Manufacturing process optimization method of stainless steel pipe fitting

Abstract

The invention provides a manufacturing process optimization method for a stainless steel pipe fitting, and relates to the technical field of pipe fitting machining, and the method comprises the steps: carrying out the nondestructive testing of a welded semi-finished product, generating an apparent defect detection instruction, obtaining a welded image array, synchronizing the welded image array to a machining defect recognition network, carrying out the defect positioning, and obtaining a to-be-processed defect array; synchronizing the to-be-processed defect array to a defect processing analysis network to execute technological procedure analysis to obtain a defect technological procedure array, executing surface processing fitness analysis according to the K technological procedure sets to obtain a target manufacturing technological procedure, and executing surface finishing of the post-welding semi-finished product according to the target manufacturing technological procedure. The technical problem that in the prior art, after-welding semi-finished product defect detection in the stainless steel pipe fitting manufacturing process is incomplete and inaccurate is solved. The technical effects of comprehensively, accurately and efficiently detecting and positioning the defects, ensuring that the defects can be correctly and efficiently treated and improving the quality of the stainless steel pipe fitting are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of pipe fitting processing, and in particular to a method for optimizing the manufacturing process of a stainless steel pipe fitting. Background Art

[0002] In modern manufacturing, especially in the production of high-end stainless steel pipe fittings, product quality control is crucial. Stainless steel pipe fittings play an indispensable role in the chemical, petroleum, pharmaceutical, food processing and other industries due to their excellent corrosion resistance, high strength and good mechanical properties. However, in the manufacturing process of stainless steel pipe fittings, welding is a key step in forming high-quality products, and it is also one of the links most prone to defects. In the manufacturing process of stainless steel pipe fittings, the existing technology is not comprehensive and accurate in the detection of defects in semi-finished products after welding, resulting in the failure to detect and repair welding defects in a timely manner, affecting the quality of the final stainless steel pipe fittings and posing a safety hazard.

[0003] The existing technology is not comprehensive and accurate in detecting defects in semi-finished products after welding during the manufacturing process of stainless steel pipe fittings, resulting in the technical problem that welding defects cannot be discovered and repaired in time, affecting the quality of the final stainless steel pipe fittings. Summary of the invention

[0004] The present application provides a method for optimizing the manufacturing process of stainless steel pipe fittings, which is used to solve the technical problem that the prior art is incomplete and inaccurate in the detection of defects in semi-finished products after welding in the manufacturing process of stainless steel pipe fittings, resulting in the failure to timely discover and repair welding defects, thus affecting the quality of the final stainless steel pipe fittings.

[0005] In view of the above problems, the present application provides a method for optimizing the manufacturing process of stainless steel pipe fittings, the method comprising: performing nondestructive testing on the semi-finished products after welding, and generating apparent defect detection instructions based on the detection results; performing image acquisition on the semi-finished products after welding according to the apparent defect detection instructions to obtain a post-weld image array; pre-constructing a processing defect recognition network, and synchronizing the post-weld image array to the processing defect recognition network to perform defect positioning to obtain a defect array to be processed, wherein the defect array to be processed includes K defect feature information of K spatial defects, the defect feature information has a defect type identifier, and K is a positive integer; pre-constructing a defect processing analysis network, and synchronizing the defect array to be processed to the defect processing analysis network to perform process procedure analysis to obtain a defect process procedure array, wherein the defect process procedure array includes K process procedure sets for the K spatial defects; performing surface treatment fitness analysis according to the K process procedure sets to obtain a target manufacturing process procedure; performing surface finishing of the semi-finished products after welding according to the target manufacturing process procedure.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The method provided in the embodiment of the present application performs nondestructive testing on the semi-finished product after welding, generates an apparent defect detection instruction based on the detection result, performs image acquisition on the semi-finished product after welding according to the apparent defect detection instruction, obtains a post-weld image array, pre-builds a processing defect recognition network, and synchronizes the post-weld image array to the processing defect recognition network to perform defect positioning, obtains a defect array to be processed, pre-builds a defect processing analysis network, and synchronizes the defect array to be processed to the defect processing analysis network to perform process specification analysis, obtains a defect process specification array, performs surface treatment fitness analysis according to the K process specification sets, obtains a target manufacturing process specification, and performs surface finishing of the semi-finished product after welding according to the target manufacturing process specification. Comprehensive, accurate and efficient defect detection and positioning are achieved, ensuring that defects can be correctly and efficiently processed and improving the technical effect of the quality of stainless steel pipe fittings. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A schematic diagram of a manufacturing process optimization method for a stainless steel pipe fitting provided in this application; Figure 2 A schematic diagram of a flow chart for generating surface defect detection instructions in a method for optimizing a manufacturing process of a stainless steel pipe provided in the present application; Figure 3 The present application provides a flow chart of a pre-constructed defect processing analysis network in a method for optimizing the manufacturing process of a stainless steel pipe fitting. DETAILED DESCRIPTION

[0009] The present application provides a method for optimizing the manufacturing process of stainless steel pipe fittings, which is used to solve the technical problem that the prior art is incomplete and inaccurate in the detection of defects in semi-finished products after welding in the manufacturing process of stainless steel pipe fittings, resulting in the failure to timely discover and repair welding defects, thus affecting the quality of the final stainless steel pipe fittings. The method achieves comprehensive, accurate and efficient defect detection and positioning, ensures that defects can be correctly and efficiently handled, and improves the technical effect of the quality of stainless steel pipe fittings.

[0010] Below, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all of them.

[0011] like Figure 1 As shown, the present application provides a method for optimizing the manufacturing process of stainless steel pipe fittings, the method comprising: Perform non-destructive testing on semi-finished products after welding, and generate apparent defect detection instructions based on the test results.

[0012] Specifically, stainless steel pipe fittings that have been processed by the welding process but have not completed the entire processing flow, that is, semi-finished products after welding, which contain welded joints but have not undergone grinding, polishing, cleaning or other subsequent processing steps. Then, non-destructive testing is performed on the semi-finished products after welding according to non-destructive testing methods, and non-destructive testing methods include X-ray testing, ultrasonic testing, magnetic particle testing, etc. Next, the test results of the non-destructive testing are analyzed. The results of the non-destructive testing include both non-destructive and destructive results of the semi-finished products after welding. Finally, based on the test results of the non-destructive testing, detection instructions for apparent defects of the semi-finished products after welding are generated. The apparent defect detection instructions are used to guide subsequent surface defect image acquisition and analysis work.

[0013] Image acquisition is performed on the post-weld semi-finished product according to the apparent defect detection instruction to obtain a post-weld image array.

[0014] Specifically, according to the apparent defect detection instructions, determine the areas of the semi-finished product after welding that need to be inspected, such as welding seams, heat-affected zones, etc. Then, based on the apparent defect detection instructions, multi-angle and multi-position image acquisition is performed on the inspection area of ​​the semi-finished product after welding to ensure that each potential defect can be clearly recorded. The image acquisition device is a high-definition camera, macro lens or other imaging device, and the image acquisition device used needs to ensure that it can capture subtle surface defects. Finally, all the collected images are integrated in sequence to form a post-weld image array. The post-weld image array is a set of high-resolution and high-quality images obtained after multi-angle and multi-position image acquisition of the semi-finished product after welding. By obtaining the post-weld image array, data support is provided for subsequent image analysis, defect detection and evaluation.

[0015] A processing defect recognition network is pre-constructed, and the post-weld image array is synchronized to the processing defect recognition network to perform defect location, so as to obtain a defect array to be processed, wherein the defect array to be processed includes K defect feature information of K spatial defects, the defect feature information has a defect type identifier, and K is a positive integer.

[0016] Specifically, in order to automatically and quickly identify processing defects in the semi-finished product image after welding, the deep learning network structure, such as the neural network structure, is trained using the defect image data with annotations to obtain a processing defect recognition network, and the post-weld image array is synchronized to the trained processing defect recognition network. The processing defect recognition network performs defect detection on each image in the post-weld image array, and outputs the position and type information of the defect to obtain a defect array to be processed. The defect array to be processed includes K defect feature information of K spatial defects. The spatial defect refers to the specific spatial position of the defect. The defect feature refers to the determination of the defect size, shape, texture, etc. The defect feature information also has a defect type identifier, and K is a positive integer. By constructing a processing defect recognition network, the automatic recognition and positioning of processing defects in the post-weld image can be realized, the efficiency and accuracy of the detection can be improved, and support can be provided for the subsequent quality control and repair of stainless steel pipe fittings.

[0017] A defect processing analysis network is pre-constructed, and the defect array to be processed is synchronized to the defect processing analysis network to perform process procedure analysis to obtain a defect process procedure array, wherein the defect process procedure array includes K process procedure sets of the K spatial defects.

[0018] Specifically, deep learning or other machine learning technologies are used to pre-build a defect processing analysis network that can process and analyze defect data, and synchronize the array of defects to be processed output by the processing defect recognition network to the defect processing analysis network. The defect processing analysis network is used to perform process procedure analysis on the synchronized array of defects to be processed, that is, the defect processing analysis network performs targeted process procedure analysis on each defect in the array of defects to be processed. After the process procedure analysis, a defect process procedure array is output, which is a data set of processing steps to be taken for each defect, including K process procedure sets for the K spatial defects, each set containing processing methods and steps for specific defects. Obtaining the defect process procedure array through defect analysis provides clear guidance for subsequent defect processing, ensuring that defects can be handled correctly and efficiently.

[0019] Surface treatment fitness analysis is performed according to the K process specification sets to obtain a target manufacturing process specification.

[0020] Specifically, according to the characteristics of the semi-finished product after welding, the surface treatment fitness analysis is performed on each process specification set in the K process specification sets. The surface treatment fitness analysis includes analyzing the technical feasibility, operation difficulty, processing time, cost, etc. of the process specification. According to the analysis results, the target manufacturing process specification is obtained. The target manufacturing process specification is the most suitable manufacturing process specification selected after comprehensive consideration of multiple factors.

[0021] Surface finishing of the welded semi-finished product is performed according to the target manufacturing process specification.

[0022] Specifically, after obtaining the target manufacturing process specifications, the surface of the semi-finished products after welding is finished in accordance with the steps and requirements of the target manufacturing process specifications, that is, the defects of the semi-finished products after welding are processed according to the target manufacturing process specifications to ensure that the surface quality of the semi-finished products after welding is effectively improved to meet the requirements of subsequent use or further processing and improve the quality of stainless steel pipe fittings.

[0023] In one embodiment, Figure 2 As shown, nondestructive testing is performed on the semi-finished product after welding, and an apparent defect detection instruction is generated based on the detection result. The method also includes: calling to obtain the pipe fitting conceptual design of the semi-finished product after welding; introducing a pipe fitting parameter adjustment model, and adjusting the size parameters of the pipe fitting parameter adjustment model according to the pipe fitting conceptual design to generate a pipe fitting conceptual model; performing ultrasonic testing on the semi-finished product after welding, and obtaining an internal defect feature set based on the obtained ultrasonic signal analysis; setting a fitting starting point in the pipe fitting conceptual model, and fitting the internal defect feature set to the pipe fitting conceptual model according to the fitting starting point to obtain a pipe fitting defect model; introducing a defect risk evaluation network to evaluate the pipe fitting defect model, and generating the apparent defect detection instruction.

[0024] Specifically, the conceptual design of the semi-finished pipe fitting after welding is called from the design database. The conceptual design of the semi-finished pipe fitting after welding refers to the design concept of the semi-finished pipe fitting after welding before manufacturing, including the geometric shape, size, material and other specific parameter information of the design of the pipe fitting. Then, a pipe fitting parameter adjustment model that can be adjusted according to the geometric shape and size parameters of the pipe fitting is introduced. The model is a parametric CAD model or a simulation model. According to the specific parameters of the conceptual design of the semi-finished pipe fitting after welding, the size parameters of the pipe fitting parameter adjustment model are adjusted to ensure that it is consistent with the original design, and the conceptual model of the pipe fitting is generated. Next, the semi-finished pipe fitting after welding is subjected to ultrasonic testing. Ultrasonic testing is a non-destructive testing technology used to detect defects inside the material. By analyzing the ultrasonic pulse sent and the reflected signal, the defect information inside the pipe fitting is obtained, including the location, size, shape and possible type of the defect, to form an internal defect feature set. A fitting starting point is set in the pipe fitting conceptual model. The fitting starting point is the reference point for combining the internal defect feature set with the model. Using the fitting starting point as a reference, the information in the internal defect feature set obtained from the ultrasonic signal is fitted into the pipe fitting conceptual model, thereby generating a pipe fitting defect model that reflects the actual defect situation. Finally, the pipe fitting defect model is evaluated based on the defect risk assessment network constructed by pre-training. The defect risk assessment network can assess the risk size of the pipe fitting defect model according to the parameters of the pipe fitting defect model, and then generate corresponding apparent defect detection instructions according to the results of the defect risk assessment network. The apparent defect detection instructions are used to suggest further analysis or scrapping of pipe fittings. For example, when the evaluation result is that the current post-weld semi-finished product has a high defect risk and there is no need to repair the defect, the pipe fitting is scrapped and the post-weld semi-finished product is not further manufactured. If the evaluation result is that the risk is not great, the instruction can be to further analyze the post-weld semi-finished product. By performing non-destructive testing on the post-weld semi-finished product, the defect information of the post-weld semi-finished product can be effectively identified and evaluated, and the apparent defect detection instructions are generated accordingly, which can further improve the production efficiency of stainless steel pipe fittings and ensure the quality and safety of the final pipe fittings.

[0025] Furthermore, the method further includes: the defect risk assessment network includes a cascaded single-dimensional risk assessment layer and a comprehensive risk assessment layer; introducing defect projection constraints and defect depth constraints in the single-dimensional risk assessment layer; introducing a comprehensive risk assessment function and a comprehensive risk constraint in the comprehensive risk assessment layer to complete the construction of the defect risk assessment network, wherein the comprehensive risk assessment function is as follows: ;in, is the comprehensive risk factor, is the maximum defect depth, is the mean defect depth, is the defect density, is the maximum defect spacing, is the minimum defect spacing.

[0026] Specifically, the defect risk assessment network includes a cascaded single-dimensional risk assessment layer and a comprehensive risk assessment layer, which work together to provide a comprehensive risk assessment result. The single-dimensional risk assessment layer includes defect projection constraints and defect depth constraints. The defect projection constraints are used to constrain the projection size of the defect, for example, larger projections have higher risks, and the defect depth constraints limit or weight the impact of defects of different depths on the overall risk by setting a depth threshold or using a depth distribution function, for example, deeper defects have higher risks.

[0027] The comprehensive risk assessment layer includes a comprehensive risk assessment function and a comprehensive risk constraint, which takes into account the comprehensive impact of multiple defect characteristics. The comprehensive risk assessment function is as follows: ;in, is the comprehensive risk factor, is the maximum defect depth, is the mean defect depth, is the defect density, is the maximum defect spacing, is the minimum defect spacing. By comprehensively considering the depth, density and spacing of defects, the risk of defects can be comprehensively assessed. In the comprehensive risk assessment layer, comprehensive risk constraints are also introduced to ensure that the final risk assessment results meet specific requirements or standards. Ultimately, by constructing such a defect risk assessment network, a comprehensive and accurate assessment of the risk of pipe defects can be achieved, and accurate and reasonable decisions can be made for subsequent repairs, scrapping or further testing.

[0028] Furthermore, a defect risk assessment network is introduced to evaluate the pipe defect model and generate the apparent defect detection instruction. The method also includes: extracting multiple defect projection parameters and multiple defect depth parameters of multiple internal defects based on the internal defect feature set; synchronizing the multiple defect projection parameters and multiple defect depth parameters to the one-dimensional risk assessment layer of the defect risk assessment network, and traversing and comparing the multiple defect projection parameters and multiple defect depth parameters using the defect projection constraints and defect depth constraints in the one-dimensional risk assessment layer; if the multiple defect projection parameters and multiple defect depth parameters all meet the defect Projection constraints and defect depth constraints, then the maximum defect depth and the mean defect depth are solved based on the multiple defect depth parameters, and the defect density, the maximum defect spacing and the minimum defect spacing are analyzed based on the pipe defect model; the maximum defect depth, the mean defect depth, the defect density, the maximum defect spacing and the minimum defect spacing are synchronized to the comprehensive risk evaluation layer, and the comprehensive risk coefficient is calculated through the comprehensive risk evaluation function, and it is determined whether the comprehensive risk coefficient falls within the comprehensive risk constraint; if the comprehensive risk coefficient falls within the comprehensive risk constraint, the apparent defect detection instruction is generated.

[0029] Specifically, from the internal defect feature set obtained by ultrasonic testing, multiple defect projection parameters and multiple defect depth parameters of multiple internal defects are extracted, wherein the defect projection parameters are such as the projection position and shape of the defect on the two-dimensional plane, and the defect depth parameters are such as the maximum depth and average depth of each defect. And the extracted multiple defect projection parameters and multiple defect depth parameters are synchronized to the single-dimensional risk assessment layer of the defect risk assessment network. In the single-dimensional risk assessment layer, the defect projection constraint is used to traverse and compare each defect projection parameter, and at the same time, the defect depth constraint is applied to traverse and compare each defect depth parameter. If the multiple defect projection parameters and multiple defect depth parameters all meet the defect projection constraint and the defect depth constraint, that is, the multiple defect projection parameters and multiple defect depth parameters in the internal defect feature set are within the defect projection constraint and the defect depth constraint, then the maximum defect depth and the defect depth mean are obtained based on the multiple defect depth parameters. Further analysis is performed using the pipe defect model to obtain the defect density, the maximum defect spacing and the minimum defect spacing. And the maximum defect depth, the defect depth mean, the defect density, the maximum defect spacing and the minimum defect spacing obtained by calculation and analysis are synchronized to the comprehensive risk assessment layer. In the comprehensive risk assessment layer, the comprehensive risk evaluation function is used to calculate the comprehensive risk coefficient. The calculated comprehensive risk coefficient is compared with the comprehensive risk constraint to determine whether the calculated comprehensive risk coefficient falls within the preset comprehensive risk constraint range. If the comprehensive risk coefficient falls within the comprehensive risk constraint range, it means that the risk level of the defect is within an acceptable range, and the defect can be repaired accordingly, and the corresponding apparent defect detection instructions can be generated to guide subsequent detection work. Through precise parameter analysis and risk assessment, a comprehensive and detailed evaluation of the defects of semi-finished pipe fittings after welding is ensured, thereby ensuring the quality and safety of subsequent stainless steel pipe fittings.

[0030] Furthermore, a defect risk assessment network is introduced to evaluate the pipe defect model and generate the apparent defect detection instruction. The method also includes: if any of the multiple defect projection parameters and the multiple defect depth parameters does not satisfy the defect projection constraint or the defect depth constraint, a scrapping and recycling instruction is generated.

[0031] Specifically, if during the constraint comparison process of the single-dimensional risk assessment layer, it is found that any defect projection parameter or defect depth parameter does not meet the preset defect projection constraint or defect depth constraint, it means that the defect is beyond the acceptable range and the defect may pose an unacceptable risk to the overall performance and safety of the pipe fittings. In this case, a scrapping and recycling instruction is generated to prevent unqualified products from entering the market and reduce potential risks.

[0032] Further, a processing defect recognition network is pre-constructed, and the post-weld image array is synchronized to the processing defect recognition network to perform defect location, and obtain a defect array to be processed. The method also includes: performing image acquisition on the post-weld semi-finished product according to the apparent defect detection instruction, and splicing the image acquisition result to the pipe conceptual model to obtain the post-weld image array; performing pipe defect integration on the post-weld semi-finished product, and performing historical data call based on the integration result to obtain M sample pipe defect image sets and M sample defect feature information sets of M types of pipe defect types, where M is a positive integer; randomly calling M image processing operators from an operator library, and using the M sample pipe defect image sets and M sample defect feature information sets to perform the M image processing operator optimization to obtain M defect recognition branches; connecting the M defect recognition branches in parallel to generate the processing defect recognition network; synchronizing the post-weld image array to the processing defect recognition network to perform defect location, and obtaining the defect array to be processed, wherein the defect array to be processed includes K defect feature information of K spatial defects, and the defect feature information has a defect type identifier.

[0033] Specifically, according to the apparent defect detection instruction, the image acquisition device is used to acquire images of the semi-finished pipe fittings after welding, and the acquired images cover all areas of the semi-finished pipe fittings. Then, the coordinates of each image are calibrated to ensure that they can accurately correspond to the corresponding positions on the pipe fitting conceptual model. Next, the image registration technology is used to align each image with the corresponding position on the pipe fitting conceptual model. After aligning all images, the image fusion algorithm is used to seamlessly splice all images together, and the spliced ​​images are combined with the conceptual model of the pipe fitting to obtain the post-weld image array. Then, the pipe fitting defect integration is performed on the semi-finished product after welding to obtain the defect types existing in the semi-finished product after welding, and the historical data is called to obtain M sample pipe fitting defect image sets and M sample defect feature information sets of M types of pipe fitting defects from the historical data. M is a positive integer. The sample pipe fitting defect image set contains different types of defect images, and the sample defect feature information set contains detailed descriptions of these defects, such as location, size, shape, etc.

[0034] M image processing operators are randomly called from the operator library. The operator library is a collection of various image processing algorithms for stainless steel pipe fittings, which can be used to perform various operations on images. Image processing operators refer to algorithms used to perform specific image processing tasks, including image processing functions such as edge detection, filtering, and segmentation. Then, the M image processing operators are optimized using the M sample pipe defect image sets and the M sample defect feature information sets to improve the accuracy and efficiency of identifying specific types of defects. For example, two image processing operators, edge detection and threshold segmentation, are randomly selected from the operator library, and then the sample pipe defect images containing two common defects, pores and cracks, are processed using edge detection and threshold segmentation. The edge detection operator detects edge information in the image, which can help highlight the contours of defects such as cracks. The threshold segmentation operator can separate defects from the background. Then, the edge detection operator is optimized using the sample pipe defect image set and the sample defect feature information set, which can better capture the fine edges of the cracks. The optimized threshold segmentation operator can ensure that the pore defects are clearly marked in the image, thereby better identifying the defect type in the image. The optimized M image processing operators form M defect recognition branches, each of which focuses on identifying a specific defect type. Next, these M defect recognition branches are connected in parallel to generate a comprehensive processing defect recognition network that can identify multiple types of defects.

[0035] Finally, the post-weld image array is input into the processing defect recognition network to perform defect location and obtain the defect array to be processed, including K defect feature information of K spatial defects, each defect feature information has a defect type identification. By efficiently and accurately detecting and locating the defects of semi-finished pipe fittings after welding, the defect location on the pipe fitting can be quickly found, and detailed defect feature information can be extracted, which improves the efficiency and accuracy of defect detection and location, and provides a basis for subsequent defect processing.

[0036] In one embodiment, Figure 3As shown, a defect processing analysis network is pre-constructed, and the method further includes: interactively obtaining multiple sample first process operations, multiple sample first process regulations and multiple sample first defect feature information of a first pipe defect type, wherein the first pipe defect type is any one of the M types of pipe defect types; aggregating the multiple sample first process operations, multiple sample first process regulations and multiple sample first defect feature information according to the process operation to obtain H sample first process regulation sets and H sample first defect feature information sets of H process operations; randomly calling to obtain H data processing operators from the operator library, and using the H sample first process regulation sets and H sample first defect feature information sets to perform the H data processing operator optimization to obtain H defect processing analysis branches; connecting the H defect processing analysis branches in parallel to obtain a first defect processing analysis channel; and so on, constructing M defect processing analysis channels of the M types of pipe defect types; connecting the M defect processing analysis channels in parallel to obtain the defect processing analysis network.

[0037] Specifically, first, a plurality of sample data about the first pipe defect type are obtained through interaction, wherein the plurality of sample data include a sample first process operation, a sample first process specification, and a sample first defect feature information, and the first pipe defect type is any one of the M pipe defect types. Next, the plurality of sample data obtained are aggregated according to different process operations. Specifically, the sample data with the same process operation, that is, the sample first process operation, the sample first process specification, and the sample first defect feature information with the same process operation are classified into one category to form H sample first process specification sets of H process operations and corresponding H sample first defect feature information sets, each set containing a set of sample data with the same process operation. Then, H data processing operators are randomly called from a pre-built operator library, and the data processing operators include statistical analysis, machine learning models, data mining technology, etc., which are used to analyze and process process specifications and defect feature data, and the H sample first process specification sets and H sample first defect feature information sets are used to execute these H data processing operators respectively, and optimize and adjust them to obtain H defect processing analysis branches for different process operations. For example, the operator library randomly calls two data processing operators, decision tree and neural network, and uses decision tree to identify the level of defects. By optimizing the decision tree, the level information of defects can be further accurately determined. The neural network is used to identify different types of welding defects from the image. After optimization, a more in-depth defect type analysis can be performed to improve the recognition accuracy. Then, through the above optimization process, two defect processing analysis branches are formed: decision tree branch and neural network branch. After obtaining H defect processing analysis branches, the H defect processing analysis branches are connected in parallel to form a defect processing analysis channel for the first pipe defect type. The defect processing analysis channel can comprehensively consider the impact of different process operations on defects and provide corresponding processing and analysis solutions.

[0038] Next, we use the same method to build defect processing and analysis channels for M different types of pipe defects. Each channel is designed and optimized for a specific pipe defect type to ensure that it can accurately and effectively process and analyze the corresponding defect problem. Finally, these M defect processing and analysis channels are connected in parallel to form a complete defect processing and analysis network, which can simultaneously process and analyze M different types of pipe defects and provide comprehensive solutions and decision support.

[0039] Furthermore, surface treatment fitness analysis is performed according to the K process procedure sets to obtain target manufacturing process procedures. The method also includes: performing shortest path fitting according to the K spatial defects in the pipe fitting conceptual model to obtain a target manufacturing process trajectory; enumerating the K process procedure sets in combination to obtain multiple groups of alternative process procedures; serializing the multiple groups of alternative process procedures according to the target manufacturing process trajectory to obtain multiple alternative process procedure series; simulating manufacturing process processing with reference to the multiple alternative process procedure series to obtain multiple alternative processing time consumptions and multiple alternative processing costs; assigning a first weight to the processing time consumption and a second weight to the processing cost, and weightedly calculating the multiple alternative processing time consumptions and multiple alternative processing costs based on the first weight and the second weight to obtain multiple alternative cost coefficients; and obtaining the target manufacturing process procedure by mapping and extracting the multiple alternative process procedure series according to the minimum value of the multiple alternative cost coefficients.

[0040] Specifically, the shortest path algorithm is used to fit the pipe conceptual model and K spatial defects to obtain the target manufacturing process trajectory, where the shortest path algorithm is such as the Dijkstra algorithm, the A* algorithm, etc. The target manufacturing process trajectory refers to the optimal path selection in the manufacturing process to ensure the processing efficiency and quality of the pipe. Then, a mathematical combinatorial algorithm is used to enumerate the K process procedure sets and generate multiple sets of alternative process procedures. The alternative process procedures cover a variety of possible process steps and parameter settings, providing a wealth of options for manufacturing process optimization.

[0041] After obtaining multiple sets of alternative process specifications, the multiple sets of alternative process specifications are arranged in order according to the target manufacturing process trajectory path to form multiple alternative process specification sequences. Next, a manufacturing simulation tool is used to simulate the manufacturing process for each alternative process specification sequence, and during the simulation process, the execution time corresponding to each alternative process specification sequence is recorded, that is, the processing time. At the same time, the direct and indirect costs of each sequence are calculated, including material costs, labor costs, energy consumption, machine depreciation, etc., that is, the processing cost, to obtain multiple alternative processing times and multiple alternative processing costs. Then, a weight is assigned to the processing time and the processing cost respectively, that is, the first weight is assigned to the processing time, and the second weight is assigned to the processing cost. The sum of the first weight and the second weight is 1. The first weight and the second weight can be adjusted according to actual needs to reflect the importance of different factors in process optimization. Based on these two weights, multiple alternative processing times and multiple alternative processing costs are weighted and calculated to obtain multiple alternative cost coefficients. Compare the alternative cost coefficients of all alternative process specifications, find the minimum value among these coefficients, and map this minimum value back to the original series of alternative process specifications to extract the corresponding process specification, which is the target manufacturing process specification. It is possible to select the best solution that meets both production efficiency and cost control among multiple alternative process specifications, so as to process the defects of semi-finished products after welding according to the target manufacturing process specification, realize effective processing of the defects of semi-finished products after welding, ensure that the surface quality of semi-finished products after welding is effectively improved, and thus improve the quality of stainless steel pipe fittings.

[0042] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0043] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for optimizing the manufacturing process of stainless steel pipe fittings, characterized in that: The method comprises: Perform non-destructive testing on semi-finished products after welding and generate apparent defect detection instructions based on the test results; Perform image acquisition on the post-weld semi-finished product according to the apparent defect detection instruction to obtain a post-weld image array; Pre-constructing a processing defect recognition network, and synchronizing the post-weld image array to the processing defect recognition network to perform defect location, and obtaining a defect array to be processed, wherein the defect array to be processed includes K defect feature information of K spatial defects, the defect feature information has a defect type identifier, and K is a positive integer; Pre-constructing a defect processing analysis network, and synchronizing the defect array to be processed to the defect processing analysis network to perform process procedure analysis to obtain a defect process procedure array, wherein the defect process procedure array includes K process procedure sets of the K spatial defects; Performing surface treatment fitness analysis according to the K process specification sets to obtain a target manufacturing process specification; Surface finishing of the welded semi-finished product is performed according to the target manufacturing process specification.

2. The method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 1, characterized in that: Performing nondestructive testing on the semi-finished product after welding, and generating an apparent defect detection instruction based on the detection result, the method also includes: Calling to obtain the conceptual design of the pipe fitting of the semi-finished product after welding; A pipe fitting parameter adjustment model is introduced, and size parameters of the pipe fitting parameter adjustment model are adjusted according to the pipe fitting concept design to generate a pipe fitting concept model; Performing ultrasonic testing on the welded semi-finished product, and obtaining an internal defect feature set based on the obtained ultrasonic signal analysis; Setting a fitting starting point in the pipe fitting conceptual model, and fitting the internal defect feature set to the pipe fitting conceptual model according to the fitting starting point to obtain a pipe fitting defect model; A defect risk assessment network is introduced to evaluate the pipe defect model and generate the apparent defect detection instruction.

3. The method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 2, characterized in that: The method further comprises: The defect risk assessment network includes a cascaded single-dimensional risk assessment layer and a comprehensive risk assessment layer; Introducing defect projection constraints and defect depth constraints into the one-dimensional risk assessment layer; A comprehensive risk assessment function and a comprehensive risk constraint are introduced into the comprehensive risk assessment layer to complete the construction of the defect risk assessment network, wherein the comprehensive risk assessment function is as follows: ； in, is the comprehensive risk factor, is the maximum defect depth, is the mean defect depth, is the defect density, is the maximum defect spacing, is the minimum defect spacing.

4. A method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 3, characterized in that: Introducing a defect risk assessment network to evaluate the pipe defect model and generate the apparent defect detection instruction, the method further includes: Extract and obtain multiple defect projection parameters and multiple defect depth parameters of multiple internal defects based on the internal defect feature set; Synchronizing the multiple defect projection parameters and the multiple defect depth parameters to the one-dimensional risk assessment layer of the defect risk assessment network, and traversing and comparing the multiple defect projection parameters and the multiple defect depth parameters using the defect projection constraints and the defect depth constraints in the one-dimensional risk assessment layer; If the multiple defect projection parameters and the multiple defect depth parameters all satisfy the defect projection constraints and the defect depth constraints, then the maximum defect depth and the defect depth mean are obtained based on the multiple defect depth parameters, and the defect density, the maximum defect spacing and the minimum defect spacing are obtained based on the pipe defect model analysis; The maximum defect depth, the mean defect depth, the defect density, the maximum defect spacing, and the minimum defect spacing are synchronized to the comprehensive risk assessment layer, a comprehensive risk coefficient is obtained by calculating the comprehensive risk assessment function, and it is determined whether the comprehensive risk coefficient falls within the comprehensive risk constraint; If the comprehensive risk coefficient falls within the comprehensive risk constraint, the apparent defect detection instruction is generated.

5. The method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 4, characterized in that: If any of the plurality of defect projection parameters and the plurality of defect depth parameters does not satisfy the defect projection constraint or the defect depth constraint, a scrapping and recycling instruction is generated.

6. The method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 2, characterized in that: Pre-constructing a processing defect recognition network, and synchronizing the post-welding image array to the processing defect recognition network to perform defect location, and obtaining a defect array to be processed, the method further includes: Perform image acquisition on the post-weld semi-finished product according to the apparent defect detection instruction, and splice the image acquisition result to the pipe conceptual model to obtain the post-weld image array; Perform pipe defect integration on the welded semi-finished product, and perform historical data call based on the integration result to obtain M sample pipe defect image sets and M sample defect feature information sets of M types of pipe defect types, where M is a positive integer; Randomly call and obtain M image processing operators from an operator library, and use the M sample pipe defect image sets and the M sample defect feature information sets to perform optimization of the M image processing operators to obtain M defect recognition branches; Connecting the M defect recognition branches in parallel to generate the processing defect recognition network; The post-weld image array is synchronized to the processing defect recognition network to perform defect location, and the defect array to be processed is obtained, wherein the defect array to be processed includes K defect feature information of K spatial defects, and the defect feature information has a defect type identification.

7. A method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 6, characterized in that: Pre-constructing a defect processing analysis network, the method further comprises: Interactively obtaining a plurality of sample first process operations, a plurality of sample first process procedures, and a plurality of sample first defect feature information of a first pipe defect type, wherein the first pipe defect type is any one of the M pipe defect types; Aggregate the plurality of sample first process operations, the plurality of sample first process procedures, and the plurality of sample first defect feature information according to the process operation to obtain H sample first process procedure sets and H sample first defect feature information sets of H process operations; Randomly calling and obtaining H data processing operators from the operator library, and using the H sample first process procedure sets and the H sample first defect feature information sets to perform optimization of the H data processing operators, to obtain H defect processing analysis branches; The H defect processing and analysis branches are connected in parallel to obtain a first defect processing and analysis channel; By analogy, M defect processing and analysis channels for the M types of pipe defects are constructed; The M defect processing and analysis channels are connected in parallel to obtain the defect processing and analysis network.

8. The method for optimizing the manufacturing process of a stainless steel pipe fitting according to claim 7, characterized in that: Performing surface treatment fitness analysis according to the K process specification sets to obtain a target manufacturing process specification, the method further comprising: Performing shortest path fitting on the pipe conceptual model according to the K spatial defects to obtain a target manufacturing process trajectory; Enumerate the K process specification sets in combination to obtain multiple sets of candidate process specifications; Sequencing the multiple groups of candidate process procedures according to the target manufacturing process trajectory to obtain multiple candidate process procedure sequences; Simulating manufacturing process processing by referring to the plurality of candidate process specification sequences to obtain a plurality of candidate processing time consumptions and a plurality of candidate processing costs; Assigning a first weight to the processing time, assigning a second weight to the processing cost, and weightedly calculating the multiple candidate processing time consumptions and multiple candidate processing costs based on the first weight and the second weight to obtain multiple candidate cost coefficients; The target manufacturing process specification is obtained by extracting and mapping the plurality of candidate process specification sequences according to the minimum values ​​of the plurality of candidate cost coefficients.

Images