Gear detection device and detection method

Abstract

The invention relates to the technical field of automatic detection, in particular to a gear detection device and method, and the method comprises the following steps: S1, adjusting a gear to an initial position through a servo motor; s2, synchronously acquiring a gear image and laser ranging data; s3, performing neural network training based on the acquired gear image, and outputting a defect classification result and position coordinates in real time; integrating a digital twinning technology to simulate stress distribution of the gear under different temperature and humidity conditions, and outputting predicted maximum deformation and fatigue life; and S4, the gear qualification condition is judged, a gear detection report is generated, and if the gear is not qualified, early warning is carried out. The method has the beneficial effects that the detection period of gear detection can be shortened, the defect classification result and the position coordinates can be output in real time, the maximum deformation and the fatigue life can be predicted, and the efficiency and the accuracy of gear detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automated detection, and particularly to a gear detection device and a detection method. Background Art

[0002] With the development of modern industry towards high precision, high efficiency, and high reliability, the manufacturing precision requirements for gears are getting higher and higher. High-precision gears can achieve smoother transmission, lower noise, and higher transmission efficiency. However, during the manufacturing process of gears, due to the influence of various factors, such as processing technology, equipment precision, material properties, etc., various errors and defects will inevitably occur, which will seriously affect the transmission performance and service life of gears.

[0003] In the prior art, gear detection technology mostly relies on manual visual inspection or contact measuring instruments. Manual visual inspection has large precision errors, and contact measuring instruments need to detect gears one by one, with low efficiency. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present application proposes a gear detection device and a detection method, which can shorten the detection cycle of gear detection, can output the defect classification results and position coordinates in real time, and predict the maximum deformation amount and fatigue life, improving the efficiency and accuracy of gear detection.

[0005] The following is the technical solution of the present invention. A gear detection method includes the following steps:

[0006] S1. Adjust the gear to the initial position through a servo motor;

[0007] S2. Synchronously collect gear images and laser ranging data;

[0008] S3. Based on the collected gear images, perform neural network training to output the defect classification results and position coordinates in real time; integrate digital twin technology to simulate the stress distribution of the gear under different temperature and humidity conditions, and output the predicted maximum deformation amount and fatigue life;

[0009] S4. Judge the qualification of the gear and generate a gear detection report. If it is unqualified, give an early warning.

[0010] As a preferred solution of the present invention, in S2, an adaptive Kalman filtering algorithm is adopted to dynamically adjust the data sampling frequency according to the gear speed. When the speed is less than or equal to 30 rpm, the sampling frequency is 100 Hz. When the speed is greater than 30 rpm, the sampling frequency linearly increases to 1000 Hz.

[0011] As a preferred solution of the present invention, in S3, based on the collected gear images, perform neural network training to output the defect classification results and position coordinates in real time, including the following steps:

[0012] S301. Data collection and annotation;

[0013] Collect more than or equal to 500 sample images for each type of defect, and annotate the defect location and category;

[0014] S302. Model training and optimization;

[0015] Adopt the MobileNetV3 lightweight network, with an input size of 640×640 pixels, select the FocalLoss as the loss function, and use transfer learning technology to load the pre-trained weights;

[0016] S303. System integration and verification;

[0017] Embed the trained model into the data processing module, link it with the laser ranging module through the API interface, and output the defect classification results and position coordinates in real time.

[0018] As a preferred solution of the present invention, in S3, integrate digital twin technology to simulate the stress distribution of the gear under different temperature and humidity conditions, and output the predicted maximum deformation amount and fatigue life, including the following steps:

[0019] S311. Use SolidWorks to build a 3D model based on the detection data;

[0020] S312. Synchronize the real-time detection data to the digital twin platform through the OPC UA protocol, and embed the finite element analysis module in the twin model;

[0021] S313. Input the temperature and humidity detection parameters, simulate the gear stress distribution, and output the maximum deformation amount and fatigue life.

[0022] As a preferred solution of the present invention, in S4, if there are defects in the gear, or any one of the data of the tooth pitch, tooth thickness, maximum deformation amount and fatigue life does not meet the gear standard data, the gear is unqualified.

[0023] A gear detection device, comprising:

[0024] An image acquisition module for acquiring gear images;

[0025] A gear placement module for placing the gear;

[0026] A motion control module for controlling the movement of the gear and connecting to the gear placement module;

[0027] A laser ranging module for measuring the tooth pitch and tooth thickness of the gear;

[0028] A data processing module for processing images and data analysis, connecting to the image acquisition module and the laser ranging module;

[0029] A communication module, for data transmission and remote monitoring, is integrated in the data processing module.

[0030] As a preferred solution of the present invention, the image acquisition module includes:

[0031] An industrial camera, for acquiring gear images;

[0032] An LED light source, for illumination to improve the imaging clarity of the industrial camera;

[0033] An ultraviolet light source, for exciting fluorescence to enhance the contrast;

[0034] A hyperspectral camera, for acquiring the reflection spectrum of the gear surface.

[0035] As a preferred solution of the present invention, the outer circumferential diameter of the gear turntable in the gear placement module is larger than the outer circumferential diameter of the gear, and several concentric positioning rings are arranged at the center of the gear turntable.

[0036] As a preferred solution of the present invention, the motion control module includes a control unit and a servo motor. The control unit controls the servo motor to drive the gear turntable to rotate, and the rotation speed range of the gear turntable is 0 - 100 rpm.

[0037] As a preferred solution of the present invention, the image acquisition module and the laser ranging module are connected to the data processing module through data lines, and the motion control module communicates with the data processing module through a servo driver.

[0038] The beneficial effects of the present invention are:

[0039] 1. In the present invention, compared with the traditional contact - type gear detection, the detection cycle is shortened to 30 seconds per piece, which can greatly improve the efficiency of gear detection;

[0040] 2. In the present invention, by comprehensively acquiring gear images, tooth pitch, tooth thickness and other data through the image acquisition module and the laser ranging module, the comprehensiveness of data acquisition can be improved. By outputting the defect classification results and position coordinates of the gear through the MobileNetV3 lightweight network, the accuracy of gear detection is improved;

[0041] 3. In the present invention, by integrating the digital twin technology to simulate the stress distribution of the gear under different temperature and humidity conditions, and outputting the predicted maximum deformation amount and fatigue life, the accuracy of gear detection is further improved. Description of the Drawings

[0042] Figure 1 It is a schematic diagram of the detection device;

[0043] Figure 2 It is a schematic diagram of the gear placement module;

[0044] Figure 3It is a diagram of the steps of the detection method;

[0045] Figure 4 It is a flowchart of the detection method;

[0046] In the figure: 1. Image acquisition module; 2. Gear placement module; 3. Motion control module; 4. Laser ranging module; 5. Data processing module; 6. Communication module; 7. Industrial camera; 8. LED light source; 9. Servo motor; 10. Laser sensor; 11. Ultraviolet light source; 12. Hyperspectral camera. Specific implementation manners

[0047] To make the technical problems solved by the present invention, the technical solutions adopted and the achieved technical effects clearer, the technical solutions of the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are some but not all of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0048] Embodiment 1:

[0049] As Figure 1 and Figure 2 shown, a gear detection device includes:

[0050] An image acquisition module 1 for acquiring gear images;

[0051] A gear placement module 2 for placing gears;

[0052] A motion control module 3 for controlling the movement of the gear and connected to the gear placement module 2;

[0053] A laser ranging module 4 for measuring the pitch and thickness of the gear teeth;

[0054] A data processing module 5 for processing images and data analysis, connected to the image acquisition module 1 and the laser ranging module 4;

[0055] A communication module 6 for data transmission and remote monitoring, integrated in the data processing module 5;

[0056] In this embodiment, the image acquisition module 1 includes a high-resolution industrial camera 7 and a ring-shaped LED light source 8. Among them, the resolution of the industrial camera 7 is ≥5 million pixels, which is used to acquire gear images, and the ring-shaped LED light source 8 is used for illumination to improve the clarity of the gear images acquired by the industrial camera 7.

[0057] In this embodiment, the gear placement module 2 is a gear turntable. Placing the gear on the gear turntable facilitates the detection of the gear. The outer diameter of the circumscribed circle of the gear turntable is larger than that of the gear, which facilitates the gear to be completely placed on the gear turntable. A number of concentric positioning rings are provided at the center of the gear turntable, which facilitates placing the gear at the center of the gear turntable.

[0058] In this embodiment, the motion control module 3 includes a control unit and a servo motor 9. The control unit controls the servo motor 9 to drive the gear turntable to rotate, thereby realizing the 360° rotation of the gear. The rotation speed range of the gear turntable is 0 - 100 rpm.

[0059] In this embodiment, the laser ranging module 4 uses a line array laser sensor 10. The pitch and thickness of the gear are measured by the laser sensor 10, and the accuracy of the laser sensor 10 is ±0.01 mm.

[0060] In this embodiment, the data processing module 5 integrates an embedded processor, with an image processing algorithm and a data analysis model built-in. The gear image is processed through the image processing algorithm, and the detection data of the gear is analyzed through the data analysis model.

[0061] In this embodiment, the communication module 6 supports Ethernet or RS485 protocol and is used for data transmission and remote monitoring.

[0062] In this embodiment, the image acquisition module 1 and the laser ranging module 4 are connected to the data processing module 5 through data lines; the motion control module 3 communicates with the data processing module 5 through a servo driver; the communication module 6 is integrated in the data processing module 5 to realize data interaction with the host computer.

[0063] Embodiment 2

[0064] As Figure 3 and Figure 4 shown, a gear detection method includes the following steps:

[0065] S1. Adjust the gear to the initial position through the servo motor 9;

[0066] S2. Synchronously collect the gear image and the laser ranging data;

[0067] S3. Based on the collected gear image, conduct neural network training, and real-time output the defect classification result and the position coordinates; integrate the digital twin technology to simulate the stress distribution of the gear under different temperature and humidity conditions, and output the predicted maximum deformation amount and fatigue life;

[0068] S4. Judge the qualification of the gear and generate a gear detection report. If it is unqualified, give an early warning.

[0069] In step S1, the gear is adjusted to the initial position by the servo motor 9. Specifically, the gear is positioned, and the gear to be measured is fixed at the center of the turntable and adjusted to the initial angle by the servo motor 9 for subsequent processing of the gear.

[0070] In step S2, the gear image and the laser ranging data are collected synchronously. Specifically, the annular LED light source 8 is started, and the industrial camera 7 collects gear images at a rate of 10 frames per second. The linear array laser sensor 10 is started to measure the pitch and thickness of the gear, and the gear image, pitch, and thickness data are transmitted to the data processing module 5.

[0071] In step S3, neural network training is performed based on the collected gear images, and the defect classification results and position coordinates are output in real time; the digital twin technology is integrated to simulate the stress distribution of the gear under different temperature and humidity conditions, and the predicted maximum deformation and fatigue life are output. Specifically, the tooth profile parameters are extracted by the image processing algorithm, and the image processing algorithm is based on OpenCV edge detection. The laser ranging module 4 synchronously obtains the pitch data, and the multi-sensor data is fused by the Kalman filtering algorithm.

[0072] Neural network training is performed based on the collected gear images, and the defect classification results and position coordinates are output in real time. A lightweight neural network, such as MobileNetV3, is introduced into the data processing module 5 to perform real-time defect classification and positioning on the gear images, supporting the automatic recognition of more than 20 defect types such as cracks and wear. Through transfer learning technology, it adapts to the detection requirements of different gear materials, such as steel and nylon, including the following steps:

[0073] S301. Data collection and annotation;

[0074] The industrial camera 7 with high resolution is used to collect gear images covering more than 20 defect types such as cracks and wear. For each type of defect, more than or equal to 500 sample images are collected, and the defect positions and categories are annotated by the image annotation tool. Combining with the optical measurement method, the surface topography data of the gear is synchronously obtained as the auxiliary training features.

[0075] S302. Model training and optimization;

[0076] The MobileNetV3 lightweight network is adopted, the input size is adjusted to 640×640 pixels, the Focal Loss function is selected to alleviate the problem of sample imbalance, and the pre-trained weights are loaded using transfer learning technology during training. By deploying the TensorRT acceleration engine, the model inference speed is less than or equal to 50 ms / frame.

[0077] S303. System integration and verification;

[0078] Embed the trained model into the data processing module 5, and link it with the laser ranging module 4 through the API interface to output the defect classification results and position coordinates in real time; during the verification phase, test 1000 gear samples, and the defect recognition accuracy reaches 98.5%.

[0079] Perform edge computing and adaptive filtering optimization, dynamically adjust the data sampling frequency according to the gear speed, and reduce the measurement error caused by vibration interference. Integrate an edge computing unit such as NVIDIA Jetson Nano in the laser ranging module 4, adopt the adaptive Kalman filtering algorithm, dynamically adjust the data sampling frequency according to the gear speed, and eliminate the measurement error caused by vibration interference. Support data caching and compensation calculation in the offline mode to ensure high-precision detection can still be completed during communication interruption.

[0080] Integrate the NVIDIA Jetson Nano edge computing unit in the laser ranging module 4, directly connect it to the laser sensor 10 through the PCIe interface to reduce data transmission latency. Based on the adaptive Kalman filtering algorithm, dynamically adjust the laser sampling frequency according to the speed of the servo motor 9. When the speed is less than or equal to 30 rpm, the sampling frequency is 100 Hz. When the speed is greater than 30 rpm, the sampling frequency linearly increases to 1000 Hz. And enable local caching in the offline mode. The storage capacity of the local cache is greater than or equal to 128 GB, and the data during communication interruption is compensated through timestamp alignment technology. Comparative experiments show that the measurement error caused by vibration interference is reduced from ±0.05 mm to ±0.01 mm, improving the precision requirements.

[0081] Integrate digital twin technology to simulate the stress distribution of gears under different temperature and humidity conditions, and output the predicted maximum deformation and fatigue life. Based on the detection data such as gear images, pitch, and tooth thickness, use SolidWorks to build a 3D model of the gear, synchronize the real-time detection data to the digital twin platform through the OPC UA protocol, embed a finite element analysis module in the twin model to simulate the stress distribution of gears under different temperature and humidity conditions, and output the maximum deformation and fatigue life. Provide a Web-based interactive interface that allows users to remotely modify detection parameters, such as light source intensity, turntable speed, etc., and preview the detection effect after parameter adjustment through the digital twin model, reducing the number of physical device debugging times.

[0082] In a possible embodiment, the surface of the gear is coated with a nano-material coating, and the nano-material coating is detected. An ultraviolet light source 11 and a hyperspectral camera 12 are added to the image acquisition module 1. The wavelength of the ultraviolet light source 11 is 365 nm, and the wavelength range of the hyperspectral camera 12 is 400 - 1000 nm. The reflection spectrum in the 400 - 1000 nm band of the gear surface is collected to generate a heat map of the coating thickness distribution, which is used to detect the uniformity and thickness of the nano-coating on the gear surface. A regression model of the coating thickness and the gear wear rate is established, and the coating performance is correlated with the gear life prediction model.

[0083] In step S4, the qualification status of the gear is judged and a gear detection report is generated. If it is unqualified, a warning is given. Specifically, data such as the pitch and tooth thickness in S3 are obtained, and the defect classification result, defect position coordinates, maximum deformation amount and fatigue life at the target temperature are organized into a detection report and saved in the form of a PDF document for subsequent reference. If a defect is detected in the gear, or any one of the data of the pitch, tooth thickness, maximum deformation amount and fatigue life does not meet the gear standard, a warning is given, and the warning methods are horn broadcast and information notification.

[0084] In the present invention, the rotational movement of the gear is controlled by the motion control module 3, and the gear image and data such as the pitch and tooth thickness are comprehensively collected by the image acquisition module 1 and the laser ranging module 4, which can improve the comprehensiveness of data collection; the defect classification result and position coordinates of the gear are output through the MobileNetV3 lightweight network, which can detect the defect classification result and position coordinates of the gear and improve the accuracy of gear detection; by integrating digital twin technology to simulate the stress distribution of the gear under different temperature and humidity conditions and output the predicted maximum deformation amount and fatigue life, the accuracy of gear detection is further improved.

[0085] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Obviously, those skilled in the art can make various changes and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the equivalent technology of the present invention, the present invention also intends to include these modifications and variations.

Claims

1. A gear detection method, characterized in that: The following steps are involved: S1, adjust the gear to the initial position through the servo motor; S2, synchronously collect gear images and laser ranging data; S3. Perform neural network training based on the collected gear images, and output defect classification results and position coordinates in real time; integrate digital twin technology to simulate the stress distribution of gears under different temperature and humidity conditions, and output the predicted maximum deformation and fatigue life; S4. Determine the gear quality and generate a gear inspection report. If it is unqualified, issue an early warning.

2. A gear detection method according to claim 1, characterized in that: In S2, an adaptive Kalman filter algorithm is used to dynamically adjust the data sampling frequency according to the gear speed. When the speed is less than or equal to 30rpm, the sampling frequency is 100Hz. When the speed is greater than 30rpm, the sampling frequency is linearly increased to 1000Hz.

3. A gear detection method according to claim 1, characterized in that: In S3, neural network training is performed based on the collected gear images, and defect classification results and position coordinates are output in real time, including the following steps: S301, data collection and annotation; Collect more than or equal to 500 sample images for each type of defect, and mark the defect location and category; S302, model training and optimization; The MobileNetV3 lightweight network is used, the input size is 640×640 pixels, the loss function is Focal Loss, and the transfer learning technology is used to load the pre-trained weights; S303, system integration and verification; The trained model is embedded in the data processing module, linked with the laser ranging module through the API interface, and the defect classification results and location coordinates are output in real time.

4. A gear detection method according to claim 1, characterized in that: In S3, the digital twin technology is integrated to simulate the stress distribution of gears under different temperature and humidity conditions, and output the predicted maximum deformation and fatigue life, including the following steps: S311, Use SolidWorks to build a 3D model based on inspection data; S312. Synchronize the real-time detection data to the digital twin platform through the OPC UA protocol and embed the finite element analysis module in the twin model; S313. Input temperature and humidity detection parameters, simulate gear stress distribution, and output maximum deformation and fatigue life.

5. A gear detection method according to claim 1, characterized in that: In S4, if there are defects in the gear, or any of the data of pitch, tooth thickness, maximum deformation and fatigue life does not meet the gear standard data, the gear is unqualified.

6. A gear detection device, applicable to a gear detection method according to any one of claims 1 to 5, characterized in that: include: An image acquisition module, used for acquiring gear images; Gear placement module, used to place gears; A motion control module, used to control the gear movement, connected to the gear placement module; Laser distance measurement module, used to measure gear pitch and tooth thickness; Data processing module, used for image processing and data analysis, connecting the image acquisition module and the laser ranging module; The communication module is used for data transmission and remote monitoring and is integrated into the data processing module.

7. A gear detection device according to claim 6, characterized in that: The image acquisition module includes: Industrial cameras for capturing images of gears; LED light source, used for lighting to improve the imaging clarity of industrial cameras; Ultraviolet light source, used to excite fluorescence and enhance contrast; Hyperspectral camera, used to collect the reflectance spectrum of the gear surface.

8. A gear detection device according to claim 6, characterized in that: The diameter of the circumscribed circle of the gear turntable in the gear placement module is larger than the diameter of the circumscribed circle of the gear, and a plurality of concentric positioning rings are arranged at the center of the gear turntable.

9. A gear detection device according to claim 8, characterized in that: The motion control module includes a control unit and a servo motor. The control unit controls the servo motor to drive the gear turntable to rotate. The speed range of the gear turntable is 0-100rpm.

10. A gear detection device according to claim 8, characterized in that: The image acquisition module and the laser distance measurement module are connected with the data processing module through a data line, and the motion control module communicates with the data processing module through a servo driver.

Images