CN112733451A - Key inspection characteristic identification and extraction method of MBD model - Google Patents
Key inspection characteristic identification and extraction method of MBD model Download PDFInfo
- Publication number
- CN112733451A CN112733451A CN202110036971.1A CN202110036971A CN112733451A CN 112733451 A CN112733451 A CN 112733451A CN 202110036971 A CN202110036971 A CN 202110036971A CN 112733451 A CN112733451 A CN 112733451A
- Authority
- CN
- China
- Prior art keywords
- inspection
- model
- characteristic
- mbd
- extracting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Abstract
The invention provides a method for identifying and extracting key inspection characteristics of an MBD (model based definition) model, which comprises the following steps of: firstly, the method comprises the following steps: analyzing the MBD model to obtain a three-dimensional labeling and inspection tool list set; II, secondly: labeling the attached feature type and the inspection feature type at the later stage, extracting the precision, and matching an inspection tool; thirdly, the method comprises the following steps: adding inspection items on the geometric model and the structural tree; fourthly, the method comprises the following steps: encoding a component chromosome; fifthly: randomly generating an initial population; sixthly, the method comprises the following steps: training an SVM prediction model; seventhly, the method comprises the following steps: calculating an individual adaptation value by using K-fold cross validation; eighthly: selecting excellent individuals to form a new population; nine: crossing and mutating to generate new individuals; ten: if the condition is met, executing the step eleven, otherwise, returning to the step seven; eleven: outputting the key inspection characteristics to an excel table; through the steps, the key inspection characteristics in the MBD model can be identified, then the marking information is extracted, the inspection tool is distributed, and finally the marking information is output to the spreadsheet, so that subsequent guidance is provided for field inspection.
Description
Technical Field
The invention provides a method for identifying and extracting key inspection characteristics based on Model-based Definition (MBD). The method can automatically extract the inspection characteristics from the input MBD model to make an inspection rule, further identifies the key inspection characteristics based on the inspection information, assists in the intelligent decision of the subsequent Process, and belongs to the field of Computer Aided Process design (CAPP).
Background
The method integrates a three-dimensional digital Model of design, manufacture and inspection information Based on Model Based Definition (MBD), completely expresses product Definition information, establishes various associated information among the design-manufacture-inspection information, improves the integration level and the sharing degree of data, and is a product full life cycle (PLM) information carrier. In the product inspection phase, the MBD technique brings a lot of convenience. The two-dimensional drawing is used as a detection basis, and detection and analysis are mostly inconvenient due to detection of irregular shapes. By adopting the MBD technology, information such as formulation of inspection processes, acquisition of tolerance data, planning of detection paths, analysis and feedback of measurement results and the like can be directly acquired from the three-dimensional model. The measured data reconstruction model is compared with the theoretical model, so that the out-of-tolerance can be judged quickly; the analysis of the measurement data and the manufacturing data can improve the efficiency of the formulation of the inspection process and the inspection efficiency. MBD has gained increasing use in the aerospace industry over the last decade of development.
In the manufacturing process of the parts, the inspection link between the working procedures is a means for ensuring the quality of the parts. Although the MBD technology has been applied to some extent in the design stage, it is also weak in the parts processing and inspection links. At present, in the part inspection stage, most of research focuses on the definition of an inspection model, the distribution of measurement points, the measurement path planning and the like, and is not considered from the perspective of an inspection object. For complex parts, the characteristics to be detected are often many, and there is great difficulty in detecting all the characteristics of each product.
Under the background, the recognition and extraction of the key inspection characteristics can automatically analyze and Design the part MBD model of a Computer Aided Design (CAD) system and automatically generate the corresponding inspection rule, and the method has important significance for improving the automation level of CAPP, so that the method is widely researched. However, in the existing research, the operating efficiency of the feature recognition algorithm is low, and meanwhile, the existing achievements are mostly directed at a CAD geometric model, and the research on an MBD model is less, the invention overcomes the defects of the research, and provides a method for recognizing and extracting the key inspection characteristics of the MBD model by carrying out secondary development on a CATIA (Computer Aided Three-dimensional Interactive Application) based on a Component Application Architecture (CAA).
Disclosure of Invention
The invention aims to provide an MBD model key inspection characteristic identification and extraction method for assisting machining inspection process design, so that the defects in the prior art are overcome, on one hand, the field inspection efficiency is improved, on the other hand, geometric information of key characteristics can be extracted, three-dimensional marking information attached to the surface of a machined feature and engineering annotation information on a model structure tree can be extracted, inspection tools are automatically distributed to the three-dimensional marking information and the engineering annotation information, and inspection procedures are generated, so that the manual participation is reduced, and the time cost of an inspection link is reduced.
(II) technical scheme
The invention mainly comprises two parts, wherein the first part is the identification of key test characteristics, and the second part is the extraction of geometric information and non-geometric information of the key test characteristics and the generation of a test procedure. The key inspection characteristic identification part adopts a genetic method (GA) and a Support Vector Machine (SVM), the core theory is that the genetic method is used for quickly searching a key inspection characteristic solution set, and an SVM classifier is used for calculating the fitness of the characteristic set, namely the classification precision, the class II error rate and the number of characteristics in the characteristic set, and the method is shown in figure 1. Genetic algorithms start with a population (population) representing a possible potential solution set to the problem, and a population is composed of a certain number of individuals (induvidual) encoded by genes (gene). Each individual is in fact a chromosome (chromosome) characterized entity; in the genetic method, after an initial population is formed by encoding, the task of genetic operation is to apply certain operation to individuals of the population according to the fitness of the individuals to the environment, so that the evolution process of high-quality and low-quality is realized; from the perspective of optimization search, genetic operations can optimize the solution of the problem, one generation after another, and approach the optimal solution; in the invention, the type II error rate, the characteristic number and the classification precision are all used as indexes for measuring the effectiveness of the characteristic set, so that the key detection characteristic identification problem can be converted into a multi-objective optimization problem, and a rapid non-dominated sorting method (NSGA-II) in a genetic method is adopted; compared with the traditional single-target optimization method (GA), the NSGA-II method does not need to determine a comprehensive fitness function, so that the difficulty that the relative weight of each sub-target in the comprehensive fitness function is difficult to determine is avoided;
in the second part, checking the three-dimensional labeling information on the basis and the model; the three-dimensional label is associated with the geometric elements (lines and surfaces) of the MBD model and contains the description of the information such as the size, tolerance and surface roughness of a specific shape; the labels are divided into three types, namely size labels, tolerance labels and other labels; the dimension labels comprise information such as linearity, diameter, radius, angle and the like, the tolerance labels comprise information such as roundness, verticality, parallelism and the like, and other labels comprise information such as surface roughness and the like, as shown in table 1; the invention is mainly obtained by extracting the three-dimensional label with the attached characteristics and the engineering annotation information on the information tree of the part structure by means of an Application Programming Interface (API) provided by CAD.
TABLE 1 three-dimensional annotation types and annotation contents
The invention discloses a method for identifying and extracting key inspection characteristics of an MBD model, which comprises the following steps:
the method comprises the following steps: resolving the MBD model to obtain a three-dimensional annotation list set (ListAnno) of the MBD model and obtain a preset inspection tool list set (Listtool) of a database;
step two: traversing the labels in ListAnno, acquiring attached features, judging whether the feature types are fillets and chamfers, if so, extracting geometric information of the features, and acquiring corresponding inspection tools in Listtool; otherwise, acquiring the inspection characteristic type, judging whether the inspection characteristic type is a tolerance mark, if so, extracting the reference of the inspection characteristic type, and then distributing an inspection tool for the inspection characteristic type; otherwise, extracting characteristic precision, and then acquiring a corresponding inspection tool from the Listtool; storing the matching result into a generated inspection properties container (vectorInstructions Characteristics), see FIG. 3;
step three: associating the inspection projects with the three-dimensional model, visually marking the model, and adding inspection information sub-nodes on the structure tree to realize storage and management of the inspection projects;
step four: acquiring the total number (N) of the test characteristic sets, and arranging and coding the test characteristic sets according to the sequence of test characteristic serial numbers to form a chromosome body;
step five: randomly generating M chromosomes to form an initial population, namely a chromosome array;
step six: dividing a data set into a training set and a testing set, and training an SVM prediction model by using a training sample;
step seven: performing K-fold cross validation, inputting the test set corresponding to each chromosome into an SVM prediction model, calculating the adaptive value f (x) of each individual in the population, and storing the adaptive value f (x) into a container (vectorFit) with generated adaptive values;
step eight: selecting individuals with excellent performance from the old population to the new population by adopting a non-dominant sorting method;
step nine: randomly generating a cross probability (P)ci) Performing cross operation, selecting a plurality of parents to pair, and generating a new individual; followed byMachine-generated variation probability (P)mi) Performing variation operation to determine variant individuals;
step ten: judging whether a preset termination condition is met, if so, terminating the technology to obtain an optimal solution, executing the step eleven, and if not, returning to the step seven;
step eleven: analyzing chromosome information of the optimal solution, deleting non-key inspection characteristics on the MBD model, and outputting the inspection characteristics to an excel table in a formatted manner, so that the field inspection work can be guided conveniently.
The MBD model in the step one is a model set which is added with abundant manufacturing semantic information on the basis of a traditional CAD geometric solid model, and a typical MBD model example and an information structure thereof are shown in an attached figure 1;
the specific method of "analyzing the MBD model" in step one is as follows: the method comprises the steps of obtaining an MBD model pointer pDOc in an editor, obtaining a part pointer pitPSDOc according to the pDOc, obtaining a part marking information list according to the part pointer and a marking set obtaining function GetSets () provided by a system, and storing the part marking information list in ListAnno.
Wherein, the step two of obtaining the attached feature is as follows: and acquiring a feature path identifier (piTPS) through three-dimensional labeling, and calling an acquisition labeling set function (GetTTRS () provided by the system to acquire a label attachment feature.
Wherein, the method for adding the inspection information child node on the structure tree in the step three is as follows: acquiring an MBD model pointer pDOc in an editor, acquiring the initialization of CATPArt through a part pointer, calling an acquisition part function GetPlart () provided by a system to retrieve part elements, storing the part elements into a part list pIPRtPlart, scanning a tree structure to acquire all geometric figure sets, storing the geometric figure sets into a geometric figure set list GSList, calling a creation geometric figure function CreateGeometricalSet () provided by the system to establish a 'test procedure' geometric set, and adding a test item into a test procedure.
The specific method of "arranging codes according to the sequence of the inspection characteristic sequence numbers" in the fourth step is that a one-dimensional Array (Array) with a length of N is generated to represent a chromosome, and in the ith position, namely, Array [ i ] can take a value of 0 or 1, and 0 represents that the ith inspection characteristic is not selected, otherwise, the ith inspection characteristic is selected.
Wherein, the step five of randomly generating M chromosomes to form an initial population, i.e. a chromosome array, is as follows: each one-dimensional array represents a chromosome, numbers of 0 and 1 are randomly generated at each position of the one-dimensional array, the value of the chromosome is calculated, and M chromosomes with different values are randomly generated according to the method.
In the step six, the SVM prediction model is trained, that is, a hyperplane is found to divide the part into a qualified part and a unqualified part, and the specific method is as follows: the input of the model is the characteristic value of the part, and the output of the model is whether the quality of the part is qualified or not; for a sample with sample number i, the input of the model is the characteristic value { X _1^ ((i)), X _2^ ((i)) … X _ n ^ ((i)) }, noted as X ^ (i)), and the part quality original label noted as L ^ ((i)); finally, the SVM model obtains a separation hyperplane, wherein one side of the hyperplane is an unqualified sample, and the other side of the hyperplane is a qualified sample; the separation hyperplane finally solved by the SVM model consists of two parameters: each characteristic weight { ω _1, ω _2, …, ω _ n } is recorded as ω and b; a hyperplane calculation formula, wherein x represents a key inspection property value;
wTx+b＝0
a part quality calculation formula, wherein y represents the part quality, +1 represents unqualified, and 0 represents qualified;
SVC (C, kernel, gamma, precision _ function _ shape); in the part quality prediction model, the SVM model mainly uses the following parameters:
(1) kernel function (kernel): the method comprises the steps of specifying the category of an application kernel function in the method, and taking values as { linear, poly, rbf, sigmoid, precomputed }; in general, the rbf kernel has higher accuracy and robustness;
(2) decision function (decision _ function _ shape): decision function classification modes take the values of { ovo (one-vs-one), ovr (one-vs-rest) and None }, wherein the two classifications, the multi-classification and the other classifications are respectively adopted;
(3) penalty term coefficient C: representing a penalty term coefficient, and influencing the distance between the support vector and the decision plane; a small value can lead to model under-fitting, and an excessive value can lead to model over-fitting;
(4) kernel function coefficient gamma: and high-dimension mapping is carried out on low-dimension samples, the larger the gamma value is, the higher the mapping dimension is,
the better the training results, but it is prone to overfitting, i.e. low generalization ability.
The specific method of the K-fold cross validation in the seventh step is that the data set is equally divided into K parts, 1 part of the data set is selected as test data in each round of flow, and the rest K-1 parts of data are used as training data; putting the data into a model to be trained and predicted respectively, and averaging the prediction results; for K-fold cross validation, the effect is better when K takes the total number n of training data, and the method is also called as leave-one-out method; but when the data volume is large, the calculation cost of using the leave-one-out method far exceeds the bearing capacity of the people; generally, K is approximately equal to log (n) and n/K is ensured to be greater than 3d, and d represents the characteristic number of the input matrix;
wherein, in the seventh step, the "adaptive value" has the following indexes: classification accuracy, class II error rate, characteristic number of solution set. Wherein type II error rate is also referred to as recall rate; the characteristic number of the solution set is the number of '1' in the chromosome array; the method for calculating the classification precision and the recall rate comprises the following steps:
the classification results are described by a matrix, as shown in table 2;
TP: actually positive, and divided into positive sample number, true number;
FP: actually negative, but divided into positive, false positive, numbers of samples;
TN is actually negative and is divided into negative sample number and true negative number;
FN: actually positive, but divided into negative sample numbers, false negative numbers;
TABLE 2 results of the classification
| Actual \ prediction | Is predicted to be positive | Prediction is negative |
| Is actually positive | TP | FN |
| Is actually negative | FP | TN |
And (3) classification precision: predicting the proportion of the correct result in the total number; the compound can be obtained by the following formula,
the recall ratio is as follows: the sample which is predicted to be positive and is predicted to be correct accounts for the proportion of all the samples which are actually positive; the high recall rate means that samples which are not easy to distinguish are divided into positive samples if the positive samples can be identified; is obtained from the following formula
Wherein, the "Non-dominant ranking" described in step eight, italian economist v.pareto proposed the concept of a solution-independent solution (Non-dominant set) of multiple targets in 1986; non-dominant solutions, assuming any two solutionsThe solution is named as S1 and S2 respectively, S1 is superior to S2 for all targets, the solution is called S1 dominating S2, and if the solution of S1 is not dominated by other solutions, the solution is called S1 as non-dominated solution; that is, when two subsets of characteristics, subset 1 has higher classification accuracy than subset 2, and the number of characteristics of subset 1 is less than that of subset 2, then subset 1 can be considered as dominating subset 2; the non-dominated sorting is detailed by traversing the population P and calculating two parameters n of each individual PpAnd spWherein n ispNumber of individuals, s, dominating individual p in a populationpA set of individuals within the population that are dominated by individual p;
(1) find all n in the populationp0 and stored in the set F
(2) For each individual i in the current set F, the set of individuals it governs is SiGo through SiIs performed with nl-1, if nl is 0, the individual l is saved in the set H
(3) Note that the individuals obtained in F are the individuals of the first non-dominant layer, and with H as the current set, the above operation is repeated until the entire population is graded.
Through the steps, the key inspection characteristics in the MBD model can be successfully identified, after the identification of the key inspection characteristics is finished, the three-dimensional labeling information can be extracted and an inspection tool can be distributed to the three-dimensional labeling information, a complete inspection item is constructed, and finally the identification and extraction results of the key inspection characteristics are structurally output to Excel (Microsoft Office Excel, electronic forms), so that follow-up guidance is provided for field inspection.
(III) advantages and benefits
The invention provides a method for identifying and extracting key inspection characteristics of an MBD (model based definition) model for a machining part inspection process design. Compared with the prior art, the effect is positive and obvious. First, in the part inspection stage, most of research focuses on defining an inspection model, distributing measurement points, planning measurement paths, and the like, and is not considered from the viewpoint of an inspection object. Aiming at the defects of complex machining inspection steps and long consumed time, the invention provides a key quality characteristic identification technology based on measured data, and reduces objects in the inspection process. Secondly, when identifying key inspection characteristics, in order to prevent errors caused by unbalanced data, the class II error rate is used as an index for measuring a characteristic set, so that the NSGA-II method is adopted, and compared with the traditional genetic method, the search efficiency is improved and the global search space is reduced. And thirdly, compared with the traditional characteristic extraction method only aiming at the CAD model, the method increases the screening of key characteristics, allocates corresponding inspection tools for the key characteristics and finally forms a complete inspection project. Finally, the whole process of the method is realized automatically, the operation is simple and convenient, the identification and extraction of key inspection characteristics can be realized conveniently, and the operation time is saved.
Drawings
FIG. 1 is a flow chart of the key verification feature identification and extraction method of the present invention.
FIG. 2(a) is an example of a typical MBD model of the present invention; FIG. 2(b) is the MBD model information structure of the present invention.
FIG. 3 is a flow chart of a method of matching inspection tools of the present invention.
The numbers, symbols and codes in the figures are explained as follows:
the symbols of FIG. 1 have the following meanings: gen is the number of population iterations, and 1 is added to the number of population iterations for each cycle.
The foreign language numbers referred to in this specification are summarized as follows:
MBD, Model-based Definition, is based on Model Definition
CAPP, Computer Aided Process Planning
CAD, Computer Aided Design
PLM, Product Lifecycle Management, Product full life cycle
CAA, a Component Application Architecture, is based on a Component Application Architecture
Computer graphics-assisted Three-dimensional Interactive Application, Computer Aided Three-dimensional Interactive Application, CATIA
Detailed Description
The invention will be further described with reference to the accompanying figures 1-3 and examples, without limiting the invention thereto.
The method is an actual workpiece of an aviation enterprise, and the key inspection characteristics of the MBD model of the actual workpiece are identified. The embodiment is implemented by taking python3.0, Microsoft Visual Studio 2005 and RADE V5R18 as development platforms based on the CATIA CAA secondary development technology, and the following steps are specific steps of the embodiment of the invention:
the invention discloses a method for identifying and extracting key inspection characteristics of an MBD model, which comprises the following specific implementation steps of:
the method comprises the following steps: parsing the MBD model (see fig. 1 for an example of the model, but not limiting the invention) obtains listano and ListTool. The specific implementation process comprises the following steps: and opening the MBD model of the part, clicking the part inspection tool bar, and starting the part inspection characteristic extraction dialogue box. And clicking a part feature extraction key, automatically acquiring a part pointer in the current editor by the program, and analyzing the MBD model corresponding to the part pointer to obtain a three-dimensional labeling list set (ListAnno). Compared with the traditional CAD geometric solid model, the MBD model adds a model set of rich manufacturing semantic information, and a typical MBD model comprises a solid model, a design reference, a three-dimensional label, other information and engineering annotation, which is shown in the attached figure 2 (b). The abundant information contained in the MBD model is the source of information for all subsequent steps.
Step two: go through ListAnno, match the verification tool. The specific implementation process comprises the following steps: and traversing ListAnno, acquiring the attachment features and the labeling types of the ListAnno, and matching the corresponding inspection tools from Listtool.
Step three: and associating the inspection projects with the three-dimensional model, visually marking the model, and adding inspection information sub-nodes on the structure tree to realize storage and management of the inspection projects.
Step four: the chromosome is encoded. In this case, the length of the individual chromosome is the length of all the examined characteristics, and a binary coding mode is adopted. In the following embodiment, the total number (N) of the listano is obtained, which is the total number of genes of one chromosome. The codes are arranged in the order of the check characteristic numbers. A "1" represents the selection of the verification feature and a "0" represents the non-selection of the verification feature.
Step five: the population size determines the selected assay characteristics in a generation. The larger the population size, the easier it is to find a global solution, but with a consequent longer run time, typically selected between 40-100, in this case 60.
Step six: and dividing the data set into a training set and a testing set, and training the SVM prediction model by using the training samples. The part quality "pass/fail" is represented by 1 and 0, respectively. For the sample with sample number i, the input of the model is the inspection characteristic value { X _1^ ((i)), X _2^ ((i)) … X _ n ^ ((i)) }, which is recorded as X ^ (i)), and the original label of the part quality is recorded as L ^ (i)). For chromosome j, n is the number of "1" s in the array of chromosome j. The separation hyperplane finally solved by the support vector machine model consists of two parameters: the check feature weights { ω _1, ω _2, …, ω _ n } are denoted as ω and b. By constructing and solving a constraint optimization problem: the final support vector machine model has the following parameters:
SVM＝SVC(kernel＝'rbf',decision_function_shape＝'ovo',C＝20)
step seven: and (5) cross validation. In this case, five-fold cross validation is employed. And storing the calculation results { f1, f2 …, fn } into a container. Wherein each calculation fi has three different attributes, classification precision (precision)i) Class II error Rate (recall)i) Total number of features (character)i)。
Step eight: non-dominant ordering. The specific implementation process is as follows: initializing a current individual serial number i, a current dominated individual serial number, a container grade serial number k, and a layered individual total size (i, j, k, size are all initialized to 0). After the initialization is finished, the following steps are executed:
(1) judging whether the number i of the current individual is more than or equal to the number m of the population, if so, indicating that the population is completely layered, ending the method, otherwise, turning to (2)
(2) Judging whether the serial number j of the currently dominated individual is less than the number m, if so, j + +, and turning to (3); otherwise the method ends
(3) Judging whether three attributes in the current individual fi are all better than fjIf yes, store j in the container of i dominating individual (S)i) Turning to (5); otherwise, turning to (4);
(4) judging that all three attributes in the fi of the current individual are different from fjIf yes, doi++；
(5) Judgment of doiIf equal to 0, if yes, ranki is equal to 0, store i in the container (V) with level 00)；
(6) i + +, turn (1)
(7) Judging whether the container with the current grade of k is empty and whether the size is smaller than the population number m, if so, the size is equal to the size plus the number of the individuals of the current container, traversing the current container, finding the dominated individual p, and giving the dominated number do of the individualpMinus 1. Otherwise, ending the method;
(8) judgment of dopAnd if the number is equal to 0, storing the data into a container with the level k +1, and otherwise, turning to (7).
Step nine: and (5) cross mutation. The binary coding crossing mode comprises single-point crossing, two-point crossing, multi-point crossing, uniform crossing and the like, the single-point crossing is simple to realize and is suitable for the condition of few chromosome digits, and therefore the single-point crossing is adopted in the scheme. Because the direction of variation is unknown, each bit variation is randomly generated and varies by more than 50%, i.e. the bit changes from 0 to 1 or from 1 to 0. .
Step ten: and judging whether the termination condition is met. Namely, the solution is not changed any more, and the solution can be obtained from iteration times to more than 55 times through multiple verification in the example of the scheme, so that the solution is optimal.
Step eleven: and analyzing chromosome information of the optimal solution, and deleting non-key test characteristics on the MBD model. The specific implementation method comprises the following steps: when the value of the ith gene of the chromosome is 1, the characteristic is kept, otherwise, the characteristic is deleted. And outputting the information on the structure tree to an excel table from the MulList.
Through the steps, the key inspection characteristics in the MBD model can be successfully and effectively identified, after the key inspection characteristics are identified, the three-dimensional marking information can be extracted and the inspection tools can be distributed to the three-dimensional marking information, a complete inspection item is constructed, and finally the key inspection characteristics are identified and the extraction result is structurally output to excel, so that follow-up guidance is provided for field inspection.
Claims (9)
1. A key inspection characteristic identification and extraction method of an MBD model is characterized in that: the method comprises the following steps:
the method comprises the following steps: analyzing the MBD model to obtain a three-dimensional labeling list set ListAnno thereof and obtain a preset inspection tool list set Listtool of the database;
step two: traversing the labels in ListAnno, acquiring attached features, judging whether the feature types are fillets and chamfers, if so, extracting geometric information of the features, and acquiring corresponding inspection tools in Listtool; otherwise, acquiring the inspection characteristic type, judging whether the inspection characteristic type is a tolerance mark, if so, extracting the reference of the inspection characteristic type, and then distributing an inspection tool for the inspection characteristic type; otherwise, extracting characteristic precision, and then acquiring a corresponding inspection tool from the Listtool; storing the matching result into a generated inspection characteristic container vectorInsectionCharacteristics;
step three: associating the inspection projects with the three-dimensional model, visually marking the model, and adding inspection information sub-nodes on the structure tree to realize storage and management of the inspection projects;
step four: acquiring the total number N of the test characteristic sets, and arranging and coding the test characteristic sets according to the sequence of test characteristic serial numbers to form a chromosome body;
step five: randomly generating M chromosomes to form an initial population, namely a chromosome array;
step six: dividing a data set into a training set and a testing set, and training an SVM prediction model by using a training sample;
step seven: performing K-fold cross validation, inputting the test set corresponding to each chromosome into an SVM prediction model, calculating the adaptive value f (x) of each individual in the population, and storing the adaptive value f (x) into a container vectorFit with the generated adaptive value;
step eight: selecting individuals with excellent performance from an old population to a new population by adopting a non-dominant sorting method;
step nine: randomly generating a cross probability PciPerforming cross operation, selecting a plurality of parents to pair, and generating a new individual; randomly generating a probability of variation PmiPerforming variation operation to determine variant individuals;
step ten: judging whether a preset termination condition is met, if so, terminating the technology to obtain an optimal solution, executing the step eleven, and if not, returning to the step seven;
step eleven: analyzing chromosome information of the optimal solution, deleting non-key inspection characteristics on the MBD model, and outputting the inspection characteristics to an excel table in a formatted manner, so that the field inspection work can be guided conveniently.
2. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: the MBD model in the step one is a model set which is added with abundant manufacturing semantic information on the basis of the traditional CAD geometric solid model; the "analytic MBD model" described in the step one is specifically as follows: the method comprises the steps of obtaining an MBD model pointer pDOc in an editor, obtaining a part pointer pitPSDOc according to the pDOc, obtaining a part marking information list according to the part pointer and a marking set obtaining function GetSets () provided by a system, and storing the part marking information list in ListAnno.
3. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: the "obtaining attached features" in step two is implemented as follows: and acquiring a characteristic path identifier (piTPS) through three-dimensional labeling, and calling an acquisition labeling set function (GetTTRS () provided by the system to acquire a label attachment characteristic.
4. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: adding a check information child node on the structure tree in the step three, which is specifically implemented as follows: acquiring an MBD model pointer pDOc in an editor, acquiring the initialization of CATPArt through a part pointer, calling an acquisition part function GetPlart () provided by a system to retrieve part elements, storing the part elements into a part list pIPRtPlart, scanning a tree structure to acquire all geometric figure sets, storing the geometric figure sets into a geometric figure set list GSList, calling a creation geometric figure function CreateGeometricalSet () provided by the system to establish a 'test procedure' geometric set, and adding a test item into a test procedure.
5. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: the specific method of "arranging codes according to the sequence of the sequence numbers of the inspection features" in the fourth step is that a one-dimensional Array representation chromosome with the length of N is generated, at the ith position, namely one of the values 0 and 1 of Array [ i ], 0 represents that the ith inspection feature is not selected, otherwise, the ith inspection feature is selected.
6. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: the random generation of M chromosomes to form the initial population, i.e., chromosome array, in step five is as follows: each one-dimensional array represents a chromosome, numbers of 0 and 1 are randomly generated at each position of the one-dimensional array, the value of the chromosome is calculated, and M chromosomes with different values are randomly generated according to the method.
7. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: in step six, the SVM prediction model is trained, namely a hyperplane is found to divide the part into a qualified part and a disqualified part, and the method comprises the following steps: the input of the model is the characteristic value of the part, and the output of the model is whether the quality of the part is qualified or not; for a sample with sample number i, the input of the model is the characteristic value { X _1^ ((i)), X _2^ ((i)) … X _ n ^ ((i)) }, noted as X ^ (i)), and the part quality original label noted as L ^ ((i)); finally, the SVM model obtains a separation hyperplane, wherein one side of the hyperplane is an unqualified sample, and the other side of the hyperplane is a qualified sample; the separation hyperplane finally solved by the SVM model consists of two parameters: each characteristic weight { ω _1, ω _2, …, ω _ n } is recorded as ω and b; a hyperplane calculation formula, wherein x represents a key inspection property value;
wTx+b＝0
a part quality calculation formula, wherein y represents the part quality, +1 represents unqualified, and 0 represents qualified;
svc (C, kernel, gamma, precision _ function _ shape); in the part quality prediction model, the SVM model uses the following parameters:
(1) kernel function kernel: the method comprises the steps of specifying the category of an application kernel function in the method, and taking values as { linear, poly, rbf, sigmoid, precomputed }; the rbf kernel has higher accuracy and robustness;
(2) decision function decision _ function _ shape: decision function classification modes take the values of { ovo (one-vs-one), ovr (one-vs-rest) and None }, wherein the two classifications, the multi-classification and the other classifications are respectively adopted;
(3) penalty term coefficient C: representing a penalty term coefficient, and influencing the distance between the support vector and the decision plane; a small value can lead to model under-fitting, and an excessive value can lead to model over-fitting;
(4) kernel function coefficient gamma: and (3) carrying out high-dimensional mapping on low-dimensional samples, wherein the larger the gamma value, the higher the mapping dimension is, the better the training result is, but overfitting is easily caused, namely, the generalization capability is low.
8. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: the specific method of the K-fold cross validation in the seventh step is that the data set is equally divided into K parts, 1 part of the K parts is selected as test data each time in turn, and the rest K-1 parts are selected as training data; putting the data into a model for training and predicting respectively, and averaging the prediction results; for K-fold cross validation, the effect is good when K takes the total number n of training data, and the method is also called as leave-one-out method; however, when the data volume is large, the calculation cost of using the leave-one-out method far exceeds the bearing capacity; selecting K ≈ log (n) and ensuring that n/K >3d, wherein d represents the characteristic number of the input matrix;
the "adaptive value" described in step seven is indicated by: classification accuracy, class II error rate, characteristic number of solution set; wherein type II error rate is also referred to as recall rate; the characteristic number of the solution set is the number of '1' in the chromosome array; the method for calculating the classification precision and the recall rate comprises the following steps:
the classification results are described by a matrix, as shown in table 2;
TP: actually positive, and divided into positive sample number, true number;
FP: actually negative, but divided into positive, false positive, numbers of samples;
TN is actually negative and is divided into negative sample number and true negative number;
FN: actually positive, but divided into negative sample numbers, false negative numbers;
TABLE 2 results of the classification
And (3) classification precision: predicting the proportion of the correct result in the total number; the compound can be obtained by the following formula,
the recall ratio is as follows: the sample which is predicted to be positive and is predicted to be correct accounts for the proportion of all the samples which are actually positive; the high recall rate means that samples which are not easy to distinguish are divided into positive samples if the positive samples can be identified; is obtained from the following formula
9. The method for identifying and extracting the key inspection characteristics of the MBD model according to claim 1, wherein: in the "non-dominated sorting" in the step eight, any two solutions are named as S1 and S2, respectively, for all targets, S1 is better than S2, and is called that S1 dominates S2, and if the solution of S1 is not dominated by other solutions, is called that S1 is a non-dominated solution; that is, when two subsets of characteristics, subset 1 has higher classification accuracy than subset 2, and the number of characteristics of subset 1 is less than that of subset 2, then subset 1 is considered to dominate subset 2; the non-dominated sorting is detailed by traversing the population P and calculating two parameters n of each individual PpAnd spWherein n ispNumber of individuals, s, dominating individual p in a populationpIs an individual set dominated by an individual p in the population;
(1) find all n in the populationp0 and stored in the set F
(2) For each individual i in the current set F, the set of individuals it governs is SiGo through SiIs performed with nl-1, if nl is 0, the individual l is saved in the set H
(3) Note that the individuals obtained in F are the first non-dominant level of individuals, and repeat the above operations with H as the current set until the entire population is graded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110036971.1A CN112733451B (en) | 2021-01-12 | 2021-01-12 | Key inspection characteristic identification and extraction method of MBD model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110036971.1A CN112733451B (en) | 2021-01-12 | 2021-01-12 | Key inspection characteristic identification and extraction method of MBD model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112733451A true CN112733451A (en) | 2021-04-30 |
| CN112733451B CN112733451B (en) | 2022-05-10 |
Family
ID=75590607
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110036971.1A Active CN112733451B (en) | 2021-01-12 | 2021-01-12 | Key inspection characteristic identification and extraction method of MBD model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112733451B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114547405A (en) * | 2022-02-22 | 2022-05-27 | 吉林大学 | Feature selection method with low cost and high accuracy |
| CN114611473A (en) * | 2022-05-11 | 2022-06-10 | 希维科技(广州)有限公司 | Generation method of inspection execution file and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105242538A (en) * | 2015-10-12 | 2016-01-13 | 南京航空航天大学 | Layer-based part multi-processing procedure MBD model and realization method thereof |
| CN105718653A (en) * | 2016-01-20 | 2016-06-29 | 西北工业大学 | Automatic checking method of annotated information completeness for MBD (Model Based Definition) process model |
| CN107180138A (en) * | 2017-06-05 | 2017-09-19 | 武汉开目信息技术股份有限公司 | A kind of part machinability analysis method and system based on MBD models |
| CN109117560A (en) * | 2018-08-17 | 2019-01-01 | 武汉理工大学 | Automobile typical machined part three-dimensional process design method and platform based on MBD |
| CN110795797A (en) * | 2019-09-26 | 2020-02-14 | 北京航空航天大学 | MBD model processing feature recognition and information extraction method |
-
2021
- 2021-01-12 CN CN202110036971.1A patent/CN112733451B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105242538A (en) * | 2015-10-12 | 2016-01-13 | 南京航空航天大学 | Layer-based part multi-processing procedure MBD model and realization method thereof |
| CN105718653A (en) * | 2016-01-20 | 2016-06-29 | 西北工业大学 | Automatic checking method of annotated information completeness for MBD (Model Based Definition) process model |
| CN107180138A (en) * | 2017-06-05 | 2017-09-19 | 武汉开目信息技术股份有限公司 | A kind of part machinability analysis method and system based on MBD models |
| CN109117560A (en) * | 2018-08-17 | 2019-01-01 | 武汉理工大学 | Automobile typical machined part three-dimensional process design method and platform based on MBD |
| CN110795797A (en) * | 2019-09-26 | 2020-02-14 | 北京航空航天大学 | MBD model processing feature recognition and information extraction method |
Non-Patent Citations (1)
| Title |
|---|
| 于勇 等: "基于MBD模型的工序模型构建方法", 《浙江大学学报（工学版）》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114547405A (en) * | 2022-02-22 | 2022-05-27 | 吉林大学 | Feature selection method with low cost and high accuracy |
| CN114611473A (en) * | 2022-05-11 | 2022-06-10 | 希维科技(广州)有限公司 | Generation method of inspection execution file and electronic equipment |
| CN114611473B (en) * | 2022-05-11 | 2022-08-12 | 希维科技(广州)有限公司 | Generation method of inspection execution file and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112733451B (en) | 2022-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109117560B (en) | Three-dimensional process design method and platform for machining parts of automotive typical machine based on MBD | |
| CN104881706B (en) | A kind of power-system short-term load forecasting method based on big data technology | |
| Shatkay et al. | Genes, themes and microarrays | |
| CN112733451B (en) | Key inspection characteristic identification and extraction method of MBD model | |
| CN109101632A (en) | Product quality abnormal data retrospective analysis method based on manufacture big data | |
| CN109902714B (en) | Multi-modal medical image retrieval method based on multi-graph regularization depth hashing | |
| Li et al. | Feature selection for high dimensional data using weighted k-nearest neighbors and genetic algorithm | |
| CN114119057B (en) | User portrait model construction system | |
| Lu et al. | An effective ABC-SVM approach for surface roughness prediction in manufacturing processes | |
| CN112463971A (en) | E-commerce commodity classification method and system based on hierarchical combination model | |
| CN106372660A (en) | Spaceflight product assembly quality problem classification method based on big data analysis | |
| CN107016416B (en) | Data classification prediction method based on neighborhood rough set and PCA fusion | |
| CN112182152A (en) | Sina microblog user emotion influence analysis method based on deep learning | |
| CN103902759B (en) | A kind of build-up tolerance Optimization Design based on genetic algorithm | |
| Liu et al. | A typical process route discovery method based on clustering analysis | |
| Khan et al. | Optimization of constrained function using genetic algorithm | |
| CN112508363A (en) | Deep learning-based power information system state analysis method and device | |
| CN116629716A (en) | Intelligent interaction system work efficiency analysis method | |
| Xu et al. | A data mining-based strategy for direct multidisciplinary optimization | |
| CN109639283A (en) | Workpiece coding method based on decision tree | |
| CN111105041A (en) | Machine learning method and device for intelligent data collision | |
| US11947491B2 (en) | Apparatus and methods for geometric searching | |
| Gavagsaz | Efficient Parallel Processing of k-Nearest Neighbor Queries by Using a Centroid-based and Hierarchical Clustering Algorithm | |
| CN105183804A (en) | Ontology based clustering service method | |
| CN116304726A (en) | Material similarity analysis method based on semantic library and knowledge graph |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |