Automobile part quality detection method based on multi-sensor fusion

Abstract

The invention relates to the technical field of automobile part detection, and particularly discloses an automobile part quality detection method based on multi-sensor fusion, which comprises the following steps: S1, placing an automobile part on a conveyor belt, and conveying the automobile part to a detection area by the conveyor belt, S2, collecting a part image through a camera, determining the specific type and model of the part through image recognition, and determining the quality of the automobile part according to the specific type and model. S3, the four-station rotary detection device drives a rotary platform to switch a corresponding sensor base to a working position according to the identified part type, after the sensor is in place, corresponding detection programs and parameters are loaded, and detection preparation is completed; and the four-station rotary detection device drives the in-place sensor to execute high-precision detection according to a preset path and parameters, and outputs a result, so that the technical problems that the existing single automobile part quality detection cannot adapt to different types of automobile parts and cannot be dynamically adjusted are solved.

Description

Technical Field

[0001] This invention relates to the field of automotive parts testing technology, and specifically discloses a method for quality testing of automotive parts based on multi-sensor fusion. Background Technology

[0002] In the automotive manufacturing industry, the quality inspection of parts is a key link in ensuring the safety and reliability of the whole vehicle. Currently, for the inspection of different types of defects (such as dimensional deviations, surface defects, and internal welding defects), the industry generally adopts a single or fixed multi-sensor solution. Common practices include: deploying industrial cameras for two-dimensional vision inspection, structured light scanners for three-dimensional dimensional measurement, and infrared thermal imagers for non-destructive testing along the production line conveyor belt or at different workstations. Each device works independently to complete its preset single type of inspection task. For complex parts, it is usually necessary to pass through multiple independent inspection stations in sequence, or to transfer the workpiece to different dedicated inspection equipment to complete a comprehensive quality assessment.

[0003] However, the existing testing methods have certain problems. First, multiple single-function testing devices or sensors need to be installed in different locations on the production line, which results in a large footprint for the entire testing system, a complex line layout, and a significant increase in equipment procurement, installation, and maintenance costs. Since the functions and positions of the sensors are fixed, when the production line switches to different types of automotive parts, such as from engine blocks that require 3D scanning to body welded parts that require thermal imaging, the existing system cannot automatically select the most suitable sensor combination, nor can it dynamically adjust the optimal testing angle and sensor collaborative working mode for specific parts. This directly results in a cumbersome and inefficient testing process, making it difficult to meet the needs of modern flexible manufacturing for rapid production changeover and intelligent testing. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a quality inspection method for automotive parts based on multi-sensor fusion, so as to solve the technical problem that existing single automotive parts quality inspection methods cannot adapt to different types of automotive parts and cannot be dynamically adjusted.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for quality inspection of automotive parts based on multi-sensor fusion, comprising the following steps:

[0006] S1: Place the automotive parts on the conveyor belt, which then transports them to the inspection area.

[0007] S2: Collect images of components using a camera, and determine their specific type and model through image recognition.

[0008] S3: The four-station rotary inspection device drives the rotary platform to switch the corresponding sensor base to the working position according to the identified part type. After the sensor is in place, the corresponding inspection program and parameters are loaded to complete the inspection preparation.

[0009] S4: The parts are conveyed to the inspection station, where the four-station rotary inspection device drives the pre-positioned sensors to perform high-precision inspection according to the preset path and parameters, and outputs the results.

[0010] This solution constructs a closed-loop process from identification to automatic switching to precise detection, which contrasts significantly with existing technologies. Common existing devices often employ fixed sensors or require manual intervention to change detection tooling, resulting in low changeover efficiency and difficulty in meeting the flexible production demands of multi-product mixed lines. This method, through real-time image recognition in S2 and automatic drive switching in S3, enables a single detection device to autonomously match the optimal detection scheme for multiple parts, achieving seamless integration from general conveying, intelligent identification, sensor adaptive positioning to high-precision detection. This effectively solves the pain points of traditional detection lines, such as insufficient flexibility, slow cycle time, and reliance on manual labor. While ensuring detection accuracy, it significantly improves the overall intelligence level and comprehensive detection efficiency of the production line.

[0011] Furthermore, the four-station rotary detection device includes an industrial camera, which is set on the conveyor belt conveying path. A robotic arm is set on the side wall of the conveyor belt, and a chassis is set on the telescopic end of the robotic arm. A protective cylinder is set on the chassis, and a first motor is set inside the protective cylinder. The first motor is fixedly connected to the chassis, and a rotating cylinder is fixedly connected to the power output shaft of the first motor. Multiple sensor fixing components are set on the circumference of the rotating cylinder.

[0012] One of the multiple sensor mounting components is used to mount a high-precision 3D structured light camera for full-size scanning of precision castings;

[0013] Another of the multiple sensor mounting assemblies is used to mount an infrared thermal imager for non-destructive testing of the welds and internal structures of the identified welded structural components;

[0014] Another of the multiple sensor mounting assemblies is used to mount a high-resolution industrial CCD camera to detect surface defects in identified stamped parts;

[0015] One of the multiple sensor fixing assemblies is used to fix a high-precision 3D line laser profile measuring instrument to measure the adhesive strip size of the identified glued part;

[0016] The chassis is equipped with an export component, which is used to export one of the corresponding multiple sensor fixing components. The chassis is also equipped with a detection module, which is used to identify and analyze the collected image data of automotive parts and generate detection control commands based on the analysis results.

[0017] In this solution, four types of sensors with different functions—a high-precision 3D structured light camera, an infrared thermal imager, an industrial CCD camera, and a 3D line laser profile measuring instrument—are integrated in a circumferentially distributed manner into a rotatable sensor fixing component. Intelligent scheduling is achieved through a unified image recognition and decision-making module. This design fundamentally differs from existing technologies that often require separate workstations for different inspection tasks or rely on manual replacement of bulky inspection modules. Existing devices often suffer from large footprints, difficulties in sensor coordination, and low switching efficiency. This device, relying on the coordinated action of the rotating cylinder and the export component, can automatically and accurately switch the corresponding sensor to the working position after identifying the part type, not only improving space utilization but also enhancing the degree of inspection automation.

[0018] Furthermore, each of the multiple sensor fixing components includes a carrier cylinder, which is slidably disposed on the outer wall of the rotating cylinder. The carrier cylinder is used to load the corresponding sensor, and a slider is provided on the carrier cylinder.

[0019] In this design, each sensor is modularly encapsulated within an independent carrier cylinder and slidably connected to the outer wall of the rotating cylinder via a slider. This integrated sliding design offers higher switching reliability and motion consistency compared to existing sensor modules that require additional robotic arms for gripping, replacement, or complete disassembly. Existing devices often face problems such as large cumulative positioning errors, repeated interface wear, or complex operation processes when switching sensors. In this structure, the carrier cylinder, as the inherent carrier of the sensor, has its radial sliding path strictly constrained, ensuring that each extension and retraction can accurately reposition itself in the same spatial position. This provides a solid mechanical foundation for high repeatability detection, greatly simplifies the transduction process, and improves the long-term stability of the equipment.

[0020] Furthermore, the export component includes a sliding frame disposed on the upper side of the base plate, a slidable movable plate disposed within the sliding frame, a fixed plate fixedly connected below the sliding frame, a second motor disposed on the fixed plate, a screw fixedly connected to the power output end of the second motor, the end of the screw away from the second motor being threadedly connected to the movable plate, a locking block fixedly connected below the movable plate, a second sliding groove provided on the slider that can slide in contact with the locking block, and a second opening provided on the protective cylinder for the bearing cylinder to pass through.

[0021] In this solution, a second motor drives a screw to convert rotational motion into precise linear motion of the moving plate. Then, through the engagement of a locking block under the moving plate with a pre-set second groove on the slider, power is transmitted to the target bearing cylinder, driving it to slide out of the protective cylinder along a predetermined path. The self-locking property of the screw drive ensures the stability of the sensor in the working position, and the simple cooperation between the locking block and the groove eliminates the need for complex active docking calibration. Thus, a simpler and more reliable mechanical method is used to achieve a smooth and accurate transfer of the sensor module from the storage position to the detection position, significantly improving the efficiency of the entire switching process.

[0022] Furthermore, each of the sensor fixing components also includes a slide bar, which is fixedly connected to the bearing cylinder. The outer wall of the rotating cylinder is provided with a first sliding groove, and the slide bar is slidably disposed in the first sliding groove.

[0023] In this design, the slide bar is fixedly connected to the support cylinder and slidably fitted into the first groove directly machined into the outer wall of the rotating cylinder. This makes the extension and retraction of the support cylinder strictly limited to the preset linear trajectory, integrating the guiding function with the rotating cylinder. No additional complex support and installation benchmarks are required. Through integrated design, redundant connection links are eliminated, which greatly improves the overall rigidity and guiding accuracy of the motion components. This ensures that the sensor has stable posture and accurate path during high-speed, repeated linear switching motion, thereby guaranteeing high repeatability accuracy and long-term operational reliability of subsequent detection tasks from a mechanical perspective.

[0024] Furthermore, the detection module includes an image recognition and decision-making submodule, a rotation drive control submodule, and a linear export control submodule. The image recognition and decision-making submodule is used to recognize the acquired images of automotive parts and generate control commands. The rotation drive control submodule is used to drive the rotating cylinder to rotate according to the control commands to switch detection sensors. The linear export control submodule is used to control the export mechanism to position the selected sensor to the detection station.

[0025] In this solution, a precise, reliable, and fully automated closed-loop control link from visual perception to physical execution is constructed through the hierarchical and specialized division of labor and collaboration among three sub-modules: image recognition and decision-making sub-module, rotation drive control sub-module, and linear derivation control sub-module.

[0026] Furthermore, the image recognition and decision-making submodule is used to receive and analyze images of automotive parts captured by the visual acquisition device, identify the current part type according to the preset part feature library, and generate control instructions containing target sensor station information accordingly.

[0027] In this solution, the image recognition and decision-making submodule compares real-time acquired part images with a preset feature library to directly and accurately determine the part type and automatically generate the necessary control commands. Compared with existing technologies that generally rely on operator visual judgment, manual selection of detection programs in the system, or preset single scanning modes for fixed parts, this solution achieves a shift from manual intervention or fixed execution to automatic recognition and decision-making. This avoids detection errors caused by human subjective judgment mistakes or incorrect program selections, ensuring that from the recognition stage onwards, all subsequent sensor switching and detection actions can accurately match the optimal process scheme for the current part. This improves the reliability, detection efficiency, and rapid response capability of the entire system for multi-variety mixed-line production from the source.

[0028] Furthermore, the rotation drive control submodule is used to calculate the target station angle according to the control command issued by the image recognition and decision-making submodule, and control the first motor to drive the rotating cylinder and sensor fixing assembly to rotate, positioning the carrier cylinder carrying the corresponding detection sensor to the position to be exported.

[0029] In this solution, the rotary drive control submodule calculates the target station angle according to the instructions and drives the rotary cylinder to automatically rotate the required sensor to the pre-determined fixed position. This design changes the cumbersome and low-precision operation mode commonly found in existing devices, which requires manual rotation of the indexing plate based on experience or pre-setting multiple independent rotation axes for different sensors. Through direct linkage with the image recognition and decision-making submodule, it not only achieves complete automation of sensor selection, but also ensures the speed and repeatability of each rotation positioning. This enables the multi-sensor integrated device to respond accurately and efficiently to the constantly changing types of parts on the production line, just like a single-function device.

[0030] Furthermore, the linear export control submodule is used to control the second motor to drive the export component after receiving the rotation into position signal, so as to stably push the aligned carrier cylinder and its detection sensor out of the protective cylinder to the preset detection position.

[0031] In this solution, the linear export control submodule, upon receiving a signal indicating that the carrier cylinder has rotated and aligned with the second opening, controls the second motor to drive the export component, stably pushing the aligned carrier cylinder and its onboard detection sensor from the protective cylinder to the preset detection station. It automatically controls the second motor to drive the screw and other export mechanisms, smoothly pushing the carrier cylinder out of the protective cylinder. This design overcomes the shortcomings of existing technologies, which often require manual pulling of the sensor after rotational selection or rely on unstable pneumatic propulsion. Through precise timing coordination with the rotation drive control submodule, it ensures that the sensor first completes circumferential selection in space and then performs the orderly linear export action, thus achieving a closed-loop automation of the entire process from intelligent identification to sensor positioning. This not only significantly improves switching efficiency but also fundamentally guarantees the positional consistency and structural stability of the sensor when it arrives at the detection station, providing a reliable foundation for subsequent high-precision detection.

[0032] Furthermore, it also includes controller testing: including a dynamometer bench for simulating different driving conditions, characterized by:

[0033] It also includes a vehicle simulation platform, a dynamics model building module, a scene simulation module, a host computer, and a real-time machine module;

[0034] The dynamics model building module is used to establish a vehicle dynamics model;

[0035] The vehicle simulation platform is used to install the intelligent driving domain controller being tested;

[0036] The scene simulation module is used to create 3D simulation scenes;

[0037] The host computer is used to receive feedback data during the testing process and control the dynamometer bench, real-time machine module, dynamic model construction module, and vehicle simulation platform according to the preset test procedures and parameters.

[0038] During testing, the real-time machine module and the scene simulation module interact to obtain sensor data. The real-time machine module transmits the sensor data to the intelligent driving domain controller. The intelligent driving domain controller performs algorithm path planning based on the sensor data and the current vehicle status information, and then feeds back the execution instructions to the real-time machine module. The dynamometer bench feeds back the operating condition information to the real-time machine module. The real-time machine module takes the operating condition information and the execution instructions, inputs them into the vehicle dynamics model, performs calculations to obtain response information, and transmits it to the scene simulation module. The scene simulation module updates the 3D simulation scene in real time based on the response information.

[0039] This solution combines a dynamometer rig, a vehicle simulation platform, and a scenario simulation module. It simulates road conditions in a 3D simulation scenario, with the dynamometer rig mimicking the resistance and inertia during driving. This allows for testing the correctness of decisions made by the intelligent driving domain controller. It eliminates the need for full-vehicle testing on actual roads, requiring only the vehicle's transmission system. This improves testing efficiency, reduces costs, and avoids the risk of vehicle damage due to malfunctions in traditional intelligent driving domain controllers. The real-time machine module inputs real-time operating information and execution commands into the vehicle dynamics model for calculation and response information. The scenario simulation module updates the 3D simulation scenario in real-time based on the response information, allowing testers to more intuitively observe the vehicle's driving process in the 3D virtual scenario. Compared to simply viewing test data, it makes it easier to identify issues such as following too closely or delayed steering.

[0040] The working principle and beneficial effects of this solution are as follows:

[0041] At the start of the inspection, automotive parts are transported on a conveyor belt. As the parts pass the mounting frame, an industrial camera automatically captures an image and sends it to the inspection module. The image recognition submodule identifies the part type by comparing it with a feature library and generates control commands. The rotation drive control submodule then instructs the first motor to rotate the carrier cylinder corresponding to the required sensor to a predetermined angle. When the carrier cylinder aligns with the second opening, the linear output control submodule starts the second motor, which smoothly pushes the carrier cylinder out of the protective cylinder via a screw mechanism. The robotic arm then moves the sensor to the main inspection station above the part. Based on the part type, a dedicated sensor, such as a 3D structured light camera, infrared thermal imager, industrial CCD camera, or 3D line laser profilometer, is called to perform specialized inspection. After inspection, all components are reset in sequence, ready for the next operation. This method solves the technical problems of existing single inspection devices being unable to adapt to multiple types of parts and unable to dynamically adjust the inspection strategy. Attached Figure Description

[0042] Figure 1 This is a flowchart of a quality inspection method for automotive parts based on multi-sensor fusion.

[0043] Figure 2 This is a schematic diagram of a four-station rotary inspection device for automotive parts quality inspection based on multi-sensor fusion.

[0044] Figure 3 This is an enlarged schematic diagram of point A of the four-station rotary testing device.

[0045] Figure 4 This is a partial structural diagram of a four-station rotary inspection device for automotive parts quality inspection based on multi-sensor fusion.

[0046] Figure 5An exploded view of the sensor fixing assembly of a four-station rotary inspection device for automotive parts quality inspection based on multi-sensor fusion.

[0047] Figure 6 The flowchart shows the control module of a multi-sensor fusion-based quality inspection method for automotive parts.

[0048] Figure 7 This is a schematic diagram of the first embodiment of a vehicle intelligent driving domain simulation test system based on a multi-sensor fusion-based automotive parts quality inspection method.

[0049] The following components are labeled in the attached diagram: 1. Fixture; 2. Industrial camera; 3. Mounting bracket; 4. Robotic arm; 5. Chassis; 6. Protective cylinder; 7. First motor; 8. Rotating cylinder; 9. Circular slide rail; 10. Detection module; 11. Slide bar; 12. Bearing cylinder; 13. Slider; 14. First slide groove; 15. Sliding frame; 16. Moving plate; 17. Limiting plate; 18. Fixing plate; 19. Second motor; 20. Screw; 21. Locking block; 22. Second slide groove; 23. First opening; 24. Second opening; 25. Image recognition and decision-making submodule; 26. Rotation drive control submodule; 27. Linear derivation control submodule. Detailed Implementation

[0050] The following detailed description illustrates the specific implementation method:

[0051] Example

[0052] like Figures 1 to 7 As shown, a quality inspection method for automotive parts based on multi-sensor fusion is disclosed. The quality inspection method for automotive parts includes:

[0053] S1: Place the automotive parts to be inspected on the conveyor belt. The conveyor belt will transport the parts sequentially into the inspection area according to a preset rhythm.

[0054] S2: Industrial camera 2, set at the front end of the conveyor belt, captures images when automotive parts arrive at the recognition station. The images are then transmitted to an image recognition algorithm to determine the specific type, model, and precise position and orientation of the automotive parts on the conveyor belt.

[0055] S3: After identification, the type and location information of the component are sent to the four-station rotary inspection device. The controller within the device automatically generates control commands based on the received information and predefined detection logic, driving the rotary platform to rotate the corresponding sensor base to the working position.

[0056] If it is identified as a precision casting, such as a cylinder or housing, it is rotated to the base that carries a high-precision 3D structured light camera for performing a full-size scan of three-dimensional dimensions and geometric tolerances.

[0057] If identified as a welded structural component, such as a vehicle body bracket, it is rotated to a base that integrates an infrared thermal imager and a pulsed thermal excitation source for non-destructive testing of the weld and internal structure.

[0058] If it is identified as a stamped exterior part, such as a car door panel, it is rotated to a base equipped with a high-resolution industrial CCD camera and a multi-angle ring light source to capture and identify surface scratches, dents and other appearance defects.

[0059] If the part is identified as being coated with adhesive, such as a car door or sunroof, it will be rotated to a base equipped with a high-precision 3D line laser profile measuring instrument. This instrument is used to perform high-speed online three-dimensional measurement of the width, height, and continuity of the adhesive coating, and to determine in real time whether there are defects such as broken adhesive or insufficient adhesive height.

[0060] After the sensor is in place, the four-station rotary detection device loads its corresponding detection program and parameters, such as 3D scanning density, thermal excitation intensity, light source combination scheme, and contour measurement accuracy, to complete the detection preparation.

[0061] S4: After the four-station rotary inspection device completes the corresponding switching, the parts are transferred to the main inspection station. Based on the precise position information obtained in S2, the device drives the positioning mechanism to make fine adjustments. Then, it instructs the sensor modules that are already in place to perform high-precision inspection on the specified features of the parts according to the preset path and parameters, and outputs the analysis results in real time.

[0062] like Figure 2 , Figure 3 and Figure 4 As shown, the four-station rotary inspection device includes a fixed frame 1, an industrial camera 2, a mounting frame 3, a robotic arm 4, a chassis 5, a protective cylinder 6, a first motor 7, a rotating cylinder 8, multiple sensor fixing components, a circular slide rail 9, an export component, and an inspection module 10. The fixed frame 1 is mounted on the conveyor belt and is located on the conveyor belt's travel path. The industrial camera 2 is mounted on the fixed frame 1, and its optical axis is perpendicular to the plane of the conveyor belt, ensuring that the main optical axis of the lens of the industrial camera 2 is aligned with the center of the necessary path of the automotive parts to be inspected. The mounting frame 3 is provided on the side wall of the conveyor belt, and the robotic arm 4 is mounted on the upper side of the mounting frame 3. The extended end of the robotic arm 4 is fixedly connected to the chassis 5, and the end of the chassis 5 away from the robotic arm 4 is fixedly connected to the protective cylinder 6. The first motor 7 is fixedly connected to the chassis 5 and is located inside the protective cylinder 6. The power output shaft of the first motor 7 is fixedly connected to the rotating cylinder 8, and the power output shaft of the first motor 7 is fixedly connected to the inner wall of the rotating cylinder 8. Multiple sensor fixing components are evenly fixedly connected along the circumference on the outer wall of the rotating cylinder 8.

[0063] As a preferred embodiment of the present invention, there are four sensor fixing components, which are distributed at 90-degree intervals around the central axis of the rotating cylinder 8. Each sensor fixing component is used to support a specific type of special detection device. Specifically, a high-precision 3D structured light camera is fixedly installed on the first sensor fixing component, which is used to perform a full-size scan of the three-dimensional dimensions and geometric tolerances of the part when the system identifies the part to be inspected as a precision casting.

[0064] The second sensor mounting assembly integrates an infrared thermal imager and a pulsed thermal excitation source that work together with it, which is dedicated to non-destructive testing of welds and internal structures of identified welded structural components.

[0065] The third sensor mounting assembly is equipped with a high-resolution industrial CCD camera and a matching multi-angle ring light source. Its function is to acquire images of the surface of the identified stamped parts in order to detect surface defects such as scratches and dents.

[0066] The fourth sensor mounting assembly is equipped with a high-precision 3D line laser profile measuring instrument. When the part to be inspected is identified as a glued part, the measuring instrument can perform high-speed online three-dimensional measurement of the width, height and continuity of the glue strip, and determine defects such as glue breakage or insufficient glue height in real time.

[0067] The chassis 5 is equipped with an annular slide rail 9, which is used to limit and fix multiple sensor fixing components. The chassis 5 is equipped with an export component, which is used to drive and position the selected sensor fixing component from the protective cylinder 6 to the inspection station for quality inspection. The chassis 5 is equipped with a detection module 10, which is used to receive and analyze the images of automotive parts captured by the industrial camera 2, generate control commands according to the identified part type, and drive the first motor 7 to rotate so that the corresponding sensor fixing component is aligned with the export component.

[0068] like Figure 5 As shown, each of the multiple sensor fixing components includes a slide bar 11, a support cylinder 12, and a slider 13. A first slide groove 14 is provided on the outer wall of the rotating cylinder 8. The first slide groove 14 passes through the end of the rotating cylinder 8 away from the chassis 5. The slide bar 11 is slidably connected in the first slide groove 14. The support cylinder 12 is fixedly connected to the slide bar 11. The support cylinder 12 is used to load the special detection sensor corresponding to each detection mode. The slider 13 is fixedly connected to the side of the support cylinder 12 near the chassis 5. The end of the slider 13 away from the rotating cylinder 8 is slidably set in the annular slide rail 9.

[0069] like Figure 4As shown, the export component includes a sliding frame 15, a movable plate 16, a limiting plate 17, a fixed plate 18, a second motor 19, a screw 20, and a locking block 21. A sliding frame 15 is mounted on the upper side of the chassis 5. A slidable movable plate 16 is mounted inside the sliding frame 15. A limiting plate 17 is fixedly connected to the side of the sliding frame 15 away from the protective cylinder 6. A fixed plate 18 is fixedly connected to the side of the sliding frame 15 closer to the protective cylinder 6. The lower side of the fixed plate 18 is fixedly connected to the protective cylinder 6. A second motor 19 is fixedly connected to the fixed plate 18. A screw 20 is fixedly connected to the power output shaft of the second motor 19. The end of the rod 20 away from the second motor 19 is threadedly connected to the moving plate 16 and then rotatably connected to the limiting plate 17. The moving plate 16 is fixedly connected to the side away from the sliding frame 15 with a locking block 21. The slider 13 is provided with a second sliding groove 22 on the side near the chassis 5. The second sliding groove 22 passes through both sides of the slider 13. The locking block 21 can slide into the second sliding groove 22. The annular slide rail 9 is provided with a first opening 23. The width of the first opening 23 is greater than the width of the slider 13. The protective cylinder 6 is provided with a second opening 24. The width of the second opening 24 is greater than the width of the bearing cylinder 12.

[0070] like Figure 6 As shown, the detection module 10 includes an image recognition and decision-making submodule 25, a rotation drive control submodule 26, and a linear derivation control submodule 27.

[0071] The image recognition and decision-making submodule 25 is used to receive and analyze images of automotive parts captured by the industrial camera 2, identify the current part type according to the preset part feature library, and generate control commands containing target sensor station information accordingly.

[0072] The rotary drive control submodule 26 is configured to control the first motor 7 to rotate clockwise. The four sensor fixing components are arranged in a clockwise order around the central axis of the bearing cylinder 12: the first station (initial 0° position) is fixed with a high-precision 3D structured light camera, which is used to inspect precision castings; the second station (clockwise 90° direction) integrates an infrared thermal imager and a pulsed thermal excitation source, which is used to inspect welded structural parts; the third station (clockwise 180° direction) is equipped with a high-resolution industrial CCD camera and a multi-angle ring light source, which is used to inspect stamped appearance parts; and the fourth station (clockwise 270° direction) is equipped with a high-precision 3D line laser profile measuring instrument, which is used to inspect glued parts.

[0073] After the image recognition and decision-making submodule 25 determines the part type, the rotation drive control submodule 26 will calculate the station angle corresponding to the target sensor and control the first motor 7 to rotate clockwise by an integer multiple of 90°, so as to accurately rotate the target sensor and stop it at the position aligned with the second opening 24 on the protective cylinder 6.

[0074] The linear output control submodule 27 is communicatively connected to the rotary drive control submodule 26 and the second motor 19. After receiving the rotation positioning signal, it controls the second motor 19 to drive the screw 20 mechanism, thereby driving the moving plate 16 to stably push the aligned bearing cylinder 12 and its detection sensor from the protective cylinder 6 to the preset detection position.

[0075] like Figure 7 As shown, it also includes the controller's detection: including a dynamometer test bench, a vehicle simulation platform, a dynamics model building module, a scene simulation module, a host computer, a real-time machine module, and an environment chamber module.

[0076] The host computer is used to receive feedback data during the testing process and control the dynamometer bench, real-time machine module, dynamic model construction module, vehicle simulation platform, and environmental chamber module according to the preset test procedures and parameters.

[0077] The dynamometer test bench is used to simulate the resistance and inertia of a vehicle during driving. It includes four motors, which can output different torques and speeds according to control signals to simulate different driving conditions, such as acceleration, deceleration, constant speed driving, and different road slopes. At the same time, it is equipped with high-precision torque and speed sensors to measure the output torque and speed of the dynamometer in real time and feed these data back to the host computer.

[0078] The vehicle simulation platform is mechanically connected to the dynamometer bench and is used to install the intelligent driving domain controller to be tested and related vehicle components (including engine control unit, transmission control unit, and chassis control module) to build a complete vehicle electrical and mechanical system simulation environment (which can also be built based on a real vehicle).

[0079] The dynamic model building module is used to establish a vehicle dynamic model, which is used to design various vehicle parameters such as body structure, chassis layout, and power system configuration, including tire characteristics, suspension stiffness, damping coefficient, and mass distribution. By simulating the dynamic response of the vehicle under different parameter combinations, the vehicle is modeled to find the optimal parameter configuration in order to improve the vehicle's handling stability, comfort, and safety.

[0080] The environmental chamber module consists of the following devices: 1. A temperature simulation device, which can simulate different ambient temperatures in the test space, from extremely cold to high temperature environments, to test the performance and reliability of the vehicle domain control system under different temperature conditions; 2. A humidity simulation device, which is used to adjust the humidity of the test environment to examine the impact of humidity on the system's electronic components and electrical performance; 3. A light simulation device, which can simulate light conditions of different intensities and angles to test the vehicle's lighting system and related sensors to ensure that they work normally under various lighting environments.

[0081] The scene simulation module is used to create 3D simulation scenes, which, combined with the real-time machine module, allow the vehicle to operate in real time in the 3D scene for visualization.

[0082] The real-time machine module includes a video injection board, a millimeter-wave radar simulator, a DSI3 board, an Ethernet media converter, and a GNSS signal simulation module; fault diagnosis testing is achieved by setting up a fault injection board.

[0083] In this embodiment, the vehicle intelligent driving domain simulation testing system based on a dynamometer bench utilizes video injection for camera image data during testing. A video injection board is configured to convert the target object signals extracted from the scene constructed within the simulation software into GMSL2 data, which is then transmitted to the real-time machine module. Video injection supports 4K resolution and various video protocols, and includes string encoding and deserialization chips, ensuring compatibility with both injection and acquisition functions. Millimeter-wave radar can transmit obstacle echo information to the intelligent driving domain controller using a millimeter-wave radar simulator (or via CAN signal transmission through millimeter-wave radar bus simulation). Ultrasonic radar injects CANFD signals to the intelligent driving domain controller using a DSI3 board. LiDAR acquires data from the simulated scene... Point cloud signals are converted to 1000base-T1 via RAD-MOON2 standard Ethernet to in-vehicle Ethernet equipment and injected into the intelligent driving domain controller. Vehicle positioning simulation is performed through GNSS signal simulation and OpenDrive format coordinate transformation of the map in the simulation scene, and the results are sent to the intelligent driving domain controller and dynamometer bench. The dynamometer bench determines the slope based on road condition information (information collected by cameras and LiDAR) and the acceleration based on the vehicle's current status information, thereby adjusting its own speed and torque, and feeding back the real-time operating condition information to the real-time machine module. The intelligent driving domain controller makes decisions based on the input of camera, millimeter-wave radar, ultrasonic radar, LiDAR and positioning information, as well as the vehicle's current status information, and sends the issued execution commands back to the real-time machine module.

[0084] The real-time machine module receives execution instructions from the intelligent driving domain controller and driving condition information from the test bench, and then uses this information to solve the vehicle dynamics model. The vehicle's response information is then sent to the scene simulation software via Ethernet communication for real-time updates of the vehicle animation.

[0085] The industrial control computer configured with the scene simulation software can be connected to an industrial display screen via an HDMI interface to display the entire vehicle simulation operation animation in real time.

[0086] Example 2:

[0087] Based on the first embodiment, the dynamometer frame includes four load motors, and two rollers are installed on the output shaft of each load motor. When the output shaft of the load motor rotates, the two rollers rotate in the same direction. A hydraulically driven sensing plate is set between the two rollers. A pressure sensor is set on the sensing plate. When the pressure value detected by the pressure sensor exceeds the preset value (i.e., when the wheel is on the sensing plate), the hydraulic rod can drive the sensing plate to descend, so that the two rollers support the wheel.

[0088] The two load motors at the front end are mounted on a fixed base, and the two load motors at the rear end are mounted on a sliding base that can slide relative to the fixed base. The output shafts of the two load motors at the front end are fixedly connected to a gearbox. The gearbox has two output shafts, namely a multi-gear output shaft and a single-gear output shaft. A first gear is mounted on the multi-gear output shaft, which meshes with an intermediate gear. The other side of the intermediate gear meshes with a second gear. The first gear and the second gear rotate coaxially with the two rollers, respectively. A third gear is mounted on the single-gear output shaft, and a rack is mounted on the fixed base. The third gear meshes with the rack. When the load motor drives the single-gear output shaft to rotate, it can drive the rack to move through the third gear, thereby driving the sliding base to move.

[0089] Before testing, the vehicle under test is driven to a designated position so that the front wheels are on the sensor plate. At this time, if there is no signal from the rear sensor plate, the transmission switches to a single gear output shaft, which drives the third gear to rotate through the front load motor, causing the rack to move the sliding seat until the rear wheels are on the sensor plate. This allows for the adjustment of the distance between the front and rear load motors, meeting the testing requirements of different vehicle models.

[0090] In practice:

[0091] When the inspection begins, the automotive parts are first placed on the conveyor belt. The conveyor belt transports the parts according to a preset rhythm. When the parts enter the field of view of the industrial camera 2 mounted on the fixed frame 1, the industrial camera 2 will automatically trigger and acquire images to ensure that the moment when the parts are located below the fixed frame 1 is accurately captured.

[0092] The collected images of the parts are sent in real time to the detection module 10 located on the chassis 5. The image recognition and decision-making submodule 25 within the module processes the images, identifies the specific type of the parts by comparing them with a preset part feature library, and generates control commands containing target sensor station information.

[0093] Subsequently, the rotary drive control submodule 26 calculates the station angle corresponding to the target sensor according to the instruction. The rotary drive control submodule 26 controls the first motor 7 to drive the rotary cylinder 8 and the four sensor fixing components fixed to its outer wall to rotate clockwise. Each sensor fixing component includes a slide bar 11, a bearing cylinder 12 and a slider 13. When rotating, the slider 13 slides in the annular slide rail 9 to form a stable circumferential limit on the bearing cylinder 12.

[0094] When the rotation stops and the selected bearing cylinder 12 is precisely aligned with the first opening 23 on the annular slide rail 9, the linear output control submodule 27 is activated. It controls the second motor 19 to drive the screw 20, which in turn moves the moving plate 16 and the fixed locking block 21 on it. At this time, the slider 13 has rotated with the rotating cylinder 8 to the first opening 23, thereby breaking free from the constraint of the annular slide rail 9, so that the locking block 21 on the moving plate 16 can be fully embedded in the second slide groove 22 at the bottom of the slider 13.

[0095] Next, the second motor 19 continues to rotate the screw 20 clockwise, pushing the moving plate 16 and the locking block 21 to move in a straight line away from the robotic arm 4. The locking block 21 pushes the slider 13 through the second slide groove 22, thereby driving the entire bearing cylinder 12 to slide radially along the first slide groove 14 on the rotating cylinder 8, and finally fully extends to the pre-test position through the second opening 24 on the protective cylinder 6.

[0096] Robotic arm 4 then moves the entire inspection device (including chassis 5, protective cylinder 6, and extended sensors) to the main inspection station above the conveyor belt, and makes fine adjustments based on the identified part position to initiate specialized inspection. The inspection rules for different parts are as follows:

[0097] If the part is identified as a precision casting (such as a cylinder block), the high-precision 3D structured light camera on the first sensor mounting assembly is invoked to perform a full-size scan of the three-dimensional dimensions and geometric tolerances.

[0098] If the component is identified as a welded structure (such as a vehicle body bracket), the infrared thermal imager and pulsed thermal excitation source integrated on the second sensor mounting assembly are activated to perform non-destructive testing on the weld and internal structure.

[0099] If the part is identified as a stamped exterior part (such as a car door panel), a high-resolution industrial CCD camera with a multi-angle ring light source is used on the third sensor mounting assembly to capture and identify surface scratches, dents and other appearance defects.

[0100] If the part is identified as being coated with adhesive (such as a door or sunroof), the high-precision 3D line laser profile measuring instrument mounted on the fourth sensor mounting assembly is activated to perform high-speed online three-dimensional measurement of the width, height, and continuity of the adhesive coating.

[0101] After the test is completed, the robotic arm 4 drives the device to reset. Then, the second motor 19 rotates counterclockwise and pulls the bearing cylinder 12 back into the protective cylinder 6 through the screw 20 mechanism. Subsequently, the first motor 7 rotates counterclockwise and drives the rotating cylinder 8 to rotate the sensor fixing assembly back to the initial position, preparing for the next test.

[0102] The above descriptions are merely embodiments of the present invention, and common knowledge regarding specific structures and characteristics in the solutions is not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the structure of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or its practicality.

Claims

1. A method for quality inspection of automotive parts based on multi-sensor fusion, characterized in that: Includes the following steps: S1: Place the automotive parts on the conveyor belt, which then transports them to the inspection area. S2: Collect images of components using a camera, and determine their specific type and model through image recognition. S3: The four-station rotary inspection device drives the rotary platform to switch the corresponding sensor base to the working position according to the identified part type. After the sensor is in place, the corresponding inspection program and parameters are loaded to complete the inspection preparation. S4: The parts are conveyed to the inspection station, where the four-station rotary inspection device drives the pre-positioned sensors to perform high-precision inspection according to the preset path and parameters, and outputs the results.

2. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 1, characterized in that: The four-station rotary detection device includes an industrial camera, which is set on the conveyor belt conveying path. A robotic arm is set on the side wall of the conveyor belt. A chassis is set on the telescopic end of the robotic arm. A protective cylinder is set on the chassis. A first motor is set inside the protective cylinder. The first motor is fixedly connected to the chassis. The power output shaft of the first motor is fixedly connected to the rotating cylinder. Multiple sensor fixing components are set on the circumference of the rotating cylinder. One of the multiple sensor mounting components is used to mount a high-precision 3D structured light camera for full-size scanning of precision castings; Another of the multiple sensor mounting assemblies is used to mount an infrared thermal imager for non-destructive testing of the welds and internal structures of the identified welded structural components; Another of the multiple sensor mounting assemblies is used to mount a high-resolution industrial CCD camera to detect surface defects in identified stamped parts; One of the multiple sensor fixing assemblies is used to fix a high-precision 3D line laser profile measuring instrument to measure the adhesive strip size of the identified glued part; The chassis is equipped with an export component, which is used to export one of the corresponding multiple sensor fixing components. The chassis is also equipped with a detection module, which is used to identify and analyze the collected image data of automotive parts and generate detection control commands based on the analysis results.

3. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 2, characterized in that: Each of the aforementioned sensor fixing components includes a carrier cylinder, which is slidably disposed on the outer wall of the rotating cylinder. The carrier cylinder is used to load the corresponding sensor, and a slider is provided on the carrier cylinder.

4. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 2, characterized in that: The export component includes a sliding frame disposed on the upper side of the base plate. A slidable movable plate is disposed inside the sliding frame. A fixed plate is fixedly connected below the sliding frame. A second motor is disposed on the fixed plate. A screw is fixedly connected to the power output end of the second motor. The end of the screw away from the second motor is threadedly connected to the movable plate. A locking block is fixedly connected below the movable plate. A second sliding groove is provided on the slider, which can slide in contact with the locking block. A second opening is provided on the protective cylinder, allowing the carrying cylinder to pass through.

5. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 4, characterized in that: Each of the sensor fixing components further includes a slide bar, which is fixedly connected to the bearing cylinder. The outer wall of the rotating cylinder is provided with a first sliding groove, and the slide bar is slidably disposed in the first sliding groove.

6. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 2, characterized in that: The detection module includes an image recognition and decision-making submodule, a rotation drive control submodule, and a linear export control submodule. The image recognition and decision-making submodule is used to recognize the acquired images of automotive parts and generate control commands. The rotation drive control submodule is used to drive the rotating cylinder to rotate according to the control commands to switch detection sensors. The linear export control submodule is used to control the export mechanism to position the selected sensor to the detection station.

7. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 6, characterized in that: The image recognition and decision-making submodule is used to receive and analyze images of automotive parts captured by the visual acquisition device, identify the current part type according to the preset part feature library, and generate control instructions containing target sensor station information accordingly.

8. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 6, characterized in that: The rotation drive control submodule is used to calculate the target station angle according to the control command issued by the image recognition and decision-making submodule, and control the first motor to drive the rotating cylinder and sensor fixing assembly to rotate, positioning the carrier cylinder carrying the corresponding detection sensor to the position to be exported.

9. The method for quality inspection of automotive parts based on multi-sensor fusion according to claim 6, characterized in that: The linear export control submodule is used to control the second motor to drive the export component after receiving the rotation into position signal, so as to stably push the aligned carrier cylinder and its detection sensor out of the protective cylinder to the preset detection position.

10. The automotive parts quality inspection method based on multi-sensor fusion according to claim 1 further includes controller detection: including a dynamometer bench for simulating different driving conditions, characterized in that: It also includes a vehicle simulation platform, a dynamics model building module, a scene simulation module, a host computer, and a real-time machine module; The dynamics model building module is used to establish a vehicle dynamics model; The vehicle simulation platform is used to install the intelligent driving domain controller to be tested; The scene simulation module is used to create 3D simulation scenes; The host computer is used to receive feedback data during the testing process and control the dynamometer bench, real-time machine module, dynamic model construction module, and vehicle simulation platform according to the preset test procedures and parameters. During testing, the real-time machine module and the scene simulation module interact to obtain sensor data. The real-time machine module transmits the sensor data to the intelligent driving domain controller. The intelligent driving domain controller performs algorithm path planning based on the sensor data and the current vehicle status information, and then feeds back the execution instructions to the real-time machine module. The dynamometer bench feeds back the operating condition information to the real-time machine module. The real-time machine module takes the operating condition information and the execution instructions, inputs them into the vehicle dynamics model, performs calculations to obtain response information, and transmits it to the scene simulation module. The scene simulation module updates the 3D simulation scene in real time based on the response information.

Images