CN116362005A - Reuse design method for similar instance processing technology of injection mold part based on instance reasoning - Google Patents
Reuse design method for similar instance processing technology of injection mold part based on instance reasoning Download PDFInfo
- Publication number
- CN116362005A CN116362005A CN202310163383.3A CN202310163383A CN116362005A CN 116362005 A CN116362005 A CN 116362005A CN 202310163383 A CN202310163383 A CN 202310163383A CN 116362005 A CN116362005 A CN 116362005A
- Authority
- CN
- China
- Prior art keywords
- instance
- similarity
- retrieval
- similar
- layer
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 162
- 238000012545 processing Methods 0.000 title claims abstract description 32
- 238000013461 design Methods 0.000 title claims abstract description 31
- 238000005516 engineering process Methods 0.000 title claims abstract description 30
- 238000002347 injection Methods 0.000 title claims abstract description 23
- 239000007924 injection Substances 0.000 title claims abstract description 23
- 238000012937 correction Methods 0.000 claims abstract description 34
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 26
- 238000004364 calculation method Methods 0.000 claims description 33
- 238000003754 machining Methods 0.000 claims description 26
- 239000000463 material Substances 0.000 claims description 21
- 230000003746 surface roughness Effects 0.000 claims description 21
- 230000000694 effects Effects 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 7
- 239000000243 solution Substances 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 229910000831 Steel Inorganic materials 0.000 description 3
- 239000010959 steel Substances 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- FFBHFFJDDLITSX-UHFFFAOYSA-N benzyl N-[2-hydroxy-4-(3-oxomorpholin-4-yl)phenyl]carbamate Chemical compound OC1=C(NC(=O)OCC2=CC=CC=C2)C=CC(=C1)N1CCOCC1=O FFBHFFJDDLITSX-UHFFFAOYSA-N 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000013278 delphi method Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000013450 outlier detection Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a reuse design method for a processing process of a similar instance of an injection mold part based on instance reasoning. The basic information and the processing technology attribute of a certain type of part of the injection mold are taken as example retrieval attributes, similar example retrieval is carried out, and the method for rapidly re-using the technology is realized by combining technology correction based on feature similarity. Clustering similar part clusters based on an improved K-means algorithm according to the basic information and other attributes of the parts to obtain a clustering result of similar part example clusters; according to the processing technology attribute of the parts in the similar part cluster, calculating the overall similarity retrieval by adopting a KNN algorithm; the method adopts a double-layer search model based on the combination of improved K-means and KNN algorithm, improves the search efficiency and accuracy of similar parts, and realizes the rapid, accurate and intelligent generation of the target part processing technology file based on the feature similarity of example parts.
Description
Technical Field
The invention relates to a reuse design method for a processing technology of a similar instance of an injection mold part based on instance reasoning, belonging to the technical analysis field of the reuse design technology in the technology CAPP. Aiming at similar structural parts of the injection mold, the processing technical scheme of the injection mold is rapidly obtained.
Background
The process reuse design is always a hot problem in the CAPP field research, along with the wide application of the digital design and manufacturing technology, a large number of process examples are accumulated by a plurality of enterprises through constructing a process example library, the existing process is effectively utilized, and the processing process of a new product target part is rapidly and accurately generated, so that the method is one of important ways for improving the process design efficiency of the enterprises and reducing the processing process design period of the parts.
The processing technology design is to make an economic and reasonable processing scheme and processing technological process according to the requirements of materials, dimension specifications, structural characteristics, precision and roughness of the parts. At present, the traditional injection mold part processing technology mainly relies on experience of a process engineer, analysis of part structural features and programming of working procedures are carried out manually, programming efficiency is low, a design period is long, historical cases of previous designs are not recorded and effectively reused, the requirements of various, small-batch and efficient intelligent production of injection molds cannot be met, and development of efficient intelligent manufacturing of mold parts is severely restricted. Therefore, it is necessary to research how to quickly retrieve similar part examples and use the process thereof as a modified process skeleton, and to fuse the local process information of multiple similar examples to perform process reuse design, so as to quickly generate a more accurate and effective target part processing process.
The core idea of the instance inference technology (CBR) is to store the experience knowledge accumulated by the past work into an instance library, and when people solve new problems, the solution process and the solution of the similar problems are obtained from the instance library by using an analogy method, and the solution of the similar problems is reapplied to the solution process of the new problems by modifying and adjusting the solution of the similar problems to a certain extent. The example reasoning technology has the characteristics of easy maintenance, self-accumulation of data, easy operation and the like, and is suitable for the development status quo of the intelligent design problem that the design knowledge of the existing injection mold part processing technology is difficult to regularize. In recent years, many studies have been made on "similar parts have similar processes" in structure, and it is considered that similar parts have similar processing methods and process routes. When different parts have the same processing characteristics, the same process can be considered for processing, which forms the basis for process reuse. The process reuse design not only can quickly obtain a solution, but also can continuously supplement successful cases to realize self-updating.
At present, a process reuse design method based on an example reasoning technology is already applied to intelligent design of a part machining process. In the aspect of process correction, rule reasoning is carried out for correction in a process rule mode, data in case of instance reasoning is not fully utilized, and evaluation indexes of local similarity are not effectively utilized for process correction. The key of the process reuse technology based on the instance reasoning is the retrieval and correction of the instance, and an injection mold part similar instance processing process reuse design method based on the instance reasoning is provided. When similar examples are searched, an improved K-means clustering algorithm is utilized to narrow the searching range of the examples; and (3) adopting a nearest neighbor algorithm (KNN) to realize instance retrieval based on local attribute similarity. After the retrieval is completed, the process corresponding to the reusable features is queried by calculating the similarity of the features in the examples, and the process skeleton is intelligently modified by utilizing the process information corresponding to the local features in the rest similar examples, so that the efficient and intelligent design of the processing process of the injection mold parts is realized. The method has important guiding significance for realizing the efficient and intelligent design of the injection mold part processing technology.
Disclosure of Invention
The invention aims to provide an example reasoning-based reuse design method for the processing technology of the injection mold part similar to the example, and the intelligent design of the target part technology is realized. The method is based on an improved K-means algorithm, basic part information such as part specifications, part materials, part types and the like is used as an example to search a first layer of search attribute to complete similar example part clustering search, then based on a KNN algorithm, the size, machining precision, surface roughness and number of main machining features of the parts are used as second layer of search attribute of the example search to complete search calculation of the overall similarity of the parts, thus a double-layer search model is established, search of the most similar example of the target part is realized, process information of the most similar example is extracted to serve as a process framework of the target part, then a process reuse correction model is constructed according to the feature similarity of the multi-phase example to serve as an index of process correction, the process framework of the target part is corrected, information replacement of a process corresponding to the local features is realized, and intelligent process design is completed.
In order to achieve the purpose, the invention takes injection mold template parts as an example, establishes the retrieval attribute of a double-layer retrieval model, reuses a correction model based on the process of feature similarity, and verifies the feasibility and effectiveness of the method in the aspect of rapid process design of the injection mold parts through experimental cases.
The specific implementation steps are as follows:
step 1: and establishing a first-layer search of a double-layer search model, and clustering the examples in the example set based on an improved K-means algorithm.
Step 1.1: the part material, the part specification and the part type are used as the first layer retrieval attributes.
Step 1.2: improving K-means, eliminating isolated points in the initial instance set, deleting isolated points far from the mean center, and avoiding the influence of isolated points on the instance clustering effect
Step 1.3: and determining the K value of the optimal cluster according to the distance cost function, calculating the K value range by using an empirical formula, calculating the distance cost function for each K value clustering result, and selecting the K value with the smallest distance evaluation function value as the optimal cluster number to realize the optimal clustering effect.
Step 1.4: inputting the optimal clustering number K value, and obtaining the clustering cluster of similar parts in the example set by using a K-means algorithm after the operation of the example set is improved.
Step 1.5: inputting the part materials, the part specifications and the part types of the target parts, calculating the distance between the clustering center points of the clustering clusters of the similar parts, and dividing the target parts into clusters closest to the clustering center points.
Step 2: and establishing a second-layer retrieval of the double-layer retrieval model, and calculating the overall similarity of the clustered examples in the example set based on a KNN algorithm.
Step 2.1: the size, machining precision, surface roughness and number of main machining characteristics of the part are used as second layer retrieval attributes.
Step 2.2: and (5) making an importance level table of the part retrieval attribute.
Step 2.3: the hierarchical analysis is used to determine the second level retrieval of individual attribute weights.
Step 2.4: and calculating the similarity of each local attribute of each instance in the similar cluster.
Step 2.5: the overall similarity of each instance to the target part is calculated.
Step 2.6: and setting a dynamic threshold value, and dynamically adjusting the threshold value according to the maximum overall similarity calculated in practice.
Step 2.7: and outputting the instances with the similarity larger than the threshold value, and extracting the process information of the instance with the maximum overall similarity as a process skeleton of the target instance.
Step 3: and establishing a process reuse correction model, taking the characteristic similarity calculation result of the similar instance into consideration as an index of local process correction, and carrying out process reuse correction on the process skeleton.
Step 3.1: the calculation attribute of the local feature similarity is exemplified by the size, machining accuracy, surface roughness, and number of local features.
Step 3.2: and comparing the characteristic information of the target part and the characteristic information of the most similar example, and calculating the corresponding characteristic similarity.
Step 3.3: and selecting the characteristics with the characteristic similarity not meeting 100 percent, and correcting the process information corresponding to the characteristics in the process framework.
Step 3.4: and inquiring the feature similarity of the rest similar instances about the feature to see whether the multiplexing condition is met.
Step 3.5: and replacing the process information corresponding to the queried reusable characteristics with the process information in the process framework, wherein the replaced process information comprises process content, process equipment and cutters.
The details of the specific steps are set forth in the following description in conjunction with the accompanying drawings.
Drawings
FIG. 1 is a diagram of a two-layer search model framework based on example reasoning
FIG. 2 is a frame diagram of a process reuse correction model based on feature similarity
FIG. 3 is a schematic diagram of outlier detection
FIG. 4a is a diagram of initial examples in an example set
FIG. 4b shows the clustering effect when the k value of the cluster is 2
FIG. 4c shows the clustering effect when the k value of the cluster is 3
FIG. 5 is a second layer search flow chart based on KNN algorithm
FIG. 6 is a flow chart of feature similarity calculation
Detailed Description
The technical scheme of the invention is described in detail below with reference to the attached drawings of the specification:
the main body of the method is divided into three modules. Firstly, clustering similar parts based on an improved K-means algorithm aiming at an example set to obtain a similar part cluster. And secondly, calculating the overall similarity based on the KNN algorithm for the examples in the similar part cluster, and retrieving similar examples larger than a threshold value. And finally, carrying out process correction based on the feature similarity of the multiple phase examples on the process skeleton of the example with the highest overall similarity. FIG. 1 is a diagram of a two-tier search model framework based on example reasoning. FIG. 2 is a process reuse correction model framework diagram based on feature similarity.
And (1) establishing a first-layer search of a double-layer search model, and clustering the examples in the example set based on an improved K-means algorithm.
Step 1.1: the part material, the part specification and the part type are used as the first layer retrieval attributes. The common materials of a certain type of parts of the injection mold parts are classified, the specifications of the parts are divided into large, medium and small parts, and the types of the parts are classified according to the functions of the parts.
Taking part specification, part material and part type as examples to search the first layer search attribute, and setting X= { X 1 ,x 2 ,...,x n An instance set containing n instances, where the ith instance may be represented as X i ＝{x i1 ,x i2 ,...,x im M is the number of first layer search attributes, x im The value of each search attribute of the ith m of each instance, k is the cluster number, Z p ＝{z p1 ,z p2 ,...,z pm The p cluster center, p E [ 1-k ]]。
Step 1.2: the K-means is improved, isolated points in the initial instance set are removed, isolated points far from the mean center are deleted, and the influence of the isolated points on the instance clustering effect is avoided.
Let d (x) i ,x j ) For example data x i And x j And d (x) i ,x j )＝d(x j ,x i ) The distance between each instance data and the other instance data in the instance set is calculated using the following formula (1).
Is provided with
For the average of the distances between all examples, d is used k Representing the spacing between two instances, calculated using equation (2), x savg For the sample center point of the sample set of the s-th attribute, calculating by adopting a formula (3), and firstly judging the distance d between certain two examples k Whether or not it is greater than
If it is greater than, compare thisAnd the two points are respectively separated from the center of the sample, and the point with larger distance is the example data of the isolated point and is removed. Fig. 3 is a schematic diagram of isolated point detection.
Step 1.3: and determining the optimal cluster k value according to the distance cost function.
The different cluster k values have a great influence on the example clustering effect, fig. 4a is an initial example graph in the example set, fig. 4b is a cluster effect when the cluster k value is 2, and fig. 4c is a cluster effect when the cluster k value is 3. The k value range is first calculated using an empirical formula. In the case of unknown specific values of the cluster number k, according to the empirical formula (4), the value range of k satisfies the following:
where n represents the total number of instances in the set of instances.
According to the possible maximum value of k
Start calculation from +.>
All the clustering results in the space, calculating the distance evaluation function value of each type of clustering results, and selecting the k value with the smallest distance evaluation function value as the optimal clustering number, wherein the calculation formula is as follows:
the inter-class distance is as follows:
intra-class distance:
distance cost function:
step 1.4: inputting the optimal clustering number K value, and obtaining the clustering cluster of similar parts in the example set by using a K-means algorithm after the operation of the example set is improved.
The method comprises the steps of firstly removing isolated points, determining an optimal k value, then carrying out clustering search, and running a k-means algorithm, wherein the specific calculation formula is as follows:
cluster allocation calculation for example:
wherein u is ip Indicating that the ith instance is divided into the p-th cluster, if u ip A value of 1 indicates that the ith instance is assigned to the p-th cluster, otherwise the ith instance is not in the p-th cluster.
Cluster center calculation:
clustering criterion function:
the flow of clustering the instances is as follows:
(1) Performing cluster analysis on the instance set, and randomly selecting k instance sample points as initial cluster center points according to the determined optimal cluster number k;
(2) Sequentially calculating the distance from each data point in the sample set to the sample clustering center by adopting formulas (8) and (10), evaluating similarity by using the distance, dividing the examples into clusters, and completing one-time clustering;
(3) Obtaining a new cluster center by adopting a cluster center calculation formula (9) according to the calculation result of the step (2), and carrying out next round of clustering and continuously iterating;
(4) Repeating the steps (2) and (3) until the positions of the k center points are not changed any more, and converging the algorithm to finish the clustering of the samples.
Step 1.5: inputting the part materials, the part specifications and the part types of the target parts, calculating the distance between the clustering center points of the clustering clusters of the similar parts, and dividing the target parts into clusters closest to the clustering center points.
And (2) establishing a second-layer search of the double-layer search model, and calculating the overall similarity of the clustered examples in the example set based on a KNN algorithm. Fig. 5 is a second layer retrieval flow chart based on KNN algorithm.
Step 2.1: the size, machining precision, surface roughness and number of main machining characteristics of the part are used as second layer retrieval attributes.
Taking a template part as an example, because of the structural characteristics of the template part, the finish machining needs to be carried out on the upper and lower matching surfaces and the precision requirement of holes on the surface is considered, and eight properties of the material type, the plane machining precision, the plane surface roughness, the main machining hole size, the number, the machining precision, the surface roughness and the type which have great influence on machining are taken as second-layer retrieval properties.
Step 2.2: and (5) making an importance level table of the part retrieval attribute.
Step 2.3: the hierarchical analysis is used to determine the second level retrieval of individual attribute weights.
The analytic hierarchy process determines weights:
the distribution scheme of the KNN algorithm retrieval attribute weight is determined by adopting an analytic hierarchy process, so that objective and practical parts retrieval can be embodied. Firstly, making an importance level table of the retrieval attributes of the parts, if the parts have one attribute to be assigned with weight, then according to the enterpriseKnowledge and experience accumulated in the process design of parts, and attribute importance degree matrix Q= [ Q ] is constructed by comparing each attribute with each other according to a ratio scale of 1 to 9 ij ] l×l Q in matrix ij Represents the importance of attribute i to attribute j, scale q ij Determination is judged by a "1 to 9 ratio scale" method in combination with the Delphi method. Normalized for each column of the matrix to obtain a search attribute weight vector w= (W) 1 ,w 2 ,…,w l ) The specific calculation formula is as follows:
wherein w is i Is the weight of the ith attribute, and
l is the total number of retrieval attributes of the case, q ij Importance scale value for ith attribute and jth attribute, and q ij ＝1/q ji 。
Step 2.4: and calculating the similarity of each local attribute of each instance in the similar cluster.
Calculating local attribute similarity:
the similarity degree of each retrieval attribute of the injection mold part is called local attribute similarity, and is an important theoretical basis for retrieving the most similar examples. The local similarity is divided into a numerical value type local similarity and a character string type local similarity, and the calculation formula is as follows:
for numerical properties such as material hardness, size, machining precision, surface roughness, etc., in order to eliminate the dimensional influence between the properties, a normalization method is used to perform data processing. And mapping the spatial distance result between the attributes representing the local characteristic similarity degree of the die part to 0-1, and taking the spatial distance result as a similarity value, as shown in a formula (12).
In the method, in the process of the invention,
and->
Respectively the target parts X M And instance library part X i Is the u-th attribute of (c).
For the text type attribute, such as material type, hole type, etc., the character string form is described. Therefore, measurement cannot be directly performed by using a distance formula, and the similarity of the attributes of the template type parts is calculated by using formula (13).
In the method, in the process of the invention,
and->
Respectively the target part X M And instance library part X i Is the u-th attribute of (c).
Step 2.5: the overall similarity of each instance to the target part is calculated.
Using eight properties of material type, plane machining precision, plane surface roughness, hole type, hole size, hole machining precision, hole surface roughness and hole number as second layer retrieval properties of template type parts, using c 1 ～c u And the influence degree of different attributes on the part machining process is different, and the similarity is used for measuring the similarity between example parts. Therefore, the overall similarity between any two parts is obtained by weighting each local attribute similarity and the corresponding weight, and an overall similarity calculation formula shown in formula (14) is constructed.
Wherein X is M And X i Respectively, the target part and the example part in the cluster, i is the total number of retrieval attributes of the template type part, i=8,
and->
Is X M And X i W (c) u ) Is the weight of the u-th attribute, sim (X M ,X i ) Is the target part X M And example part X i Overall similarity of->
Is->
And->
Is a local feature similarity of (c).
Step 2.6: and setting a dynamic threshold value, and dynamically adjusting the threshold value according to the maximum overall similarity calculated in practice.
When the instance screening is carried out, a system threshold value of the overall similarity needs to be determined, the system threshold value is set according to experience, and a dynamic overall similarity threshold value determining method is adopted to reduce the influence of the fact that the part instances in the actual production instance library are continuously amplified, so that the rapidity and the stability of the retrieval model are influenced. The dynamic system threshold value is calculated by using similarity information of corresponding examples in the example library, so that dependence on expert experience can be reduced, and example library resources are fully utilized.
Obtaining the maximum similarity Sim between the example part and the target part in the example library through calculation max Comparing with the preset system threshold value, and selecting a smaller value as the overall similarity threshold value Sim of the system α By means of this adaptive adjustment, the control unit,the retrieval system does not have a case where similar examples are not retrieved, as shown in equation (15).
In the formula, sim max Maximum similarity is calculated; eta is the preset system threshold value, obtained empirically, and generally taken as X α When β is equal to or less than the threshold, the search range coefficient determined from the maximum similarity calculated in practice is generally β=0.1 to 0.15.
Step 2.7: and outputting the instances with the similarity larger than the threshold value, and extracting the process information of the instance with the maximum overall similarity as a process skeleton of the target instance.
And (3) establishing a process reuse correction model, taking the characteristic similarity calculation result of the similar instance into consideration, and taking the characteristic similarity calculation result as an index of local process correction to perform process reuse correction on the process framework.
Step 3.1: the calculation attribute of the local feature similarity is exemplified by the size, machining accuracy, surface roughness, and number of local features.
Comparing the characteristics in the target part with the corresponding characteristic information on the process framework, comparing the size, the number of the characteristics, the processing precision and the surface roughness of the characteristics, calculating the characteristic similarity of a certain characteristic by using a formula (16), comparing the characteristic similarity, selecting the characteristic with large characteristic similarity, and checking whether the corresponding process information can be directly reused; if not, calculating the feature similarity of the rest examples, inquiring the process information corresponding to the replaceable features, finishing the correction process of the process skeleton until all the features are compared, and outputting the process file information of the target part. Fig. 6 is a flowchart of feature similarity calculation, and the specific formula is as follows:
in the formula TZ M And TZ j A feature, f, of the target part and of the jth example part in the cluster, respectively u Is the value of the u-th attribute corresponding to the feature, f 1 ～f 4 Respectively representing the size, the number, the processing precision and the surface roughness of the feature, wherein 0.25 is the weight of the four attributes, and the importance of the four attributes is consistent, the fixed value is 0.25, sim (TZ M ,TZ j ) Is the feature similarity of a feature of the target part and the example part,
is a feature related to attribute f u Attribute similarity of (c).
Step 3.2: and comparing the characteristic information of the target part and the characteristic information of the most similar example, and calculating the corresponding characteristic similarity.
Step 3.3: and selecting the characteristics with the characteristic similarity not meeting 100 percent, and correcting the process information corresponding to the characteristics in the process framework.
Step 3.4: and inquiring the feature similarity of the rest similar instances about the feature to see whether the multiplexing condition is met.
Step 3.5: and replacing the process information corresponding to the queried reusable characteristics with the process information in the process framework, wherein the replaced process information comprises process content, process equipment and cutters.
Example 1
Fig. 1 shows a two-layer search model based on example reasoning for the first part of the implementation case, fig. 2 shows a process reuse correction model based on feature similarity after the search is finished, and experiments are carried out in an example library containing 24 examples by taking a certain injection mold template type part as an example, wherein the example library comprises template type parts of movable templates, fixed templates, ejector pin base plates, ejector pin plates and movable fixed die base plates with different sizes and different material types.
Basic information of a target part is input firstly, wherein the basic information comprises a top pillow base plate with a part type of 45 steel and a large specification. And feature information of the part, wherein the main machining plane precision is IT10, the surface roughness is Ra0.8μm, the types of main machining holes are guide sleeve holes, the sizes are 50mm, the number is 4, the machining precision is IT6, and the surface roughness is Ra1.6μm.
Table 1 results of first tier cluster search
The first layer search is carried out by inputting the part specification, the part material and the part type of the target part in the double-layer search model, the first layer cluster search is carried out, the first layer search-clustering result is shown in table 1, the example set is divided into three clusters, the first cluster is a movable die base plate and a fixed die base plate and is made of 45 steel in material type, the second cluster is a fixed die plate and a movable die plate and is made of 45Cr in material type, the second cluster is a thimble plate and a thimble base plate and is made of 45 steel in material type, the cluster to which the target part belongs is 3, and all examples in the cluster 3 are extracted to be used as the example set of the second layer search.
As shown in FIG. 5, a second layer searching flow chart based on KNN algorithm is shown, when searching the second layer, eight attributes C, namely material type, material hardness, plane dimensional accuracy, plane surface roughness, hole type, hole size, hole machining accuracy and hole surface roughness are firstly obtained according to an analytic hierarchy process 1 -C 8 The attribute importance degree matrix is constructed as shown in table 2, and the weight of each attribute is calculated.
Table 2 judgment matrix of retrieval attributes of template type parts
C is calculated according to a formula 1 Weight w is obtained 1 。
The same principle can be obtained: w (w) 2 ＝0.1589，w 3 ＝0.1468，w 4 ＝0.1047，w 5 ＝0.0846，w 6 ＝0.0800，w 7 ＝0.0506，w 8 = 0.0440, and pass the consistency check.
Calculating the overall similarity of the examples in the similar part cluster according to an overall similarity calculation formula (14), wherein table 3 is a target part second layer similarity retrieval attribute value, calculating the local attribute similarity of each attribute of each example in the similar example set according to an attribute similarity calculation formula, and carrying out weighted calculation on all the local attribute similarities to obtain the overall similarity of each example, and table 4 is a target part second layer retrieval-overall similarity calculation result.
TABLE 3 target part similarity retrieval attributes
TABLE 4 second level search-overall similarity calculation results for target parts
According to the overall similarity calculation result, the similarity sorting is carried out, so that the example X can be seen 20 And example X 18 The overall similarity is larger than 90%, and the example part X with the maximum similarity is selected 20 The process file is used as a process skeleton of the target part, and further used as a basis for process reuse correction. Table 5 shows characteristic information of the target parts.
TABLE 5 characterization information of target parts
According to the feature similarity calculation flow, a feature similarity calculation flow chart shown in fig. 6 is shown. The most similar part is compared for feature similarity to the target part.
TABLE 6 most similar part X 20 Feature similarity calculation result with target part
Table 6 shows the most similar part X 20 And (3) according to the feature similarity calculation result, the local features to be corrected can be seen, the correction model is reused based on the process of the feature similarity, the process corresponding to the features with the feature similarity not meeting 100% is corrected, namely, the process information in the process skeleton is corrected, the feature similarity calculation is carried out on the features in other similar examples based on the feature similarity on the basis of the correction, the available process information is queried, and the query flow is shown in figure 2.
TABLE 7 Process information of Process skeletons
Table 8 queries reusable process information corresponding to local features of the remaining similar instances
Table 9 Process skeleton information before and after Process reuse correction
Table 7 shows process information of the process framework, table 8 shows reusable process information corresponding to local features of other similar examples, and it can be seen that example X 18 The feature upper and lower surfaces in (a) satisfy the condition of multiplexing, example X 21 The characteristic holes 1, 4 and 6 in the process framework meet the multiplexing condition, and the process information corresponding to the characteristics is extracted and replaced to the corresponding positions in the process framework, and the process framework information before and after the process reuse correction is shown in the table 9.
Claims (4)
1. The reuse design method for the processing technology of the similar instance of the injection mold part based on instance reasoning is characterized by establishing a double-layer retrieval model, taking part materials, part specifications and basic part information of the part type as instance retrieval attributes of a first layer, realizing the clustering of similar parts, reducing the retrieval range of the similar parts, taking the size, processing precision, surface roughness and quantity of main processing features of the parts as retrieval attributes of a second layer of instance retrieval, realizing the accurate retrieval of the similar parts, establishing a technology reuse correction model, calculating the feature similarity according to the size, processing precision, surface roughness and quantity of the features, mapping the feature similarity to technology information by using the feature similarity as an index of local technology correction, extracting the technology information corresponding to the features, and finishing the intelligent design of the technology; the method comprises the following steps:
firstly, establishing a first layer of retrieval of a double-layer retrieval model, and clustering examples in an example set based on an improved K-means algorithm to obtain similar part clusters;
establishing a second-layer search of a double-layer search model, accurately searching the similarity of the example set where the clustered target parts are located based on a KNN algorithm, determining local attribute similarity weight according to a hierarchical analysis method, setting a dynamic threshold, dynamically adjusting the threshold of a general similarity screening similar example, and sequencing from large to small according to the general similarity value;
and (3) establishing a process reuse correction model, extracting process information of a part example with highest overall similarity as a process skeleton of the target part, correcting the process skeleton, taking the characteristic similarity calculation result as an index of local process correction according to the size, machining precision, surface roughness and number of local characteristics as attributes of characteristic similarity calculation, extracting reusable process information corresponding to the local characteristics, and replacing the process information on the process skeleton to obtain corrected process information of the target part.
2. The reuse design method for the similar instance processing technology of the injection mold part based on instance reasoning of claim 1, which is characterized in that: the method comprises the steps of (1) clustering examples in an example set based on an improved K-means algorithm, wherein the K-means algorithm is improved, firstly, before clustering is started, isolated points are removed from an initial example set, isolated points far from the center of an average value are deleted, the influence of the isolated points on the example clustering effect is avoided, secondly, the optimal clustering K value is determined according to a distance cost function, the optimal clustering effect is realized, the first layer of example retrieval clusters an example cluster with the farthest distance between classes and the nearest distance in the class, the example retrieval range is reduced, and the example retrieval speed is improved; and (3) taking the part materials, the part specifications and the part type basic part information as examples to search the first layer of search attributes, so that the clustering of similar parts is realized, and the search range of the similar parts is reduced.
3. The reuse design method for the similar instance processing technology of the injection mold part based on instance reasoning of claim 1, which is characterized in that: step (2) based on a KNN algorithm, taking the size, machining precision, surface roughness and quantity of main machining characteristics of the parts as second-layer retrieval attributes for example retrieval, and realizing accurate retrieval of similar parts; firstly, determining a second-layer retrieval attribute weight by using an analytic hierarchy process, more objectively calculating the weight, weighting and calculating the overall similarity of the instance and the target part by using the local attribute similarity of the instance, setting a dynamic threshold, dynamically adjusting the threshold according to the maximum overall similarity calculated in practice, avoiding the situation that the instance cannot be retrieved, outputting the instance larger than the threshold similarity, extracting the process information of the instance with the maximum overall similarity as the process skeleton of the target instance, and carrying out process correction.
4. The reuse design method for the similar instance processing technology of the injection mold part based on instance reasoning of claim 1, which is characterized in that: step (3), a process reuse correction model is established, characteristics in the target part are compared with corresponding characteristic information on a process framework, the size, the characteristic quantity, the machining precision and the surface roughness of the characteristics are compared, and the characteristic similarity is calculated; according to the feature similarity calculation result of the multi-phase example, the feature similarity is compared as an index of local process correction, and a process with the maximum feature similarity is selected to check whether process information can be directly reused; if the process framework is not reusable, the feature similarity of the other examples is continuously inquired, the feature similarity is used as an index of process correction, process information corresponding to the features on the process framework is replaced, the replaced process information comprises process content, process equipment and cutters, the correction process of the process framework is further completed, and process information of the target part is output, so that intelligent correction of the process framework is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310163383.3A CN116362005A (en) | 2023-02-24 | 2023-02-24 | Reuse design method for similar instance processing technology of injection mold part based on instance reasoning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310163383.3A CN116362005A (en) | 2023-02-24 | 2023-02-24 | Reuse design method for similar instance processing technology of injection mold part based on instance reasoning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116362005A true CN116362005A (en) | 2023-06-30 |
Family
ID=86938933
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310163383.3A Pending CN116362005A (en) | 2023-02-24 | 2023-02-24 | Reuse design method for similar instance processing technology of injection mold part based on instance reasoning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116362005A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116680839A (en) * | 2023-08-02 | 2023-09-01 | 长春设备工艺研究所 | Knowledge-driven-based engine intelligent process design method |
-
2023
- 2023-02-24 CN CN202310163383.3A patent/CN116362005A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116680839A (en) * | 2023-08-02 | 2023-09-01 | 长春设备工艺研究所 | Knowledge-driven-based engine intelligent process design method |
| CN116680839B (en) * | 2023-08-02 | 2023-12-08 | 长春设备工艺研究所 | Knowledge-driven-based engine intelligent process design method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106371427B (en) | Industrial process Fault Classification based on analytic hierarchy process (AHP) and fuzzy Fusion | |
| CN109158953B (en) | A kind of cutting-tool wear state on-line monitoring method and system | |
| CN112464638B (en) | Text clustering method based on improved spectral clustering algorithm | |
| CN105975488A (en) | Method for querying keyword based on topic cluster unit in relational database | |
| CN111160750A (en) | Distribution network analysis and investment decision method based on association rule mining | |
| CN108197647B (en) | Rapid clustering method for automobile starter endurance test data | |
| CN116362005A (en) | Reuse design method for similar instance processing technology of injection mold part based on instance reasoning | |
| CN104598586A (en) | Large-scale text classifying method | |
| CN106815447B (en) | Intelligent defining and classifying method for machining characteristics of complex structural part based on historical data | |
| CN112308298B (en) | Multi-scenario performance index prediction method and system for semiconductor production line | |
| CN110069630A (en) | A kind of improved mutual information feature selection approach | |
| CN109271427A (en) | A kind of clustering method based on neighbour's density and manifold distance | |
| CN112289391A (en) | Anode aluminum foil performance prediction system based on machine learning | |
| CN107944487B (en) | Crop breeding variety recommendation method based on mixed collaborative filtering algorithm | |
| CN115169453A (en) | Hot continuous rolling width prediction method based on density clustering and depth residual error network | |
| CN116662521B (en) | Electronic document screening and inquiring method and system | |
| CN107220320B (en) | Emerging technology identification method based on patent citation | |
| CN110866317A (en) | Design tool kit reconstruction method based on product historical information and design behavior analysis | |
| CN111950652A (en) | Semi-supervised learning data classification algorithm based on similarity | |
| CN112232885A (en) | Multi-mode information fusion-based warehouse rental price prediction method | |
| CN113523904A (en) | Cutter wear detection method | |
| CN116822985A (en) | Intelligent process decision method | |
| CN113159220B (en) | Random forest based concrete penetration depth empirical algorithm evaluation method and device | |
| CN113033683B (en) | Industrial system working condition monitoring method and system based on static and dynamic joint analysis | |
| CN115238965A (en) | Technical trend prediction method, device, equipment and medium based on patent information |
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 |