A transfer equipment parts inspection and maintenance system

Abstract

The present invention relates to the technical field of parts testing and maintenance, and in particular to a parts maintenance system for transfer equipment. The present invention proposes the following scheme, including a parts data acquisition module, a parts data testing module and a parts maintenance module. The parts data acquisition module converts the optical signal of the transfer equipment parts into an image through an energy conversion strategy; the parts data testing module detects the image using image processing technology to obtain the defect state of the parts, and is equipped with a defect testing strategy of image preprocessing, feature extraction, Hough transform and deep learning network; the parts maintenance module generates a defect assessment report and maintenance suggestions based on the detection results. The system provides detailed defect assessment and maintenance suggestions, optimizes maintenance strategies and predicts future needs.

Description

Technical Field

[0001] The invention relates to the technical field of parts testing and repairing, and in particular to a parts repairing system for transfer equipment. Background Art

[0002] During the operation of transfer equipment, the loss of parts is one of the main causes of equipment failure and shutdown. In order to ensure the normal operation of the equipment, parts need to be inspected and repaired regularly, and workpiece loss detection is an indispensable management method to ensure product quality. However, traditional parts detection methods usually rely on manual inspection, which is inefficient, inaccurate, and difficult to monitor the status of parts in real time. Therefore, how to efficiently and accurately detect the loss of transfer equipment parts and provide scientific maintenance suggestions has become a technical problem that needs to be solved in this field. The use of image processing technology to test parts is efficient and timely, and is especially suitable for large-scale testing and maintenance methods.

[0003] For example, a Chinese patent with authorization announcement number CN116067655B provides a part testing device, a part testing equipment and a part testing method, which relates to the field of part testing technology, to solve the problem that the applied static load affects the dynamic load, thereby affecting the test results when the dynamic load is applied to the bearing. The part testing device includes: a base, a connecting shaft, a sleeve, a fixing and a test fastening assembly. The connecting shaft and the sleeve are both arranged on the base, and the connecting shaft is arranged through the sleeve. The connecting shaft and the inner wall of the sleeve are spaced apart to form a accommodating space, and the accommodating space is used to accommodate the parts. The fixing is located in the sleeve, the fixing is connected to the sleeve, and abuts against the part, and the part is located between the base and the fixing. The two ends of the test fastening assembly are respectively connected to the sleeve and the connecting shaft, and the test fastening assembly is used to adjust the distance between the sleeve and the connecting shaft to compress or release the part.

[0004] The above patents all have the problems raised by this background technology: the traditional method relies on manual inspection and judgment, which is inefficient and prone to errors. Currently, the inspection of transfer equipment parts requires shutdown, affecting daily production activities. To solve the above problems, this application designs a transfer equipment parts inspection and maintenance system. Summary of the invention

[0005] The technical problem to be solved by the present invention is to address the deficiencies of the prior art and provide a transfer equipment parts maintenance system, including a parts data acquisition module, a parts data testing module and a parts maintenance module. The parts data acquisition module converts the optical signal of the transfer equipment parts into an image through an energy conversion strategy. The parts data testing module uses image processing technology to detect images and obtain the defect status of the parts, and is equipped with a defect testing strategy of image preprocessing, feature extraction, Hough transform and deep learning network. The parts maintenance module generates a defect assessment report and maintenance suggestions based on the detection results. The system provides detailed defect assessments and maintenance suggestions, optimizes maintenance strategies and predicts future needs.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A transfer equipment parts maintenance system, the system comprising a parts data acquisition module, a parts data testing module and a parts maintenance module;

[0008] The part data acquisition module is provided with an energy conversion strategy, and the energy conversion strategy is used to convert the transfer equipment part optical signal into the transfer equipment part image;

[0009] The part data testing module is used to detect the image of the transfer equipment part through image processing technology to obtain the defect status of the transfer equipment part. The part data testing module is configured with a defect testing strategy. The defect testing strategy evaluates the part defect status through image preprocessing, feature extraction, Hough transform and part defect recognition network. The defect status includes part defect category and part defect rate;

[0010] The parts inspection and repair module is used to generate a parts defect assessment report and inspection and repair suggestions according to the defect status of the transfer equipment parts.

[0011] The part data testing module includes a part data processing unit and a part defect detection unit. The part data processing unit is used to pre-process the image of the transfer equipment part to generate a loss feature vector. The part defect detection unit is used to calculate the center loss feature according to the loss feature vector, and identify the defect status of the transfer equipment part according to the center loss feature. The defect testing strategy includes data processing logic and part detection logic. The data processing logic is configured in the part data processing unit, and the part detection logic is configured in the part defect detection unit.

[0012] The data processing logic includes:

[0013] De-noising the image of the parts of the transfer equipment. The image noise of the parts of the transfer equipment includes AD quantization noise, dark current noise and parts reflection noise;

[0014] Performing histogram statistics on the grayscale values ​​of the denoised transfer equipment part image, calculating an optimal segmentation threshold, and extracting a target area according to the optimal segmentation threshold;

[0015] Connecting the pixels of the target area, obtaining an image of the target area, extracting key points of the image of the target area, and calculating the reflection intensity of the key points;

[0016] The reflection intensity of the key point is compared with the feature threshold. If it is greater than or equal to the feature threshold, the key point is marked as a feature point, and the loss feature vector is calculated based on the feature point.

[0017] The part detection logic includes a feature extraction sub-logic and a defect detection sub-logic. The feature extraction sub-logic is used to process the loss feature vector through Hough transform to obtain the center loss feature. The defect detection sub-logic is used to build a part defect recognition network, use the center loss feature as the input parameter of the part defect recognition network, train the input parameter through the part defect recognition network, and output the defect status of the transfer equipment parts.

[0018] The feature extraction sub-logic includes:

[0019] Randomly select edge pixels from the edge pixel set of the loss feature vector and project them into the parameter space accumulator, and delete the selected edge pixels;

[0020] Determine whether the extreme value in the parameter space accumulator is greater than the accumulation threshold. If it is less than or equal to the accumulation threshold, put the edge pixel back into the edge pixel set and reselect the edge pixel.

[0021] If it is greater than the accumulation threshold, the peak values ​​formed by the extreme values ​​in the parameter space accumulator are concentrated to find the largest curve segment that is continuous or has a gap not greater than the gap threshold, and the largest curve segment is compared with the limited length;

[0022] If it is greater than or equal to the specified length, the remaining edge pixels in the edge pixel set are projected as the center loss feature. If it is less than the specified length, the edge pixels are reselected.

[0023] The part defect recognition network comprises:

[0024] An input layer, which establishes a feature sequence according to input parameters and calculates the fitting value of the feature sequence;

[0025] A hidden layer converts the fitting value of the feature sequence of the input layer into a high-dimensional space, and calculates the output value of the hidden layer in the high-dimensional space;

[0026] The output layer is used to output the defect status of the transfer equipment parts.

[0027] The step of training the input parameters through the part defect recognition network and outputting the defect status of the transfer equipment parts includes:

[0028] Initializing the part defect recognition network according to the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer of the part defect recognition network, and determining the connection weights and node thresholds of the input layer and the hidden layer;

[0029] The input parameters are biased by the connection weights of the input layer and the hidden layer to obtain the output of the hidden layer, and the output of the hidden layer is updated by the node threshold to obtain the output layer input;

[0030] Calculating the error of the output layer input according to the error function of the output layer, determining whether the error of the output layer input meets the convergence requirement of the output layer, and if not, updating the connection weight and the node threshold;

[0031] If the convergence requirement is met, the defect status of the transfer equipment parts is output, wherein the calculation formula of the part defect rate is as follows:

[0032] ,

[0033] in, Indicates the defect rate of parts of transfer equipment parts, represents the learning rate, u represents a single node in the hidden layer, U represents the total number of nodes in the hidden layer, v represents a single node in the output layer, and V represents the total number of nodes in the output layer. represents the output of the u-th node in the hidden layer, represents the connection weights between the hidden layer and the output layer nodes, represents the node threshold of the output layer, and E represents the error of the output layer input.

[0034] The part data acquisition module includes an optical lens unit, a pixel array unit and an image sensor readout circuit unit. The optical lens unit is used to store a CMOS image sensor. The pixel array unit is used to read the fluorescent light signal of the transfer equipment parts and convert it into photocurrent voltage information and part surface information. The image sensor readout circuit unit is used to current bias the source follower of the pixel, dynamically adjust the exposure according to the change of light in the scene, convert the voltage information into a digital code value, and output the transfer equipment part image. The energy conversion strategy includes photoelectric conversion logic and image generation logic. The photoelectric conversion logic is configured in the pixel array unit, and the image generation logic is configured in the image sensor readout circuit unit.

[0035] The photoelectric conversion logic configuration includes:

[0036] The CMOS image sensor is illuminated from the front by an LED to obtain the light radiation energy of the parts of the transfer equipment reaching the CMOS image sensor;

[0037] Photoelectric conversion is performed through diodes to generate photocurrent, and pixel gain is performed on the photocurrent through the pixel array to extract photons and part surface information;

[0038] The photons and part surface information are converted into voltage information and transmitted to the COMS image sensor readout circuit through the metal bus.

[0039] The parts maintenance module includes a defect assessment report generation unit, a maintenance suggestion unit, a data storage unit and a data analysis unit. The defect assessment report generation unit is used to generate a parts defect assessment report to describe the current defect status of the part. The maintenance suggestion unit is used to generate maintenance suggestions based on the parts defect assessment report. The maintenance suggestions include parts that need to be replaced, maintenance priorities and maintenance operation steps. The data storage unit is used to store the parts defect assessment report and the maintenance suggestions in a transfer equipment maintenance database. The data analysis unit is used to perform data analysis on the transfer equipment maintenance database, optimize maintenance strategies and predict future maintenance needs.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention integrates the influence of image processing technology on the process of parts loss analysis, improves the parts defect analysis technology, and the improved technology as a whole has the required real-time performance and easy implementation advantages, and improves the accuracy of defect detection.

[0042] 2. Through the deep learning network, the present invention can adapt to different types of transfer equipment parts, can quickly and accurately identify the defect status of transfer equipment parts, avoid errors in manual detection, and has universality and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0044] Figure 1 This is a module diagram of a transfer equipment parts maintenance system according to Embodiment 1 of the present invention;

[0045] Figure 2 This is a schematic diagram of calculating the reflection intensity of key points in Example 1 of the present invention;

[0046] Figure 3 This is a schematic diagram of a key point encoder according to Embodiment 1 of the present invention;

[0047] Figure 4This is a feature extraction flow chart of Example 1 of the present invention;

[0048] Figure 5 This is a diagram of a network structure for part defect recognition in Example 1 of the present invention;

[0049] Figure 6 This is a flow chart of a method for repairing parts of a transfer equipment according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0051] Embodiment 1:

[0052] See also Figure 1 , an embodiment provided by the present invention: a transfer equipment parts maintenance system, the system comprising a parts data acquisition module, a parts data testing module and a parts maintenance module;

[0053] The part data acquisition module is provided with an energy conversion strategy, and the energy conversion strategy is used to convert the optical signal of the transfer equipment part into the image of the transfer equipment part;

[0054] The part data testing module is used to detect the image of the transfer equipment part through image processing technology to obtain the defect status of the transfer equipment part. The part data testing module is configured with a defect testing strategy. The defect testing strategy evaluates the part defect status through image preprocessing, feature extraction, Hough transform and part defect recognition network. The defect status includes part defect category and part defect rate;

[0055] The parts inspection and repair module is used to generate a parts defect assessment report and inspection and repair suggestions according to the defect status of the transfer equipment parts.

[0056] The part data testing module includes a part data processing unit and a part defect detection unit. The part data processing unit is used to pre-process the image of the transfer equipment part to generate a loss feature vector. The part defect detection unit is used to calculate the center loss feature according to the loss feature vector, and identify the defect status of the transfer equipment part according to the center loss feature. The defect testing strategy includes data processing logic and part detection logic. The data processing logic is configured in the part data processing unit, and the part detection logic is configured in the part defect detection unit.

[0057] The data processing logic includes:

[0058] De-noising the image of the parts of the transfer equipment. The image noise of the parts of the transfer equipment includes AD quantization noise, dark current noise and parts reflection noise;

[0059] Perform histogram statistics on the grayscale values ​​of the denoised transfer equipment part image, use the peak-to-valley grayscale value between the two peaks as the segmentation threshold 1, calculate the segmentation threshold 2 according to the probability density function of the transfer equipment part image, obtain the optimal segmentation threshold by weighted summing the segmentation threshold 1 and the segmentation threshold 2, compare the pixel point of the transfer equipment part image with the optimal segmentation threshold, if it is greater than or equal to the optimal segmentation threshold, mark the pixel point as the target area, if it is less than the optimal segmentation threshold, mark the pixel point as the background area;

[0060] Connecting the pixels of the target area to obtain an image of the target area, and smoothing the image of the target area by edge detection, wherein the edge detection includes image Gaussian filtering, calculating the gradient amplitude and direction by differential method, performing non-maximum suppression on the gradient amplitude, and double threshold edge connection;

[0061] Calculate the Gaussian difference based on the smoothed target area image, perform multi-scale processing on the image, generate a series of images of different scales, calculate the difference image between adjacent scales, detect local extreme points in the Gaussian difference image through the SIFT algorithm, mark the local extreme points as the key points of the image, extract the key points of the target area image through edge response, divide the key points into four components of R, G, B and Deep for pixel decomposition, and calculate the reflection intensity of the key points;

[0062] See also Figure 2 , a schematic diagram of calculating the reflection intensity of key points in an embodiment of the present invention, the extracted key points are decomposed into four components of RGB and Deep, the RGB component represents the red, green and blue color channels, and the Deep component represents the depth information, and the response values ​​of the R, G, B and Deep channels are calculated respectively. According to the response values ​​of the four components, the reflection intensity of the key points is calculated by weighted average, and the calculation formula of the response value is:

[0063] ,

[0064] in, represents the response value of the kth component, n represents the single pixel position of the kth component, and N represents the total number of pixel positions of the kth component. represents the average intensity value of the kth component, Represents the intensity value corresponding to the nth pixel position in the kth component, represents the maximum intensity value in the kth component, Represents the norm-finding function.

[0065] The reflection intensity of the key point is compared with the feature threshold. If it is greater than or equal to the feature threshold, the key point is marked as a feature point and the feature point is described using the BRIEF descriptor. The BRIEF descriptor generates a binary feature vector by performing a simple comparison operation on the pixel pairs around the key point to describe the local features of the feature point. The feature point descriptor of the feature point is calculated and the feature point is converted into high-dimensional data using the MLP network. The MLP network consists of an input layer, a hidden layer, and an output layer. The feature point descriptor is converted into a higher-dimensional feature space through nonlinear mapping of the feature point descriptor and added to the feature point descriptor to generate a loss feature vector. This fusion method combines the local information of the feature point with the high-level features extracted by deep learning, improving the expressive power of the feature vector.

[0066] See also Figure 3 , a schematic diagram of a key point encoder of an embodiment of the present invention, which is used to increase the dimension of feature points, so that the feature points and feature point descriptors are added to form a feature vector, each feature point is combined with its coordinates and confidence, and a multi-layer perceptron is used to embed the feature points into a high-dimensional vector. First, a five-layer fully connected perceptron is used to embed the feature points into a high-dimensional vector. Converted into 256-dimensional data, the number of channels of the five-layer perceptron are 3, 32, 64, 128 and 256 respectively.

[0067] The part detection logic includes a feature extraction sub-logic and a defect detection sub-logic. The feature extraction sub-logic is used to process the loss feature vector through Hough transform to obtain the center loss feature. The defect detection sub-logic is used to build a part defect recognition network, use the center loss feature as the input parameter of the part defect recognition network, train the input parameter through the part defect recognition network, and output the defect status of the transfer equipment parts.

[0068] The feature extraction sub-logic includes:

[0069] Randomly select edge pixels from the edge pixel set of the loss feature vector and project them into the parameter space accumulator, and delete the selected edge pixels;

[0070] Calculate the extreme value in the parameter space accumulator, determine whether the extreme value in the parameter space accumulator is greater than the accumulation threshold, if it is less than or equal to the accumulation threshold, put the edge pixel back into the edge pixel set, and reselect the edge pixel;

[0071] If it is greater than the accumulation threshold, the peak values ​​formed by the extreme values ​​in the parameter space accumulator are concentrated to find the largest curve segment that is continuous or has a gap not greater than the gap threshold, and the largest curve segment is compared with the limited length;

[0072] If it is greater than or equal to the specified length, the remaining edge pixels in the edge pixel set are projected as the center loss feature. If it is less than the specified length, the edge pixels are reselected.

[0073] See also Figure 4 , the feature extraction flow chart of the embodiment of the present invention, randomly selects some edge pixels from the edge pixel set of the feature vector and projects them into the parameter space accumulator. After multiple selections and projections, the extreme points in the accumulator gradually form obvious peaks. After analyzing these peaks, the largest curve segment with a continuous gap not greater than the preset gap threshold is found. The curve segment is compared with the preset limited length, and it is found that its length is greater than the limited length, so the edge pixels in the curve segment are extracted as the center loss feature. The feature extraction includes the following four steps:

[0074] A1: Edge pixel selection and projection:

[0075] Randomly select edge pixels from the edge pixel set of the loss feature vector, and project the selected edge pixels into the parameter space accumulator. The parameter space accumulator is used to accumulate the projection results of edge pixels to form extreme points, and randomly select some pixels from the edge pixel set to ensure the diversity and randomness of the selection process;

[0076] A2: Accumulator extreme value judgment:

[0077] It is determined whether the extreme value in the parameter space accumulator is greater than the preset accumulation threshold. If the extreme value in the accumulator is less than or equal to the accumulation threshold, these edge pixels are put back into the edge pixel set, and new edge pixels are reselected for projection. The projection results of edge pixels are accumulated by the parameter space accumulator, which effectively enhances the aggregation effect of edge features.

[0078] A3: Extreme peak analysis:

[0079] If the extreme value in the parameter space accumulator is greater than the accumulation threshold, the peak value formed by these extreme values ​​is found in the accumulator, and the largest curve segment that is continuous or has a gap not greater than the preset gap threshold is found from the peak value. By analyzing the extreme value peak value in the accumulator, the continuous curve segment with significant features is extracted;

[0080] A4: Comparison of curve segment lengths:

[0081] The maximum curve segment found is compared with the preset limit length. If the length of the maximum curve segment is greater than or equal to the limit length, the remaining edge pixels are projected and extracted as the center loss feature. If the length of the maximum curve segment is less than the limit length, new edge pixels are reselected to perform the above steps until a curve segment that meets the conditions is found. Dynamic adjustments are made according to the length of the curve segment to ensure that the extracted center loss feature has high stability and accuracy.

[0082] The part defect recognition network comprises:

[0083] The input layer establishes a feature sequence based on the input parameters, processes the input parameters using a linear or nonlinear transformation function, calculates the fitting value of the feature sequence, and provides a preliminary feature representation for subsequent layers;

[0084] The hidden layer maps the feature sequence fitting values ​​of the input layer to a high-dimensional space through multiple neuron nodes. Each node applies an activation function to process the input signal. In the high-dimensional space, the hidden layer nodes calculate the output values ​​based on the input features. These output values ​​provide richer feature information and help distinguish different types of defects.

[0085] The output layer is used to output the defect status of the transfer equipment parts.

[0086] Through the multi-layer structure of input layer, hidden layer and output layer, feature information is extracted and converted layer by layer to improve feature expression ability and classification accuracy. The hidden layer converts the feature sequence into a high-dimensional space, which enhances the richness and discrimination of feature representation. Through error feedback and weight update during training, the network parameters are dynamically adjusted to ensure the accuracy and stability of the output results.

[0087] Initializing the part defect recognition network according to the number of nodes in the input layer, the number of nodes in the hidden layer, and the number of nodes in the output layer of the part defect recognition network, and determining the connection weights and node thresholds of the input layer and the hidden layer;

[0088] The input parameters are biased by the connection weights of the input layer and the hidden layer to obtain the output of the hidden layer, and the output of the hidden layer is updated by the node threshold to obtain the output layer input;

[0089] Calculating the error of the output layer input according to the error function of the output layer, determining whether the error of the output layer input meets the convergence requirement of the output layer, and if not, updating the connection weight and the node threshold;

[0090] If the convergence requirement is met, the defect status of the transfer equipment parts is output, wherein the defect categories of the transfer equipment parts include wear defects, corrosion defects, deformation defects and mixed defects, wherein the calculation formula of the part defect rate is as follows:

[0091] ,

[0092] in, Indicates the defect rate of parts of transfer equipment parts, represents the learning rate, u represents a single node in the hidden layer, U represents the total number of nodes in the hidden layer, v represents a single node in the output layer, and V represents the total number of nodes in the output layer. represents the output of the u-th node in the hidden layer, represents the connection weights between the hidden layer and the output layer nodes, represents the node threshold of the output layer, and E represents the error of the output layer input.

[0093] See also Figure 5 , the structure diagram of the part defect recognition network of the embodiment of the present invention, specifically, the part defect recognition network is provided with an input layer, a hidden layer and an output layer, the input layer is used to establish a feature sequence according to the input parameters of the part defect recognition network, accumulate the feature sequence, establish a grey differential equation, calculate the least squares parameter of the grey differential equation, and calculate the fitting value of the feature sequence by discretizing the least squares parameter solution. The hidden layer is used to transform the feature sequence fitting value of the input layer into a high-dimensional space by mapping according to the connection weight between the input layer and the hidden layer, sum the feature sequence fitting value in the high-dimensional space after nonlinear weighting, and calculate the output value of the hidden layer by bias. The output layer is used to obtain the output layer input, calculate the average error of the input value of the output layer, and judge whether the average error meets the convergence condition. If it does not meet the convergence requirement, the connection weight and the node threshold are updated. If it meets the convergence requirement, the defect state of the part is output.

[0094] The part data acquisition module includes an optical lens unit, a pixel array unit and an image sensor readout circuit unit. The optical lens unit is used to store a CMOS image sensor. The pixel array unit is used to read the fluorescent light signal of the transfer equipment parts and convert it into photocurrent voltage information and part surface information. The image sensor readout circuit unit is used to current bias the source follower of the pixel, dynamically adjust the exposure according to the change of light in the scene, convert the voltage information into a digital code value, and output the transfer equipment part image. The energy conversion strategy includes photoelectric conversion logic and image generation logic. The photoelectric conversion logic is configured in the pixel array unit, and the image generation logic is configured in the image sensor readout circuit unit.

[0095] The photoelectric conversion logic configuration includes:

[0096] The CMOS image sensor is illuminated from the front by an LED, which can stably provide uniform illumination, ensuring that the image sensor can receive sufficient light signals and obtain the light radiation energy of the transfer equipment parts reaching the CMOS image sensor. The calculation formula of the light radiation energy of the transfer equipment parts is:

[0097] ,

[0098] in, Indicates the light radiation energy of the transfer equipment parts reaching the CMOS image sensor, Indicates the LED lighting fluorescent light radiation energy, Indicates the fluorescent diffuse reflectance of the transfer equipment parts. represents the spectral transmittance of CMOS image sensor, Represents the diffuse reflectivity of the CMOS image sensor surface, Indicates the internal interference light radiation energy, It represents the cosine of the angle between the front-illuminated fluorescence and the CMOS image sensor.

[0099] Photoelectric conversion is performed through diodes to generate photocurrent, and pixel gain is performed on the photocurrent through the pixel array. The signal strength is improved through the gain amplification of each pixel to ensure that weak light signals can also be accurately detected and extract photon and part surface information. Photon information reflects the surface reflection characteristics of the part, while surface information includes the material, texture and loss characteristics of the part. The calculation formula for the photocurrent pixel gain coefficient is:

[0100] ,

[0101] in, G represents the photocurrent pixel gain coefficient, Indicates the maximum current of the pixel circuit, represents the photocurrent, Indicates the rated voltage of the pixel circuit, represents the photodiode built-in voltage, It represents the built-in voltage of the transistor. represents the pixel circuit fixed noise voltage, T represents the capacitance integration time, represents the photodiode capacitance, represents the transistor capacitance;

[0102] The photons and part surface information are converted into voltage information and transmitted to the COMS image sensor readout circuit through the metal bus. The readout circuit processes and digitally converts the voltage signal to generate digital image data that can be used for subsequent image processing.

[0103] The image generation logic includes current biasing the source follower of the pixel, dynamically adjusting the exposure according to the change of the light in the scene, amplifying the voltage information, converting the voltage information into a digital code value through an analog-to-electric converter, and receiving the digital signal output data by the image processor to obtain the image of the parts of the transfer equipment;

[0104] The CMOS image sensor readout circuit includes: an analog front-end circuit, an analog-to-digital converter and a digital signal output interface;

[0105] In a factory working environment, the on-site environment is relatively complex, and overexposure or underexposure problems are prone to occur in low-light conditions and high dynamic range scenarios. Photoelectric conversion and gain control can process optical signals with a higher dynamic range, ensure image quality, effectively enhance signals, and improve detection sensitivity.

[0106] The parts maintenance module plays a key role in the inspection and maintenance system of transfer equipment. It can effectively evaluate the parts wear and tear, provide maintenance suggestions, and optimize the maintenance strategy through data analysis.

[0107] The parts maintenance module includes a defect assessment report generation unit, a maintenance suggestion unit, a data storage unit and a data analysis unit. The defect assessment report generation unit is used to generate a part defect assessment report to describe the current defect status of the part. The maintenance suggestion unit is used to generate a maintenance suggestion based on the part defect assessment report. The maintenance suggestion includes the parts that need to be replaced, maintenance priority and maintenance operation steps. The data storage unit is used to store the part defect assessment report and the maintenance suggestion in the transfer equipment maintenance database. Using cloud computing technology, the part defect assessment report and the maintenance suggestion are stored in the cloud to ensure data security and accessibility. The data analysis unit is used to perform data analysis on the transfer equipment maintenance database, optimize maintenance strategies and predict future maintenance needs. Using big data technology, in-depth analysis is performed on the part defect assessment reports and maintenance suggestions stored in the database.

[0108] Embodiment 2:

[0109] See also Figure 6 The present invention provides an embodiment: a method for repairing parts of a transfer equipment, comprising the following steps:

[0110] S1: The optical signal of the transfer equipment parts is obtained through the industrial camera, and the optical signal is converted into an electrical signal according to the pixel array. The electrical signal is processed by the analog-to-digital converter after passing through the polarization circuit, and the image of the transfer equipment parts is output;

[0111] S2: preprocessing the transfer equipment part image, detecting feature points on the preprocessed transfer equipment part image, calculating feature descriptors, and generating a loss feature vector;

[0112] S3: Detect the loss feature vector through Hough transform to obtain the center loss feature;

[0113] S4: construct a part defect recognition network, use the center loss feature as an input parameter of the part defect recognition network, train the input parameter through the part defect recognition network, and output the defect status of the transfer equipment parts.

[0114] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A transfer equipment parts maintenance system, characterized in that: The system includes a parts data acquisition module, a parts data testing module and a parts maintenance module; The part data acquisition module is provided with an energy conversion strategy, and the energy conversion strategy is used to convert the optical signal of the transfer equipment part into the image of the transfer equipment part; The part data testing module is used to detect the image of the transfer equipment part through image processing technology to obtain the defect status of the transfer equipment part. The part data testing module is configured with a defect testing strategy. The defect testing strategy evaluates the part defect status through image preprocessing, feature extraction, Hough transform and part defect recognition network. The defect status includes part defect category and part defect rate; The parts inspection and repair module is used to generate a parts defect assessment report and inspection and repair suggestions according to the defect status of the transfer equipment parts; The part data testing module includes a part data processing unit and a part defect detection unit. The part data processing unit is used to pre-process the image of the transfer equipment part to generate a loss feature vector. The part defect detection unit is used to calculate the center loss feature according to the loss feature vector, and identify the defect state of the transfer equipment part according to the center loss feature. The defect testing strategy includes data processing logic and part detection logic. The data processing logic is configured in the part data processing unit, and the part detection logic is configured in the part defect detection unit. The part detection logic includes a feature extraction sub-logic and a defect detection sub-logic. The feature extraction sub-logic is used to process the loss feature vector through Hough transform to obtain the center loss feature. The defect detection sub-logic is used to build a part defect recognition network, use the center loss feature as an input parameter of the part defect recognition network, train the input parameter through the part defect recognition network, and output the defect status of the transfer equipment part. The feature extraction sub-logic includes: Randomly select edge pixels from the edge pixel set of the loss feature vector and project them into the parameter space accumulator, and delete the selected edge pixels; Determine whether the extreme value in the parameter space accumulator is greater than the accumulation threshold. If it is less than or equal to the accumulation threshold, put the edge pixel back into the edge pixel set and reselect the edge pixel. If it is greater than the accumulation threshold, the peak values ​​formed by the extreme values ​​in the parameter space accumulator are concentrated to find the largest curve segment that is continuous or has a gap not greater than the gap threshold, and the largest curve segment is compared with the limited length; If it is greater than or equal to the specified length, the remaining edge pixels in the edge pixel set are projected as the center loss feature. If it is less than the specified length, the edge pixels are reselected. The part defect recognition network comprises: An input layer, which establishes a feature sequence according to input parameters and calculates the fitting value of the feature sequence; A hidden layer converts the fitting value of the feature sequence of the input layer into a high-dimensional space, and calculates the output value of the hidden layer in the high-dimensional space; An output layer, used to output the defect status of the transfer equipment parts; Initializing the part defect recognition network according to the number of input layer nodes, the number of hidden layer nodes and the number of output layer nodes of the part defect recognition network, and determining the connection weights and node thresholds of the input layer and the hidden layer; The input parameters are biased by the connection weights of the input layer and the hidden layer to obtain the output of the hidden layer, and the output of the hidden layer is updated by the node threshold to obtain the output layer input; Calculating the error of the output layer input according to the error function of the output layer, determining whether the error of the output layer input meets the convergence requirement of the output layer, and if not, updating the connection weight and the node threshold; If the convergence requirement is met, the defect status of the transfer equipment parts is output, wherein the calculation formula of the part defect rate is as follows: ， in, Indicates the defect rate of parts of transfer equipment parts, represents the learning rate, u represents a single node in the hidden layer, U represents the total number of nodes in the hidden layer, v represents a single node in the output layer, and V represents the total number of nodes in the output layer. represents the output of the u-th node in the hidden layer, represents the connection weights between the hidden layer and the output layer nodes, represents the node threshold of the output layer, and E represents the error of the output layer input.

2. According to claim 1, a transfer equipment parts maintenance system is characterized in that: The data processing logic includes: De-noising the image of the parts of the transfer equipment. The image noise of the parts of the transfer equipment includes AD quantization noise, dark current noise and parts reflection noise; Performing histogram statistics on the grayscale values ​​of the denoised transfer equipment part image, calculating an optimal segmentation threshold, and extracting a target area according to the optimal segmentation threshold; Connecting the pixels of the target area, obtaining an image of the target area, extracting key points of the image of the target area, and calculating the reflection intensity of the key points; The reflection intensity of the key point is compared with the feature threshold. If it is greater than or equal to the feature threshold, the key point is marked as a feature point, and the loss feature vector is calculated based on the feature point.

3. According to claim 1, a transfer equipment parts maintenance system is characterized in that: The part data acquisition module includes an optical lens unit, a pixel array unit and an image sensor readout circuit unit. The optical lens unit is used to store a CMOS image sensor. The pixel array unit is used to read the fluorescent light signal of the transfer equipment parts and convert it into photocurrent voltage information and part surface information. The image sensor readout circuit unit is used to current bias the source follower of the pixel, dynamically adjust the exposure according to the change of light in the scene, convert the voltage information into a digital code value, and output the transfer equipment part image. The energy conversion strategy includes photoelectric conversion logic and image generation logic. The photoelectric conversion logic is configured in the pixel array unit, and the image generation logic is configured in the image sensor readout circuit unit.

4. A transfer equipment parts maintenance system according to claim 3, characterized in that: The photoelectric conversion logic includes: The CMOS image sensor is illuminated from the front by an LED to obtain the light radiation energy of the parts of the transfer equipment reaching the CMOS image sensor; Photoelectric conversion is performed through diodes to generate photocurrent, and pixel gain is performed on the photocurrent through the pixel array to extract photons and part surface information; The photons and part surface information are converted into voltage information and transmitted to the COMS image sensor readout circuit through the metal bus.

5. A transfer equipment parts maintenance system according to claim 1, characterized in that: The parts maintenance module includes a defect assessment report generation unit, a maintenance suggestion unit, a data storage unit and a data analysis unit. The defect assessment report generation unit is used to generate a parts defect assessment report to describe the current defect status of the part. The maintenance suggestion unit is used to generate maintenance suggestions based on the parts defect assessment report. The maintenance suggestions include parts that need to be replaced, maintenance priorities and maintenance operation steps. The data storage unit is used to store the parts defect assessment report and the maintenance suggestions in a transfer equipment maintenance database. The data analysis unit is used to perform data analysis on the transfer equipment maintenance database, optimize maintenance strategies and predict future maintenance needs.

Images