Engine component matching method and system based on multi-gene multi-objective optimization

Abstract

The application relates to an engine part matching method based on multi-gene multi-target optimization. First, an initial multi-gene population is obtained; the initial multi-gene population comprises multiple multi-gene individuals, each multi-gene individual is a coded representation of m sets of engine part matching schemes, the multi-gene individual comprises multiple genes, each gene is a coded representation of an engine, the coded representation of the engine is composed of coded representations of components constituting the engine, and the coded representation of each component is a sub-gene; the coded representations of different components are different, and m represents the number of engines required to be assembled; then, the initial multi-gene population is input into a multi-target optimization model for iterative optimization, and an optimal matching scheme of engine parts is determined according to an optimal individual obtained through the last iteration. The application uses a multi-gene coding strategy to enhance the convergence ability of the multi-target optimization model, and provides a more excellent matching result for the engine part matching method.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of multi-objective optimization matching, and particularly relates to an engine part matching method and system based on multi-gene multi-objective optimization. BACKGROUND

[0002] At present, engine part matching is one of the most critical links in the engine manufacturing process. In the process of continuously manufacturing and updating engines, the difficulty of engine part matching is increasing dramatically. First, the optimization target dimension increases, and it cannot be limited to the "successful assembly" of the engine, and further requires the engine performance to reach the "optimal assembly process" obtained according to experience. Second, the internal structure is more complex, and there is a "one-to-many" and "many-to-many" relationship between engine parts, which greatly increases the complexity of the matching constraints. Finally, the matching scale is expanded, and the modern production process requires engine part matching to no longer take "assembly" as the smallest matching unit, but to take "system" as the smallest matching unit, and dozens of matching units need to be matched at the same time, and the number of parts involved increases by tens of times accordingly. Therefore, the engine part matching has the characteristics of multi-dimensional optimization target, complex matching relationship and large-scale matching.

[0003] In the prior art, engine part matching is a multi-objective optimization problem with complex internal logic and complex constraint conditions. Traditional multi-objective optimization algorithms are usually used to match the engine matching requirements of "assembly" matching with simple matching relationship. However, when the matching relationship is complex or the matching scale is expanded, the performance of the traditional multi-objective optimization algorithm is significantly reduced, and even cannot meet the minimum requirement of "successful matching". Therefore, when the traditional method is applied to the multi-objective optimization matching in the engine part assembly process, it cannot produce excellent matching results. SUMMARY

[0004] The purpose of the present application is to provide an engine part matching method and system based on multi-gene multi-objective optimization with excellent matching results.

[0005] To achieve the above purpose, the present application provides the following solutions:

[0006] An engine part matching method based on multi-gene multi-objective optimization, comprising the following steps:

[0007] Step 11, obtaining an initial multi-gene population; the initial multi-gene population comprises a plurality of multi-gene individuals, each of the multi-gene individuals being an encoded representation of an m-set engine component matching scheme, the multi-gene individual comprising a plurality of genes, each of the genes being an encoded representation of an engine, the encoded representation of the engine being composed of encoded representations of components constituting the engine, each of the encoded representations of the components being a sub-gene; the encoded representations of different components are different, and m represents the number of engines to be assembled;

[0008] Step 12, inputting the initial multi-gene population into a multi-objective optimization model for iterative optimization, and determining an optimal matching scheme of the engine components according to an optimal individual obtained in the last iteration.

[0009] Optionally, the inputting the initial multi-gene population into the multi-objective optimization model for iterative optimization specifically comprises:

[0010] Step 21, calculating an optimization target value of each of the multi-gene individuals in the initial multi-gene population;

[0011] Step 22, non-dominantly sorting the multi-gene individuals according to the optimization target values;

[0012] Step 23, determining a first multi-gene population participating in evolution according to the non-dominant sorting result, the first multi-gene population being composed of elite reservations of each of the multi-gene individuals;

[0013] Step 24, performing gene evolution on the first multi-gene population to obtain a second multi-gene population;

[0014] Step 25, merging the first multi-gene population and the second multi-gene population to obtain a merged multi-gene population;

[0015] Step 26, calculating an optimization target value of each of the multi-gene individuals in the merged multi-gene population;

[0016] Step 27, non-dominantly sorting each of the multi-gene individuals in the merged multi-gene population according to the optimization target values;

[0017] Step 28, judging whether the merged multi-gene population meets an output condition, if yes, proceeding to step 29, otherwise returning to step 23;

[0018] Step 29, performing elite reservation on the merged multi-gene population, and outputting an optimal individual.

[0019] Optionally, the multi-objective optimization model comprises the following constraint conditions: a matching size constraint, a matching deviation constraint, and a matching part repetition constraint.

[0020] Optionally, the measurement formula of the assembly matching consistency MC is as follows:

[0021]

[0022]

[0023] In the formula, C(module i ) is the matching consistency of assembly matching; MC is the assembly matching consistency; module i is the ith type of assembly; moduleNum i is the matching number of the ith type of assembly; realRc i,j is the actual assembly radial clearance of the jth assembly of the ith type; is the average value of the actual assembly radial clearance of the ith type of assembly; moduleKind is the number of assembly types in the assembly scheme; i ∈ (1, 2, …, moduleKind) and j ∈ (1, 2, …, moduleNum).

[0024] Optionally, the calculation formula of the assembly matching success rate is as follows:

[0025]

[0026] In the formula, SR is the matching success rate of the matched assembly in the assembly scheme; n s is the number of qualified assemblies in the matching of the assembly scheme; n t is the total number of assemblies in the assembly scheme.

[0027] Optionally, the measurement formula of the assembly matching optimal error is as follows:

[0028]

[0029]

[0030] In the formula, module i is the ith type of assembly; moduleNum i is the matching number of the ith type of assembly; realRc i,j is the actual assembly radial clearance of the jth assembly of the ith type; bestRc i is the optimal radial clearance of the ith type of assembly; lowbias i is the lower deviation of the ith type of assembly; upbias iis the upper deviation of the i-th kind of component; A is a penalty value when the component assembly exceeds the deviation range; moduleKind is the number of component kinds in the assembly scheme; i∈(1, 2, …, moduleKind); j∈(1, 2, …, moduleNum).

[0031] Optionally, the genetic evolution on the first multi-gene population specifically includes:

[0032] According to the crossover probability, the sub-genes of the multi-gene individual are subjected to crossover evolution to obtain a crossover variation individual;

[0033] According to the mutation probability, the sub-genes in the crossover variation individual are subjected to mutation evolution to obtain the second multi-gene population.

[0034] Optionally, the mutation probability is determined by a mutation probability adaptive strategy, and the mutation probability adaptive strategy is that the mutation probability of the sub-gene is determined according to the proportion of the number of parts represented by the sub-gene in the total number of parts in the part library.

[0035] Optionally, the output condition includes whether the maximum iteration number is reached.

[0036] The application further provides an engine part matching system based on multi-gene multi-objective optimization, which is characterized by comprising a memory and a processor, the memory stores a computer program, and the processor realizes a method for matching engine parts based on multi-gene multi-objective optimization when executing the computer program.

[0037] According to the specific embodiments provided in the application, the following technical effects are disclosed: the application provides an engine part matching method and system based on multi-gene multi-objective optimization, an initial multi-gene population is obtained; the initial multi-gene population includes a plurality of multi-gene individuals, each multi-gene individual is an encoded representation of m sets of engine part matching schemes, the multi-gene individual includes a plurality of genes, each gene is an encoded representation of an engine, the encoded representation of the engine is composed of encoded representations of components constituting the engine, and each encoded representation of the component is a sub-gene; the encoded representations of different components are different, and m represents the number of engines required to be assembled; the initial multi-gene population is input into a multi-objective optimization model for iterative optimization, and the optimal matching scheme of the engine parts is determined according to the optimal individual obtained in the last iteration, wherein the optimization objectives of the multi-objective optimization model include component matching success rate, component matching optimal error and component matching consistency. The application corresponds the number of engines required to be assembled to the multi-gene individual, and corresponds the components required for assembling the engine to each gene in the multi-gene individual, enhances the convergence ability of the multi-objective optimization model by using the multi-gene coding strategy, searches for the global optimal solution, so that a more suitable part matching mode is found in the actual engine assembly process application, and the multi-objective optimization requirements in the engine part matching process are better met. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0039] Figure 1 The algorithm flowchart of the classical Pareto-based multi-objective evolutionary algorithm provided in the application;

[0040] Figure 2 The engine part matching method flowchart provided in the application;

[0041] Figure 3 The engine part matching model diagram based on the multi-objective optimization model provided in the application;

[0042] Figure 4 The multi-objective optimization model evolution process flowchart provided in the application;

[0043] Figure 5 The engine part matching evolution process diagram of a certain type of two-shaft engine provided in the application; wherein Figure 5 (a), Figure 5 (d),Figure 5 (g) are schematic diagrams of three optimization objectives for low-voltage systems: component matching success rate, optimal component matching error, and component matching consistency. Figure 5 (b) Figure 5 (e) Figure 5 (h) is a schematic diagram of three optimization objectives for high-voltage systems: component matching success rate, optimal component matching error, and component matching consistency. Figure 5 (c) Figure 5 (f) Figure 5 (i) is a schematic diagram of three optimization objectives for the entire engine: component assembly success rate, component assembly optimal error, and component assembly consistency.

[0044] Figure 6 This is a schematic diagram of the non-dominated sorting process provided in this application. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] like Figure 1 As shown, a classic Pareto-based multi-objective evolutionary algorithm (MOEA) process is presented. The main processes of MOEA can be classified into: multi-objective process and evolutionary process.

[0047] The classic Pareto-based multi-objective evolutionary algorithm (MOEA) process can be summarized as follows:

[0048] 1) Generate an initial population P. The generation of the initial population is directly related to the user-defined evolutionary parameters, including population size, encoding method, and individual length.

[0049] 2) The initial population P is evolved to obtain a new population R. The method of population evolution determines the optimization ability of the algorithm. Classic evolution methods include: genetic algorithm, particle swarm optimization algorithm and differential evolution algorithm.

[0050] 3) Construct the non-dominated set NDSet of P∪R, and select the next generation of population based on the non-dominated relation;

[0051] 4) Adjust the size of NDSet to meet the distribution requirements. The selected next generation population needs to constrain the number of individuals to ensure that the number of individuals in each generation is consistent.

[0052] 5) judge whether the iteration evolution termination condition is met, (the judgment condition is generally set as: whether the current iteration number is greater than the required iteration number), yes, go to 2), otherwise go to 6).

[0053] 6) output the result and end the iteration evolution.

[0054] However, when the multi-objective evolutionary algorithm is used to select and match the matching relationship, the evolutionary gene is too long, and when the matching relationship is complex or the selection and matching scale is expanded, the performance of the traditional multi-objective optimization algorithm is significantly reduced, and even the minimum requirement for successful selection and matching cannot be met. Therefore, when the traditional method is applied to the multi-objective optimization and selection and matching of engine parts in the assembly process, an excellent selection and matching result cannot be obtained.

[0055] The purpose of the present application is to provide a multi-gene multi-objective optimization based engine part selection and matching method and system with excellent selection and matching results.

[0056] In order to make the above-mentioned purposes, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0057] The present application provides a multi-gene multi-objective optimization based engine part selection and matching method, as shown in Figure 2 The method comprises the following steps:

[0058] Step 11: obtaining an initial multi-gene population; the initial multi-gene population comprises a plurality of multi-gene individuals, each of the multi-gene individuals is a coded representation of m sets of engine part selection and matching schemes, the multi-gene individual comprises a plurality of genes, each of the genes is a coded representation of an engine, the coded representation of the engine is composed of coded representations of components constituting the engine, and each of the coded representations of the components is a sub-gene; the coded representations of different components are different, and m represents the number of engines required to be assembled.

[0059] Step 12: inputting the initial multi-gene population into a multi-objective optimization model for iterative optimization, and determining the optimal selection and matching scheme of the engine parts according to the optimal individual obtained in the last iteration, wherein the optimization objectives of the multi-objective optimization model include component matching success rate, component matching optimal error and component matching consistency.

[0060] In some embodiments, the obtained initial multi-gene population is composed of a plurality of coded multi-gene individuals, each of which contains a plurality of small genes, and each of the small genes contains a plurality of gene individuals. The method for obtaining the initial multi-gene population can be as follows:

[0061] 1) initializing selection and matching parameters; the selection and matching parameters include: selection and matching component data, selection and matching engine data.

[0062] The optional component data includes the number of optional parts, the number of optional components, the type of optional parts, and the type of optional components.

[0063] The optional engine data includes the model of the optional engine, the system of the optional engine, and the number of the optional engine.

[0064] 2) Initialize the algorithm parameters, including the number of populations, the number of genes, the crossover probability, the mutation probability, and the maximum number of iterations.

[0065] The number of populations is crucial to the evolution of the algorithm, and the initialized population is randomly generated by the constraint conditions. The size of the population number ensures the distribution of the population.

[0066] The definition of the number of genes depends on the number of parts involved in the selection. Each part corresponds to a gene.

[0067] The definition of the crossover probability, the crossover operation is one of the key factors for the optimization of the algorithm based on genetic algorithm. The crossover operation will evolve in pairs within the population.

[0068] The definition of the mutation probability range, the mutation probability range should not be too large, otherwise the mutation degree will be too large and the selection result will not be successful.

[0069] The definition of the maximum number of iterations determines the evolution degree of the model.

[0070] 3) Initialize the single-gene population, which is composed of the number of all parts used for selection as the column number of the population matrix, and the number of populations as the row number of the population matrix.

[0071] Specifically, the single-gene population of each set of engine systems is initialized, and each individual in the population represents all the components required for assembling each set of engine systems. At this time, each individual has only one gene, and all the required component types and different systems have not been divided. Each single-gene individual is divided according to the assembly characteristics of each set of engine systems, i.e., the required component types. First, the gene of each system is distinguished, and then the system gene is arranged to distinguish the different types of components required for assembling each set of systems.

[0072] 4) Divide the single-gene individuals in the initialized single-gene population into multiple genes (multiple gene coding) to obtain multiple-gene individuals, and multiple multiple-gene individuals form the initial multiple-gene population.

[0073] Specifically, multiple gene coding further subdivides the gene corresponding to each individual into smaller genes to form multiple-gene individuals, and multiple multiple-gene individuals form a multiple-gene population.

[0074] For example, when selecting the cooling system of the engine, the components of the cooling system that need to be selected include the water temperature gauge, the water temperature sensor, the radiator, the fan, the water pump, and the water jacket thermostat. The cooling system of the engine is regarded as a multi-gene individual, and the water temperature gauge, the water temperature sensor, the radiator, the fan, the water pump, and the water jacket thermostat are regarded as small gene strips of the multi-gene individual. Taking the water temperature sensor as an example, the water temperature sensor is composed of a thermistor, an ignition switch, a bimetallic strip, and an indicator rod, i.e., the thermistor, the ignition switch, the bimetallic strip, and the indicator rod are regarded as the smallest individual of the multi-gene population.

[0075] In some embodiments, before the initial multi-gene population is input into the multi-objective optimization model for iterative optimization, the constraint conditions of the multi-objective optimization model are determined, which can specifically include:

[0076] The selection size constraint, the selection deviation constraint, and the selection part repetition constraint.

[0077] The selection size constraint is a part theoretical size constraint: due to the characteristics of irregularity, multiple part numbers, and multiple sizes of the selected parts, the theoretical size constraint of the selected parts must be considered during selection.

[0078] The selection deviation constraint is a radial gap deviation constraint: for each group of components, the requirement for successful selection is that the actual radial gap meets the upper and lower deviation constraints of the component radial gap.

[0079] The selection part repetition constraint refers to the problem of whether the selected parts are repeatedly used during selection.

[0080] In some embodiments, before the initial multi-gene population is input into the multi-objective optimization model for iterative optimization, the optimization objectives of the multi-objective optimization model are determined, which include the component matching success rate, the component matching optimal error, and the component matching consistency.

[0081] Specifically, the component matching success rate is calculated in units of components, representing the ratio of the number of successfully selected components to the total number of planned selected components in one selection.

[0082]

[0083] In the formula: SR is the component matching success rate in this assembly scheme; n s is the number of qualified components in the selection of this assembly scheme; n t is the total number of components in the selection of this assembly scheme.

[0084] The component matching optimal error refers to the error degree between the actual value selected by the defined component for each assembly and the "optimal radial gap" of the component. The definition function Q(modulei ) to evaluate the fit degree of the selected components of a certain type, and the MBE is used to measure the optimal error of the fit of the selected components.

[0085]

[0086]

[0087] In the formula: module i is the i-th type of component; moduleNum i is the number of selected components of the i-th type; realRc i,j is the actual assembly radial clearance of the j-th component of the i-th type; bestRc i is the optimal radial clearance of the i-th type of component; lowbias i is the lower bias of the i-th type of component; upbias i is the upper bias of the i-th type of component; A is the penalty value when the assembly exceeds the deviation range; moduleKind is the number of component types in the assembly scheme, i∈(1, 2, …, moduleKind), j∈(1, 2, …, moduleNum).

[0088] For example, when selecting the cooling system of the engine, the component types that need to be selected for the cooling system include the water temperature gauge type, the water temperature sensor type, the radiator type, the fan type, the water pump type, and the water jacket thermostat type, so module i represents one of the above component types, for example, the i-th type is the water temperature sensor type, and the water temperature sensor type includes multiple types of water temperature sensors, such as water temperature sensor a, water temperature sensor b, and water temperature sensor c, and the j-th component is the water temperature sensor a, so realRc i,j is the water temperature sensor a component of the water temperature sensor type.

[0089] The component fit consistency refers to the fact that the radial clearances of multiple selected components of the same type need to be kept as small as possible within the deviation constraint. For the case of multiple selections of the same component in the engine, the “optimal assembly process” requires that the radial clearances of multiple selected components of the same type need to be kept as small as possible within the deviation constraint. The function C(module i ) is used to evaluate the fit consistency of the selected components of a certain type, and MC is used to measure the fit consistency of the selected components.

[0090]

[0091]

[0092] In the formula: module iis the i-th component; moduleNum i is the i-th component; realRc i,j is the i-th component; realRc is the i-th component; realRc

[0093] In some embodiments, the initial multi-gene population is input into a multi-objective optimization model for iterative optimization, as shown in Figure 3 Specifically, the step 21 can include the following steps.

[0094] Step 21, calculate the optimization target value of each multi-gene individual in the initial multi-gene population.

[0095] Step 22, non-dominated sorting of the multi-gene individuals according to the optimization target value.

[0096] Step 23, according to the non-dominated sorting result, determine the first multi-gene population participating in evolution, which is composed of each multi-gene individual for elite preservation.

[0097] Step 24, genetic evolution of the first multi-gene population to obtain the second multi-gene population.

[0098] Step 25, merging the first multi-gene population and the second multi-gene population to obtain the merged multi-gene population.

[0099] Step 26, calculate the optimization target value of each multi-gene individual in the merged multi-gene population.

[0100] Step 27, non-dominated sorting of each multi-gene individual in the merged multi-gene population according to the optimization target value.

[0101] Step 28, determine whether the merged multi-gene population meets the output condition, if yes, proceed to step 29, otherwise return to step 3.

[0102] Step 29, elite preservation of the merged multi-gene population, output the optimal individual.

[0103] Specifically, in step 21, the purpose of calculating the optimization target value of each multi-gene individual is to perform non-dominated sorting based on the Pareto concept in step 22.

[0104] Specifically, the non-dominated sorting in step 22 can include the following steps. Figure 6As shown in the multi-objective optimization problem, since there may be conflicts between each objective, the optimal solution is not necessarily only one, at this time it is necessary to define the "Pareto optimal solution", also known as "non-dominated solution". Let's solve a multi-objective minimization problem, with k objective functions denoted as f i (x), where i∈{1,2,…k}, j∈{1,2,…k}, i≠j. If individuals x1 and x2 have f i (x1)<f j (x2) for any objective function, then individual x1 is non-dominant (non-dominated), and individual x2 is dominated (dominated); if f i (x1)≤f j (x2) for any objective function, and at least one objective function makes f j (x1)<f j (x2) is true, then x1 is non-weakly dominated (non-dominated), and x2 is weakly dominated (dominated); if there is an objective function that makes f i (x1)≤f j (x2) is true, and there is an objective function that satisfies f i (x1)>f j (x2), then individuals x1 and x2 are not dominated.

[0105] Therefore, non-dominated sorting is to classify non-dominated individuals according to non-dominated strength, and the stronger the non-dominated degree, the lower the level. The unclassified individuals in the individual set are compared with the classified individuals, and the classification of the individuals is obtained in turn and added to the classification set until all individuals in the set are classified.

[0106] Specifically, the method of elite reservation in step 23 is tournament selection, which selects multiple gene individuals participating in evolution from the initial multi-gene population.

[0107] Specifically, the gene evolution in step 24 can include cross evolution and mutation evolution. Assuming that m sets of components are selected to assemble an engine or a system, the evolution processing method of MG-MOEA is as shown in Figure 4 Specifically, it can be as follows:

[0108] Firstly, each individual in the original "system" single-gene population represents all components required for assembling m sets of "system", the individuals of this single gene are divided according to the "system" assembly characteristics, and each set of system genes is distinguished, and then the system genes are arranged to distinguish different component groups required for m sets of system assembly; then, the original m sets of "system" single-gene population is decomposed and mapped into m sets of "system" multi-gene population with part types as multi-genes; then, each multi-gene corresponding to a part type is given a corresponding evolution operator, including selection, crossover and mutation operations; finally, the final selection result is iterated through evolution.

[0109] For the "i-th m set of system" in Figure 4 , here taking 10 sets of engines assembled in the system as an example, X i，1 to X i，mn in the figure represent all components required for assembling the 10 sets of engines, for example, there are 100 components in total, among which X i，1 to X i，n represent the components used for assembling the first set of engines, such as X i，1 in the figure represents a cover, X i，2 represents a fuselage, "..." represents other components, and X i，n represents a rotor, so these small cells constitute a set of engine system. Since the component types used for assembling the 10 sets of engines are the same, the individuals of the single-gene population are divided into multi-genes, and the multi-gene individuals after division are as shown in the "m sets of "system" multi-gene population" in Figure 4 , that is, each column represents a component of the same type, and each row represents all components constituting a set of engine system.

[0110] Specifically, when performing crossover evolution operation, crossover operation is performed according to the crossover probability defined in the initialization algorithm, partial match crossover PMX for permutation encoding is adopted, two crossover points are randomly selected to determine the crossover region, and the genes in the crossover region are exchanged. In order to repair the chromosome, PMX establishes a matching rule for each gene in the crossover region, and then applies the matching rule to the repeated genes outside the crossover region to eliminate invalid genes.

[0111] Specifically, when performing mutation evolution operation, first, the mutation probability of the multi-gene corresponding to different types of parts is adaptively mapped according to the mutation probability range defined in the algorithm parameters. Then, each gene decides whether to mutate the elements on the gene according to the mutation probability, the elements will be replaced with other values in the range, and if the replaced value conflicts with other elements on the gene, the elements conflicting with it will be matched and replaced.

[0112] The mutation probability is determined by a mutation probability adaptive strategy. The mutation probability adaptive strategy is to determine the mutation probability of the subgene according to the proportion of the number of parts represented by the subgene in the total number of parts in the parts library. The adaptive strategy of the mutation probability mainly draws on the idea of data normalization. Data normalization is a basic work of data mining. Different attribute variables often have different value ranges, and the difference between values may be large. If not handled, it may affect the result of data calculation. In order to eliminate the difference between indicators due to the value range, normalization processing is needed. The data is scaled according to the proportion so as to fall into a specific area, which is convenient for comprehensive analysis.

[0113] In practical application, an adaptive mutation probability is set for each part gene. Since the mutation probability determines the probability of mutation of the gene in the evolution process, and mutation can make the gene obtain combinations that do not appear in the original gene, in combination with the actual selection process, gene mutation corresponds to part replacement in the production process, and the mutation probability determines the probability of part replacement.

[0114] For example, when selecting a product, three parts A, B and C are needed, and the part library has numA A parts, numB B parts and numC C parts, and numA ≠ numB ≠ numC. When selecting a product, if the product is unqualified, A or B or C parts are considered to be replaced, at this time, part replacement probability is needed to determine which part to replace, that is, the mutation probability in the algorithm. For probability, the inherent range is 0-1, and in combination with the actual selection of parts, it means not replacing the part (mutation probability is 0) or necessarily replacing the part (mutation probability is 1). In combination with the production process and the algorithm evolution process, the mutation probability cannot be 0 or 1, that is, the selection cannot always replace a certain part or always replace a certain part, which results in that the selection and the algorithm evolution process fall into a local optimal solution or fall into randomness. At this time, we need to artificially set a reasonable range [minRate, maxRate] for the mutation probability according to experience.

[0115] The mutation probability is realized by using the maximum and minimum value normalization method to realize the self-adaptation of the mutation probability. It is assumed that there are N kinds of parts in the parts library, and the number of various parts is cmpNum = [num1, num2…num i …num N ] The part quantity of cmpNum is converted into part proportion cmpProp by the following formula:

[0116]

[0117] The obtained cmpProp=[prop1, prop2…prop i …prop N ] represents the proportion of the number of various parts in the parts library. The minimum value in cmpProp is minProp, and the maximum value is maxProp. The cmpProp part proportion is converted into cmpRate variation probability by the following formula:

[0118]

[0119] The obtained cmpRate=[rate1, rate2…rate i …rate N ] represents the variation probability of each part corresponding to the gene.

[0120] When selecting the above product, taking the variation of part A as an example, it is assumed that part A has n, then all the numbers of part A are represented as cmpNo A ={a1, a2, a3, a4, a5…, a n}. If the gene corresponding to part A at this time is: gene A =(a2, a1, a3), then during the variation operation, a1 needs to be varied to a i , wherein a i ∈cmpNo A . Since a i represents the different numbers of the same A part, a i corresponds to the actual existing A type part one by one, so the same number cannot appear in the same gene, which can be understood as the same part cannot be assembled in two places in actual selection. For the gene corresponding to the varied part A: gene A =(a2, a i , a3), wherein

[0121] Specifically, in steps 25-27, since the evolved second multi-gene population may produce solutions that are better or not better, the first multi-gene population and the second multi-gene population are combined, and the multi-gene individuals in the combined multi-gene population are calculated for the optimization target value, and the optimal solution in the combined multi-gene population is obtained through non-dominated sorting.

[0122] Specifically, the output condition of the combined multi-gene population in step 28 can be set as whether the iteration number of the current combined multi-gene population reaches the maximum iteration number.

[0123] In some embodiments, to verify that the MG-MOEA model is superior to the classic MOEA algorithm, a certain type of two-shaft engine is taken as an example, and the matching task is divided into "low-pressure system matching", "high-pressure system matching" and "whole engine matching". Among them, assembling a "low-pressure system" requires 11 kinds of 12 parts, and 11 groups of 12 groups of components need to be defined; assembling a "high-pressure system" requires 18 kinds of 19 parts, and 18 groups of 18 groups of components need to be defined; assembling a "whole engine" requires 26 kinds of 33 parts, and 28 groups of 33 groups of components need to be defined.

[0124] The experiment will use the real data set collected in the part library when planning to produce 13 sets of a certain type of two-shaft whole engine as data support, and use the MG-MOEA model to compare another 7 kinds of classic MOEA models under different system and matching quantity conditions. The experimental settings are shown in Table 1.

[0125] Table 1 Experimental settings

[0126]

[0127] The experimental optimization results are shown in Table 2, and the MOEA evolution process for matching 13 sets of whole engines is shown in Figure 5 For convenience of description, the following will refer to the component matching success rate as SR, the component matching optimal error as MBE, and the component matching consistency as MC.

[0128] Table 2 Experimental optimization results

[0129]

[0130]

[0131]

[0132] Among them, Table 2 is the experimental optimization results under different conditions. In the table, not only the values of the three optimization targets under different conditions are recorded, but also the time required for matching.

[0133] From the data in Table 2, it can be seen that MG-MOEA is not worse than any other compared MOEA algorithm in all indicators: "component matching success rate", "component matching optimal error", "component matching consistency" and "matching time" in matching 3, 5, 7, 9, 11 and 13 sets of "low-pressure system", "high-pressure system" and "whole engine". From the component matching success rate, MG-MOEA can always achieve full matching success, while the matching success rate of the compared algorithm gradually decreases with the increase of the matching task quantity and the complexity of the matching system.

[0134] From the perspective of optimal component matching error, the optimization results of MG-MOEA are always better than other MOEA algorithms, and the improvement is always above 30%. As the number of components selected and the system complexity increase, the improvement gradually increases to about 50%.

[0135] In terms of component fit consistency, the optimization results of MG-MOEA are generally no worse than those of other MOEA algorithms. Component fit consistency measures the distribution of the selected component values. Combined with the analysis of the component fit optimal error, it can be seen that the optimization results of MG-MOEA are far superior to those of other MOEA algorithms in both aspects.

[0136] In terms of mating time, MG-MOEA greatly simplifies the complex task of judging constraints and representing component matching relationships through multi-gene encoding and adaptive mutation probability strategies, resulting in a significant reduction in mating time compared to other MOEA algorithms.

[0137] like Figure 5 As shown, Figure 5 This paper describes the optimization iterative process for a certain type of dual-shaft engine under three optimization objectives: "component fit success rate," "optimal component fit error," and "component fit consistency," with 13 sets of "low-pressure system," "high-pressure system," and "complete engine" configurations. The complexity of the three configuration scenarios increases progressively, involving more types of parts, a greater number of parts, and a corresponding increase in the number of components. Regarding the component fit success rate (SR), all algorithms compared with MG-MOEA converge to a non-zero value, and their initial SR values ​​are lower, while MG-MOEA's initial SR value is higher. This indicates that under complex conditions, the algorithm requires significant computational power to handle constraints, compared to algorithms such as MOEAD / DE, NSGA2, NSGA2 / DE, NSGA3, NSGA3 / DE, RVEA, and RVEA / RES. Figure 5 The algorithms labeled in the small figures (from left to right) are MOEAD / DE, NSGA2, NSGA2 / DE, NSGA3, NSGA3 / DE, RVEA, RVEA / RES, and MG-MOEA. MG-MOEA simplifies a large amount of homogeneous work through a multi-gene approach. For the component fit with the most error (MBE), MG-MOEA's final convergence value is lower than the other comparative methods. This shows that by using different mutation methods on multiple genes, excellent genes can be well preserved without gene fluctuations caused by excessively high or low mutation probabilities. For component fit consistency (MC), the evolutionary process of MG-MOEA eventually shows a trend of convergence, but the final MC value is still significantly improved compared to other algorithms.

[0138] The application also provides an engine part matching system based on multi-gene multi-objective optimization, comprising a memory and a processor, the memory storing a computer program, and the processor realizing the method of engine part matching based on multi-gene multi-objective optimization when executing the computer program.

[0139] To sum up, the application has the following advantages: the application uses the multi-gene coding and mutation probability adaptive strategy to improve the classical MOEA model, and obtains the engine part matching model based on MG-MOEA. The model uses the multi-gene coding strategy to enhance the convergence ability of the model, and uses the mutation probability adaptive strategy to enhance the ability of the model to search for the global optimal solution. The experimental results show that the comprehensive ability of the MG-MOEA model is stronger than that of the classical MOEA model, the MG-MOEA model can find a more suitable part matching mode in the actual engine assembly process application, and can better meet the multi-objective optimization requirements in the engine part matching process.

[0140] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0141] The principles and implementation manners of the application are described by using specific examples in the present application, and the above embodiment description is only used to help understand the method of the application and its core idea; meanwhile, for the general technical personnel in the art, the specific implementation manner and application range of the application will be changed according to the idea of the application. To sum up, the content of the specification should not be understood as the limitation of the application.

Claims

1. A method for selecting engine components based on multi-gene, multi-objective optimization, characterized in that, Includes the following steps: Step 11, obtain the initial multi-gene population; the initial multi-gene population includes multiple multi-gene individuals, each multi-gene individual is a coded representation of m sets of engine component selection schemes, each multi-gene individual includes multiple genes, each gene is a coded representation of an engine, the coded representation of the engine is composed of the coded representations of the components constituting the engine, and the coded representation of each component is a sub-gene; different components have different coded representations, and m represents the number of engines to be assembled; Step 12: Input the initial multi-gene population into the multi-objective optimization model for iterative optimization, and determine the optimal matching scheme of the engine components based on the optimal individual obtained in the last iteration. The optimization objectives of the multi-objective optimization model include component matching success rate, component matching optimal error, and component matching consistency. The formula for measuring component fit consistency (MC) is as follows: In the formula: C(module) i ) represents the consistency of component selection and matching; MC represents the consistency of component matching; module i For the i-th type of component; moduleNum i The number of components selected for the i-th type; realRc i,j The actual assembly radial clearance of the j-th component of the i-th type; Let i be the average radial clearance of the actual assembly of the i-th type of component; moduleKind is the number of assembly component classes in this assembly scheme; i∈(1,2,...,moduleKind) and j∈(1,2,...,moduleNum); The formula for calculating the success rate of component integration is as follows: Where: SR represents the success rate of component selection for this assembly scheme; n s Therefore, the number of qualified components selected for this assembly scheme; n t Therefore, the total number of components in this assembly scheme; The formula for measuring the optimal error of the component fit is as follows: In the formula: bestRc i The optimal radial clearance for the i-th type of component; lowbias i The upbias is the lower bias of the i-th type of component. i Let A be the upper deviation of the i-th type of component; A is the penalty value when the component assembly exceeds the deviation range.

2. The engine component selection method based on multi-gene multi-objective optimization according to claim 1, characterized in that, The step of inputting the initial multi-gene population into a multi-objective optimization model for iterative optimization specifically includes: Step 21: Calculate the optimization target value for each of the multi-gene individuals in the initial multi-gene population; Step 22: Perform non-dominated sorting on the multi-gene individuals according to the optimization target value; Step 23: Based on the non-dominated sorting results, determine the first multi-gene population participating in evolution. The first multi-gene population is composed of elite individuals of each multi-gene population. Step 24: Perform gene evolution on the first multi-gene population to obtain a second multi-gene population; Step 25: Merge the first multigene population and the second multigene population to obtain a merged multigene population; Step 26: Calculate the optimization target value for each multi-gene individual in the merged multi-gene population; Step 27: Perform non-dominated sorting of each multi-gene individual in the merged multi-gene population according to the optimization target value; Step 28: Determine whether the merged multi-gene population meets the output conditions. If it does, proceed to step 29; otherwise, return to step 23. Step 29: Elite retention is performed on the merged multi-gene population to output the optimal individual.

3. The engine component selection method based on multi-gene multi-objective optimization according to claim 1, characterized in that, The multi-objective optimization model includes the following constraints: optional size constraints, optional deviation constraints, and optional part repetition constraints.

4. The engine component selection method based on multi-gene multi-objective optimization according to claim 2, characterized in that, The genetic evolution of the first multi-gene population specifically includes: Based on the crossover probability, the subgenes of the multi-gene individuals are subjected to crossover evolution to obtain crossover variant individuals; Based on the mutation probability, the sub-genes in the cross-mutated individuals are subjected to mutation evolution to obtain the second multi-gene population.

5. The engine component selection method based on multi-gene multi-objective optimization according to claim 4, characterized in that, The mutation probability is determined using an adaptive mutation probability strategy, which is to determine the mutation probability of the sub-gene based on the proportion of the number of parts represented by the sub-gene in the total number of parts in the parts library.

6. The engine component selection method based on multi-gene multi-objective optimization according to claim 2, characterized in that, The output conditions include whether the maximum number of iterations has been reached.

7. An engine component selection system based on multi-gene, multi-objective optimization, characterized in that, include: A memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the method according to any one of claims 1-6.

Images