Omnibearing intelligent defect detection system and method for workpiece surface

Abstract

The invention discloses a workpiece surface all-dimensional intelligent defect detection system and method in the technical field of workpiece detection, and aims to solve the technical problems of single illumination direction, limited shooting visual angle, unstable image quality, detection result feedback lagging and data management dispersion of existing detection equipment. The system comprises a detection table module, a synchronous belt sliding table vision module, a workpiece side face detection module, a side face light source, a six-axis mechanical arm, an annular light source, a control computer, a three-color lamp, a main rack, a baffle and a control button. Automatic workpiece positioning, uniform illumination, full-surface imaging, real-time defect recognition and standardized data archiving are achieved through cooperation of the multiple modules. The detection coverage rate is larger than or equal to 99%, the defect recognition accuracy rate is larger than or equal to 98%, the single-workpiece detection period is smaller than or equal to 30 seconds, the method is suitable for surface defect detection of workpieces such as metal and castings, and the high-beat and high-precision detection requirements of the modern manufacturing industry are met.

Description

Technical Field

[0001] This invention relates to the field of workpiece surface inspection, specifically to a comprehensive intelligent defect detection system and method for workpiece surfaces. Background Technology

[0002] In modern manufacturing processes, surface defect detection is a crucial step in ensuring product quality. Common defects include corrosion spots, cracks, indentations, porosity, foreign matter, and scratches. Failure to identify these defects in a timely manner can directly affect product performance and lifespan, and even pose safety hazards.

[0003] Traditional inspection methods rely heavily on manual visual inspection or handheld camera photography to capture images of each item, followed by manual judgment to determine if defects exist. This approach has significant drawbacks: low inspection efficiency, with each workpiece typically taking more than 2 minutes to inspect; poor stability, affected by the experience and fatigue of the inspectors, resulting in a false judgment rate of over 5%; and a large amount of repetitive labor, making it difficult to meet the high-speed, automated inspection requirements of modern production lines, which typically require "≥2 pieces per minute".

[0004] To improve automation levels, existing machine vision inspection equipment is gradually replacing manual labor, but many technical bottlenecks still exist:

[0005] 1. Single illumination direction and poor image quality: Most devices use a fixed top light source, and the illumination direction cannot be adjusted. Complex curved surfaces or reflective surfaces of workpieces are prone to forming shadows and highlights, resulting in incomplete image features. For example, dark areas often appear on the curved surface of metal castings, and small defects such as pores and cracks are obscured, resulting in low recognition accuracy of detection algorithms, with an accuracy rate generally ≤90%.

[0006] 2. Limited shooting angle and blind spots in inspection: Most cameras use a fixed single-camera installation structure, requiring manual or motor-driven workpiece rotation to obtain multi-angle images. This is cumbersome and prone to positioning errors, with an inspection coverage of only ≤85%. Defects on the sides, bevels, and grooves of the workpiece are easily missed, making it impossible to achieve full-surface inspection.

[0007] 3. Delayed detection feedback and insufficient real-time performance: After image acquisition, it needs to be manually imported into the computer for offline analysis. The delay from acquisition to output of detection results exceeds 5 minutes, which cannot achieve real-time feedback on the production line. Defective workpieces are likely to flow into the next process, increasing rework costs.

[0008] 4. Dispersed data management and poor traceability: Detection images and results are mostly stored in scattered folders, lacking a unified archiving and classification mechanism. When reviewing historical detection data, manual searching is required, resulting in low retrieval efficiency and failing to meet the requirements of standardized and traceable production management.

[0009] Among the existing publicly available technologies, patent CN223006101U (authorization announcement date June 20, 2025) discloses a metal surface defect detection device. Although it achieves automatic loading and unloading through a conveyor belt, the detection components are fixed and can only detect the upper surface of the workpiece. Side and inclined surface defects cannot be covered. Moreover, the light source is fixed, and complex curved surfaces are prone to shadows, resulting in insufficient image clarity. Data still needs to be manually transmitted after detection, and there is no real-time recognition function.

[0010] Patent CN116519714B (authorization announcement date August 29, 2023) discloses a metal substrate inspection device that achieves double-sided inspection through a flipping mechanism. However, the light source and camera are fixed, making it impossible to image the sides or complex curved surfaces. The illumination angle and brightness are not adjustable, and blind spots are easily generated on reflective surfaces. The inspection results need to be manually imported and analyzed, and the data storage is scattered and lacks standardized management.

[0011] In summary, existing detection equipment still has significant shortcomings in terms of illumination uniformity, detection comprehensiveness, imaging clarity, and real-time detection. Therefore, those skilled in the art provide a comprehensive intelligent defect detection system and method for workpiece surfaces to address the problems mentioned in the background section. Summary of the Invention

[0012] The purpose of this invention is to provide a comprehensive intelligent defect detection system and method for workpiece surfaces, addressing the shortcomings of existing detection equipment mentioned in the background section, such as single illumination direction, limited shooting angle, unstable image quality, delayed detection feedback, and fragmented data management. Through multi-module collaborative design, the following objectives are achieved:

[0013] 1. Provides multi-angle adjustable lighting to eliminate shadows and reflections on the workpiece surface, ensuring image brightness uniformity deviation ≤5%;

[0014] 2. Achieve full-surface imaging of the workpiece's top surface, sides, bevels, and edges, with a detection coverage rate ≥99%;

[0015] 3. Real-time image processing and defect recognition, single workpiece inspection cycle ≤30 seconds, recognition accuracy ≥98%;

[0016] 4. Inspection data is automatically categorized and archived by "workpiece number / batch", supporting quick retrieval and cloud backup to meet traceability requirements.

[0017] To achieve the above objectives, the present invention provides the following technical solution:

[0018] A comprehensive intelligent defect detection system and method for workpiece surface includes a main frame, and an inspection table module, a synchronous belt slide vision module, a workpiece side inspection module, a side light source, and a six-axis robotic arm integrated in the main frame. It also includes a control computer and a three-color light that are matched with the main frame.

[0019] The main frame is equipped with a baffle on the inner side and control buttons on the surface;

[0020] The testing station module includes an electric rotary table, a fixture, a workpiece guide block, a rotary platform drive motor, and a power interface. The rotary platform drive motor is located below the electric rotary table, and the fixture and workpiece guide block are located above the electric rotary table.

[0021] The synchronous belt slide vision module includes drive motor one, drive motor two, foot, industrial camera one, synchronous belt guide rail, XY adapter, coupling, motor base, and drive motor three. Drive motor one, drive motor two, and drive motor three are fixed by the motor base, and industrial camera one is mounted on the synchronous belt guide rail by the XY adapter.

[0022] The workpiece side inspection module includes a lead screw guide rail drive motor, an industrial camera two, a Z-axis dovetail slide, a camera fixing component, an X-axis dovetail slide, a slide, a lead screw guide rail and an alloy plated handwheel. The industrial camera two is installed on the Z-axis dovetail slide through the camera fixing component, and the Z-axis dovetail slide is slidably connected to the X-axis dovetail slide.

[0023] The six-axis robotic arm is equipped with a ring light source at its end and a side light source is located inside the main frame;

[0024] The control computer is electrically connected to the inspection table module, the synchronous belt slide vision module, the workpiece side inspection module, the side light source, the six-axis robotic arm, the ring light source, and the three-color lamp.

[0025] As a further aspect of the present invention: the electric rotary table of the testing platform module can rotate 360° along the Z-axis, with a speed range of 0-10 rpm. The drive motor of the rotary platform is a servo motor with a power of 400W and a power interface of 220V AC.

[0026] As a further embodiment of the present invention: the drive motor 1 and drive motor 2 of the synchronous belt slide vision module are connected to the synchronous belt guide rail through a coupling, driving the industrial camera 1 to move along the X-axis (travel 0-800mm) and Y-axis (travel 0-600mm); drive motor 3 drives the industrial camera 1 to move along the Z-axis (travel 0-300mm), and the industrial camera 1 has a resolution of 16 million pixels.

[0027] As a further embodiment of the present invention: the lead screw guide rail drive motor of the workpiece side detection module drives the X-axis dovetail slide to move through the lead screw guide rail, thereby driving the second industrial camera to move along the X-axis (travel 0-500mm); the alloy plated handwheel is linked with the X-axis dovetail slide to realize the fine adjustment of the second industrial camera in the Y-axis direction, with a fine adjustment accuracy of 0.01mm and a resolution of 20 million pixels.

[0028] As a further aspect of the present invention: the side light source is an LED surface light source with an adjustable light intensity of 0-5000 lux; the ring light source is a flexible LED light source with an adjustable color temperature of 3000-6500K; the six-axis robotic arm has a repeatability accuracy of ≤0.05mm and can drive the ring light source to adjust the illumination angle from 0-360°.

[0029] As a further aspect of the present invention: the control computer has a built-in deep learning-based defect recognition algorithm, which is trained with 100,000+ defect samples and can identify defects such as corrosion spots, cracks, indentations, pores, foreign objects, and scratches; it also includes a data storage module that can automatically create folders according to "workpiece number / inspection date" to store the original images, processed images, and inspection results collected by industrial camera one and industrial camera two.

[0030] 7. The workpiece surface all-round intelligent defect detection system according to claim 1, characterized in that the three-color light has an audible and visual alarm function: when the detection is qualified, the green light is always on; when the detection is unqualified, the red light flashes and a buzzer sounds (volume 60-80dB); when the equipment is abnormal, the yellow light flashes; the control buttons include a start button, a pause button, and an emergency stop button.

[0031] 8. The workpiece surface all-round intelligent defect detection system according to claim 1, characterized in that the clamping method of the fixture is one of mechanical clamping, pneumatic clamping, vacuum adsorption or electromagnetic clamping, and the clamping force is adjustable from 50 to 500 N; the workpiece guide block positioning deviation is ≤0.1 mm to ensure that the workpiece is placed in a consistent position.

[0032] A method for omnidirectional intelligent defect detection on workpiece surface includes the following steps:

[0033] Step S1: Loading and positioning: The workpiece is placed on the electric rotary table by a person or a robot. The workpiece guide block is positioned in the center. The fixture clamps the workpiece under the command of the control computer. The rotary platform drive motor drives the electric rotary table to rotate to the initial angle of 0°. The control computer detects the positioning status. If the workpiece is not clamped or there is no workpiece, the three-color yellow light alarm is triggered.

[0034] Step S2: Lighting adjustment: The control computer turns on the side light source and adjusts the light intensity to 3000 lux; the six-axis robotic arm moves the ring light source to 100mm above the workpiece, and adjusts the angle (30-45°) and brightness (3500-4000 lux) of the ring light source according to the initial image captured by the industrial camera to eliminate shadows and reflections;

[0035] Step S3: Image Acquisition: The drive motors 1 and 2 of the synchronous belt slide vision module move the industrial camera 1 along the X and Y axes, acquiring one image of the top surface of the workpiece every 20mm, for a total of 30-40 images; every time the electric rotary table rotates 90°, the lead screw guide motor of the workpiece side detection module moves the industrial camera 2 along the X axis, acquiring 20 side images; the drive motor 3 adjusts the Z-axis height of the industrial camera 1, acquiring 10-15 images of the inclined surface;

[0036] Step S4: Defect Identification: The computer performs Gaussian filtering for noise reduction and histogram equalization on the image. The workpiece area is segmented using a U-Net network. Defect features are extracted based on a ResNet50 network. The defect type is identified and the result is determined as qualified or unqualified (crack length > 0.5 mm and pore diameter ≥ 0.3 mm are considered unqualified).

[0037] Step S5: Result Feedback and Archiving: The control computer transmits the inspection results to the display module. If the result is unqualified, a red alarm light is triggered. The data storage module creates a folder named "Workpiece Number_Inspection Date" to archive all images and inspection results (including workpiece number, inspection time, and defect type).

[0038] Step S6: System Reset: The fixture is released, the electric rotary table is reset to 0°, industrial camera one and industrial camera two return to their initial positions, the side light source and ring light source are turned off, and the system enters standby mode.

[0039] As a further aspect of the present invention: In step S3, industrial camera one and industrial camera two acquire images at a frequency of 10 frames / second, and the electric rotary table rotates at a speed of 5 rpm, ensuring that one image is acquired every 20° rotation of the workpiece; In step S4, the defect identification accuracy is ≥98%, and the single workpiece inspection cycle is ≤30 seconds; In step S2, the control computer monitors the image brightness in real time, and automatically adjusts the brightness of the side light source and the ring light source when the brightness deviation exceeds ±5%; In step S5, the inspection results are synchronously uploaded to the cloud for backup, supporting retrieval by workpiece number, with a retrieval response time of ≤1 second.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] 1. Uniform illumination and high imaging quality: Through the synergy of side light source + six-axis robotic arm + ring light source, the shadows and reflections on the workpiece surface are eliminated, and the image brightness deviation is ≤5%, providing high-quality input for defect recognition and improving the recognition accuracy to ≥98%, which is more than 8 percentage points higher than the existing equipment (≤90%).

[0042] 2. Comprehensive viewing angle, no blind spots in inspection: The synchronous belt sliding table vision module and the workpiece side inspection module, together with the electric rotary table, can achieve full surface coverage of the top, side, inclined and edge of the workpiece, with an inspection coverage rate of ≥99%, which reduces the inspection blind area by 14 percentage points compared with the existing equipment (≤85%).

[0043] 3. Real-time detection and high efficiency: The entire process from image acquisition to result output is automated, with a single workpiece detection cycle of ≤30 seconds, which is 4 times faster than traditional manual (≥2 minutes) and 1 times faster than existing machine vision equipment (≥1 minute), making it suitable for the high-speed production line requirement of "≥2 pieces per minute".

[0044] 4. Standardized data and strong traceability: Data is automatically categorized and stored according to "workpiece number / inspection date", and supports quick retrieval by workpiece number (response time ≤ 1 second). Cloud backup ensures data security and solves the problems of scattered data and difficulty in traceability of existing equipment.

[0045] 5. Wide adaptability and strong practicality: The clamping method and light source parameters of the fixture can be adjusted according to the workpiece material (metal, plastic) and size (50-500mm). There is no need to replace the core components. It can cover more than 80% of industrial workpiece inspection scenarios and reduce the equipment investment cost of enterprises. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the structure of the present invention;

[0047] Figure 2 This is a schematic diagram of the workpiece side detection module in this invention;

[0048] Figure 3 This is a schematic diagram of the detection stage module in this invention;

[0049] Figure 4 This is a schematic diagram of the two synchronous belt slide vision modules in this invention.

[0050] In the diagram: 1. Main frame; 2. Synchronous belt slide vision module; 201. Drive motor one; 202. Drive motor two; 203. Foot; 204. Industrial camera one; 205. Synchronous belt guide rail; 206. XY adapter; 207. Coupling; 208. Motor mount; 209. Drive motor three; 210. Drive motor four; 3. Workpiece side inspection module; 301. Lead screw guide rail drive motor; 302. Industrial camera two; 303. Z-axis dovetail. 304. Slide table; 305. Camera mounting bracket; 306. X-axis dovetail slide table; 307. Slide table; 308. Lead screw guide rail; 4. Alloy-plated handwheel; 4. Inspection table module; 401. Electric rotary table; 402. Fixture; 403. Workpiece guide block; 404. Rotary platform drive motor; 405. Power interface; 5. Side light source; 6. Six-axis robotic arm; 7. Ring light source; 8. Control computer; 9. Three-color lamp; 10. Baffle; 11. Control button. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] Please see Figures 1-4 In this embodiment of the invention, a comprehensive intelligent defect detection system and method for workpiece surfaces adopts an integrated technical solution of "multi-module collaboration + intelligent algorithm". Through the deep linkage of the detection table module, vision acquisition module, illumination module and control module, fully automatic detection of workpiece surface defects is achieved, as detailed below:

[0053] All-round intelligent defect detection system for workpiece surface

[0054] Main Frame 1: Overall Support and Protection Unit

[0055] The machine uses an aluminum profile frame with dimensions of 1500×1200×1800mm. An inner baffle 10 forms a closed detection space to avoid interference from external light. Three control buttons 11 are set on the surface: start button, pause button, and emergency stop button for operation control. Unmarked feet are provided at the bottom to ensure the stability of the frame.

[0056] Inspection Platform Module 4: Workpiece Positioning and Attitude Adjustment Unit

[0057] Electric rotary table 401: a loading platform with a diameter of 500mm, which can rotate 360° along the Z-axis and has an adjustable speed of 0-10rpm. It is used to drive the workpiece to adjust the inspection posture.

[0058] Fixture 402: Located above the electric rotary table 401, it adopts pneumatic clamping method and can be mechanical or vacuum adsorption, with a clamping force of 200N, to prevent workpiece rotation or displacement during shooting;

[0059] Workpiece guide blocks 403: Two are symmetrically arranged on the edge of the electric rotary table 401, with a positioning deviation of ≤0.1mm, to ensure that the workpiece is placed in the same position every time;

[0060] Rotary platform drive motor 404: servo motor, power 400W, located below the electric rotary table 401, drives the rotary table to rotate;

[0061] Power interface 405: 220V AC interface, which supplies power to the rotary platform drive motor 404.

[0062] Visual acquisition module: Full surface image acquisition unit

[0063] It includes a synchronous belt slide vision module 2 and a workpiece side inspection module 3, which are responsible for top / sloping surface and side imaging, respectively:

[0064] Synchronous Belt Slide Vision Module 2:

[0065] Drive motor 1 201 and drive motor 2 202: stepper motors with a speed of 2000 rpm, connected to the synchronous belt guide rail 205 via coupling 207, driving industrial camera 1 204 to move along the X-axis (0-800mm) and Y-axis (0-600mm);

[0066] Drive motor 3209: Stepper motor, speed 2000rpm, drives industrial camera 1204 to move along the Z-axis 0-300mm;

[0067] Industrial camera 204: 16 megapixel resolution, 10 frames per second frame rate, mounted on synchronous belt guide rail 205 via XY adapter 206, to acquire images of the top surface and inclined surface of the workpiece;

[0068] Foot 203: Four foot feet are located at the bottom of the module to secure it.

[0069] Workpiece side inspection module 3:

[0070] Lead screw guide drive motor 301: Stepper motor, speed 1500rpm, drives the X-axis dovetail slide 305 to move through lead screw guide 307;

[0071] Industrial camera 2 302: 20 million pixels resolution, 10 frames per second frame rate, mounted on the Z-axis dovetail slide 303 via camera fixture 304, to acquire side images of the workpiece;

[0072] Z-axis dovetail slide 303: Adjusts the height of the Z-axis of industrial camera 2 302, travel 0-200mm;

[0073] Alloy-plated handwheel 308: Linked with the X-axis dovetail slide 305, it enables fine adjustment of the Y-axis of the industrial camera 302 with an accuracy of 0.01mm, ensuring accurate focus.

[0074] Lighting Module: Adaptive Uniform Illumination Unit

[0075] Side light source 5: LED surface light source, located inside the main rack 1, with adjustable light intensity from 0-5000 lux, providing basic uniform illumination;

[0076] The six-axis robotic arm 6 has a repeatability accuracy of ≤0.05mm and is equipped with a ring light source 7 at the end, which can move the light source to any shaded area of ​​the workpiece.

[0077] Ring Light Source 7: Flexible LED light source, color temperature adjustable from 3000-6500K, brightness adjustable from 0-4000 lux, providing ring-shaped fill light to eliminate shadows and reflections.

[0078] Control and Feedback Module: Core Control and Intelligent Analysis Unit

[0079] Control Computer 8: Industrial control computer, CPU i7-12700, 32GB memory, built-in deep learning defect recognition algorithm and data storage module; the algorithm is trained with 100,000+ defect samples and can identify more than 6 types of defects; the data storage module uses a 1TB hard drive + cloud backup, and automatically classifies and stores data according to "workpiece number / inspection date";

[0080] Three-color light 9: Located on the top of main rack 1, the green light is always on when it is qualified, the red light flashes and a buzzer sounds (volume 70dB) when it is unqualified, and the yellow light flashes when there is an abnormality, so as to realize status feedback.

[0081] (II) Comprehensive Intelligent Defect Detection Method for Workpiece Surface

[0082] Based on the above system, the detection method includes the following steps:

[0083] S1: Loading and positioning (time ≤ 5 seconds)

[0084] The workpiece (such as a 200×150×100mm metal casting) is placed on the electric rotary table 401 by a person or a robot. The workpiece guide block 403 automatically centers and positions the workpiece. The control computer 8 sends a command, and the clamp 402 pneumatically clamps the workpiece. The rotary platform drive motor 404 drives the electric rotary table 401 to rotate to the initial angle of 0°. The control computer 8 detects the positioning status through sensors. If the workpiece is not clamped or there is no workpiece, the three-color light 9 yellow light alarm is triggered, and the process is paused.

[0085] S2: Lighting adjustment (time ≤ 3 seconds)

[0086] The control computer 8 turns on the side light source 5 and adjusts the light intensity to 3000 lux; the six-axis robotic arm 6 moves the ring light source 7 to 120 mm above the workpiece, and the industrial camera 2 302 acquires the initial image and transmits it to the control computer 8; the computer analyzes the image brightness, and if there are shadows or highlights, it adjusts the angle of the ring light source 7 to 35° and the brightness to 3800 lux until the uniformity deviation of the illumination is ≤5%.

[0087] S3: Image acquisition (time ≤ 15 seconds)

[0088] Top surface imaging: The drive motor 201 and drive motor 202 of the synchronous belt slide vision module 2 drive the industrial camera 204 to move along the X and Y axes, and acquire one top surface image every 20mm, for a total of 35 images;

[0089] Side imaging: The electric rotary table 401 rotates 90° under the drive of the rotary platform drive motor 404. The lead screw guide motor 301 of the workpiece side detection module 3 drives the industrial camera 2 302 to move along the X-axis and acquire 20 first side images. This action is repeated 3 times to complete 4 side imaging.

[0090] Inclined plane imaging: Drive motor 3209 drives industrial camera 1204 to move along the Z-axis to the height of the inclined plane of the workpiece, and acquire 12 inclined plane images;

[0091] All images are transmitted to the control computer 8 in real time via gigabit Ethernet, with a transmission delay of ≤100ms.

[0092] S4: Defect identification (time ≤ 5 seconds)

[0093] The computer 8 controls the image preprocessing: noise reduction through 3×3 Gaussian filtering and histogram equalization to enhance contrast; the workpiece area is segmented using a U-Net network to eliminate background interference; defect features are extracted based on a ResNet50 network, compared with a sample library, defect types (such as porosity and cracks) are identified and their location coordinates are marked; the workpiece is judged to be unqualified according to a preset threshold (crack length > 0.5 mm, porosity diameter ≥ 0.3 mm), otherwise it is judged to be qualified.

[0094] S5: Results Feedback and Archiving (Time ≤ 2 seconds)

[0095] The control computer 8 transmits the inspection results to the 10.1-inch touch display module. A "√" and a defect-free image are displayed for qualified products, while a red box marks the defect location for unqualified products. When a product is unqualified, the three-color light 9 flashes red and a buzzer sounds. The data storage module automatically creates a folder named "ZT20240501_001_20240520" (ZT20240501 is the workpiece number, and 20240520 is the inspection date), archives all images and inspection results (including workpiece number, inspection time, and defect type), and synchronizes them to the cloud for backup.

[0096] S6: System reset (time ≤ 2 seconds)

[0097] The fixture 402 is released, and the electric rotary table 401 is reset to 0°; the industrial camera 204 and the industrial camera 302 return to their initial positions under the drive of their respective motors; the side light source 5 and the ring light source 7 are turned off; the display module switches to the standby interface, waiting for the next workpiece to be inspected.

[0098] Example: Surface Defect Detection of Metal Castings

[0099] System parameter configuration

[0100] Workpiece specifications: Gray cast iron casting, dimensions 200×150×100mm; common defects: porosity (diameter ≥0.3mm), cracks (length ≥0.5mm), sand holes;

[0101] Inspection module 4: Electric rotary table 401 with a rotation speed of 5 rpm, clamp 402 with pneumatic clamping (200N), workpiece guide block 403 with a positioning deviation of 0.08 mm;

[0102] Visual acquisition module: Industrial camera 1 204 with a resolution of 16 million pixels, Industrial camera 2 302 with a resolution of 20 million pixels, acquisition frequency of 10 frames / second.

[0103] Lighting module: Side light source 5 with a brightness of 3000 lux, Ring light source 7 with a color temperature of 5000K and a brightness of 3800 lux, Six-axis robotic arm 6 with a positioning accuracy of 0.05mm;

[0104] Control computer 8: CPU i7-12700, memory 32GB, defect recognition algorithm based on ResNet50 training (sample library contains 120,000+ defect images).

[0105] Implementation process

[0106] S1: Loading and positioning: The robotic arm places the casting on the electric rotary table 401, the workpiece guide block 403 is centered and positioned, and the fixture 402 clamps it; the rotary platform drive motor 404 drives the rotary table 401 to rotate to 0°, the control computer 8 detects that the status is normal, and proceeds to the next step.

[0107] S2: Lighting adjustment: Side light source 5 is turned on (3000 lux), and the six-axis robotic arm 6 drives the ring light source 7 to move 120 mm above the casting, adjust the angle to 35° and the brightness to 3800 lux, and eliminate the top surface reflection;

[0108] S3: Image acquisition: Synchronous belt slide vision module 2 acquires 35 top surface images; electric rotary table 401 rotates 90°, 180°, 270°, and 360°, and workpiece side detection module 3 acquires 20 images of each of the four sides; drive motor 3 209 adjusts the Z-axis height of industrial camera 1 204 and acquires 12 inclined surface images.

[0109] S4: Defect identification: The control computer processes the image and identifies two 0.4mm diameter pores, which are judged as unqualified.

[0110] S5: Result Feedback: The display module marks the location of the vents with a red box, and the three-color light 9 red light alarms; the data storage module creates a folder "ZT20240501_001_20240520" to archive all images and detection results;

[0111] S6: System Reset: Fixture 402 is released, rotary table 401 is reset, camera returns to initial position, system goes into standby mode.

[0112] Implementation effect

[0113] Test metrics This invention Existing machine vision equipment Traditional manual Detection coverage (%) 99.5 82 75 Defect identification accuracy (%) 98.2 89 80 Single workpiece inspection cycle (seconds) 28 65 150 Data retrieval response time (seconds) 0.8 30 (Manual search) 120 (manual search) Defective product alarm delay (seconds) 0.5 300 (Offline Analysis) 60

[0114] in conclusion

[0115] In the inspection of metal castings, this invention achieves full-surface, blind-angle inspection, real-time accurate identification, and standardized data management through multi-module collaboration. The inspection efficiency is 1.3 times higher than that of existing equipment, and the accuracy is increased by 9.2 percentage points. It fully meets the high-speed and high-quality inspection requirements of production lines and can be extended to surface defect inspection scenarios for workpieces made of other materials such as plastic and alloy parts.

[0116] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A comprehensive intelligent defect detection system for workpiece surfaces, characterized in that, It includes a main frame (1) and an inspection table module (4), a synchronous belt slide vision module (2), a workpiece side inspection module (3), a side light source (5), a six-axis robotic arm (6) integrated in the main frame (1), as well as a control computer (8) and a tri-color light (9) that are matched with the main frame (1). The main frame (1) has a baffle (10) on its inner side and a control button (11) on its surface. The testing station module (4) includes an electric rotary table (401), a fixture (402), a workpiece guide block (403), a rotary platform drive motor (404), and a power interface (405). The rotary platform drive motor (404) is located below the electric rotary table (401), and the fixture (402) and the workpiece guide block (403) are located above the electric rotary table (401). The synchronous belt slide vision module (2) includes a drive motor 1 (201), a drive motor 2 (202), a foot (203), an industrial camera 1 (204), a synchronous belt guide rail (205), an XY adapter (206), a coupling (207), a motor mount (208), and a drive motor 3 (209). The drive motor 1 (201), drive motor 2 (202), and drive motor 3 (209) are fixed by the motor mount (208), and the industrial camera 1 (204) is mounted on the synchronous belt guide rail (205) by the XY adapter (206). The workpiece side inspection module (3) includes a lead screw guide rail drive motor (301), an industrial camera two (302), a Z-axis dovetail slide (303), a camera fixing component (304), an X-axis dovetail slide (305), a slide (306), a lead screw guide rail (307), and an alloy plated handwheel (308). The industrial camera two (302) is mounted on the Z-axis dovetail slide (303) through the camera fixing component (304). The Z-axis dovetail slide (303) and the X-axis dovetail slide (305) are slidably connected. The six-axis robotic arm (6) is equipped with a ring light source (7) at its end and a side light source (5) is located inside the main frame (1). The control computer (8) is electrically connected to the detection table module (4), the synchronous belt slide vision module (2), the workpiece side detection module (3), the side light source (5), the six-axis robotic arm (6), the ring light source (7), and the tri-color lamp (9).

2. The omnidirectional intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The electric rotary table (401) of the testing platform module (4) can rotate 360° along the Z-axis with a speed range of 0-10 rpm. The rotary platform drive motor (404) is a servo motor with a power of 400W, and the power interface (405) is a 220V AC interface.

3. The omnidirectional intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The drive motor 1 (201) and drive motor 2 (202) of the synchronous belt slide vision module (2) are connected to the synchronous belt guide rail (205) through the coupling (207), driving the industrial camera 1 (204) to move along the X-axis (travel 0-800mm) and Y-axis (travel 0-600mm); drive motor 3 (209) drives the industrial camera 1 (204) to move along the Z-axis (travel 0-300mm), and the industrial camera 1 (204) has a resolution of 16 million pixels.

4. The omnidirectional intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The lead screw guide motor (301) of the workpiece side detection module (3) drives the X-axis dovetail slide (305) to move through the lead screw guide (307), thereby driving the industrial camera (302) to move along the X-axis (stroke 0-500mm); the alloy plated handwheel (308) is linked with the X-axis dovetail slide (305) to realize the fine adjustment of the industrial camera (302) in the Y-axis direction, with a fine adjustment accuracy of 0.01mm and a resolution of 20 million pixels.

5. The omnidirectional intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The side light source (5) is an LED surface light source with an adjustable light intensity of 0-5000 lux; the ring light source (7) is a flexible LED light source with an adjustable color temperature of 3000-6500K; the six-axis robotic arm (6) has a repeatability accuracy of ≤0.05mm and can drive the ring light source (7) to adjust the irradiation angle from 0 to 360°.

6. The omnidirectional intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The control computer (8) has a built-in deep learning-based defect recognition algorithm. The algorithm is trained with 100,000+ defect samples and can identify defects such as corrosion spots, cracks, indentations, pores, foreign objects, and scratches. It also includes a data storage module that can automatically create folders according to "workpiece number / inspection date" to store the original images, processed images, and inspection results collected by industrial camera one (204) and industrial camera two (302).

7. The all-around intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The three-color light (9) has an audible and visual alarm function: when the test is qualified, the green light is always on; when the test is unqualified, the red light flashes and a buzzer sounds (volume 60-80dB); when the equipment is abnormal, the yellow light flashes. The control button (11) includes a start button, a pause button, and an emergency stop button.

8. The omnidirectional intelligent defect detection system for workpiece surfaces according to claim 1, characterized in that, The clamping method of the fixture (402) is one of mechanical clamping, pneumatic clamping, vacuum adsorption or electromagnetic clamping, and the clamping force is adjustable from 50 to 500 N; the positioning deviation of the workpiece guide block (403) is ≤0.1 mm to ensure that the workpiece is placed in a consistent position.

9. A method for omnidirectional intelligent defect detection of workpiece surface based on the system described in any one of claims 1-8, characterized in that, Includes the following steps: Step S1: Loading and positioning: The workpiece is placed on the electric rotary table (401) by a person or a robot. The workpiece guide block (403) is positioned in the center. The fixture (402) clamps the workpiece under the instruction of the control computer (8). The rotary platform drive motor (404) drives the electric rotary table (401) to rotate to the initial angle of 0°. The control computer (8) detects the positioning status. If the workpiece is not clamped or there is no workpiece, the three-color light (9) yellow light alarm is triggered. Step S2: Lighting adjustment: The control computer (8) turns on the side light source (5) and adjusts the light intensity to 3000 lux; the six-axis robotic arm (6) moves the ring light source (7) to 100 mm above the workpiece, and adjusts the angle (30-45°) and brightness (3500-4000 lux) of the ring light source (7) according to the initial image collected by the industrial camera (302) to eliminate shadows and reflections; Step S3: Image acquisition: The drive motor 1 (201) and drive motor 2 (202) of the synchronous belt slide vision module (2) drive the industrial camera 1 (204) to move along the X-axis and Y-axis, acquiring one image of the top surface of the workpiece every 20mm, for a total of 30-40 images; every time the electric rotary table (401) rotates 90°, the lead screw guide motor (301) of the workpiece side detection module (3) drives the industrial camera 2 (302) to move along the X-axis, acquiring 20 side images; drive motor 3 (209) adjusts the Z-axis height of the industrial camera 1 (204), acquiring 10-15 images of the inclined plane; Step S4: Defect identification: The control computer (8) performs Gaussian filtering for noise reduction and histogram equalization enhancement on the image, segments the workpiece area through the U-Net network, extracts defect features based on the ResNet50 network, identifies the defect type and determines whether it is qualified or unqualified (crack length > 0.5 mm and pore diameter ≥ 0.3 mm are determined to be unqualified). Step S5: Result Feedback and Archiving: The control computer (8) transmits the test results to the display module. If the result is not qualified, the three-color light (9) will trigger a red alarm. The data storage module creates a folder named "Workpiece Number_Test Date" to archive all images and test results. Step S6: System Reset: The clamp (402) is released, the electric rotary table (401) is reset to 0°, the industrial camera one (204) and the industrial camera two (302) return to their initial positions, the side light source (5) and the ring light source (7) are turned off, and the system enters standby mode.

10. The omnidirectional intelligent defect detection method for workpiece surface according to claim 9, characterized in that, In step S3, industrial camera 1 (204) and industrial camera 2 (302) acquire images at a frequency of 10 frames / second, and the electric rotary table (401) rotates at a speed of 5 rpm, ensuring that one image is acquired every 20° rotation of the workpiece; in step S4, the defect identification accuracy is ≥98%, and the single workpiece detection cycle is ≤30 seconds; in step S2, the control computer (8) monitors the image brightness in real time, and automatically adjusts the brightness of the side light source (5) and the ring light source (7) when the brightness deviation exceeds ±5%; in step S5, the detection results are synchronously uploaded to the cloud for backup, and can be retrieved by workpiece number with a retrieval response time of ≤1 second.

Images