A complex parts process route planning method based on re-optimization genetic algorithm

Abstract

The present invention discloses a method for process route planning of complex parts based on a re-optimization genetic algorithm, comprising the following steps: Step 1, based on the minimum processing cost, process route planning is performed and a multi-objective optimization model is established; Step 2, the optimal process route of complex parts is solved by re-optimizing the genetic algorithm, specifically: based on the process step matrix, the traditional crossover and mutation links are re-optimized, and multi-fragment crossover and local recombination mutation are proposed, so as to efficiently update the population and obtain the optimal process route. The present invention re-optimizes the traditional crossover and mutation links based on the process step matrix, and multi-fragment crossover and local recombination mutation are proposed, so as to efficiently update the population and obtain the optimal process route. Case studies and analyses of typical cabin parts effectively prove the effectiveness and superiority of the algorithm.

Description

Technical Field

[0001] The invention relates to a process route planning method, and in particular to a complex parts process route planning method based on a re-optimized genetic algorithm. Background Art

[0002] Process planning is an indispensable part of product design and manufacturing. Its purpose is to standardize the sequencing of processing procedures and make the overall processing process of the product economical and mature. Process planning is a complex decision-making process, including process priority determination, processing feature solution selection, process sequencing problems, and the selection of manufacturing resources such as machine tools, cutting tools and fixtures. Since the planning process involves a large number of process sequencing optimizations and the configuration of various manufacturing resources, process planning is a complex combinatorial optimization problem and is an NP-hard problem. There is currently no definite solution to NP-hard problems. How to obtain the approximate optimal solution in process planning more quickly and accurately has become a topic of focus for scholars.

[0003] The difficulty of process planning lies mainly in the sorting of processing procedures and the allocation of manufacturing resources. For the problem of sorting processing procedures, due to the existence of a processing sequence, each process must satisfy a specific priority constraint relationship. If the priority constraint relationship is destroyed, the resulting process route will be infeasible. In this regard, scholars mainly use three processing methods. First, topological sorting and encoding. The process is encoded according to the priority, and the feasible solution space of the associated process sequence is searched to obtain the best feasible solution. Second, the penalty method elimination method. By designing a penalty method, the route that violates the priority relationship in the process planning is eliminated, so as to achieve the optimization purpose. Third, the repair operator adjustment. In the process of process planning, the infeasible solution is repaired into a feasible solution through the repair operator, so as to further optimize it. Among the above three methods, the penalty method is simple to operate and easy to handle, but it will lead to an expansion of the search space and increase the calculation time; the repair operator reduces the search space, but increases the difficulty of algorithm operation; the topological sorting method can relatively better balance the search space, operation difficulty and calculation time, and has been widely used in recent years. On the other hand, for manufacturing resource allocation, after the process is sorted, the most suitable machine tools, cutting tools and fixtures should be selected for each process, so that the process route can achieve the purpose of economy and high quality. The widely used methods include edge selection method and intelligent search method.

[0004] For the above-mentioned combinatorial optimization problems, optimization techniques are commonly used to solve them. At present, optimization techniques have been widely used in practical engineering fields, such as steel structure optimization, image processing, path planning, PID control and other problems. Among various optimization techniques, the optimization performance of heuristic algorithms has been widely recognized. At present, genetic algorithms (GA), particle swarm optimization algorithms (PSO), ant colony algorithms (ACO), simulated annealing algorithms (SA), etc. have been widely used in process planning. These algorithms and their improved algorithms are oriented to different product parts and have their own advantages under different process conditions. For products with certain processing characteristics and certain process conditions, they can effectively obtain their optimal process routes and have achieved very impressive results.

[0005] Although the above-mentioned heuristic optimization algorithm can solve the process route planning problem, for parts with numerous features and complex processes, the current conventional algorithm does not work well due to the large number of process steps, large search space, and high calculation difficulty. There is still room for improvement in optimization speed and optimal results. Summary of the invention

[0006] In order to solve the shortcomings of the above-mentioned technology, for parts with numerous features and complex processes in process planning, the present invention provides a complex part process route planning method based on a re-optimized genetic algorithm to promote the process optimization process.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a complex parts process route planning method based on a re-optimized genetic algorithm, comprising the following steps:

[0008] Step 1: Based on the minimum processing cost, process route planning is carried out and a multi-objective optimization model is established;

[0009] Step 2: Solve the multi-objective optimization model through re-optimization genetic algorithm to solve the optimal process route for complex parts. Specifically: re-optimize the traditional crossover and mutation links based on the process step matrix, and propose multi-fragment crossover and local recombination mutation, so as to efficiently update the population and obtain the optimal process route.

[0010] Preferably, in step 1, a multi-objective optimization model is established by taking processing cost as the objective function and considering various costs of various manufacturing resources;

[0011] The total processing cost of a part includes the cost of using and replacing manufacturing resources;

[0012] When calculating the replacement cost, it is necessary to compare the machine tool, tool and TAD in pairs and define two comparison operators as follows:

[0013]

[0014] Note: On the same machine tool, the parts have the same feed direction TAD after one clamping.

[0015] Preferably, in step 1, the total use cost UC includes two parts: the total machine tool use cost and the total tool use cost, as shown in formula (5):

[0016]

[0017] Wherein, step[i] represents step i, i is a positive integer; step[i].macID represents the machine tool number selected for step i; step[i].toolID represents the tool number selected for step i; step[i].tadID represents the TAD number selected for step i;

[0018] The total replacement cost CC includes three parts: total machine tool replacement cost, total tool replacement cost and total fixture replacement cost, as shown in the following formula:

[0019]

[0020] Where:

[0021] nmc is the total number of machine tool changes in the process route, and the calculation formula is as follows:

[0022]

[0023] NTC is the total number of tool changes in the process route, and the calculation formula is as follows:

[0024]

[0025] nsc is the number of TAD changes in the process route, and the calculation formula is as follows:

[0026]

[0027] The total processing cost TC is the sum of the total use cost and the total replacement cost;

[0028] TC=UC+CC (10)

[0029] Wherein, MCI[m] represents the cost index of machine tool m, where m represents the machine tool number and is a positive integer; TCI[t] represents the cost index of tool t, where t represents the tool number and is a positive integer; MCCI, TCCI and SCCI represent the cost index of machine tool, tool and fixture change, respectively, i.e., the cost required to change the machine tool, tool or fixture between two working steps.

[0030] Preferably, in step 2, reoptimizing the genetic algorithm comprises the following steps:

[0031] Step S1, set basic parameters and initialize gene population;

[0032] Step S2, calculating the fitness values ​​of individuals in the initialized gene population one by one, and selecting high-quality individuals as the parent population according to the roulette method;

[0033] Step S3, updating and optimizing the population through multi-fragment crossover and local recombination mutation behaviors until the maximum number of iterations is met;

[0034] Step S4: output the best process route and its corresponding resource allocation.

[0035] Preferably, in step S1, the work step priority graph is converted into an adjacency matrix for storage and calculation, which is recorded as a 0-1 work step matrix and is expressed as:

[0036] D={d ij} (11)

[0037] In the formula, i and j are positive integers, indicating the process step number, d ij It represents the priority relationship between step i and step j as follows:

[0038]

[0039] After the work step matrix is ​​generated, the processing priority relationship between each work step can be determined. Then, the work steps can be topologically sorted according to the work step matrix and the process route can be randomly generated.

[0040] Preferably, the specific steps of initializing the gene population are shown in Algorithm 1:

[0041] Algorithm 1: Process route generation

[0042] Step 1: Input the process step matrix D;

[0043] Step 2: Search the work step matrix column by column, find the full 0 column, and add it to the list list_temp;

[0044] Step 3: Randomly select element i from list_temp and add it to list list_result. At the same time, delete element i from list_temp.

[0045] Step 4: Search for the elements in the i-th row, find all the columns corresponding to 1, and add them to list_temp;

[0046] Step 5: Repeat steps 3 and 4, and perform deduplication every time an element is added to list_result until list_result contains all the work steps;

[0047] Step 6: Output the process route corresponding to list_result. After multiple iterations, the initial population can be generated.

[0048] Preferably, in step S2, the roulette wheel method is used for elite selection to ensure that the individuals to be crossed are high-quality individuals. The specific steps are shown in Algorithm 2:

[0049] Algorithm 2: Roulette method to generate parent population

[0050] Step 1: Calculate the fitness value f(x i ), determine the proportion pi and cumulative proportion p of each fitness acc ;

[0051] Step 2: Generate a random number rnd∈(0,1) as the selection threshold;

[0052] Step 3: According to the cumulative proportion of fitness p acc , find the individual whose fitness value is closest to and greater than the random number rnd;

[0053] Step 4: Repeat the above steps until the population size is met;

[0054] f(x i The specific calculation formula of ) is shown in formula (15):

[0055]

[0056] In the formula, C is a constant and a positive number;

[0057]

[0058] In the formula, p i represents the probability of the i-th chromosome being selected, x i represents the i-th chromosome, i is a positive integer, N represents the population size, f(x i ) represents the fitness of the ith individual.

[0059] Preferably, in step S3, the multi-segment crossover comprises the following steps:

[0060] Step 1: Select any individual in the parent population as parent 1, and any individual in the mother population as mother 1;

[0061] Step 2: Select N in Parent 1 c crossover segments, each containing L c Number of individuals;

[0062] Step 3: Find the individuals in parent generation 1 that correspond to parent generation 1, and rearrange the fragments in parent generation 1 according to the order of individuals in parent generation 1;

[0063] Step 4: Output the new feasible offspring 1.

[0064] Preferably, in step S3, the local recombination mutation comprises the following steps:

[0065] 1) In the population that has just completed multi-segment crossover, randomly select a segment containing the mutant gene;

[0066] 2) extracting the process step matrix of the selected segment based on the initial process step matrix;

[0067] 3) forming new fragments according to the process step matrix and obtaining a new process step sequence;

[0068] 4) For new segments, obtain existing manufacturing resources;

[0069] 5) Reorganize manufacturing resources and optimize resource allocation;

[0070] 6) Complete the mutation process.

[0071] Preferably, completing the mutation process includes the following processes: completing the process step sequencing of the mutant gene; completing the resource recombination optimization of the mutant gene; and forming a new process route.

[0072] The present invention proposes a re-optimized genetic algorithm (ROGA) with an efficient update mechanism, focusing on multi-segment crossover and local recombination mutation links, and planning the optimal process route for multi-feature complex parts. First, based on the minimum processing cost, a multi-objective optimization model for process route planning is established. Secondly, a re-optimized genetic algorithm is designed. Based on the process step matrix, the traditional crossover and mutation links are re-optimized, and multi-segment crossover and local recombination mutation are proposed, so as to efficiently update the population and obtain the optimal process route. Finally, case studies and analysis are carried out through typical cabin parts to effectively prove the effectiveness and superiority of the algorithm.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1) Adopting multi-fragment crossover and local recombination mutation: In the proposed re-optimization genetic algorithm, a novel population update mechanism is designed, which takes into account the number of crossover fragments Nc and the length of crossover fragments Lc in the traditional crossover link, and the length of mutation fragments Lm in the traditional mutation link. The comprehensive and extensive considerations improve the population update speed and algorithm performance, and also speed up the algorithm calculation time.

[0075] 2) Obtain process planning for complex parts with multiple features: A typical cabin part with numerous features and complex processes is studied. By comparing it with different parameters, traditional processes, and different algorithms, the optimal process route and the lowest cost for the output parts under different conditions are obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A process diagram is generated for the work step priority diagram of the present invention.

[0077] Figure 2 This is an example of the process step priority diagram of the present invention.

[0078] Figure 3 It is a schematic diagram of the process route coding of the present invention.

[0079] Figure 4 Schematic diagram of the multi-segment crossover process of the present invention.

[0080] Figure 5 It is a diagram of the local recombination variation process of the present invention.

[0081] Figure 6 ROGA flow chart of the present invention.

[0082] Figure 7 Schematic diagram of typical cabin parts according to an embodiment of the present invention.

[0083] Figure 8 This is a typical cabin parts process step priority diagram of an embodiment of the present invention.

[0084] Fig. 9 This is a typical 0-1 cabin parts process step matrix diagram of an embodiment of the present invention.

[0085] Fig.10 ROGA key parameter determination process diagram of an embodiment of the present invention.

[0086] Fig.11 This is a diagram of the iterative process comparing the present invention with other algorithms. DETAILED DESCRIPTION

[0087] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0088] 1. Establishment of multi-objective optimization model

[0089] Part process route planning is the core content of CAPP (computer-aided process planning) and has an important impact on product manufacturing. Generally, a part can contain m manufacturing features, and each feature is processed through n processes (steps), as shown in formula (1-2). The processing of each process requires the use of corresponding machine tools, tools, and fixtures. For a complex part, its process route is often diverse, and how to sort the processes and select resources has become a major problem in part process design.

[0090]

[0091] In the existing process route planning, process engineers often determine the priority of each processing operation based on the constraints of the process sequence rules and prior knowledge, so as to plan the process route of the part more quickly. After induction and summary, the commonly used process sequence rules are mainly: (1) rough first, then fine; (2) benchmark first;

[0092] (3) The primary one comes first, then the secondary one; (4) The surface one comes first, then the hole; (5) The curved surface one comes first, then the surface one, etc.

[0093] The above process constraints are represented by a directed acyclic graph, such as Figure 1 As shown in the figure, a) represents the processing features in the physical part, b) represents the processing plan corresponding to the processing features after abstraction, and c) represents the process step priority diagram formed according to the specific process sorting rules.

[0094] The work step priority graph is a directed graph consisting of two parts: nodes (work steps) and directed edges (direction relationships), which can clearly represent the priority relationship of work steps. Figure 1 In the figure, if there is a directed edge from node a to node b, node a takes precedence over node b, that is, step a must be processed before step b. If there is no directed edge between the two nodes, it means that there is no clear priority relationship between the two steps. The step priority diagram can determine the order of processing of specific nodes, which is the basis for determining the subsequent part process route.

[0095]

Establishment of multi-objective function

[0096] The actual processing and manufacturing situation shows that enterprises and factories can improve production efficiency and flexibility, reduce processing costs and ensure processing quality by rationally and flexibly planning the process routes of products. The essence of process route planning is a combinatorial optimization problem under rule constraints. The objective function is to take the processing cost, which is the most concerned by most enterprises, as the objective function, and consider the various costs of various manufacturing resources to establish a multi-objective function model, which is then solved through ROGA to achieve the goal of minimum processing cost.

[0097] The total machining cost (TC) of a part includes two aspects: the usage cost (UC) of manufacturing resources and the change cost (CC). In the calculation process of various costs, the relevant symbol variables are defined:

[0098] step[i] represents step i, i is a positive integer; step[i].macID represents the machine tool number selected for step i; step[i].toolID represents the tool number selected for step i; step[i].tadID represents the TAD number selected for step i;

[0099] Here is a special explanation: on the same machine tool, if the parts have the same feed direction (Tool Approach Direction, TAD) after one clamping, the fixture change can be converted into a simpler feed direction change. MCI[m] represents the cost index of machine tool m, where m represents the machine tool number and is a positive integer;

[0100] TCI[t] represents the usage cost index of tool t, where t represents the tool number and is a positive integer; MCCI, TCCI and SCCI represent the change cost index of machine tools, tools and fixtures, respectively, that is, the cost required to change the machine tool, tool or fixture between two work steps.

[0101] At the same time, when calculating the replacement cost, it is necessary to compare the machine tool, tool and TAD in pairs. Hereby two comparison operators are defined as follows:

[0102]

[0103] Table 1 Description of changes in machine tools, tools and clamping

[0104]

[0105] (1) Total cost of use UC, including total machine tool cost and total tool cost, as shown in formula (5):

[0106]

[0107] (2) Total replacement cost CC, including total machine tool replacement cost, total tool replacement cost and total fixture replacement cost, as shown in formula (6):

[0108]

[0109] Where: nmc is the total number of machine tool changes in the process route, and the calculation formula is as follows:

[0110]

[0111] NTC is the total number of tool changes in the process route, and the calculation formula is as follows:

[0112]

[0113] nsc is the number of TAD changes in the process route, and the calculation formula is as follows:

[0114]

[0115] (3) Total processing cost TC is the sum of total use cost and total replacement cost.

[0116] TC=UC+CC (10)

[0117] 2. Re-optimization of genetic algorithm design

[0118] 1. Genetic concepts and mapping relationships

[0119] Genetic algorithm is an optimization algorithm that simulates natural selection and genetic mechanism, and is widely used to solve complex optimization problems. Its basic idea is derived from Darwin's theory of evolution, and it simulates the process of "survival of the fittest" to evolve the optimal solution generation by generation.

[0120] Genetic algorithms mainly include four basic operations: encoding, selection, crossover, and mutation. First, encoding is performed to represent the solution in the problem space as chromosomes. Then, the better individuals are selected according to the fitness function. The optimal solution is finally obtained by continuously iterating and optimizing the chromosomes through crossover and mutation operations. In the iterative evolution process, genetic algorithms can search for the global optimal solution or a solution close to the optimal solution. They have strong global search capabilities and parallelism, and are suitable for complex, high-dimensional optimization problems.

[0121] As a typical heuristic algorithm, genetic algorithm and its improved algorithm have made breakthroughs in many fields, such as traveling salesman problem, shop scheduling, machine learning parameter optimization, etc. Similarly, in the part process route optimization problem, there are many processing features, complex process step sequencing, and each process step corresponds to a non-unique machine tool, tool, and fixture, which leads to the process route obtained by conventional methods being not ideal or the result does not meet the rule constraints. Therefore, this solution uses a re-optimization genetic algorithm to establish a ROGA algorithm model to solve the optimal process route for complex parts.

[0122] The following defines the concept mapping of process planning in the ROGA algorithm:

[0123] (1) Chromosome: a process route;

[0124] (2) Gene: a process (step) in the process route;

[0125] (3) Population: a collection of multiple process routes;

[0126] (4) Selection, crossover, and variation: various behaviors that change the process route.

[0127] 2. Algorithm design process

[0128] 2.1 Coding Scheme

[0129] In process planning, a process route consists of multiple processes or steps, and each step involves different machine tools, tools, and fixtures. Therefore, process route planning is a complex combinatorial optimization problem. To facilitate subsequent algorithm design and calculation, we refine the process route planning problem into two sub-problems:

[0130] 1) Process step sequencing problem: By reasonably sequencing the process steps, the best process route can be selected.

[0131] 2) Resource selection problem: Determine the appropriate manufacturing resources (machine tools, tools and feed direction) for each process step to achieve the best state.

[0132] Since all the numbers involved in the calculation process of the above problem are positive integers, integer encoding is used for encoding. Figure 2 The work step priority diagram for machining a certain part consists of 10 work steps in total. Figure 3 In the example, the 4×10 matrix represents a chromosome and is composed of Figure 2 The process route code consists of 10 steps. Horizontally, the first row represents the sequence of steps, which consists of 10 steps in random order. The last three rows together constitute the resource matrix of the process route. The values ​​in the second, third, and fourth rows represent the machine tool number, tool number, and TAD number selected for each step, respectively. Vertically, each column represents a gene, which contains the steps and their corresponding manufacturing information. The permutations and combinations of each gene are varied, and the ordering between genes is complex and changeable, which ultimately leads to different total processing costs of the process route. Based on this, the algorithm will be used to optimize the optimal process route of the part and its corresponding minimum total processing cost.

[0133] 2.2 Initial population generation

[0134] The work step priority graph is an intuitive representation of the processing sequence between specific work steps in part processing, and is a form of expression that facilitates human-computer interaction. In order to facilitate the underlying operation of computer programs, the work step priority graph is purposefully converted into mathematical language. This solution converts the work step priority graph into an adjacency matrix for storage and operation, recorded as a 0-1 work step matrix, which can be expressed as:

[0135] D={d ij} (11)

[0136] In the formula, i and j are positive integers, indicating the process step number, d ij It represents the priority relationship between step i and step j as follows:

[0137]

[0138] Obviously, D is a 0-1 matrix. Figure 2 In the figure, the work step matrix represented by the work step priority diagram is:

[0139]

[0140] After the process step matrix is ​​generated, the processing priority relationship between each process step can be determined. Then, the process steps can be topologically sorted according to the process step matrix and the process route can be randomly generated. The specific steps are shown in Algorithm 1:

[0141] Algorithm 1: Process route generation

[0142] Step 1: Input the process step matrix D;

[0143] Step 2: Search the work step matrix column by column, find the full 0 column, and add it to the list list_temp;

[0144] Step 3: Randomly select element i from list_temp and add it to list list_result. At the same time, delete element i from list_temp.

[0145] Step 4: Search for the elements in the i-th row, find all the columns corresponding to 1, and add them to list_temp;

[0146] Step 5: Repeat steps 3 and 4, and perform deduplication every time an element is added to list_result until list_result contains all the work steps;

[0147] Step 6: Output the process route corresponding to list_result. After multiple iterations, the initial population can be generated.

[0148] 2.3 Elite Selection

[0149] After the initial population is generated, the chromosomes are continuously optimized through the crossover operator and mutation operator to improve the population quality and find the best process route. In order to accelerate the convergence of the algorithm and provide better chromosomes for the crossover and mutation stages, the roulette method is used for elite selection to ensure that the individuals to be crossed are high-quality individuals. The specific formula is shown in formula (14):

[0150]

[0151] In the formula, xi represents the i-th chromosome, i is a positive integer. N represents the population size, f(xi ) represents the fitness of the i-th individual, p i represents the probability of the i-th chromosome being selected. Generally, the greater the individual fitness, the better, and the lower the total processing cost, the better. Therefore, this scheme f(x i ) is a positive constant divided by the total processing cost, so as to ensure that the greater the fitness, the greater the probability of being selected. The specific calculation formula is shown in formula (15):

[0152]

[0153] Where C is a constant and is positive.

[0154] The specific steps of elite selection are as follows:

[0155] Algorithm 2: Roulette method to generate parent population

[0156] Step 1: Calculate the fitness value f(x i ), determine the proportion pi and cumulative proportion p of each fitness acc ;

[0157] Step 2: Generate a random number rnd∈(0,1) as the selection threshold;

[0158] Step 3: According to the cumulative proportion of fitness p acc , find the individual whose fitness value is closest to and greater than the random number rnd;

[0159] Step 4: Repeat the above steps until the population size is met.

[0160] After the above algorithm, the parent population contains more high-quality individuals with large fitness values, and the population quality has been significantly improved. In order to avoid premature algorithm maturation, ordinary individuals are still preserved in the parent population, and the subsequent crossover mutation process will have a better algorithm optimization effect.

[0161] 2.4 Multi-fragment crossover

[0162] After elite selection, in order to further optimize the population, crossover operations are often required, which can be divided into self-crossover and mutual crossover. For the process route planning problem, the process of step sequence crossover must meet two basic requirements: first, the offspring must still meet the basic processing constraints after crossover; second, the fitness of the offspring after crossover is not worse than that of the parent. Since process route planning needs to meet rule constraints, self-crossover will disrupt the priority order of the steps, so its crossover is a mutual crossover between two chromosomes, as shown in formula (16-18).

[0163] C f1 =(g f1 ,g f2 ,…,g fi ,…,gfj ,…,g fn ) (16)

[0164] C m1 =(g m1 ,g m2 ,…,g mi ,…,g mj ,…,g mn ) (17)

[0165] C o1 =(g f1 ,g f2 ,…,g mp ,…,g mq ,…,g fn ) (18)

[0166] Among them, C f , C m and C o They represent the paternal chromosome, the maternal chromosome, and the daughter chromosome formed after crossing. g represents the gene in the chromosome, and the corresponding g f , g m Represent the paternal chromosome gene and the maternal chromosome gene respectively. Among them, i, j, n, p, q are all positive integers. It should be noted here that since the crossover process must meet the rule constraints, the paternal and maternal generations do not exchange corresponding positions during the crossover.

[0167] In order to improve the crossover efficiency and accelerate the convergence speed, compared with the traditional single-segment and multi-point crossover, this scheme adopts the multi-segment crossover method, such as Algorithm 3 and Figure 4 As shown. Algorithm 3: Multi-fragment crossover

[0168] Step 1: Select any individual in the parent population as parent 1, and any individual in the mother population as mother 1;

[0169] Step 2: Select N in Parent 1 c crossover segments, each containing L c Number of individuals;

[0170] Step 3: Find the individuals in parent generation 1 that correspond to parent generation 1, and rearrange the fragments in parent generation 1 according to the order of individuals in parent generation 1;

[0171] Step 4: Output the new feasible offspring 1.

[0172] 2.5 Local recombination variation

[0173] Mutation and crossover are both unique links in genetic algorithms. After completing multi-fragment crossover, the algorithm may fall into the dilemma of local optimality. In order to reduce the interference of local optimality, the conventional operation method is mutation. In the process of process optimization, the mutation operation can be defined as: selecting a certain segment in the process route, changing the order of steps and resource combination in the segment through mutation behavior, so as to achieve the purpose of optimizing the process route. Traditional mutation methods are often relatively simple, and only mutate a single type of process sorting or resource combination. This scheme mutates the process sorting and resource combination at the same time, optimizes the step sorting based on the extractable step matrix on the basis of satisfying local constraint rules, and then integrates the manufacturing resources corresponding to the step, accelerates the convergence of the algorithm, and improves the process optimization capability.

[0174] The specific performance is as follows: for the population that has just completed multi-segment crossover, randomly determine whether a chromosome needs to mutate based on the mutation probability. Secondly, for the chromosome that meets the mutation probability, randomly select a segment, extract the feasible process step matrix of the mutated segment based on the initial process step matrix, and form a new segment; then combine and optimize the machine tool, tool and TAD of the new segment, so as to achieve the effect of optimizing the overall process route. The procedure of the local recombination mutation process is as follows, Figure 5 shown.

[0175]

[0176]

[0177]

[0178] 2.6 Algorithm Flow

[0179] The algorithm flow chart of ROGA is as follows: Figure 6 As shown. First, set the basic parameters of ROGA and initialize the gene population. Secondly, calculate the fitness values ​​of individuals in the population one by one, and select high-quality individuals as the parent population according to the roulette method. Then, update and optimize the population through multi-fragment crossover and local recombination mutation until the maximum number of iterations is met. Finally, output the optimal process route and its corresponding resource configuration.

[0180] The invention is further described below in conjunction with specific embodiments.

[0181] [Example]

[0182] Cabin parts generally have many features, complex processing, and variable process routes, which are good models for studying process route planning of parts. This section will further take the process route planning of typical cabin parts as an example to verify the effectiveness and practicality of ROGA. Figure 7As shown in Figure 2, it contains 14 different types of features. The specific feature information is shown in Table 2. The available machine tools and tool information in the workshop are shown in Table 3 and Table 4 respectively.

[0183] Table 2 Part features and available processing solutions

[0184]

[0185]

[0186] Table 3 Available machine tools and their use costs

[0187]

[0188] Table 4 Available tools and their usage costs

[0189]

[0190]

[0191] According to Table 2-4, the effective part feature information, available processing schemes, available machine tool tools and TAD are extracted and integrated to obtain the process step information table, as shown in Table 5. The process step priority diagram of the steering gear cabin body parts is formed by the process step information table and constraint relationships, as shown in Figure 8 According to the work step priority diagram, the work step matrix is ​​formed, as shown in Fig. 9 As shown, it is used for computer bottom-level calculation. The basic program of ROGA is written in Python environment, and the calculation is performed using part feature information, process step information table and process step matrix.

[0192] Table 5 Typical cabin parts process step information table

[0193]

[0194]

[0195] 1. Parameter determination

[0196] In order to obtain the best performance of ROGA in the process planning of typical cabin parts of spacecraft, it is necessary to determine the core parameters of the key links of the algorithm, including the population size N, the maximum number of iterations Gen_max, the crossover probability P c and mutation probability P m , the number of crossover fragments N c and the crossover segment length L c and the length of the variant fragment Lm. Design a single variable comparative test, set multiple sets of values ​​for important parameters, and finally determine the optimal parameters of the indicator set through a large number of experimental comparisons and analyses, such as Fig.10 shown.

[0197] Fig.10 a) In order to determine the optimal population size, a comparative study was conducted with N = 100, N = 200 and N = 300. The results show that when N = 100, the algorithm's optimization performance is insufficient, while the other two cases can obtain the optimal solution quickly. Compared with the case of N = 200, when N = 300, the number of populations increases and the algorithm calculation time increases, so the population size of N = 200 is the most appropriate.

[0198] Fig.10 b) Set up three groups of comparative experiments to study the crossover probability P c and mutation probability P m The results show that in P c =0.8, P m = 0.2, the optimization performance of the algorithm is better than the other two. At the same time, in this case, the convergence speed of the algorithm in the initial stage is faster, and the convergence results in the subsequent stages are better. Therefore, the crossover probability and mutation probability are P c =0.8,P m =0.2 is the most reasonable.

[0199] Fig.10 c) Set up 4 groups of comparative experiments to study the crossover fragment number N in multi-fragment crossover. c and a single segment length L c From the optimization effect, N c =2,L c =5, the optimal value found by the algorithm is slightly smaller than the other three, and the result is the most ideal. c =2,L c = 5, the algorithm converges to the optimal value in about 35 generations, and the convergence speed is faster than the other three. Therefore, the number of crossover segments and the length of a single segment are selected as N c =2,L c =5 is the most reasonable.

[0200] Fig.10 d) Set up 4 groups of comparative experiments to study the variation of the length of the variant fragment L during the local recombination variation process. m The results show that L m =5,8,10, the optimal value that the algorithm can find gradually becomes smaller, and the convergence effect gradually becomes better. m Continue to increase, L m =15, the convergence performance is reduced. In addition, in terms of iteration speed, L m =10 is better than the other three. Therefore, the variant fragment length L m =10 is the most reasonable.

[0201] Based on the above iterative process, the algorithm can generally converge in about 50 generations, so the maximum number of iterations Gen_max=200 is more appropriate.

[0202] 2. Comparative study

[0203] 1) Comparative study with traditional technology

[0204] In order to verify the effectiveness of ROGA, based on Figure 7 The process route of typical cabin parts is planned. The process route and various indicators are obtained through ROGA optimization, and compared with the existing process to draw conclusions. According to the parameter values ​​of the ROGA indicator set determined in the previous article, the initial parameters of the algorithm are set, as shown in Table 6.

[0205] Table 6 ROGA initial parameter settings

[0206]

[0207]

[0208] Table 7 Optimal process route obtained by ROGA

[0209]

[0210] Table 8 Existing process routes

[0211]

[0212] Table 9 Comparison of ROGA calculation results with existing processes

[0213]

[0214] Tables 7-8 show the optimal process route and the existing process route obtained by ROGA respectively, and Table 9 shows the comparison of the results of the two process routes. It should be noted here that due to the complex features of typical cabin parts, numerous process steps, and variable process route combinations, the optimal process route corresponding to the lowest processing cost solved by the algorithm is not unique. The results in Table 9 show that the machine tool usage cost and tool usage cost of the process route obtained by ROGA are the same as those of the existing process, while the number of machine tool replacements and TAD transformations are lower than those of the existing process, resulting in a lower replacement cost for the former. The total processing cost finally optimized by ROGA is 3606, which is 10.87% lower than the total cost of the existing process (4046), which effectively verifies the high efficiency of the ROGA proposed in this scheme.

[0215] 2) Comparative study with different algorithms

[0216] In order to prove the superiority of ROGA, it is compared with traditional genetic algorithm, particle swarm algorithm and immune algorithm. The parameters of ROGA are consistent with those in Table 6, and other algorithms select appropriate parameters. Fig.11 The iterative images of each algorithm are shown, and Table 10 shows the final results of each algorithm.

[0217] Table 10 Comparison results of different algorithms

[0218]

[0219] Table 10 shows that in solving the typical cabin parts process route problem, compared with algorithms such as GA, PSO and IA, the ROGA solution of the present invention has more ideal results, and the average solution, high-frequency solution and solution interval correspond to 3620, 3606 and [3606, 3646] respectively, which are better than the other three algorithms.

[0220] Fig.11 The iterative process shows that the optimization performance of ROGA is better than other algorithms, which is specifically manifested as follows: a) In the initial stage (within 25 generations), the graphs of ROGA and PSO are the steepest, reflecting the fast optimization speed of the algorithm, while GA and IA are relatively inferior.

[0221] b) In the middle stage (25-100 generations), each algorithm is constantly iterating to find the optimal value. Obviously, ROGA converges the fastest, finding the optimal value around 40 generations and stabilizing. GA and PSO converge slightly slower, and their convergence values ​​are worse than ROGA. IA converges even more slowly, and it continues to iterate before the 75th generation, and finds the optimal value after the 75th generation.

[0222] c) In the final stage (after 100 generations), the iterations of each algorithm tend to be stable and the graph no longer decreases.

[0223] Comparative studies of different algorithms show that ROGA exhibits better convergence speed and optimization performance in solving process route optimization problems, saving computing time and effectively reducing processing costs. Compared with other algorithms, ROGA has better optimization capabilities, which effectively proves the superiority and efficiency of ROGA.

[0224] The present invention proposes a re-optimization genetic algorithm (ROGA) with an efficient update mechanism to promote the process optimization process. In order to determine the process priority relationship, a 0-1 process step matrix is ​​obtained according to the process step priority diagram to ensure the priority relationship and generate a feasible process route. Based on this, the process design of ROGA is carried out to optimize the update mechanism and improve the algorithm performance. For the crossover link, on the basis of the traditional single-fragment crossover, the number of crossover fragments is increased and the length of the crossover fragments is controlled, and a multi-fragment crossover behavior is proposed. For the mutation link, process sorting and resource adjustment are performed simultaneously, and the length of the mutation fragment is effectively controlled, and a local recombination mutation behavior is proposed. The optimization of the ROGA update mechanism improves the crossover and mutation efficiency of the algorithm, and accelerates the algorithm iteration efficiency and convergence speed. Finally, ROGA is applied to a typical cabin case study and compared with GA, PSO, and IA. By comparing with the existing process, the minimum total cost is reduced by 10.87%, which effectively proves the efficiency and superiority of the ROGA algorithm of the present invention.

[0225] The above implementation modes are not limitations of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the scope of the technical solution of the present invention also belong to the protection scope of the present invention.

Claims

1. A complex parts process route planning method based on a reoptimized genetic algorithm, characterized by: The following steps are involved: Step 1: Based on the minimum processing cost, process route planning is carried out and a multi-objective optimization model is established; Step 2: Solve the multi-objective optimization model through re-optimization genetic algorithm to solve the optimal process route of complex parts. Specifically, re-optimize the traditional crossover and mutation links based on the process step matrix, propose multi-fragment crossover and local recombination mutation, so as to efficiently update the population and obtain the optimal process route; The multi-segment crossover includes the following steps: Step 1: Select any individual in the parent population as parent I, and any individual in the mother population as mother I; Step 2. Select N in parent I c crossover segments, each containing L c Number of individuals; Step 3: Find the individuals in the parent generation I that correspond to the parent generation I, and rearrange the fragments in the parent generation I according to the order of the individuals in the parent generation I; Step 4: Output new feasible offspring I; Local recombination mutation includes the following steps: 1) In the population that has just completed multi-segment crossover, randomly select a segment containing the mutant gene; 2) extracting the process step matrix of the selected segment based on the initial process step matrix; 3) forming new fragments according to the process step matrix and obtaining a new process step sequence; 4) For new segments, obtain existing manufacturing resources; 5) Reorganize manufacturing resources and optimize resource allocation; 6) Complete the mutation process.

2. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 1 is characterized by: In the step 1, a multi-objective optimization model is established by taking the processing cost as the objective function and considering various costs of various manufacturing resources; The total processing cost of a part includes the cost of using and replacing manufacturing resources; When calculating the replacement cost, it is necessary to compare the machine tool, tool and TAD in pairs and define two comparison operators as follows: On the same machine tool, the parts have the same feed direction TAD after one clamping.

3. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 2 is characterized by: In step 1, the total cost of use UC includes two parts: the total machine tool cost of use and the total tool cost of use, as shown in formula (5): Wherein, step[i] represents step i, i is a positive integer; step[i].macID represents the machine tool number selected for step i; step[i].toolID represents the tool number selected for step i; step[i].tadID represents the TAD number selected for step i; The total replacement cost CC includes three parts: total machine tool replacement cost, total tool replacement cost and total fixture replacement cost, as shown in the following formula: Where: nmc is the total number of machine tool changes in the process route, and the calculation formula is as follows: NTC is the total number of tool changes in the process route, and the calculation formula is as follows: nsc is the number of TAD changes in the process route, and the calculation formula is as follows: The total processing cost TC is the sum of the total use cost and the total replacement cost; TC=UC+CC (10) Wherein, MCI[m] represents the use cost index of machine tool m, and m represents the machine tool number, which is a positive integer; TCI[t] represents the usage cost index of tool t, where t represents the tool number and is a positive integer; MCCI, TCCI and SCCI represent the change cost index of machine tools, tools and fixtures, respectively, that is, the cost required to change the machine tool, tool or fixture between two work steps.

4. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 1 is characterized by: In the step 2, reoptimizing the genetic algorithm comprises the following steps: Step S1, set basic parameters and initialize gene population; Step S2, calculating the fitness values ​​of individuals in the initialized gene population one by one, and selecting high-quality individuals as the parent population according to the roulette method; Step S3, updating and optimizing the population through multi-fragment crossover and local recombination mutation behaviors until the maximum number of iterations is met; Step S4: output the best process route and its corresponding resource allocation.

5. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 4 is characterized in that: In step S1, the work step priority graph is converted into an adjacency matrix for storage and calculation, which is recorded as a 0-1 work step matrix and is expressed as: D={d ij } (11) In the formula, i and j are positive integers, indicating the process step number, d ij It represents the priority relationship between step i and step j as follows: After the work step matrix is ​​generated, the processing priority relationship between each work step can be determined. Then, the work steps can be topologically sorted according to the work step matrix and the process route can be randomly generated.

6. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 5 is characterized by: The specific steps to initialize the gene population are shown in Algorithm 1: Algorithm 1: Process route generation Step 1: Input the process step matrix D; Step 2: Search the work step matrix column by column, find the full 0 column, and add it to the list list_temp; Step 3: Randomly select element i from list_temp and add it to list list_result. At the same time, delete element i from list_temp. Step 4: Search for the elements in the i-th row, find all the columns corresponding to 1, and add them to list_temp; Step 5: Repeat steps 3 and 4, and perform deduplication every time an element is added to list_result until list_result contains all the work steps; Step 6: Output the process route corresponding to list_result. After multiple iterations, the initial population can be generated.

7. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 4 is characterized by: In step S2, the roulette wheel method is used for elite selection to ensure that the individuals to be crossed are high-quality individuals. The specific steps are shown in Algorithm 2: Algorithm 2: Roulette method to generate parent population Step 1: Calculate the fitness value f(x i ), determine the proportion pi and cumulative proportion p of each fitness acc ; Step 2: Generate a random number rnd∈(0,1) as the selection threshold; Step 3: According to the cumulative proportion of fitness p acc , find the individual whose fitness value is closest to and greater than the random number rnd; Step 4: Repeat the above steps until the population size is met; f(x i The specific calculation formula of ) is shown in formula (15): In the formula, C is a constant and a positive number; In the formula, p i represents the probability of the i-th chromosome being selected, x i represents the i-th chromosome, i is a positive integer, N represents the population size, f(x i ) represents the fitness of the ith individual.

8. The complex parts process route planning method based on re-optimized genetic algorithm according to claim 1 is characterized by: Completing the mutation process includes the following processes: completing the process step sequencing of the mutant gene; completing the resource reorganization optimization of the mutant gene; and forming a new process route.

Images