CN113592772B - Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics - Google Patents

Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics Download PDF

Info

Publication number
CN113592772B
CN113592772B CN202110710104.1A CN202110710104A CN113592772B CN 113592772 B CN113592772 B CN 113592772B CN 202110710104 A CN202110710104 A CN 202110710104A CN 113592772 B CN113592772 B CN 113592772B
Authority
CN
China
Prior art keywords
smr
cutting
contour
sub
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.)
Active
Application number
CN202110710104.1A
Other languages
Chinese (zh)
Other versions
CN113592772A (en
Inventor
黄瑞
杨昌尧
崔成
费铭涛
蒋俊锋
陈正鸣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hohai University HHU
Original Assignee
Hohai University HHU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hohai University HHU filed Critical Hohai University HHU
Priority to CN202110710104.1A priority Critical patent/CN113592772B/en
Publication of CN113592772A publication Critical patent/CN113592772A/en
Application granted granted Critical
Publication of CN113592772B publication Critical patent/CN113592772B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component

Abstract

The invention discloses a complex part contour self-adaptive finish machining method based on dynamic machining characteristics. And predicting cutting layers of the sub-processing areas according to the extracted mapping modes, and judging the mergence between the sub-processing areas through the self-adaptive adjustment of the cutting layers of the sub-processing areas. And finally, adopting a genetic algorithm to calculate an optimized processing path for different sub-processing areas with the same cutting layer, thereby constructing a process situation merging relation capable of reflecting the process design intention among the sub-processing areas and realizing the identification of dynamic characteristics. The invention can effectively construct the optimized merging relation between the sub-processing areas under different process situations, and improves the numerical control processing efficiency, thereby supporting the model guiding and data driving self-adaptive numerical control process design method.

Description

Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics
Technical Field
The invention belongs to the field of numerical control process design based on characteristics and mining and reusing process data in manufacturing industry, and particularly relates to a complex part contour self-adaptive finish machining method based on dynamic machining characteristics.
Background
In recent years, how to analyze, mine and reuse process knowledge contained in process data is becoming increasingly important in the rapid manufacturing process of parts, especially for single-piece, small-lot complex parts in the fields of aviation, aerospace, etc. The traditional process knowledge reuse mainly adopts a process data characterization and multiplexing method based on static characteristics. However, once the static characteristics are defined, the associated processing area is kept unchanged in the dynamic processing process of the part, so that it is difficult to reflect that a designer performs dynamic combination optimization processing on different processing areas according to different process situations (including processing geometry, processing resources and processing stages), and reuse of process knowledge guided by process design intent is not supported.
Aiming at the defect of the static feature in the aspect of reuse of process knowledge, the dynamic feature is proposed to perform multi-level structural characterization on the process data so as to effectively reflect and capture the process design intention of a designer under different process situations. In addition, the existing process knowledge is directly reused mainly through the adoption of a similarity evaluation method by a manufactured single part, and potential rules cannot be mined from process data, so that the existing process knowledge still stays in a personalized instance reuse stage, and is difficult to generalize effectively. In order to realize data driving and dynamic characteristic-based adaptive numerical control process design, the following key problems also need to be solved:
(1) How to learn and mine implicit process knowledge from process data, and apply the implicit process knowledge to process parameter decisions (such as cutting layers, cutting depths, feeding rates and the like) under different process situations; (2) And identifying dynamic characteristics according to different process situations so as to perform combined and optimized processing on different processing areas and improve numerical control processing efficiency.
Disclosure of Invention
The invention aims to: aiming at the problems, the invention provides a complex part contour self-adaptive finish machining method based on dynamic machining characteristics. Then, the cutting layers of the sub-processing regions having a geometric dependency relationship are adaptively adjusted based on the f-predicted cutting layers of the sub-processing regions, and the mergence between the sub-processing regions is determined based on the cutting layers. And finally, calculating an optimized processing path for different sub-processing areas with the same cutting layer by adopting a genetic algorithm, thereby constructing a process situation merging relation between the sub-processing areas and realizing the identification of dynamic characteristics.
The technical scheme adopted for solving the technical problems is as follows: a complex part contour self-adaptive finish machining method based on dynamic machining features comprises the following steps:
(a) Extracting contour finishing process data from a structural process database, carrying out characterization and deep learning on the contour finishing process data to obtain a contour finishing cutting layer decision model, and extracting mapping modes f between different process situations and cutting layers in the process data;
(b) Predicting cutting layers of the sub-processing areas according to the decision model, and performing self-adaptive adjustment of the cutting layers of the sub-processing areas, and judging the same cutting layers and optimizing paths among the sub-processing areas;
(c) Calculating process situation merging relations among different sub-processing areas with the same cutting layer, realizing the identification of dynamic characteristics, and constructing a driving geometry;
the dynamic characteristics are defined as a common machining area DF where the tool can continuously cut under a given technological situation, and are expressed as
Wherein z is s 、z e Respectively a top surface and a bottom surface with dynamic characteristics, DG is a driving geometry with dynamic characteristics, SMR i G () is a driving geometry building function for sub-processing regions with geometry dependencies;
under a given process situation, SMR i At [ z ] s ,z e ]With the same cutting layer therebetween, and the cutter has a pair of SMRs at each cutting layer i Non-interfering continuous machining is performed, i=1, …, n, n being the number of sub-machining regions.
Further, the step (a) specifically includes:
1.1, under the situation of a contour finishing process, characterizing contour finishing process data;
1.2 according to the contour finish machining process data characterized in 1.1, a deep neural network classifier model is adopted to learn the implicit judgment rule of the cutting layer under the given contour finish machining process situation, and a contour finish machining cutting layer decision model is obtained.
Further, in the step 1.1, under the circumstance of the contour finishing process, the contour finishing process data is characterized as follows:
in the contour finishing process context, according to the cutting layer of the contour finishing operation of the geometric decision to be processed and the cutter T, the finishing of the contour surface is described as follows:
given one at z e Deep, side machining allowance is a e Side with wall height h, z s ＝z e +h, when a tool T is used, n cutting layers are required; z s Is the height, z, corresponding to the top surface of the dynamic feature e Is the height corresponding to the dynamic feature bottom surface;
for any one contour finish sample in the structuring process data, it is expressed as contour finish process context x i And cutting layer n i A mapping between them, denoted as
x i ＝{D,L,FL,z s ,z e ,a e }→x_label i ＝n i
Wherein D is the diameter of the cutter, L is the length of the cutter, FL is the edge length of the cutter, and x_label i The cutting layer is predefined for the i-th.
Further, in the step 1.2, a deep neural network classifier model is adopted to learn a cutting layer implicit decision rule under a given contour finish machining process situation, and the method specifically comprises the following steps:
learning a cutting layer implicit judgment rule under a given contour finish machining process situation by adopting a four-layer deep neural network classifier model to obtain a contour finish machining cutting layer decision model;
the input layer of the model is a process situation parameter X= { X related to contour finishing cutting layer calculation i -a }; the hidden layer is 2 full-connection layers with 20 nodes;
the output layer of the model is x i Mapping to a predefined cutting layer x_labels= { n j Each element n in } j Probability p of (2) ij Represents x i Requiring n j The possibility of contour finishing of the individual cutting layers.
Further, the cutting layer of the sub-processing area is adaptively adjusted, specifically as follows:
obtaining a sub-processing region SMR according to the mapping pattern f i Is not limited to the cutting layer n i The expression is
Wherein a is p For depth of cut, h is the difference in height of the top and bottom surfaces of the dynamic feature,for the effective depth of each layer, a max Maximum cutting depth allowed for tool []Is a rounding operation; by pairing SMR i Is +.>Performing adaptive adjustment to enable n i Remain unchanged;
given 2 sub-process regions SMR with geometrical dependence i And SMR (SMR) j Wherein, SMR i For the target sub-process region, SMR j To sub-process regions that need to be merged, SMR j Depending on SMR i ，SMR i And SMR (SMR) j Respectively n is needed for contour finishing of (c) i And n j Cutting layers, n i ≤n j ；
If SMR i And SMR (SMR) j Merging, SMR j At the position ofProfile surface between->And SMR (SMR) i With the same cutting layer->SMR respectively j Top surface height of SMR i Is a bottom surface height of (2); />Is SMR j Is a bottom surface height of (2); while SMR j At->Profile surface between->Is n j -n i The following equation needs to be satisfied:
wherein the contour surfaceThe cutting depth of each cutting layer is defined by +.>Adjust to-> Is SMR j Effective cutting depth of each layer of the cutting layer, < >>Is SMR i Effective cutting depth of each layer of the cutting layer; contour surface->The cutting depth of each cutting layer is defined by +.>Adjust to-> To adjust the sub-process region SMR j Effective cutting depth of each layer of the cutting layer; h is a j Is SMR j Height difference between top and bottom surfaces of (a);
according to SMR i And SMR (SMR) j The self-adaptive adjustment of the cutting layer of the sub-processing area is divided into the following 2 cases:
1) When n is i ＝n j If at the timeThen->Is defined by->Reduced to-> At least 1 cutting layer is required, SMR j At least need n j +1 cutting layers, SMR i Is not matched with SMR j Merging; if->Then SMR i And SMR (SMR) j Merging, marked as->
2) When n is i ≠n j When according toAnd->The relationship between them is divided into the following two types:
2a) A group II of the type which is known as such,
SMR i and SMR (SMR) j At the position ofWith the same cutting layer in between, combined, denoted +.>
2b) The group III of the type,
prediction from mapping pattern fWhether or not the cutting layer of (2) is n j -n i A plurality of; if so, SMR i And SMR (SMR) j Merging, marked as->Otherwise, thenAnd not merging.
Further, calculating the merging relation between different sub-processing areas with the same cutting layer, and identifying the dynamic characteristics, wherein the method specifically comprises the following steps:
obtaining the SMR of each target sub-processing area from top to bottom according to the geometrical dependency relationship between the sub-processing areas i Sub-process regions to be consolidated SMR having a combinable relationship j Denoted as S i ；
Given S i Any one of the sub-processing regions SMR i Which consists of N profile chains C ij Each profile chain is composed of a set of adjacent profile surfaces, denoted as SMR i ＝{C ik K is more than or equal to 1 and less than or equal to N; constructing a virtual continuous cutting and discontinuous cutting judgment rule between profile chains;
s, constructing according to a virtual continuous cutting and discontinuous cutting judgment rule between profile chains i Is a contour chain adjacency attribute graph G i Converting contour finishing dynamic feature recognition into G i Vertex and edge access path optimization;
under the situation of the contour finishing process, the solution with the shortest time in all access paths is the optimal dynamic characteristic, and the optimal dynamic characteristic is solved through a genetic algorithm.
Further, a virtual continuous cutting and discontinuous cutting judgment rule between profile chains is constructed, and the method specifically comprises the following steps:
if the sub-processing region SMR i And SMR (SMR) j Merge into one dynamic feature, SMR i And SMR (SMR) j At least one pair (C) in ,C jm ) Make the cutter at C im And C jn Continuously cutting along the constructed virtual contour;
conversely, if SMR i And SMR (SMR) j Is not in existence (C) in ,C jm ) SMR then i And SMR (SMR) j Not merging into a dynamic feature, indicating that the machining time of the discontinuous cut by the tool is lower than the continuous cut time of the tool;
given S i Any two non-adjacent profile chains C in And C jm If C in And C jm One of the following conditions is satisfied:
1)C in and C jm An edge which can be co-straight or co-arc exists between the two edges;
2)C in and C jm Not more than 2 profile surfaces are adjacent;
then at C in And C jm Constructing a virtual contour connection C between in And C jm So that the cutter is at C in And C jm Is cut continuously, and is marked as VCM (C) in ,C jm )＝1。
Further, construct S i Is a contour chain adjacency attribute graph G i Converting contour finishing dynamic feature recognition into G i The vertex and edge access path optimization of (1) is as follows:
s, constructing based on virtual continuous cutting and discontinuous cutting judgment rules among profile chains i Is a contour chain adjacency attribute graph G i The vertex corresponds to the contour chain, the side corresponds to the relation between the contour chains, including adjacent, virtual continuous cutting and discontinuous cutting, and multiple paths can exist between the vertices;
conversion of contour finishing dynamic feature recognition to G i Vertex and edge access path optimization through each set of vertex and edge access paths L i Obtaining S i Contour finishing time t of neutron machining region i The method comprises the steps of carrying out a first treatment on the surface of the The solution with the shortest time in all access paths is the optimal dynamic feature.
Further, the solution with the shortest time in all access paths, namely the optimal dynamic characteristics, is solved through a genetic algorithm, and the method is specifically as follows:
the optimal dynamic characteristics are recorded as
Where T is the shortest access time, T (x) represents the path access time,according to L i New contour chain constructed by medium virtual continuous cutting edge R ij Is L i The non-cutting edges of (a), N and M are respectively +.>And R is R ij Is the number of (3);
solving the formula by adopting a genetic algorithm, wherein each individual c is composed of a virtual edge sequence between vertexes;
wherein each gene c i Indicating whether or not cutting is continued between two vertices, if c i 1, representing continuous cutting between two vertices; otherwise, the cutter needs to perform non-cutting movement;
from each individual c, an optimal profile finishing time based on that individual is calculated by reconstructing the driving geometry of the profile finishing.
Further, based on a genetic algorithm, the optimization steps of the profile chain finish machining path between different sub-machining areas are as follows:
step1: randomly generating a designated number of initial populations according to the individual representation method, and taking the initial populations as current populations;
step2: calculating the fitness of each individual in the current population, and marking the individual with the largest fitness as the optimal individual in the current population;
step3: comparing with the optimal individual searched in the previous round, and if the optimal individual in the current population is better than the optimal individual, updating the optimal individual;
step4: selecting two individuals from the current population based on the individual fitness value of the current population by adopting a roulette method, and executing crossover and mutation operations on the two individuals to generate new individuals;
step5: repeating step4 until a next generation population is generated;
step6: repeating step 2-step 5 until convergence tolerance or maximum iteration number is reached, and selecting individuals with optimal fitness from all populations as optimal solutions;
step7: and extracting the contour finish machining process situation merging relation among different sub-machining areas according to the optimal individual, constructing a driving geometry, and realizing dynamic feature identification under the contour finish machining process situation.
The beneficial effects are that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
(1) The dynamic characteristics are provided, the performance of processing resources can be fully considered, so that the merging relation of sub-processing areas under different process situations can be effectively reflected, and the flexibility of numerical control process design is improved.
(2) Through deep learning and mining of process data, implicit judgment rules of the cutting layer under a given process situation can be effectively extracted, so that the defect that the existing method based on the predefined rules is difficult to dynamically expand is overcome, and the generalization performance of the cutting layer decision is improved.
(3) By integrating optimization, the combination calculation among different sub-processing areas with the same cutting layer under different process situations can be realized, so that the numerical control processing efficiency is improved, and the defect that the traditional method adopts a conservation layering processing strategy is overcome.
Drawings
FIG. 1 is a general framework of the method of the present invention;
FIG. 2 is a schematic illustration of the contour finishing of the method of the present invention;
FIG. 3 is a schematic view of a contour finishing cutting layer decision model of the method of the present invention;
FIG. 4 is a schematic view of the adaptive calculation of the cutting layer in the sub-processing area of the method of the present invention;
FIG. 5 is a schematic diagram of the calculation of dynamic feature combination of sub-processing regions in the method of the present invention;
FIG. 6 is a diagram of an exemplary virtual connection of non-adjacent outline chains in accordance with the method of the present invention;
FIG. 7 is a schematic representation of a portion of a contour chain adjacency attribute of the method of the present invention;
FIG. 8 is a schematic representation of the genetic algorithm of the method of the present invention.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
FIG. 1 is a general block diagram of a complex part contour adaptive finishing method based on dynamic machining features, wherein contour finishing process data is firstly extracted from a structured process database, and is characterized and deeply learned to obtain a contour finishing cutting layer decision model; secondly, according to a contour finish machining cutting layer decision model, cutting layer self-adaptive adjustment, cutting layer judgment and path optimization are carried out between sub-machining areas; and finally, calculating the process situation merging relation between the sub-processing areas, thereby realizing the identification of the dynamic characteristics.
The invention discloses a complex part contour self-adaptive finish machining method based on dynamic machining characteristics, which comprises the following steps of:
(a) And extracting contour finishing process data from a structural process database, carrying out characterization and deep learning on the contour finishing process data to obtain a contour finishing cutting layer decision model, and extracting mapping modes f between different process situations and cutting layers in the process data.
(a1) In the contour finishing process context, characterizing contour finishing process data specifically includes:
as shown in fig. 2, in the context of the contour finishing process, the finishing of the contour surface is described as:
firstly, defining dynamic characteristics as a common machining area DF where a cutter can continuously cut under a given technological situation, which is expressed as
Wherein z is s 、z e Respectively a top surface and a bottom surface with dynamic characteristics, DG is a driving geometry with dynamic characteristics, SMR i G () is a driving geometry building function for sub-processing regions with geometry dependencies;
thus, in a given process scenario, the SMR i At [ z ] s ,z e ]With the same cutting layer therebetween, and the cutter can efficiently cut the SMR at each cutting layer i Performing non-interference continuous processingI=1, …, n, n being the number of sub-machining regions;
given one at z e Deep, side machining allowance is a e Side with wall height h, z s ＝z e +h, when a tool T is used, n cutting layers are required; z s Is the height, z, corresponding to the top surface of the dynamic feature e Is the height corresponding to the dynamic feature bottom surface;
for any one contour finish sample in the structuring process data, it is expressed as contour finish process context x i And cutting layer n i A mapping between them, denoted as
x i ＝{D,L,FL,z s ,z e ,a e }→x_label i ＝n i
Wherein D is the diameter of the cutter, L is the length of the cutter, FL is the edge length of the cutter, and x_label i The cutting layer is predefined for the i-th.
(a2) According to the represented contour finish machining process data, a depth neural network classifier model is adopted to learn a cutting layer implicit judgment rule under a given contour finish machining process situation, and a contour finish machining cutting layer decision model is obtained, and specifically comprises the following steps:
the deep learning of the contour finishing process data is to construct a trained neural network, and the process situation data X= { X can be effectively memorized and input from training samples i And the corresponding predefined cutting layer x_labels= { n i A mapping f between }, denoted n i ＝f(x i ) The method comprises the steps of carrying out a first treatment on the surface of the In view of the strong characteristic self-extraction and nonlinear mapping capability of deep learning, and because X_labels are discrete variable sets, a four-layer deep neural network classifier model is adopted to learn the implicit judgment rule of a cutting layer under the given contour finish machining process situation;
the input layer of the model is a process situation parameter X= { X related to contour finishing cutting layer calculation i -a }; the hidden layer is 2 full connection layers (20×20) with 20 nodes;
the output layer of the model is x i Mapping to a predefined cutting layer x_labels= { n j Each element in }n j Probability p of (2) ij Represents x i Requiring n j The possibility of contour finishing of the individual cutting layers is designated as
In the formula, the process situation data X= { X i And the corresponding predefined cutting layer x_labels= { n j Mapping f between }, denoted n j ＝f 2 (x i ). Last x_label i Equal to n j With maximum probability p ij The contour finishing cutting layer decision model is shown in fig. 3.
(b) And predicting the cutting layers of the sub-processing areas according to the decision model, and carrying out self-adaptive adjustment of the cutting layers of the sub-processing areas, and judging the same cutting layers and optimizing paths among the sub-processing areas.
The cutting layer of the sub-processing area is adaptively adjusted, and the method specifically comprises the following steps:
obtaining a sub-processing region SMR according to the mapping pattern f i Is not limited to the cutting layer n i The effective cutting depth of each layer isHowever, in the actual numerical control process design, different designers often set different depths of cuts a p But finally the same n can be obtained i The expression is
Wherein a is p For depth of cut, h is the difference in height of the top and bottom surfaces of the dynamic feature,for the effective depth of each layer, a max Maximum cutting depth allowed for tool []Is a rounding operation;
however, due to SMR i Geometric factors of (2) in actual processingA not according to the setting in the process p Processing is performed, so that the SMR can be processed i Is formed by cutting each cutting layer ofPerforming adaptive adjustment to enable n i Remain unchanged;
given 2 sub-process regions SMR with geometrical dependence i And SMR (SMR) j Wherein, SMR i For the target sub-process region, SMR j To sub-process regions that need to be merged, SMR j Depending on SMR i From f, we know SMR i And SMR (SMR) j Respectively n is needed for contour finishing of (c) i And n j Cutting layers, n i ≤n j ；
If SMR i And SMR (SMR) j Merging, SMR j At the position ofProfile surface between->And SMR (SMR) i With the same cutting layer->SMR respectively j Top surface height of SMR i Is a bottom surface height of (2); />Is SMR j Is a bottom surface height of (2); while SMR j At->Profile surface between->Is n j -n i The following equation needs to be satisfied:
wherein the contour surfaceThe cutting depth of each cutting layer is defined by +.>Adjust to-> Is SMR j Effective cutting depth of each layer of the cutting layer, < >>Is SMR i Effective cutting depth of each layer of the cutting layer; contour surface->The cutting depth of each cutting layer is defined by +.>Adjust to-> To adjust the sub-process region SMR j Effective cutting depth of each layer of the cutting layer; h is a j Is SMR j Height difference between top and bottom surfaces of (a);
according to SMR i And SMR (SMR) j The self-adaptive adjustment of the cutting layer of the sub-processing area is divided into the following 2 cases:
1) When n is i ＝n j If at the timeThen->Is defined by->Reduced to-> At least 1 cutting layer is required, SMR j At least need n j +1 cutting layers, SMR at this time j Will decrease the profile finishing efficiency of the SMR and therefore i Is not matched with SMR j Merging; however, if->Then SMR i And SMR (SMR) j Merging, marked as->
2) When n is i ≠n j When according toAnd->The relationship between them is divided into the following two types:
2a) A group II of the type which is known as such,
is defined by +.>Increase to-> Is defined by +.>Reduced to->By self-adaptive adjustment of the cutting depth, the deeper profile surface adopts smaller cutting depth, is relatively safer, and is particularly suitable for small cutters; in addition, a->Can still be based on->Calculating the cutting layer, thus SMR j May be smaller than n j The number of cutting layers can be reduced by combining sub-processing areas; therefore, SMR i And SMR (SMR) j At->With identical cutting layers in between, which may be combined, denoted +.>
2b) The group III of the type,
is defined by +.>Reduced to->But->Is defined by +.>Increase to->Make SMR j Is not limited to the cutting layer n j Remain unchanged. However, when a deeper profile surface adopts a larger cutting depth, a smaller tool is easily broken, and thus it is necessary to predict +_ according to the mapping pattern f>Whether or not the cutting layer of (2) is n j -n i If satisfied, that is, the following formula holds:
then SMR i And SMR (SMR) j Can be combined and recorded asOtherwise, it is not possible to combine.
Fig. 4 is a schematic view of adaptive calculation of a cutting layer in a sub-machining region, and fig. 4 (b), 4 (c) and 4 (e), 4 (f) show two types of adaptive adjustment examples of the cutting layer in the sub-machining region, respectively. From the figure, SMR 2 The upper layer cut depth of (2) is increased from 6.67 to 8, and the lower layer cut depth is decreased from 6.67 to 4. Therefore, from the cutter angle SMR 2 The lower profile finish of (c) is relatively conservative but also safer. SMR (SMR) 3 The upper layer cutting depth of (2) is reduced from 9 to 8, and the lower layer is increased from 9 to 11. Therefore, from the cutter angle SMR 3 The lower layer profile finishing of (2) is more aggressive and needs to be advanced by fAnd judging the rationality of the test piece.
(c) Calculating a process situation merging relation between different sub-processing areas with the same cutting layer, realizing the identification of dynamic characteristics, and constructing a driving geometry, wherein the method specifically comprises the following steps of:
(c1) Based on the sub-processing area self-adaptive adjustment strategy, according to the geometric dependence relationship among the sub-processing areas, obtaining the SMR of each target sub-processing area from top to bottom i Sub-process regions SMR to be consolidated having a class I, II, III mergeable relationship j Denoted as S i ：
However, SMR i And SMR (SMR) j Whether or not a dynamic feature can be incorporated is related to the geometric relative position of its contour.
(c2) Given S i Any one of the sub-processing regions SMR i Which consists of N profile chains C ij Each profile chain is composed of a set of adjacent profile surfaces, denoted as SMR i ＝{C ik K is more than or equal to 1 and less than or equal to N; constructing a virtual continuous cutting and discontinuous cutting judgment rule between profile chains; the method comprises the following steps:
according to the dynamic feature definition, if the sub-process region SMR i And SMR (SMR) j Can be combined into a dynamic feature, then the SMR i And SMR (SMR) j At least one pair (C) in ,C jm ) Make the cutter at C im And C jn Continuously cutting along the newly constructed virtual contour according to the feeding speed without cutting movement (including cutting, lifting, quick and the like), and merging and calculating the dynamic characteristics of the sub-processing area as shown in fig. 5;
conversely, if SMR i And SMR (SMR) j Is not in existence (C) in ,C jm ) SMR then i And SMR (SMR) j Not being combinable into a dynamic feature, indicating that the machining time of the discontinuous cut by the tool is lower than the continuous cutting time of the tool;
typically, virtual attachment of toolsThe continuous cut exists between two profile chains of co-straight/co-circular or co-simple profile, thus giving S i Any two non-adjacent profile chains C in And C jm If C in And C jm One of the following conditions is satisfied:
1)C in and C jm An edge which can be co-straight or co-arc exists between the two edges;
2)C in and C jm Not more than 2 profile surfaces are adjacent;
then at C in And C jm Constructing a virtual contour connection C between in And C jm So that the cutter is at C in And C jm Is cut continuously, and is marked as VCM (C) in ,C jm )＝1。
FIG. 6 is a diagram showing an example of virtual connections of non-adjacent profile chains, from which it can be seen that SMR 1 ～ 6 Co-arc of SMR 7,8 、SMR 9,10 And SMR (SMR) 11,12 Collinear, SMR 11,12 And SMR (SMR) 13 ～ 16 Belonging to the 1-profile-surface adjacency, a virtual profile surface can thus be constructed between them, enabling the tool to cut continuously.
(c3) S, constructing according to a virtual continuous cutting and discontinuous cutting judgment rule between profile chains i Is a contour chain adjacency attribute graph G i Converting contour finishing dynamic feature recognition into G i The vertex and edge access path optimization of (1) is as follows:
s, constructing based on virtual continuous cutting and discontinuous cutting judgment rules among profile chains i Is a contour chain adjacency attribute graph G i The vertex corresponds to the contour chain, the side corresponds to the relation between the contour chains, including adjacent, virtual continuous cutting and discontinuous cutting, and multiple paths can exist between the vertices; a schematic of a partial contour chain adjacency attribute graph is shown in fig. 7;
conversion of contour finishing dynamic feature recognition to G i Vertex and edge access path optimization through each set of vertex and edge access paths L i Obtaining S i Contour finishing time t of neutron machining region i ComprisesContinuous cutting and non-cutting; in the context of the contour finishing process, the solution with the shortest time in all access paths is the optimal dynamic feature.
(c4) In view of the global searching advantage of the genetic algorithm, solving the solution with the shortest time in all access paths, namely, the optimal dynamic characteristics by the genetic algorithm is as follows:
the optimal dynamic characteristics are recorded as
Where T is the shortest access time, T (x) represents the path access time,according to L i New contour chain constructed by medium virtual continuous cutting edge R ij Is L i The non-cutting edges of (a), N and M are respectively +.>And R is R ij Is the number of (3);
solving the above formula by adopting a genetic algorithm, wherein fig. 8 shows a genetic algorithm individual representation, and each individual c is composed of a virtual edge sequence between vertexes;
wherein each gene c i Indicating whether or not cutting is continued between two vertices, if c i 1, representing continuous cutting between two vertices; otherwise, the cutter needs to perform non-cutting movement;
calculating an optimal contour finishing time based on each individual c by reconstructing a driving geometry of contour finishing; the individual fitness function f (c) is defined as the reciprocal of the access time t (c) of each individual, i.e
According to basic steps of a genetic algorithm, the optimization steps of the profile chain finish machining path between different sub-machining areas are as follows:
step1: randomly generating a designated number of initial populations according to the individual representation method, and taking the initial populations as current populations;
step2: calculating the fitness of each individual in the current population, and marking the individual with the largest fitness as the optimal individual in the current population;
step3: comparing with the optimal individual searched in the previous round, and if the optimal individual in the current population is better than the optimal individual, updating the optimal individual;
step4: selecting two individuals from the current population based on the individual fitness value of the current population by adopting a roulette method, and executing crossover and mutation operations on the two individuals to generate new individuals;
step5: repeating step4 until a next generation population is generated;
step6: repeating step 2-step 5 until convergence tolerance or maximum iteration number is reached, and selecting individuals with optimal fitness from all populations as optimal solutions;
step7: and extracting the contour finish machining process situation merging relation among different sub-machining areas according to the optimal individual, constructing a driving geometry, and realizing dynamic feature identification under the contour finish machining process situation.
The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be appreciated by persons skilled in the art that the above embodiments are not intended to limit the invention in any way, and that all technical solutions obtained by means of equivalent substitutions or equivalent transformations fall within the scope of the invention.

Claims (6)

1. The self-adaptive finish machining method for the complex part contour based on the dynamic machining characteristics is characterized by comprising the following steps of: the method comprises the following steps:
(a) Extracting contour finishing process data from a structural process database, carrying out characterization and deep learning on the contour finishing process data to obtain a contour finishing cutting layer decision model, and extracting mapping modes f between different process situations and cutting layers in the process data;
(b) Predicting cutting layers of the sub-processing areas according to the decision model, and performing self-adaptive adjustment of the cutting layers of the sub-processing areas, and judging the same cutting layers and optimizing paths among the sub-processing areas;
the cutting layer of the sub-processing area is adjusted in a self-adaptive way, and the method comprises the following steps:
obtaining a sub-processing region SMR according to the mapping pattern f i Is not limited to the cutting layer n i The expression is
Wherein a is p For depth of cut, h is the difference in height of the top and bottom surfaces of the dynamic feature,for the effective depth of each layer, a max Maximum cutting depth allowed for tool []Is a rounding operation; by pairing SMR i Is +.>Performing adaptive adjustment to enable n i Remain unchanged;
given 2 sub-process regions SMR with geometrical dependence i And SMR (SMR) j Wherein, SMR i For the target sub-process region, SMR j To sub-process regions that need to be merged, SMR j Depending on SMR i ，SMR i And SMR (SMR) j Respectively n is needed for contour finishing of (c) i And n j Cutting layers, n i ≤n j ；
If SMR i And SMR (SMR) j Merging, SMR j At the position ofProfile surface between->And SMR (SMR) i Having the same cutting layer as that of the cutting layer,SMR respectively j Top surface height of SMR i Is a bottom surface height of (2); />Is SMR j Is a bottom surface height of (2); while SMR j At->Profile surface between->Is n j -n i The following equation needs to be satisfied:
wherein the contour surfaceThe cutting depth of each cutting layer is defined by +.>Adjust to-> Is SMR j Effective cutting depth of each layer of the cutting layer, < >>Is SMR i Effective cutting depth of each layer of the cutting layer; contour surface->The cutting depth of each cutting layer is defined by +.>Adjust to-> To adjust the sub-process region SMR j Effective cutting depth of each layer of the cutting layer; h is a j Is SMR j Height difference between top and bottom surfaces of (a);
according to SMR i And SMR (SMR) j The self-adaptive adjustment of the cutting layer of the sub-processing area is divided into the following 2 cases:
1) When n is i ＝n j If at the timeThen->Is defined by->Reduced to-> At least 1 cutting layer is required, SMR j At least need n j +1 cutting layers, SMR i Is not matched with SMR j Merging; if->Then SMR i And SMR (SMR) j Merging and recording as
2) When n is i ≠n j When according toAnd->The relationship between them is divided into the following two types:
2a) A group II of the type which is known as such,
SMR i and SMR (SMR) j At the position ofWith the same cutting layer in between, combined, denoted +.>
2b) The group III of the type,
prediction from mapping pattern fWhether or not the cutting layer of (2) is n j -n i A plurality of; if so, SMR i And SMR (SMR) j Merging, marked as->Otherwise, not merging;
(c) Calculating process situation merging relations among different sub-processing areas with the same cutting layer, realizing the identification of dynamic characteristics, and constructing a driving geometry;
the dynamic characteristics are defined as a common machining area DF where the tool can continuously cut under a given technological situation, and are expressed as
Wherein z is s 、z e Respectively a top surface and a bottom surface with dynamic characteristics, DG is a driving geometry with dynamic characteristics, SMR i G () is a driving geometry building function for sub-processing regions with geometry dependencies;
under a given process situation, SMR i At [ z ] s ,z e ]With the same cutting layer therebetween, and the cutter has a pair of SMRs at each cutting layer i Performing interference-free continuous processing, wherein i=1, …, n and n are the number of sub-processing areas;
obtaining the SMR of each target sub-processing area from top to bottom according to the geometrical dependency relationship between the sub-processing areas i Sub-process regions to be consolidated SMR having a combinable relationship j Denoted as S i ；
Given S i Any one of the sub-processing regions SMR i Which consists of N profile chains C ik Each profile chain is composed of a set of adjacent profile surfaces, denoted as SMR i ＝{C ik K is more than or equal to 1 and less than or equal to N; the virtual continuous cutting and discontinuous cutting judgment rules between profile chains are constructed, and the method specifically comprises the following steps:
if the sub-processing region SMR i And SMR (SMR) j Merge into one dynamic feature, SMR i And SMR (SMR) j At least one pair (C) in ,C jm ) Make the cutter at C im And C jn Continuously cutting along the constructed virtual contour;
conversely, if SMR i And SMR (SMR) j Is not in existence (C) in ,C jm ) SMR then i And SMR (SMR) j Not combined into a dynamic feature, indicating discontinuous cutting by the toolThe machining time is lower than the continuous cutting time of the cutter;
given S i Any two non-adjacent profile chains C in And C jm If C in And C jm One of the following conditions is satisfied:
1)C in and C jm An edge with a common straight line or a common arc exists between the two edges;
2)C in and C jm Not more than 2 profile surfaces are adjacent;
then at C in And C jm Constructing a virtual contour connection C between in And C jm So that the cutter is at C in And C jm Is cut continuously, and is marked as VCM (C) in ,C jm )＝1；
S, constructing according to a virtual continuous cutting and discontinuous cutting judgment rule between profile chains i Is a contour chain adjacency attribute graph G i Converting contour finishing dynamic feature recognition into G i The vertex and edge access path optimization of (1) is as follows:
s, constructing based on virtual continuous cutting and discontinuous cutting judgment rules among profile chains i Is a contour chain adjacency attribute graph G i The vertex corresponds to the contour chain, the side corresponds to the relation between the contour chains, including adjacent, virtual continuous cutting and discontinuous cutting, and multiple paths can exist between the vertices;
conversion of contour finishing dynamic feature recognition to G i Vertex and edge access path optimization through each set of vertex and edge access paths L i Obtaining S i Contour finishing time t of neutron machining region i The method comprises the steps of carrying out a first treatment on the surface of the The solution with the shortest time in all the access paths is the optimal dynamic characteristic;
under the situation of the contour finishing process, the solution with the shortest time in all access paths is the optimal dynamic characteristic, and the optimal dynamic characteristic is solved through a genetic algorithm.
2. The complex part contour adaptive finishing method based on dynamic machining features according to claim 1, wherein: the step (a) specifically comprises the following steps:
1.1, under the situation of a contour finishing process, characterizing contour finishing process data;
1.2 according to the contour finish machining process data characterized in 1.1, a deep neural network classifier model is adopted to learn the implicit judgment rule of the cutting layer under the given contour finish machining process situation, and a contour finish machining cutting layer decision model is obtained.
3. The complex part contour adaptive finishing method based on dynamic machining features according to claim 2, wherein: in the step 1.1, under the circumstance of the contour finishing process, the contour finishing process data are characterized as follows:
in the contour finishing process context, according to the cutting layer of the contour finishing operation of the geometric decision to be processed and the cutter T, the finishing of the contour surface is described as follows:
given one at z e Deep, side machining allowance is a e Side with wall height h, z s ＝z e +h, when a tool T is used, n cutting layers are required; z s Is the height, z, corresponding to the top surface of the dynamic feature e Is the height corresponding to the dynamic feature bottom surface;
for any one contour finish sample in the structuring process data, it is expressed as contour finish process context x i And cutting layer n i A mapping between them, denoted as
x i ＝{D,L,FL,z s ,z e ,a e }→x_label i ＝n i
Wherein D is the diameter of the cutter, L is the length of the cutter, FL is the edge length of the cutter, and x_label i The cutting layer is predefined for the i-th.
4. The complex part contour adaptive finishing method based on dynamic machining features according to claim 2, wherein: in the step 1.2, a deep neural network classifier model is adopted to learn the implicit judgment rule of the cutting layer under the given contour finish machining process situation, and the method concretely comprises the following steps:
learning a cutting layer implicit judgment rule under a given contour finish machining process situation by adopting a four-layer deep neural network classifier model to obtain a contour finish machining cutting layer decision model;
the input layer of the model is a process situation parameter X= { X related to contour finishing cutting layer calculation i -a }; the hidden layer is 2 full-connection layers with 20 nodes;
the output layer of the model is x i Mapping to a predefined cutting layer x_labels= { n j Each element n in } j Probability p of (2) ij Represents x i Requiring n j The possibility of contour finishing of the individual cutting layers.
5. The complex part contour adaptive finishing method based on dynamic machining features according to claim 1, wherein: solving the optimal dynamic characteristics through a genetic algorithm, wherein the optimal dynamic characteristics are as follows:
the optimal dynamic characteristics are recorded as
Where T is the shortest access time, T (x) represents the path access time,according to L i New contour chain constructed by medium virtual continuous cutting edge R ij Is L i The non-cutting edges of (a), N and M are respectively +.>And R is R ij Is the number of (3);
solving the formula by adopting a genetic algorithm, wherein each individual c is composed of a virtual edge sequence between vertexes; wherein each gene c i Indicating whether or not cutting is continued between two vertices, if c i 1, representing a continuous cut between two verticesCutting; otherwise, the cutter needs to perform non-cutting movement; from each individual c, an optimal profile finishing time based on that individual is calculated by reconstructing the driving geometry of the profile finishing.
6. The method for adaptively finishing the contour of a complex part based on dynamic machining features according to claim 5, wherein the method comprises the following steps: based on a genetic algorithm, the optimization steps of the contour chain finish machining path between different sub-machining areas are as follows:
step1: randomly generating a designated number of initial populations according to the individual representation method, and taking the initial populations as current populations;
step2: calculating the fitness of each individual in the current population, and marking the individual with the largest fitness as the optimal individual in the current population;
step3: comparing with the optimal individual searched in the previous round, and if the optimal individual in the current population is better than the optimal individual, updating the optimal individual;
step4: selecting two individuals from the current population based on the individual fitness value of the current population by adopting a roulette method, and executing crossover and mutation operations on the two individuals to generate new individuals;
step5: repeating step4 until a next generation population is generated;
step6: repeating step 2-step 5 until convergence tolerance or maximum iteration number is reached, and selecting individuals with optimal fitness from all populations as optimal solutions;
step7: and extracting the contour finish machining process situation merging relation among different sub-machining areas according to the optimal individual, constructing a driving geometry, and realizing dynamic feature identification under the contour finish machining process situation.
CN202110710104.1A 2021-06-25 2021-06-25 Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics Active CN113592772B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110710104.1A CN113592772B (en) 2021-06-25 2021-06-25 Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110710104.1A CN113592772B (en) 2021-06-25 2021-06-25 Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics

Publications (2)

Publication Number Publication Date
CN113592772A CN113592772A (en) 2021-11-02
CN113592772B true CN113592772B (en) 2023-11-14

Family

ID=78244447

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110710104.1A Active CN113592772B (en) 2021-06-25 2021-06-25 Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics

Country Status (1)

Country Link
CN (1) CN113592772B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587348A (en) * 2008-05-23 2009-11-25 蔡玉俊 Method for processing spiral line interpolation of three-dimensional cutting edge contour
WO2015096511A1 (en) * 2013-12-24 2015-07-02 沈阳飞机工业（集团）有限公司 Intelligent numerical control machining programming system and method for aircraft structural parts
EP3345713A1 (en) * 2017-01-05 2018-07-11 Robert Bosch GmbH Method and device for process-oriented beam shape adjustment and beam orientation
CN111797469A (en) * 2020-06-24 2020-10-20 华中科技大学无锡研究院 Aeroengine casing technological parameter optimization method based on machining cutter back-off deformation constraint

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101587348A (en) * 2008-05-23 2009-11-25 蔡玉俊 Method for processing spiral line interpolation of three-dimensional cutting edge contour
WO2015096511A1 (en) * 2013-12-24 2015-07-02 沈阳飞机工业（集团）有限公司 Intelligent numerical control machining programming system and method for aircraft structural parts
EP3345713A1 (en) * 2017-01-05 2018-07-11 Robert Bosch GmbH Method and device for process-oriented beam shape adjustment and beam orientation
CN111797469A (en) * 2020-06-24 2020-10-20 华中科技大学无锡研究院 Aeroengine casing technological parameter optimization method based on machining cutter back-off deformation constraint

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
NC编程系统中切削模型的统一表示;禹涌, 唐荣锡, 徐希;计算机辅助设计与图形学学报(第03期);全文 *
关联工艺引导的型腔类零件局部结构检索方法;黄瑞;蒋俊锋;张树生;;计算机辅助设计与图形学学报(第04期);全文 *

Also Published As

Publication number Publication date
CN113592772A (en) 2021-11-02

Similar Documents

Publication Publication Date Title
WO2020034632A1 (en) Mbd-based three-dimensional process designing method and platform for typical automobile machined part
Wu et al. A data mining approach for spatial modeling in small area load forecast
CN107578149B (en) Power grid enterprise key data analysis method
Cateni et al. A multivariate fuzzy system applied for outliers detection
CN112464366B (en) Multi-fidelity shape optimization method of autonomous underwater vehicle based on data mining
Wang et al. Feature weighting fuzzy clustering integrating rough sets and shadowed sets
Deng et al. A data-drivenparameter planning method for structural Parts NC machining
Jang et al. Managing approximation models in collaborative optimization
CN114611379A (en) Machining process energy-saving planning method based on data driving
CN113592772B (en) Self-adaptive finish machining method for complex part contour based on dynamic machining characteristics
CN112861433B (en) Product low-carbon design method based on multi-level integrated framework
Bhuiyan et al. Optimization of cutting parameters in turning process
Wang et al. Performance-based predictive models and optimization methods for turning operations and applications: Part 3—optimum cutting conditions and selection of cutting tools
Büche Multi-objective evolutionary optimization of gas turbine components
JP4857228B2 (en) Plant operation optimization device
Ivakhnenko et al. Problems of further development of GMDH algorithms: Part 2
Wang et al. Transfer optimization in accelerating the design of turbomachinery cascades
CN112796747B (en) Oil and gas drilling strategy prediction method and device based on multiple targets
CN112596391B (en) Deep neural network large time lag system dynamic modeling method based on data driving
CN115510517A (en) Intelligent design method for high-rise building structure
CN114330005A (en) Global optimization and decision-making method for three-dimensional development well pattern
CN116933449B (en) Transmission gear robust design optimization method based on gradient auxiliary learning function
Shang A Study of Deep Learning Neural Network Algorithms and Genetic Algorithms for FJSP
CN115467649B (en) Well pattern optimizing deployment method and system for small well spacing three-dimensional development and three-dimensional well pattern
CN116822971B (en) Well wall risk level prediction method

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