Intelligent process design method and system for part machining

Abstract

The invention relates to an intelligent process design method and system for part machining, and belongs to the technical field of mechanical engineering, and the method comprises the steps: firstly preprocessing an MBD three-dimensional model of a part, and obtaining a feature view, PMI information and a process picture; a processing feature recognizer based on a deep neural network YOLOv5 is used to recognize processing features; extracting process information on the processing features according to a PMI information extraction method; matching processing feature processing elements in the process knowledge base through a comprehensive weighting algorithm; sorting the processing elements by adopting a processing element intelligent sorting and optimization algorithm to generate an optimal process route; other basic information of the process is collected in a man-machine interaction mode, and a process card is generated. According to the method, part machining feature recognition and intelligent process design are combined, so that design information and manufacturing information are communicated, the purpose of effectively utilizing a three-dimensional design model to carry out process design is achieved, meanwhile, a reasonable machining process route with high efficiency can be rapidly obtained, and the intelligence of the process design is improved.

Description

Technical Field

[0001] The present invention relates to the field of mechanical engineering technology, and in particular to a method and system for intelligent process design of parts processing. Background Art

[0002] The deep integration of digital technology and manufacturing technology is triggering far-reaching industrial changes, forming new manufacturing models, production methods and industrial forms.

[0003] Process design is one of the main tasks of the technical department of manufacturing enterprises. Its quality and design efficiency have a great impact on production organization, product quality, product cost, productivity, production cycle, etc. With the development of computer technology, computer-aided process design systems have been widely used in process design. The advantages of applying process design systems are mainly: with the help of process design systems, process personnel can quickly compile complete and detailed process plans and process documents, greatly improve work efficiency, shorten process preparation time, and thus speed up the production of new products. In addition, automatic recognition of processing features is a core technology and key support in the field of intelligent design and manufacturing. In all computer-aided design and manufacturing systems that use features for analysis and decision-making, feature recognition is an indispensable basic component.

[0004] However, the current process design system still has problems such as lack of intelligence, digitization and automation. In particular, the existing process design system cannot effectively use three-dimensional design models for process design. Summary of the invention

[0005] In view of the problems existing in the existing process design system, the present invention proposes an intelligent process design method and system for part processing. The method and system intelligently generate the processing technology of the part through the feature information extracted by feature recognition and the process resource database, thereby realizing the intelligent generation of the part processing technology.

[0006] To solve the above problems, the present invention adopts the following technical solutions:

[0007] The intelligent process design method for parts processing comprises the following steps:

[0008] Step 1: Obtain the MBD 3D model of the part in the 3D software UG, and pre-process the MBD 3D model to obtain the processing feature view, PMI information and process pictures of the MBD 3D model;

[0009] Step 2: Use a processing feature identifier to identify processing features of the processing feature view, and display the identification results in the form of images and text, where the processing feature identifier is obtained by training the deep neural network YOLOv5;

[0010] Step 3: Use the PMI information extraction method to extract process information on the processing features, including the size information, roughness information and tolerance information of the processing features;

[0011] Step 4: According to the PMI information, processing features and process information, a comprehensive weighted algorithm is used to match processing elements in the process knowledge base, and the processing element with the highest matching degree is used as the processing element corresponding to the processing feature, and the processing element of the processing feature is displayed and saved. The formula of the comprehensive weighted algorithm is as follows:

[0012]

[0013] Among them, x i Represents the evaluation object, which includes processing feature type, feature size, surface roughness, tolerance grade and part material. i Represents the evaluation object x i The corresponding weight coefficient;

[0014] Step 5: Use the intelligent sorting and optimization algorithm of processing elements to sort the processing elements of all processing features and generate the optimal process route. The intelligent sorting and optimization algorithm of processing elements includes the following steps:

[0015] Step 5.1: Determine the number of optimization iterations to start;

[0016] Step 5.2: Divide all the processing elements of the processing features into a benchmark group and a non-benchmark group according to the benchmark-first rule;

[0017] Step 5.3: According to the rule of roughing first and finishing later and the unique rule of processing element sequence, the reference group and the non-reference group are further subdivided into roughing group, semi-finishing group and finishing group respectively;

[0018] Step 5.4: Divide each group into several level groups according to the rule of first surface and then hole and the rule of layer-by-layer processing in the processing procedure;

[0019] Step 5.5: Randomly exchange the processing elements in each level group, and reassemble them according to the process rules. Each time the tool, machine tool, and fixture are changed, one point is scored. The iteration score is calculated and the sorted combination is saved.

[0020] Step 5.6: Subtract 1 from the number of optimization iterations and return to step 5.2 until the number of optimization iterations is completed. The sorted combination with the lowest score is selected as the optimal process route.

[0021] Step 6: Obtain basic process information, including manufacturing unit, part code, and part name, through human-computer interaction, and generate a process card.

[0022] Accordingly, the present invention also proposes an intelligent process design system for parts processing, the system comprising:

[0023] The preprocessing module is used to obtain the MBD 3D model of the part in the 3D software UG, and preprocess the MBD 3D model to obtain the processing feature view, PMI information and process pictures of the MBD 3D model;

[0024] The recognition module is used to recognize the processing features of the processing feature view using the processing feature recognizer and display the recognition results in the form of images and texts, wherein the processing feature recognizer is obtained by training the deep neural network YOLOv5;

[0025] An extraction module is used to extract process information on processing features using a PMI information extraction method, including size information, roughness information, and tolerance information of the processing features;

[0026] The matching module is used to match processing elements in the process knowledge base using a comprehensive weighted algorithm based on PMI information, processing features and process information, and to use the processing element with the highest matching degree as the processing element corresponding to the processing feature, and to display and save the processing element of the processing feature. The formula of the comprehensive weighted algorithm is as follows:

[0027]

[0028] Among them, x i Represents the evaluation object, which includes processing feature type, feature size, surface roughness, tolerance grade and part material. i Represents the evaluation object x i The corresponding weight coefficient;

[0029] The sorting module is used to sort the processing elements of all processing features using the processing element intelligent sorting and optimization algorithm to generate the optimal process route. The processing element intelligent sorting and optimization algorithm includes a determination submodule, a benchmark division submodule, a subdivision submodule, a level division submodule, a calculation submodule and an output submodule:

[0030] The determination submodule is used to determine the number of optimization iterations to start, and the number is reduced by 1 after each optimization iteration;

[0031] The benchmark division submodule is used to divide the processing elements of all processing features into benchmark groups and non-benchmark groups according to the benchmark precedence rule;

[0032] The subdivision submodule is used to further subdivide the reference group and the non-reference group into a roughing group, a semi-finishing group and a finishing group according to the rule of roughing first and finishing later and the unique rule of the processing element sequence;

[0033] The level division submodule is used to divide each group into several level groups according to the rule of first surface and then hole and the rule of layer-by-layer processing in the processing procedure;

[0034] The calculation module is used to randomly exchange the processing elements in each level group. After recombining according to the process rules, each change of the tool, machine tool, and clamping will get one point, calculate the iteration score, and save the sorting combination of this time;

[0035] The output submodule is used to take the sorting combination with the least score as the optimal process route after the optimization iteration cycle is completed;

[0036] The generation module is used to obtain basic information of the process through human-computer interaction, including the manufacturing unit, part code, and part name, and generate a process card.

[0037] The gain effects of the present invention are:

[0038] (1) The present invention combines part processing feature recognition with intelligent process design, achieving the purpose of intercommunication between design information and manufacturing information, and realizing the effective use of three-dimensional design models for process design;

[0039] (2) The present invention effectively integrates the intelligent generation of feature processing methods based on part processing feature information and the intelligent sorting of process routes, thereby improving the intelligence of process design;

[0040] (3) The present invention uses intelligent sorting and optimization algorithms to sort processing elements, introduces process rules and processing efficiency scoring mechanisms, and can quickly obtain efficient and reasonable processing routes. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of the intelligent process design method for parts processing according to an embodiment of the present invention;

[0042] Figure 2 This is the network structure diagram of the deep neural network YOLOv5;

[0043] Figure 3 Flowchart for deep neural network YOLOv5 training and processing feature recognition;

[0044] Figure 4 is a flow chart of the PMI information extraction method;

[0045] Figure 5 Flowchart of intelligent sorting and optimization algorithm for processing elements;

[0046] Figure 6 This is the MBD 3D model diagram of a typical part;

[0047] Figure 7 This is an image of a typical part after feature recognition. DETAILED DESCRIPTION

[0048] In order to more clearly describe the technical problems, technical solutions and advantages to be solved by the present invention, a detailed description is given below in conjunction with the accompanying drawings and specific embodiments.

[0049] The part processing intelligent process design method proposed in the present invention first pre-processes the part MBD three-dimensional model through UG secondary development technology to obtain the feature view, PMI information and process picture of the part; secondly, the YOLOv5 algorithm is used to identify the processing features of the pre-processed part MBD three-dimensional model, and the identification results are displayed in the form of pictures and texts; then, according to the processing features and the process information of the processing features, the comprehensive weighted algorithm is used to realize the intelligent recommendation of process processing elements, that is, the comprehensive weighted algorithm is used to match the examples in the process knowledge base, and the optimal matching results are displayed and modified to realize the design of the process steps, machine tools, tools, fixtures and measuring tools in the process; finally, the processing elements are used as the sorting target, and the processing elements are reasonably sorted and optimized by using the processing element intelligent sorting and optimization algorithm based on the process rules and the processing efficiency scoring mechanism, so that the sorted process route meets the basic process constraints and certain processing quality requirements, and finally the optimal process processing route, that is, the optimal process route, is generated; finally, some basic information on the process card is obtained by human-computer interaction, including the manufacturing unit, part code, part name and other information, and finally the process card is generated with one click.

[0050] like Figure 1 As shown, the present invention provides an intelligent process design method for parts processing based on feature recognition, which mainly includes the following steps:

[0051] Step 1: Preprocess the MBD 3D model of the part, obtain the feature view, PMI information and process pictures of the MBD 3D model, and generate them into .png files and .avi files respectively;

[0052] Step 2: Automatically identify processing features using a processing feature identifier based on the deep neural network YOLOv5;

[0053] Step 3: Obtain process information on processing features based on the PMI information extraction method;

[0054] Step 4: Based on the constructed process knowledge base, the process information of the part processing feature is used to obtain the processing element of the feature in the process knowledge base through a comprehensive weighted algorithm, that is, intelligently matching the processing element corresponding to the processing feature;

[0055] Step 5: Use intelligent sorting and optimization algorithms to sort all processing elements of processing features reasonably and effectively, and generate the optimal process route;

[0056] Step 6: Collect other basic process information through human-computer interaction and generate a process card with one click.

[0057] Step 1 mainly provides an information basis for other steps. First, the MBD 3D model of the part is obtained in the 3D software UG, and then the MBD 3D model is preprocessed to obtain the processing feature view, PMI information and process pictures of the MBD 3D model.

[0058] The MBD 3D model of a part refers to the 3D modeling technology used in the product design and manufacturing process. It can create a virtual model of the product through computer software to realize the digitization of design, analysis, simulation and documentation. This model can not only visualize the product, but also contain product specifications, dimensions and other design information.

[0059] Feature view refers to the capture of 3D model processing feature views in sequence based on UG secondary development technology, including front view, left view, right view, top view, bottom view and rear view. Figure 6 A basic view.

[0060] The PMI information and process pictures are obtained through UG secondary development technology, including the PMI information such as size, roughness, and tolerance on the part MBD three-dimensional model, which provides an information basis for subsequent functional operations. The screenshot function is used to capture the process pictures of the MBD three-dimensional model. The process pictures provide an information basis for generating process cards in step 6.

[0061] In step 1, the process of preprocessing the MBD 3D model by the 3D software UG specifically includes the following steps:

[0062] Firstly, the machining feature views of the MBD 3D model are captured in sequence through UG secondary development technology. The machining feature views include the front view, left view, right view, top view, bottom view and back view.

[0063] Secondly, the PMI information of the MBD three-dimensional model is obtained through UG secondary development technology. The PMI information includes size, roughness and tolerance, and the screenshot function is used to capture the process pictures of the MBD three-dimensional model.

[0064] In step 2, a processing feature identifier is used to identify processing features of the processing feature view, and the identification result is displayed in the form of image and text, wherein the processing feature identifier is obtained by training the deep neural network YOLOv5.

[0065] YOLOv5 is a single-stage target detection algorithm based on a deep neural network. It is a new algorithm that improves the YOLOv4 algorithm and has significantly improved detection speed and accuracy. YOLOv5 has four network models of different sizes, among which YOLOv5s is the smallest network model, and the network depth and width of the other three network models are gradually increased on the basis of YOLOv5s, and the number of parameters and calculations also increase accordingly.

[0066] The YOLOv5 network structure diagram in this embodiment is as follows Figure 2 As shown in the figure, the YOLOv5 network structure consists of four parts: input, Backbone, Neck and Head output.

[0067] At the input end, the input image is processed by using the Mosaic data augmentation operation and the anchor box adaptive calculation method to improve the training speed and detection accuracy of the model. At the same time, the input image is scaled by the adaptive image scaling method to adapt to target detection tasks of different sizes and proportions.

[0068] A Focus module is added to the feature extraction network, namely the Backbone structure, to perform two operations, namely slicing and channel splicing, on the input image to reduce the number of network parameters and the amount of calculation.

[0069] In the Neck structure, the FPN (Feature Pyramid Network) module and the PAN (Pyramid Attention Network) module are used to achieve feature fusion of feature maps by upsampling and downsampling to enhance the feature fusion performance of the network.

[0070] Different loss functions are used at the head output to calculate position loss, classification loss and confidence loss, and feature prediction and classification are achieved through NMS non-maximum suppression, which improves the network detection accuracy.

[0071] To use the deep neural network YOLOv5 for part processing feature recognition, it is necessary to first train the deep neural network YOLOv5 in order to obtain a better processing feature identifier. Figure 3 As shown, the process of training the deep neural network YOLOv5 includes the following steps:

[0072] Step 2.1: First, obtain several machining feature views through UG secondary development technology to generate a machining feature data set, where the machining feature views can be obtained from the MBD three-dimensional models of different parts;

[0073] Step 2.2: Use LabelImg software to label each processing feature view in the processing feature data set. After labeling, it is automatically saved as an XML file, thereby generating a standard XML format data set, completing the preprocessing of the data set, and obtaining the preprocessed data set;

[0074] Step 2.3: Build a deep neural network YOLOv5 and use the preprocessed data set as the training set of the deep neural network YOLOv5;

[0075] Step 2.4: Use the training set to continuously train the deep neural network YOLOv5, update parameters and optimize the network structure. After training, the deep neural network structure with the best recognition effect is obtained, and the deep neural network YOLOv5 with the best recognition effect is used as the processing feature identifier.

[0076] After obtaining the processing feature identifier, the processing feature identifier is used to perform processing feature recognition. Figure 3 , through UG secondary development technology, the processing feature views of the MBD 3D model in step 1 are captured in sequence, and the processing feature views are input into the processing feature recognizer, the processing feature recognition is performed, the processing feature recognition results are output, and the recognition results are displayed in the form of images and texts. When the recognition results are displayed in the form of images and texts, the functions of adding, deleting, modifying, and viewing are added.

[0077] In step 3, the process information on the processing features is extracted using a PMI information extraction method, where the process information on the processing features includes the size information, roughness information, and tolerance information of the processing features. The PMI information extraction method specifically includes the following steps:

[0078] Step 3.1: traverse all target information in the MBD three-dimensional model and determine whether the target information has been extracted. If the extraction is completed, end the information extraction, otherwise execute step 3.2, where the target information is any one of size information, roughness information and tolerance information;

[0079] Step 3.2: Extract all target information on a processing feature;

[0080] Step 3.3: Determine whether the target information of the extracted processing features is saved in the txt document. If so, return to step 3.2 to extract all the target information on the next processing feature. Otherwise, save the target information of the extracted processing features to the txt document. The loop stops after the target information of all processing features has been extracted.

[0081] Extract the process information on the processing features, including size, roughness, tolerance and other information. The following takes size information as an example to explain the size information extraction process in detail. The extraction process of roughness, tolerance and other information is similar. Figure 4 As shown:

[0082] First, all the dimension information in the MBD 3D model is traversed and it is determined whether the information has been extracted. If all the dimension information has been extracted, the information extraction is terminated, otherwise the next step is executed;

[0083] Extract all dimensional information on a machining feature;

[0084] Determine whether the dimension information of the extracted processing features has been saved in the txt document. If the dimension information of the extracted processing features has not been saved in the txt document, save it to the txt document. If the dimension information of the extracted processing features has been saved in the txt document, return to the previous step to re-extract all the target information on the next processing feature until the dimension information of all processing features has been extracted and stop the loop.

[0085] In step 4, according to the PMI information, processing features and process information, a comprehensive weighted algorithm is used to match processing elements in the process knowledge base, and the processing element with the highest matching degree is used as the processing element corresponding to the processing feature, and the processing element of the processing feature is displayed and saved.

[0086] The comprehensive weighted algorithm is to assign weights to the part processing features, PMI, materials and other information obtained from the upstream steps through the hierarchical analysis method, and then use the comprehensive weighted formula to match the processing elements in the process knowledge base, and display the processing elements with the highest matching degree. When displaying the processing elements of the processing features, add the function of manual modification.

[0087] Intelligent process generation technology matches feature information and part information in the existing process knowledge base based on a comprehensive weighted algorithm to obtain the feature machining method with the highest matching degree, so as to quickly determine the process step information. The formulas of the comprehensive weighted algorithm are shown in formula (1) and formula (2):

[0088]

[0089] Among them, x i represents the evaluation object, w i Represents the evaluation object x iThe corresponding weight coefficient, n represents the total number of evaluation objects. The evaluation objects in the above formula include processing feature type, feature size, surface roughness, tolerance grade and part material, and the evaluation objects are weighted according to the expert experience method. The weight coefficients of processing feature type, feature size, surface roughness, tolerance grade and part material are 50%, 15%, 15%, 10% and 10% respectively. Compare the values ​​of the evaluation results Y. The larger the value, the more the corresponding processing method meets the processing requirements, which is the result recommended by the system.

[0090] In step 5, the processing elements of all processing features are sorted using intelligent sorting and optimization algorithms to generate the optimal process route.

[0091] The intelligent sorting and optimization algorithm of processing elements mainly adopts process rules and adds a scoring mechanism. The process rules include benchmark first, rough first then fine, unique processing element sequence, surface first then hole, etc. The processing efficiency is measured by minimizing the number of tool changes, machine tool changes, and clamping changes. A scoring mechanism is added, and one point is given for each change. The lower the score, the higher the processing efficiency.

[0092] The reasonable and effective sorting of processing elements is achieved through the intelligent sorting and optimization algorithm of processing elements. The flowchart of the algorithm is as follows: Figure 5 As shown, the main steps include:

[0093] Step 5.1: Determine the number N of optimization iterations to start;

[0094] Step 5.2: Divide the processing elements of all processing features into a benchmark group and a non-benchmark group according to the benchmark-first rule (Rule 1);

[0095] Step 5.3: According to the rule of roughing first and finishing later (Rule 2) and the unique rule of processing element sequence (Rule 3), the reference group and non-reference group are further subdivided into roughing group, semi-finishing group and finishing group respectively;

[0096] Step 5.4: According to the rule of first face and then hole (Rule 4) and the rule of layer-by-layer processing in the processing procedure (Rule 5), each group is divided into several level groups. For example, the level of the feature of face is higher than that of the feature of hole.

[0097] Step 5.5: Randomly exchange the processing elements in each level group, and reassemble them according to the process rules. Each time the tool, machine tool, and fixture are changed, one point is scored. The iteration score is calculated, and the sorting combination of this time, i.e., the process route, is saved;

[0098] Step 5.6: Subtract 1 from the number of optimization iterations i, then return to step 5.2 and repeat the above steps until the number of optimization iterations is completed. The sorting combination with the lowest score is taken as the optimal process route.

[0099] In step 6, based on the process picture, some basic information on the process card is obtained through human-computer interaction, including the manufacturing unit, part code, part name and other information. Based on the completion of the above steps, a process card in PDF format is generated, including a process route card and a process card, and the process route card and the process card are saved in the local database.

[0100] The present invention will be further explained below with reference to specific examples.

[0101] When verifying the present invention by examples, an intelligent process design system based on feature recognition can be developed based on the Visual Studio software platform using C++ as the programming language to verify the relevant technologies and systems by examples.

[0102] In Windows 11.0 operating system, the system functional module was developed under Microsoft Visual Studio 2017 programming software. The C / C++ programming language was mainly used for program writing, Qt Toolkit was used as the system application development framework, and the 3D software UG10.0 was combined as the auxiliary tool of the system, and the MySQL database was used as the underlying database. In order to obtain the basic information of the parts, the dynamic link library file (.DLL) generated by the internal development model was embedded in the UG software using UG secondary development technology. By calling the generated DLL file, the specific function of obtaining the basic information of the parts was realized.

[0103] Below Figure 6 The technical solution of the present invention is described in detail by taking the MBD three-dimensional model of a typical part as an example.

[0104] Step 1: Enter the 3D software UG10.0, select the MBD 3D model to be preprocessed, and then use the UG secondary development technology to capture the 3D model processing feature views in sequence, including the front view, left view, right view, top view, bottom view and rear view. Figure 6 The UG secondary development technology is used again to obtain the PMI information such as size, roughness, tolerance, etc. of the MBD 3D model, and the screenshot function is used to capture the process pictures of the 3D model to provide information basis for subsequent functional operations.

[0105] Step 2: Enter the processing feature recognition page, and display the six basic views of the 3D model, i.e. .PNG files. Then, perform processing feature detection on the feature view, and display the detection results in the form of images and text. Figure 7 The image is obtained after the typical part processing features are identified, and the part processing features are saved in the database.

[0106] Step 3: Load the .AVI file, the 3D model will automatically perform operations such as rotation and flipping, and then use the processing feature identifier to automatically identify the features and display the processing feature detection results on the 3D model. Enter the feature recognition result display page and integrate all upstream information about the processing features. The feature information includes name, size, surface roughness, and upper and lower tolerances, and has the functions of adding, deleting, modifying, and checking.

[0107] Step 4: Use human-computer interaction to fill in the processing information about the part and save the information into the database. On the one hand, it provides information sources for the design of the processing element, and on the other hand, it provides content assistance for the generation of process cards. Secondly, the feature information in the part processing feature database is displayed. Then right-click each feature information, and the processing element design page will pop up. Finally, after the processing element design is completed, close the page, and the feature information will automatically turn blue to remind the operator that this feature processing element has been designed to avoid repeated design.

[0108] Step 5: Entering the processing element design page will automatically display feature information and part size information, calculate feature tolerances, and provide information basis for the design of feature processing elements. And display the process steps with the highest matching degree. Then match the tool with the highest similarity for each process step, display the tool model, tool name and tool details. The matching operation of machine tools, fixtures, and gauges is the same as that of tools. Finally, the feature processing element design is completed, and all processing element information is saved in the database.

[0109] Step 6: Fill in some basic information about part generation, the production program will be automatically calculated, and the intelligent assistant will display the specific information of the part and recommend the selection of process principles.

[0110] Step 7: Sort the processing elements according to the process rules to form a reasonable process route. Enter the number of iterations and use the scoring mechanism to re-sort the process steps to make the process route more efficient. In addition, add manual modification operations to make the process route more complete. Finally, integrate the process steps into processes and automatically label the processes. The process dispersion operation is the same as this operation.

[0111] Step 8: The system will automatically fill in the process card content and output the process route chart, arrange all process cards in order of process steps, and finally save the process cards locally. The file name can be modified.

[0112] In another embodiment, the present invention further provides an intelligent process design system for parts processing, the system comprising a preprocessing module, a recognition module, an extraction module, a matching module, a sorting module and a generation module.

[0113] The preprocessing module is used to obtain the MBD three-dimensional model of the part in the three-dimensional software UG, and preprocess the MBD three-dimensional model to obtain the processing feature view, PMI information and process pictures of the MBD three-dimensional model.

[0114] The recognition module is used to perform processing feature recognition on the processing feature view using a processing feature recognizer, and display the recognition result in the form of image and text, wherein the processing feature recognizer is obtained after training by the deep neural network YOLOv5.

[0115] The extraction module is used to extract process information on processing features using the PMI information extraction method, including the size information, roughness information and tolerance information of the processing features.

[0116] The matching module is used to match processing elements in the process knowledge base using a comprehensive weighted algorithm based on PMI information, processing features and process information, and to use the processing element with the highest matching degree as the processing element corresponding to the processing feature, and to display and save the processing element of the processing feature. The formula of the comprehensive weighted algorithm is as follows:

[0117]

[0118] Among them, x i represents the evaluation object, w i Represents the evaluation object x i The corresponding weight coefficient, n represents the total number of evaluation objects. The evaluation objects in the above formula include processing feature type, feature size, surface roughness, tolerance grade and part material, and the evaluation objects are weighted according to the expert experience method. The weight coefficients of processing feature type, feature size, surface roughness, tolerance grade and part material are 50%, 15%, 15%, 10% and 10% respectively. Compare the values ​​of the evaluation results Y. The larger the value, the more the corresponding processing method meets the processing requirements, which is the result recommended by the system.

[0119] The sorting module is used to sort the processing elements of all processing features using the processing element intelligent sorting and optimization algorithm to generate the optimal process route. The processing element intelligent sorting and optimization algorithm includes a determination submodule, a benchmark division submodule, a subdivision submodule, a level division submodule, a calculation submodule and an output submodule:

[0120] The determination submodule is used to determine the number of optimization iterations to start, and the number is reduced by 1 after each optimization iteration;

[0121] The benchmark division submodule is used to divide the processing elements of all processing features into benchmark groups and non-benchmark groups according to the benchmark precedence rule;

[0122] The subdivision submodule is used to further subdivide the reference group and the non-reference group into a roughing group, a semi-finishing group and a finishing group according to the rule of roughing first and finishing later and the unique rule of the processing element sequence;

[0123] The level division submodule is used to divide each group into several level groups according to the rule of first surface and then hole and the rule of layer-by-layer processing in the processing procedure;

[0124] The calculation module is used to randomly exchange the processing elements in each level group. After recombining according to the process rules, each change of the tool, machine tool, and clamping will get one point, calculate the iteration score, and save the sorting combination of this time;

[0125] The output submodule is used to take the sorting combination with the least score as the optimal process route after the optimization iteration cycle is completed.

[0126] The generation module is used to obtain basic process information, including manufacturing unit, part code, part name, etc., through human-computer interaction, generate process cards, and save the process cards to the local database.

[0127] The implementation method of the specific functions of each module in the intelligent process design system for parts processing of the present invention can refer to the implementation method described in the above-mentioned embodiment of the intelligent process design method for parts processing, and will not be repeated here.

[0128] The present invention combines part processing feature recognition with intelligent process design, achieves the purpose of intercommunication between design information and manufacturing information, and realizes effective use of three-dimensional design models for process design.

[0129] The present invention effectively integrates the intelligent generation of feature processing methods based on part processing feature information and the intelligent sorting of process routes, thereby improving the intelligence of process design.

[0130] The present invention uses intelligent sorting and optimization algorithms to sort processing elements, introduces process rules and processing efficiency scoring mechanisms, and can quickly obtain efficient and reasonable processing routes.

[0131] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. An intelligent process design method for parts processing, characterized in that: The following steps are involved: Step 1: Obtain the MBD 3D model of the part in the 3D software UG, and pre-process the MBD 3D model to obtain the processing feature view, PMI information and process pictures of the MBD 3D model; Step 2: Use a processing feature identifier to identify processing features of the processing feature view, and display the identification results in the form of images and text, where the processing feature identifier is obtained by training the deep neural network YOLOv5; Step 3: Use the PMI information extraction method to extract process information on the processing features, including the size information, roughness information and tolerance information of the processing features; Step 4: According to the PMI information, processing features and process information, a comprehensive weighted algorithm is used to match processing elements in the process knowledge base, and the processing element with the highest matching degree is used as the processing element corresponding to the processing feature, and the processing element of the processing feature is displayed and saved. The formula of the comprehensive weighted algorithm is as follows: Among them, x i Represents the evaluation object, which includes processing feature type, feature size, surface roughness, tolerance grade and part material. i Represents the evaluation object x i The corresponding weight coefficient; Step 5: Use the intelligent sorting and optimization algorithm of processing elements to sort the processing elements of all processing features and generate the optimal process route. The intelligent sorting and optimization algorithm of processing elements includes the following steps: Step 5.1: Determine the number of optimization iterations to start; Step 5.2: Divide all the processing elements of the processing features into a benchmark group and a non-benchmark group according to the benchmark-first rule; Step 5.3: According to the rule of roughing first and finishing later and the unique rule of processing element sequence, the reference group and the non-reference group are further subdivided into roughing group, semi-finishing group and finishing group respectively; Step 5.4: Divide each group into several level groups according to the rule of first surface and then hole and the rule of layer-by-layer processing in the processing procedure; Step 5.5: Randomly exchange the processing elements in each level group, and reassemble them according to the process rules. Each time the tool, machine tool, and fixture are changed, one point is scored. The iteration score is calculated and the sorted combination is saved. Step 5.6: Subtract 1 from the number of optimization iterations and return to step 5.2 until the number of optimization iterations is completed. The sorted combination with the lowest score is selected as the optimal process route. Step 6: Obtain basic process information, including manufacturing unit, part code, and part name, through human-computer interaction, and generate a process card.

2. The intelligent process design method for parts processing according to claim 1 is characterized in that: The process of preprocessing the MBD 3D model includes: Through UG secondary development technology, the processing feature views of the MBD 3D model are captured in sequence, including the front view, left view, right view, top view, bottom view and back view; The PMI information of the MBD 3D model, including size, roughness and tolerance, is obtained through UG secondary development technology, and the screenshot function is used to capture the process pictures of the MBD 3D model.

3. The intelligent process design method for parts processing according to claim 1 or 2, characterized in that: The deep neural network YOLOv5 includes input, Backbone structure, Neck structure and Head output; At the input end, the Mosaic data enhancement operation and the anchor box adaptive calculation method are used to process the input image, and the input image is also scaled by the adaptive image scaling method; Add the Focus module in the Backbone structure to perform two operations on the input image: slicing and channel splicing. In the Neck structure, the FPN module and the PAN module are used to achieve feature fusion of feature maps by upsampling and downsampling. Different loss functions are used at the head output to calculate position loss, classification loss and confidence loss, and feature prediction and classification are achieved through NMS non-maximum suppression.

4. The intelligent process design method for parts processing according to claim 3 is characterized in that: The process of training the deep neural network YOLOv5 consists of the following steps: Step 2.1: Obtain several processing feature views through UG secondary development technology and generate a processing feature data set; Step 2.2: Use LabelImg software to label each processing feature view in the processing feature data set. After labeling, it is automatically saved as an XML file, thereby generating a standard XML format data set and obtaining a preprocessed data set. Step 2.3: Build a deep neural network YOLOv5 and use the preprocessed data set as the training set of the deep neural network YOLOv5; Step 2.4: Use the training set to continuously train the deep neural network YOLOv5, update parameters and optimize the network structure, and use the deep neural network YOLOv5 with the best recognition effect after training as the processing feature identifier.

5. The intelligent process design method for parts processing according to claim 1 or 2, characterized in that: The PMI information extraction method includes the following steps: Step 3.1: traverse all target information in the MBD three-dimensional model and determine whether the target information has been extracted. If the extraction is completed, end the information extraction, otherwise execute step 3.2, where the target information is any one of size information, roughness information and tolerance information; Step 3.2: Extract all target information on a processing feature; Step 3.3: Determine whether the target information of the extracted processing features is saved in the txt document. If so, return to step 3.2 to extract all the target information on the next processing feature. Otherwise, save the target information of the extracted processing features to the txt document. The loop stops after the target information of all processing features has been extracted.

6. The intelligent process design method for parts processing according to claim 1 or 2, characterized in that: The weight coefficients of machining feature type, feature size, surface roughness, tolerance grade and part material are 50%, 15%, 15%, 10% and 10% respectively.

7. The intelligent process design method for parts processing according to claim 1 or 2, characterized in that: The processing feature views of the MBD 3D model include front view, left view, right view, top view, bottom view and back view.

8. The intelligent process design method for parts processing according to claim 1 or 2, characterized in that: When the recognition results are displayed in the form of images and text, the functions of adding, deleting, modifying and viewing are added.

9. The intelligent process design method for parts processing according to claim 1 or 2, characterized in that: When displaying the processing elements of the processing features, add the function of manual modification.

10. Intelligent process design system for parts processing, characterized by: include: The preprocessing module is used to obtain the MBD 3D model of the part in the 3D software UG, and preprocess the MBD 3D model to obtain the processing feature view, PMI information and process pictures of the MBD 3D model; The recognition module is used to recognize the processing features of the processing feature view using the processing feature recognizer and display the recognition results in the form of images and texts, wherein the processing feature recognizer is obtained by training the deep neural network YOLOv5; An extraction module is used to extract process information on processing features using a PMI information extraction method, including size information, roughness information, and tolerance information of the processing features; The matching module is used to match processing elements in the process knowledge base using a comprehensive weighted algorithm based on PMI information, processing features and process information, and to use the processing element with the highest matching degree as the processing element corresponding to the processing feature, and to display and save the processing element of the processing feature. The formula of the comprehensive weighted algorithm is as follows: Among them, x i Represents the evaluation object, which includes processing feature type, feature size, surface roughness, tolerance grade and part material. i Represents the evaluation object x i The corresponding weight coefficient; The sorting module is used to sort the processing elements of all processing features using the processing element intelligent sorting and optimization algorithm to generate the optimal process route. The processing element intelligent sorting and optimization algorithm includes a determination submodule, a benchmark division submodule, a subdivision submodule, a level division submodule, a calculation submodule and an output submodule: The determination submodule is used to determine the number of optimization iterations to start, and the number is reduced by 1 after each optimization iteration; The benchmark division submodule is used to divide the processing elements of all processing features into benchmark groups and non-benchmark groups according to the benchmark precedence rule; The subdivision submodule is used to further subdivide the reference group and the non-reference group into a roughing group, a semi-finishing group and a finishing group according to the rule of roughing first and finishing later and the unique rule of the processing element sequence; The level division submodule is used to divide each group into several level groups according to the rule of first surface and then hole and the rule of layer-by-layer processing in the processing procedure; The calculation module is used to randomly exchange the processing elements in each level group. After recombining according to the process rules, each change of the tool, machine tool, and clamping will get one point, calculate the iteration score, and save the sorting combination of this time; The output submodule is used to take the sorting combination with the least score as the optimal process route after the optimization iteration cycle is completed; The generation module is used to obtain basic information of the process through human-computer interaction, including the manufacturing unit, part code, and part name, and generate a process card.

Images