Intelligent management method for hard alloy cutter welding visual inspection data

Abstract

The invention discloses a hard alloy cutter welding visual inspection data intelligent management method, and relates to the technical field of welding quality detection, and the method comprises the following steps: obtaining image data in a welding process through a visual inspection system; preprocessing the image data; analyzing defect characteristics of a welding joint area based on an image recognition algorithm, and recognizing a welding quality problem; generating a defect report based on the identified defect features; establishing a welding quality prediction model, and automatically generating a quality evaluation result; and storing the data to the cloud. Through high-resolution image capture and automatic defect analysis based on an image recognition algorithm, defects, such as cracks, air holes and weld irregularity, of a welding joint area can be accurately recognized in real time, the detection precision and efficiency are greatly improved, personal errors are reduced, a welding quality prediction model is established based on a machine learning algorithm, and the welding quality prediction accuracy is improved. Therefore, the quality of the current welding operation is predicted, and the reliability of the detection process is improved.

Description

Technical Field

[0001] The invention relates to the technical field of welding quality detection, and in particular to an intelligent management method for visual detection data of cemented carbide tool welding. Background Art

[0002] Carbide tools are important tools for high-intensity processing tasks such as metal cutting and drilling. Welding, as a key process in the manufacture of carbide tools, is often used to connect carbide and base metal to form high-strength tools. The quality of welding directly affects the service life and processing performance of the tool. Therefore, it is crucial to accurately monitor and evaluate the quality of the welding process.

[0003] Traditional welding quality inspection methods rely on manual inspection and conventional physical inspection methods, such as visual observation, ultrasound, X-ray and other methods. Although these methods are effective, they have problems such as low inspection efficiency, complex operation, and large manual errors. Especially in large-scale production processes, manual inspection cannot meet the requirements of real-time, accuracy and consistency. In addition, traditional welding quality control usually focuses on sampling inspection of terminal products, which makes it difficult to detect quality problems in the welding process in a timely manner, thus affecting production efficiency and product qualification rate. Summary of the invention

[0004] In order to solve the above technical problems, a method for intelligent management of carbide tool welding visual inspection data is provided. This technical solution solves the above problems.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for intelligent management of carbide tool welding visual inspection data, comprising:

[0007] Acquiring image data of the welding process by a visual inspection system, wherein the image data includes a high-resolution image of the weld joint area;

[0008] Preprocess the image data, including image denoising, contrast enhancement and image grayscale processing;

[0009] Analyze the defect characteristics of the weld joint area based on image recognition algorithms to identify welding quality problems;

[0010] Generate defect reports based on identified defect characteristics and correlate reports with real-time production data;

[0011] Analyze historical inspection data based on machine learning algorithms, establish a welding quality prediction model, predict welding quality based on current inspection data, and automatically generate quality assessment results;

[0012] Store all test results, defect reports, quality assessment results and related data in the cloud.

[0013] Preferably, the image data obtained during the welding process by the visual inspection system includes a high-resolution image of the welding joint area, specifically including:

[0014] Use a camera to capture high-resolution images, configure the camera with a lens, use an LED light source to illuminate the welding area, capture images of the welding joint area in real time during the welding process, and set the camera's resolution;

[0015] According to the dynamic changes of the welding process, the image capture frequency is set to periodically capture multiple frames of images to obtain complete welding process data.

[0016] Preferably, the preprocessing of the image data, including image denoising, contrast enhancement and image grayscale processing, specifically includes:

[0017] Obtain image data, and based on the mean filter algorithm, average the neighborhood pixels around each pixel in the image through a sliding window. The mean filter algorithm formula is:

[0018]

[0019] Where j(x,y) is the denoised image, G and H are the sizes of the filter windows, f(x+i,y+j) is the pixel value at position (x+i,y+j) in the input image, and i and K are variables used to traverse the neighborhood range in the convolution operation;

[0020] Based on histogram equalization, the gray levels in the image are redistributed to make the frequency of all gray levels uniform and enhance the contrast of the image. The histogram equalization formula is:

[0021]

[0022] Where S is the new gray level, r is the gray level of the original image, CDF(r) is the cumulative distribution function corresponding to gray level r, and N is the number of all pixels in the image;

[0023] The grayscale values ​​of the three RGB channels are weighted averaged to convert the color image RGB image into a grayscale image. The image weighted average formula is:

[0024] Gray=0.2989×R+0.5870×G+0.1140×B

[0025] Where Gray is the output grayscale value, R is the pixel value of the red channel in the image, G is the pixel value of the green channel in the image, B is the pixel value of the blue channel in the image, 0.2989 is the weighting coefficient of the red channel R, 0.5870 is the weighting coefficient of the green channel G, and 0.1140 is the weighting coefficient of the blue channel B.

[0026] Preferably, analyzing the defect characteristics of the weld joint area based on the image recognition algorithm and identifying the welding quality problems specifically includes:

[0027] The welding joint area is identified by the threshold segmentation method in image segmentation, a fixed threshold is selected, and the image is divided into foreground and background by comparing with the threshold. The threshold segmentation formula is:

[0028]

[0029] Where I(X,Y) represents the pixel value of the input image at position (X,Y), T is the threshold, I(X,Y) represents the pixel value of the output image after threshold processing, 255 represents the pixel value of the foreground, and 0 represents the pixel value of the background;

[0030] For the segmented welding area, different defect features are extracted, including welding cracks, pores, and weld irregularities;

[0031] Classify the extracted defect features using machine learning algorithms;

[0032] The defect locations and types in welding images are identified through the output of the classification model.

[0033] Preferably, the classifying the extracted defect features using a machine learning algorithm specifically includes:

[0034] In the segmented welding area, different defect features are extracted, including:

[0035] Welding crack characteristics: crack length, width, shape and edge complexity;

[0036] Pore ​​characteristics: diameter, shape and location of pores;

[0037] Weld irregularity characteristics: including uneven weld width, thickness variation and abnormal shape;

[0038] Feature extraction is performed based on shape features, and the extracted features are normalized by Z-score;

[0039] Defect feature classification is performed based on support vector machine.

[0040] Preferably, generating a defect report based on the identified defect features and associating the report with real-time production data specifically includes:

[0041] When generating a defect report, the identified defect data is associated with real-time production data, which includes:

[0042] Production batch number: a unique identifier for each production, used to track and associate defect data from different production batches;

[0043] Welding parameters: such as current, voltage, speed and temperature, which may affect the welding quality;

[0044] Production time: Timestamp of each welding process helps to track the moment when the defect occurred;

[0045] Operator information: information about the operator who performed the welding task, used to analyze whether it is related to a specific operator;

[0046] Equipment status: whether the equipment is in normal working condition, whether there are any faults or instability;

[0047] Material information: welding materials used, such as type and specifications of welding rods, welding wires, and base materials;

[0048] These real-time production data need to be combined with defect data and integrated through the production information management system;

[0049] Generate defect reports including report header, defect overview, defect details, defect location, defect type, defect size, confidence score, welding parameters, equipment status, and quality assessment;

[0050] In the defect report, fill in the real-time production data and defect data;

[0051] Associate defect data with corresponding production data through timestamps;

[0052] Association based on batch number: Each production batch corresponds to a set of welding operations and defect records. The defect data is associated with other data of the production line through the batch number;

[0053] Based on the association of equipment and operator information: Defect data is associated with production equipment and operator information to analyze whether there are problems with specific equipment and operators, thereby improving the stability of the welding process;

[0054] All generated defect reports need to be archived according to production batches and time.

[0055] Preferably, the historical test data is analyzed based on a machine learning algorithm, a welding quality prediction model is established, welding quality is predicted according to current test data, and quality assessment results are automatically generated, specifically including:

[0056] Based on historical data sets, including welding parameters, image features and defect label training data sets;

[0057] Establish a welding quality prediction model based on regression task algorithm;

[0058] Input the current inspection data: including the current welding parameters, environmental data and image data, perform feature extraction and preprocessing on the current inspection data, input the processed data into the trained machine learning model, and output the prediction results;

[0059] Based on the prediction results, a quality assessment report is automatically generated.

[0060] Preferably, establishing a welding quality prediction model based on a regression task algorithm specifically includes:

[0061] Among them, the welding quality prediction model expression is:

[0062]

[0063] In the formula, The predicted output value of the model, f represents the activation function of the output layer, w2 is the weight matrix of the second layer, σ is the activation function, w1 is the weight matrix of the first layer, x is the input feature vector, b1 is the bias term of the first layer, and b2 is the bias term of the output layer.

[0064] Preferably, the automatically generating a quality assessment report according to the prediction results specifically includes:

[0065] Predicted weld quality: pass / fail, and quality score;

[0066] Prediction confidence: the confidence of the model in the prediction result;

[0067] Recommended improvement measures: predict failure and generate improvement measures, including adjustment of welding parameters and equipment maintenance suggestions.

[0068] Preferably, storing all test results, defect reports, quality assessment results and related data in the cloud specifically includes: transferring data to cloud storage through API, designing storage strategies, setting data indexes, and ensuring data redundancy and backup, setting access control, encrypted storage and transmission, and enabling audit logs, processing data in the cloud, generating reports, and using dashboards to display results.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] The present invention proposes that through high-resolution image capture and automatic defect analysis based on image recognition algorithms, defects in the welding joint area, such as cracks, pores and weld irregularities, can be accurately identified in real time. Compared with traditional manual inspection, this method greatly improves the accuracy and efficiency of inspection and reduces human errors. Based on machine learning algorithms, the system can analyze historical inspection data and establish a welding quality prediction model to predict the quality of the current welding operation. This automated quality assessment and prediction capability not only improves the reliability of the inspection process, but also can promptly detect potential quality problems and avoid quality fluctuations caused by operational errors or equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a step flow framework diagram of the present invention. DETAILED DESCRIPTION

[0072] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0073] Reference Figure 1 As shown, a method for intelligent management of carbide tool welding visual inspection data includes:

[0074] Step 1:

[0075] Use a camera to capture high-resolution images, configure the camera with a lens, use an LED light source to illuminate the welding area, capture images of the welding joint area in real time during the welding process, and set the camera's resolution;

[0076] According to the dynamic changes of the welding process, the image capture frequency is set to periodically capture multiple frames of images to obtain complete welding process data.

[0077] Step 2:

[0078] Obtain image data, and average the neighborhood pixels around each pixel in the image through a sliding window based on the mean filtering algorithm. The mean filtering algorithm formula is:

[0079]

[0080] Where j(x,y) is the denoised image, G and H are the sizes of the filter windows, f(x+i,y+j) is the pixel value at position (x+i,y+j) in the input image, and i and K are variables used to traverse the neighborhood range in the convolution operation;

[0081] Based on histogram equalization, the gray levels in the image are redistributed to make the frequency of all gray levels uniform and enhance the contrast of the image. The histogram equalization formula is:

[0082]

[0083] Where S is the new gray level, r is the gray level of the original image, CDF(r) is the cumulative distribution function corresponding to gray level r, and N is the number of all pixels in the image;

[0084] The grayscale values ​​of the three RGB channels are weighted averaged to convert the color image RGB image into a grayscale image. The image weighted average formula is:

[0085] Gray=0.2989×R+0.5870×G+0.1140×B

[0086] Where Gray is the output grayscale value, R is the pixel value of the red channel in the image, G is the pixel value of the green channel in the image, B is the pixel value of the blue channel in the image, 0.2989 is the weighting coefficient of the red channel R, 0.5870 is the weighting coefficient of the green channel G, and 0.1140 is the weighting coefficient of the blue channel B.

[0087] Step 3:

[0088] The welding joint area is identified by the threshold segmentation method in image segmentation, a fixed threshold is selected, and the image is divided into foreground and background by comparing with the threshold. The threshold segmentation formula is:

[0089]

[0090] Where I(X,Y) represents the pixel value of the input image at position (X,Y), T is the threshold, I(X,Y) represents the pixel value of the output image after threshold processing, 255 represents the pixel value of the foreground, and 0 represents the pixel value of the background;

[0091] For the segmented welding area, different defect features are extracted, including welding cracks, pores, and weld irregularities;

[0092] In the segmented welding area, different defect features are extracted, including:

[0093] Welding crack characteristics: crack length, width, shape and edge complexity;

[0094] Pore ​​characteristics: diameter, shape and location of pores;

[0095] Weld irregularity characteristics: including uneven weld width, thickness variation and abnormal shape;

[0096] Feature extraction is performed based on shape features, and the extracted features are normalized by Z-score;

[0097] Classification of defect features based on support vector machine;

[0098] The defect locations and types in welding images are identified through the output of the classification model.

[0099] Step 4:

[0100] When generating a defect report, the identified defect data is associated with real-time production data, which includes:

[0101] Production batch number: a unique identifier for each production, used to track and associate defect data from different production batches;

[0102] Welding parameters: such as current, voltage, speed and temperature, which may affect the welding quality;

[0103] Production time: Timestamp of each welding process helps to track the moment when the defect occurred;

[0104] Operator information: information about the operator who performed the welding task, used to analyze whether it is related to a specific operator;

[0105] Equipment status: whether the equipment is in normal working condition, whether there are any faults or instability;

[0106] Material information: welding materials used, such as type and specifications of welding rods, welding wires, and base materials;

[0107] These real-time production data need to be combined with defect data and integrated through the production information management system;

[0108] Generate defect reports including report header, defect overview, defect details, defect location, defect type, defect size, confidence score, welding parameters, equipment status, and quality assessment;

[0109] In the defect report, fill in the real-time production data and defect data;

[0110] Associate defect data with corresponding production data through timestamps;

[0111] Association based on batch number: Each production batch corresponds to a set of welding operations and defect records. The defect data is associated with other data of the production line through the batch number;

[0112] Based on the association of equipment and operator information: Defect data is associated with production equipment and operator information to analyze whether there are problems with specific equipment and operators, thereby improving the stability of the welding process;

[0113] All generated defect reports need to be archived according to production batches and time.

[0114] Step 5:

[0115] Based on historical data sets, including welding parameters, image features and defect label training data sets;

[0116] The welding quality prediction model is established based on the regression task algorithm, where the welding quality prediction model expression is:

[0117]

[0118] In the formula, The predicted output value of the model, f represents the activation function of the output layer, w2 is the weight matrix of the second layer, σ is the activation function, w1 is the weight matrix of the first layer, x is the input feature vector, b1 is the bias term of the first layer, and b2 is the bias term of the output layer;

[0119] Input the current inspection data: including the current welding parameters, environmental data and image data, perform feature extraction and preprocessing on the current inspection data, input the processed data into the trained machine learning model, and output the prediction results;

[0120] Predicted weld quality: pass / fail, and quality score;

[0121] Prediction confidence: the confidence of the model in the prediction result;

[0122] Recommended improvement measures: predict failure and generate improvement measures, including adjustment of welding parameters and equipment maintenance suggestions.

[0123] Step 6:

[0124] Transfer data to cloud storage through APIs, design storage strategies, set up data indexes, and ensure data redundancy and backup, set up access controls, encrypt storage and transmission, and enable audit logs, process data in the cloud, generate reports, and present results using dashboards.

[0125] In summary, the advantages of the present invention are:

[0126] Through high-resolution image capture and automatic defect analysis based on image recognition algorithms, defects in the weld joint area, such as cracks, pores, and weld irregularities, can be accurately identified in real time. Compared with traditional manual inspection, this method greatly improves the accuracy and efficiency of inspection and reduces human errors.

[0127] Based on machine learning algorithms, the system can analyze historical inspection data and establish a welding quality prediction model to predict the quality of the current welding operation. This automated quality assessment and prediction capability not only improves the reliability of the inspection process, but also can detect potential quality problems in a timely manner, avoiding quality fluctuations caused by operational errors or equipment failures.

[0128] By associating defect data with real-time production data, the system can generate detailed defect reports and help production personnel trace the root cause of the problem. This data association capability provides an important basis for subsequent welding quality improvement and helps optimize the production process.

[0129] By storing all test results, defect reports, quality assessment results and related data in the cloud, centralized management and remote access are facilitated. The cloud's storage and computing capabilities ensure efficient processing and real-time analysis of large-scale data, and support data redundancy backup and encrypted storage to ensure data security.

[0130] The cloud dashboard displays real-time data and analysis results of welding quality inspection, allowing production managers to easily monitor the entire welding production process, adjust welding parameters in a timely manner, and optimize production operations. This visual quality monitoring method provides strong support for the company's production management and helps improve production efficiency and product quality.

[0131] By associating with the equipment status information, the system can detect possible abnormal conditions of the equipment during the welding process, and provide real-time feedback on equipment failure or unstable status, which helps to maintain and service the equipment in a timely manner, avoid production quality problems caused by equipment failure, thereby reducing downtime and production costs;

[0132] Through continuous training and optimization of machine learning models, the system can continuously improve the accuracy of welding quality predictions, automatically generate improvement measures, and promote the development of the production process towards intelligence and automation. Through intelligent feedback from machine learning, it can optimize production parameters and operating processes, thereby continuously improving welding quality.

[0133] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for intelligent management of carbide tool welding visual inspection data, characterized in that: include: Acquiring image data of the welding process by a visual inspection system, wherein the image data includes a high-resolution image of the weld joint area; Preprocess the image data, including image denoising, contrast enhancement and image grayscale processing; Analyze the defect characteristics of the weld joint area based on image recognition algorithms to identify welding quality problems; Generate defect reports based on identified defect characteristics and correlate reports with real-time production data; Analyze historical inspection data based on machine learning algorithms, establish a welding quality prediction model, predict welding quality based on current inspection data, and automatically generate quality assessment results; Store all test results, defect reports, quality assessment results and related data in the cloud.

2. The intelligent management method for carbide tool welding visual inspection data according to claim 1 is characterized in that: The image data obtained by the visual inspection system during the welding process includes a high-resolution image of the welding joint area, specifically including: Use a camera to capture high-resolution images, configure the camera with a lens, use an LED light source to illuminate the welding area, capture images of the welding joint area in real time during the welding process, and set the camera's resolution; According to the dynamic changes of the welding process, the image capture frequency is set to periodically capture multiple frames of images to obtain complete welding process data.

3. The intelligent management method for carbide tool welding visual inspection data according to claim 2 is characterized in that: The preprocessing of the image data, including image denoising, contrast enhancement and image grayscale processing, specifically includes: Obtain image data, and average the neighborhood pixels around each pixel in the image through a sliding window based on the mean filtering algorithm. The mean filtering algorithm formula is: Where j(x,y) is the denoised image, G and H are the sizes of the filter windows, f(x+i,y+j) is the pixel value at position (x+i,y+j) in the input image, and i and K are variables used to traverse the neighborhood range in the convolution operation; Based on histogram equalization, the gray levels in the image are redistributed to make the frequency of all gray levels uniform and enhance the contrast of the image. The histogram equalization formula is: Where S is the new gray level, r is the gray level of the original image, CDF(r) is the cumulative distribution function corresponding to gray level r, and N is the number of all pixels in the image; The grayscale values ​​of the three RGB channels are weighted averaged to convert the color image RGB image into a grayscale image. The image weighted average formula is: Gray=0.2989×R+0.5870×G+0.1140×B Where Gray is the output grayscale value, R is the pixel value of the red channel in the image, G is the pixel value of the green channel in the image, B is the pixel value of the blue channel in the image, 0.2989 is the weighting coefficient of the red channel R, 0.5870 is the weighting coefficient of the green channel G, and 0.1140 is the weighting coefficient of the blue channel B.

4. The intelligent management method for carbide tool welding visual inspection data according to claim 3 is characterized in that: The method of analyzing the defect characteristics of the weld joint area based on the image recognition algorithm and identifying the welding quality problems specifically includes: The welding joint area is identified by the threshold segmentation method in image segmentation, a fixed threshold is selected, and the image is divided into foreground and background by comparing with the threshold. The threshold segmentation formula is: Where I(X,Y) represents the pixel value of the input image at position (X,Y), T is the threshold, I(X,Y) represents the pixel value of the output image after threshold processing, 255 represents the pixel value of the foreground, and 0 represents the pixel value of the background; For the segmented welding area, different defect features are extracted; Classify the extracted defect features using machine learning algorithms; The defect locations and types in welding images are identified through the output of the classification model.

5. The intelligent management method for carbide tool welding visual inspection data according to claim 4 is characterized in that: The method of using a machine learning algorithm to classify the extracted defect features specifically includes: In the segmented welding area, different defect features are extracted, including welding crack features, pore features and weld irregularity features; Feature extraction is performed based on shape features, and the extracted features are normalized by Z-score; Defect feature classification is performed based on support vector machine.

6. The method for intelligent management of carbide tool welding visual inspection data according to claim 5, characterized in that: Generating a defect report based on the identified defect features and associating the report with real-time production data specifically includes: When generating defect reports, the identified defect data is correlated with real-time production data; Generate defect reports including report header, defect overview, defect details, defect location, defect type, defect size, confidence score, welding parameters, equipment status, and quality assessment; In the defect report, fill in the real-time production data and defect data; Associate defect data with corresponding production data through timestamps; Association based on batch number: Each production batch corresponds to a set of welding operations and defect records, and the defect data is associated with other data of the production line through the batch number.

7. The intelligent management method for carbide tool welding visual inspection data according to claim 6 is characterized in that: The method of analyzing historical test data based on machine learning algorithm, establishing a welding quality prediction model, predicting welding quality according to current test data, and automatically generating quality assessment results specifically includes: Based on historical data sets, including welding parameters, image features and defect label training data sets; Establish a welding quality prediction model based on regression task algorithm; Input the current inspection data: including the current welding parameters, environmental data and image data, perform feature extraction and preprocessing on the current inspection data, input the processed data into the trained machine learning model, and output the prediction results; Based on the prediction results, a quality assessment report is automatically generated.

8. The method for intelligent management of carbide tool welding visual inspection data according to claim 7, characterized in that: Establishing welding quality prediction model based on regression task algorithm include: Among them, the welding quality prediction model expression is: In the formula, The predicted output value of the model, f represents the activation function of the output layer, w2 is the weight matrix of the second layer, σ is the activation function, w1 is the weight matrix of the first layer, x is the input feature vector, b1 is the bias term of the first layer, and b2 is the bias term of the output layer.

9. The intelligent management method for carbide tool welding visual inspection data according to claim 8, characterized in that: The automatic generation of a quality assessment report based on the prediction results specifically includes: Predicted weld quality: pass / fail, and quality score; Prediction confidence: the confidence of the model in the prediction result; Recommended improvement measures: predict failure and generate improvement measures, including adjustment of welding parameters and equipment maintenance suggestions.

10. The method for intelligent management of carbide tool welding visual inspection data according to claim 9, characterized in that: Storing all test results, defect reports, quality assessment results and related data in the cloud specifically includes: transferring data to cloud storage through API, designing storage strategies, setting data indexes, and ensuring data redundancy and backup, setting access control, encrypted storage and transmission, and enabling audit logs, processing data in the cloud, generating reports, and using dashboards to display results.

Images