the difference in light intensity at different locations through a light attenuation algorithm to ensure the consistency of scene lighting effects with the physical environment.

[0029] The defect sample library module generates parametric defect models based on common defect types in automotive cylinder blocks. The parametric defect modeling unit simulates scratch-type defects by perturbing the surface normals with noise textures. It calls the `vec3perturbNormal` function to calculate the normal perturbation value. In the function parameter settings, the position vector `pos` is determined based on the cylinder block surface coordinate distribution, and the noise scaling factor `noiseScale` is dynamically adjusted according to the thickness of the scratch. For dent-type defects, the SDF deformation method is used for modeling, and the perturbed normals are calculated using the following formula: The disturbance coefficient The threshold range is selected based on the defect level: a smaller threshold range for minor indentations and a larger threshold range for severe indentations, along with the noise function gradient. Generated using the Perlin noise algorithm, the defect distribution management unit automatically generates multiple sets of defect samples on the virtual cylinder surface according to the actual defect probability distribution based on cylinder production defect statistics. These samples cover combinations of scratches and dents of different locations, sizes, and severity, forming a complete defect test set.

[0030] After the multiphysics simulation engine module is started, each unit collaborates to complete the simulation of the robot arm's motion and visual imaging. The rigid body dynamics unit uses the Bullet physics engine kernel of ammo.js to build the robot arm's dynamics model, creating a btRigidBody object for each link of the robot arm, and establishing rotational joint constraints for adjacent links through btHingeConstraint. The rotation axis vector is configured to be along the link axis direction, and the limit angle parameters are set according to the actual range of motion of the robot arm. During motion control, the joint rotation angle is calculated according to the following formula: , of which joint Angle of time From the initial angle Composed of harmonic superposition terms, harmonic amplitude angular frequency and phase shift The harmonic order was determined through offline fitting of robotic arm joint kinematic calibration data. Based on the requirements for motion smoothness, when the end effector of the robotic arm approaches the cylinder, the btCollisionDispatcher component processes the collision interaction between the workpiece and the fixture in real time, and calculates the contact force and collision position through a collision detection algorithm to ensure the physical realism of the virtual operation.

[0031] The vision sensor simulation unit builds an industrial camera model based on the THREE.js extended rendering pipeline, creates a WebGLRenderTarget rendering target and sets it to THREE.FloatType to support HDR imaging. When configuring camera parameters, the focal length is set according to the detection distance requirements, the field of view covers the detection area on the cylinder surface, and the depth of field parameter is simulated using a formula: Among them, the pixel values ​​after depth processing From the original pixel values The pixel depth value is calculated using the Gaussian attenuation term. Obtained via depth buffer, focusing on depth. Set as cylinder block surface distance, ambiguity coefficient Based on the camera aperture parameters, the ray tracing rendering unit uses the Raycaster tool in THREE.js to create a raycaster object and set the firstHitOnly property to true. It then emits a sampling ray from the center of the virtual camera pixel to perform intersection detection on the cylinder surface, accurately recording the coordinates of the intersection point of the ray and the defect area, as well as the normal vector information, to provide data support for defect identification.

[0032] The collaborative control module realizes closed-loop interaction between visual inspection and motion control. The data interaction unit builds a real-time communication link. When the visual inspection module identifies a defect on the cylinder surface through image processing algorithms, it immediately outputs the defect coordinates to the trajectory planning module. The trajectory planning module generates the optimal joint angle command based on the defect location and the current posture of the robotic arm, and sends it to the physics engine through the data interface. After receiving the command, the rigid body dynamics unit drives the robotic arm to adjust its posture so that the visual sensor focuses on the defect area. At the same time, the btCollisionDispatcher component monitors the collision status between the robotic arm and surrounding equipment. If a collision risk occurs, the physics engine sends the collision feedback information back to the visual inspection module in real time, triggering the detection parameter update mechanism to adjust parameters such as camera focal length and exposure time for re-detection. The real-time control unit adopts thread separation technology, which allocates visual image processing and physical simulation tasks to different threads for parallel execution. Efficient data interaction is achieved through shared memory to ensure that the response delay of the closed-loop control meets the real-time requirements.

[0033] During simulation testing, the system can quantitatively output performance indicators such as defect recognition rate and positioning accuracy of the visual detection algorithm. At the same time, it can simulate extreme working conditions such as robotic arm movement jitter and sudden changes in lighting to verify the robustness of the algorithm. Through repeated testing of multiple sets of defect samples, the detection success rate of different defect types can be statistically analyzed, providing data basis for algorithm parameter optimization.

[0034] In summary, this embodiment achieves high-fidelity simulation of automotive cylinder block surface defect detection by fully utilizing the functions of each module in the system. It uses a physics engine to replace physical debugging, avoiding equipment wear and tear caused by repeated trial and error of the robotic arm. Through rapid conversion of CAD models and lightweight loading technology, the construction time of the production line digital twin environment is shortened. The parametric defect model and high-fidelity optical simulation accurately reproduce the characteristics of industrial defects. The real-time collaborative mechanism of visual inspection and motion control effectively verifies the robustness of the algorithm in dynamic scenarios.

[0035] Example 2

[0036] This embodiment focuses on the quality inspection scenario of a precision soldering production line for electronic components. It utilizes a robotic arm simulation system supporting industrial quality inspection to construct a virtual inspection platform, enabling automated simulation detection of minute defects such as cold solder joints, pinholes, and solder buildup on circuit boards. In traditional production, electronic component solder joint inspection relies on high-precision physical robotic arms and vision systems. Frequent physical trial and error during debugging can easily damage components, and the optical imaging effect of minute defects is greatly affected by ambient light, making algorithm verification difficult. (See [link to previous document]). Figure 1 , Figure 2 and Figure 3This simulation system can reproduce the microstructural features, optical reflection characteristics, and robotic arm detection trajectory of weld points in a virtual environment, completing the full-link verification of the high-precision detection process and significantly reducing the cost and technical risks of physical debugging.

[0037] After the system starts, the digital twin interface module first completes the digital reconstruction of the production scene. The CAD model import unit receives the STEP format 3D model of the circuit board and the IGES model of the welding fixture, calls the Blender tool to perform the format conversion process, and converts the models to GLTF format in batches. During the conversion, the geometric details and material properties of the solder joint area are preserved. After the conversion is completed, the model data is imported through the GLTFLoader loader of THREE.js, and the dracoCompression function is enabled to compress the model vertex data to reduce memory usage. The scene construction unit builds a virtual inspection scene based on the imported model data, establishes the transformation matrix between the world coordinate system and the robot arm base coordinate system, and realizes the spatial position mapping of the circuit board, fixture and robot arm. In order to simulate the actual lighting conditions in the workshop, the system loads HDR panoramic lighting map and simulates the reflection and refraction effects of different wavelengths of light on the surface of the solder joint through ray tracing algorithm to ensure that the optical characteristics of the scene are consistent with the physical environment.

[0038] The defect sample library module generates parametric models for different solder joint defect types. The parametric defect modeling unit simulates cold solder joint defects by perturbing the surface normals with noise textures. It calls the `vec3perturbNormal` function to calculate the normal perturbation value. The position vector `pos` in the function parameters is set based on the solder joint coordinate distribution, and the noise scaling factor `noiseScale` is dynamically adjusted according to the area of ​​the cold solder joint region. For pinhole-type dent defects, modeling is achieved through the SDF deformation method. The formula for calculating the perturbed normal is as follows: The disturbance coefficient The threshold range is selected based on the pinhole depth: a smaller threshold range for minor pinholes and a larger threshold range for severe pinholes, along with the noise function gradient. The simplex noise algorithm is used to simulate random defect distribution. Based on the statistical data of electronic component welding defects, the defect distribution management unit automatically generates multiple sets of test samples on the virtual solder joint surface according to the actual defect probability distribution, covering different forms of cold solder joints, pinholes and solder pile combinations.

[0039] After the multiphysics simulation engine module is started, each unit works together to simulate the detection process. The rigid body dynamics unit uses the Bullet physics engine kernel of ammo.js to build the dynamic model of the robotic arm. A btRigidBody object is created for each link, and the rotational joint constraints of adjacent links are established through btHingeConstraint. The rotation axis vector and limit angle parameters are configured to match the motion characteristics of the precision robotic arm. The calculation of the joint rotation angle of the robotic arm follows the formula: When the end effector of the robotic arm approaches the circuit board, the btCollisionDispatcher component performs collision detection in real time to avoid device interference during virtual operation.

[0040] The visual sensor simulation unit constructs a microscopic vision system model based on the THREE.js extended rendering pipeline, creates a WebGLRenderTarget rendering target and sets it to THREE.FloatType to support HDR imaging, configures the focal length and field of view parameters of the high-magnification lens to meet the needs of detecting minute defects, and achieves depth-of-field rendering effects through the following formula: Among them, focusing on depth Set as the distance between solder joint surfaces, fuzziness coefficient Based on the aperture parameters of the microscope lens, the virtual imaging effect is determined to be consistent with that of the physical microscope. The ray tracing rendering unit emits sampling rays through the Raycaster tool of THREE.js to accurately record the intersection information of the rays and the solder joint defects, providing high-precision data support for defect identification.

[0041] The collaborative control module realizes closed-loop control of the inspection process. The data interaction unit builds a real-time communication link between visual inspection, trajectory planning and physics engine. When the visual inspection module identifies a weld defect, it immediately outputs the defect coordinates to the trajectory planning module. The trajectory planning module generates the optimal joint angle command and sends it to the physics engine to drive the robotic arm to adjust its posture for secondary inspection. If a collision risk occurs, the physics engine will send the collision feedback back to the visual inspection module to trigger the parameter adjustment mechanism. The real-time control unit adopts thread separation technology to achieve parallel processing, ensuring that the closed-loop control delay meets the requirements of precision inspection.

[0042] In summary, this embodiment achieves high-fidelity simulation of electronic component solder joint defect detection scenarios through the coordinated operation of various system modules. The physics engine replaces physical debugging to avoid damage to precision components. The rapid conversion technology of CAD models shortens the construction time of digital twin scenarios. Parametric defect modeling and high-fidelity optical simulation improve the accuracy of defect recognition. The real-time collaborative mechanism of vision and control effectively verifies the robustness of the algorithm in the detection of minute defects.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A robotic arm simulation system supporting industrial quality inspection, characterized in that, The system includes: a multiphysics simulation engine module, a digital twin interface module, a defect sample library module, and a collaborative control module; The multiphysics simulation engine module is used to simulate robotic arm motion, visual imaging, and light propagation. The multiphysics simulation engine module includes a rigid body dynamics unit, a visual sensor simulation unit, and a ray tracing rendering unit. The digital twin interface module is used to import CAD models and build simulation scenes. The digital twin interface module includes a CAD model import unit and a scene construction unit. The defect sample library module is used to generate parameterized defect models and implement defect distribution management. The defect sample library module includes a parameterized defect modeling unit and a defect distribution management unit. The collaborative control module is used to realize data interaction and real-time control between visual detection, trajectory planning and physics engine. The collaborative control module includes a data interaction unit and a real-time control unit. The multiphysics simulation engine module, digital twin interface module, defect sample library module, and collaborative control module interact sequentially to construct a complete virtual simulation environment for industrial quality inspection.

2. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The rigid body dynamics unit of the multiphysics simulation engine module adopts the Bullet physics engine kernel of ammo.js. It creates a btRigidBody object for each link of the robotic arm, establishes rotational joint constraints for adjacent links using btHingeConstraint, configures rotation axis vectors and limit angle parameters, and handles collision interactions between the workpiece and the fixture using the btCollisionDispatcher component. The joint rotation angle calculation satisfies the formula: ,in For the first joint From the perspective of time, Initial angle, harmonic amplitude angular frequency and phase shift Determined through offline fitting of robotic arm joint kinematic calibration data. It represents the harmonic order.

3. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The vision sensor simulation unit of the multiphysics simulation engine module is based on the THREE.js extended rendering pipeline. It creates a WebGLRenderTarget rendering target, whose type is set to THREE.FloatType to support HDR imaging. It configures the focal length, field of view, and depth parameters of the industrial camera, and outputs image data containing color and depth information. The depth rendering effect satisfies the formula: ,in These are the pixel values ​​after depth-of-field processing. These are the original pixel values. The pixel depth value. To focus on depth, is the fuzzy coefficient.

4. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The ray tracing rendering unit of the multiphysics simulation engine module is implemented based on the Raycaster tool of THREE.js. It creates a raycaster object and sets the firstHitOnly property to true, emits a sampling ray from the center of the virtual camera pixel, performs intersection detection on objects in the scene, and records the intersection information.

5. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The CAD model import unit of the digital twin interface module executes the conversion process from STEP / IGES files to GLTF format. After the format conversion is completed by the Blender tool, the GLTFLoader of THREE.js is used to load the model, and the dracoCompression function is enabled to optimize the model loading efficiency.

6. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The scene construction unit of the digital twin interface module establishes a transformation matrix between the world coordinate system and the robotic arm base coordinate system, uses HDR panoramic lighting maps to simulate ambient lighting, and realizes the spatial position mapping of each object in the scene through the coordinate transformation matrix.

7. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The parameterized defect modeling unit of the defect sample library module simulates scratches by perturbing the surface normal with noise texture. It uses the `vec3perturbNormal` function to calculate the normal perturbation value, with parameters including the position vector `pos` and the noise scaling factor `noiseScale`. It simulates dent-type defects using the SDF deformation method, and the normal perturbation satisfies the formula: ,in For the perturbated normal, For the original normal, The disturbance coefficient is... The gradient of the noise function.

8. The robotic arm simulation system supporting industrial quality inspection according to claim 1, characterized in that, The data interaction unit of the collaborative control module constructs a closed-loop data link between visual detection, trajectory planning and physics engine. The visual detection module outputs defect coordinates to the trajectory planning module, the trajectory planning module generates joint angle commands and sends them to the physics engine, and the physics engine sends collision feedback information back to the visual detection module.

Images