CN113987718B - Product change scheme design method based on multi-objective particle swarm optimization - Google Patents

Product change scheme design method based on multi-objective particle swarm optimization Download PDF

Info

Publication number
CN113987718B
CN113987718B CN202111361319.3A CN202111361319A CN113987718B CN 113987718 B CN113987718 B CN 113987718B CN 202111361319 A CN202111361319 A CN 202111361319A CN 113987718 B CN113987718 B CN 113987718B
Authority
CN
China
Prior art keywords
change
node
product
service performance
influence
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111361319.3A
Other languages
Chinese (zh)
Other versions
CN113987718A (en
Inventor
张勇
郑瑞钊
孙晓燕
王法广
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Mining and Technology CUMT
Original Assignee
China University of Mining and Technology CUMT
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China University of Mining and Technology CUMT filed Critical China University of Mining and Technology CUMT
Priority to CN202111361319.3A priority Critical patent/CN113987718B/en
Publication of CN113987718A publication Critical patent/CN113987718A/en
Application granted granted Critical
Publication of CN113987718B publication Critical patent/CN113987718B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computer Hardware Design (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a product change scheme design method based on multi-objective particle swarm optimization, aiming at any given product to be changed, firstly establishing a complex product multi-network model introducing service performance; secondly, generating a multi-target solving model containing 3 evaluation indexes of service performance influence degree, change cost and change time by decoupling the influence of the strongly associated node on the service performance in change effect propagation; then, solving the multi-target model by using an integer multi-target particle swarm optimization algorithm to obtain a group of optimal change schemes; and finally, obtaining an optimal change scheme meeting the user requirements by using a fuzzy decision method. The obtained change scheme has stronger operability, can simultaneously cover key nodes on a plurality of propagation paths, and better meets the actual design requirement. By setting appropriate constraint conditions, the feasibility of the change scheme can be ensured to the maximum extent.

Description

Product change scheme design method based on multi-objective particle swarm optimization
Technical Field
The invention belongs to the field of product design, and particularly relates to a product change scheme design method based on multi-objective particle swarm optimization.
Background
The design change is an inevitable link in the product research and development process, and an unreasonable design change scheme not only can improve the research and development cost of the product, but also can obviously influence the service performance of the product. In the real production, the product parts are often redesigned or selected under the influence of subjective and objective reasons such as product iteration upgrade, supply chain breakage and blockage, law and regulation modification and the like. More importantly, changes to one critical part propagate and affect other associated parts in the product. For the problem of Complex Product Design (CPD) with a large number of parts and close connection, if the product design change process is not analyzed and controlled effectively, the change will have a great influence on the service performance, cost, construction period, etc. of the product, and even cause the change failure. Therefore, it is of great significance to study product design changes.
The research on the complex product change problem firstly needs to establish an accurate product part correlation model. In recent years, the rise of complex network technology provides a new idea for product change research. By taking the product parts as nodes and taking the relation between the parts as edges, a complex product network can be established, and further the problem of product design change can be analyzed and solved by means of complex network knowledge. Compared with other models, the complex product network model can better reflect the coupling strength between different parts. In the prior art, a product structure matrix (DSM) is mainly adopted to establish a complex product network model, and the dependency relationship among product parts is quantitatively expressed. Yu et al [1] take customer satisfaction as an optimization target, and propose a complex product change design optimization model responding to user demand changes. Jiang et al [2] established a function-behavior-structure network model suitable for product design, and given a determination strategy for changing propagation paths by predicting the change of functions in the design process. Ma et al [3] propose a mathematical programming method for quantitatively predicting change propagation influence, and establish a change analysis model (DCAM) based on a design attribute network on the basis of a change propagation prediction result. Zheng et al [4] comprehensively evaluate the influence of product change on the whole product through four stages, and provide a staged product change influence evaluation model, thereby improving the model evaluation efficiency. The achievements lay a good foundation for effectively analyzing a product change effect propagation mechanism, but the achievements do not consider the service performance index directly related to user experience when establishing a complex product network model, and the established network model hardly meets the actual requirement.
After the product part association model is established, appropriate evaluation indexes are determined to evaluate the node importance or the change scheme. To evaluate the importance of part nodes, Ma, etc. [5] the change propagation was estimated using indices such as propagation probability, node degree, long-chain connection, and design margin, and the degree of influence of the change propagation was quantitatively analyzed. In order to measure the influence of the network change degree on a product system, Yu and other [6] define two indexes of network change scale, extra network change cost and the like. In order to evaluate the change influence more accurately, Yang et al [7] propose a product engineering change influence evaluation index system based on a multi-level complex network, and provide an engineering change influence comprehensive evaluation strategy based on a combined weighted three-parameter gray correlation model. However, due to the limitation of the complex product network model itself, the indexes cannot intuitively reflect the influence of the product change on the service performance. How to reveal and quantify the influence of part change on the service performance of a product is short of corresponding research results so far.
In order to effectively control the influence of the change effect propagation on the product performance, an optimal product change propagation path or scheme needs to be searched. In recent years, researchers have begun to use evolutionary optimization techniques with global search capabilities to solve product design change optimization problems. Ma et al [5] use ant colony algorithm to minimize cumulative change propagation strength and find the optimal change propagation path. The above work can find out an ideal propagation path with the minimum change risk for a decision maker, but because the influence of the change on the associated nodes except the optimal propagation path is neglected, the obtained path is difficult to truly reflect the real change situation. In addition, in practical problems, the decision of design change often needs to consider multiple indexes such as change time, cost and the influence thereof on the product performance. Currently, the research work of simultaneously considering a plurality of indexes is less. Ren et al [8] simultaneously considers indexes such as change cost, strength and path length, and proposes a non-dominated sorting genetic algorithm II (NSGA-II) based on a parallel search strategy, which is used for searching a plurality of optimal change paths of a problem. On the one hand, the work does not consider the impact of the change on the service performance when finding the optimal change path; on the other hand, because only the traditional NSGA-II algorithm is used for solving the optimal path, the provided algorithm still has the defects of low convergence rate and easy falling into local optimization.
The existing common product design change modeling method mainly adopts a change propagation path to depict the propagation of a change effect in a product network. They are all based on the following assumptions: after the initial node is changed, the change effect propagates along a continuous path, and the node on the path needs to be changed due to the influence of the change effect. Based on this, these approaches model the product design change problem as a class of propagation path optimization problems. The existing path planning model only considers the propagation degree of the changed path in the product network, seeks an optimal path with the minimum change propagation degree, and does not consider the influence of the change on the service performance. Existing path planning models ignore the impact of changes on nodes outside of their selected path. When a change propagation path is obtained by using these models, only the nodes on the path are changed. However, in many practical product design processes, the propagation of the alteration effect is often web-shaped and cannot be confined to a single path. At this time, if the key nodes outside the optimal path are not changed, the service performance of the product is greatly reduced. After the optimal change propagation path is determined, the existing modeling method considers that all affected nodes in the path need to be changed. However, in reality, some parts are difficult to change or even impossible. If the selected optimal path includes exactly such parts, the path may be rendered impractical or cost prohibitive.
[1]Y.Guodong,Y.Yu,and L.Aijun,“Joint optimization of complex product variant design responding to customer requirement changes,”IFS,vol.30,no.1,pp.397–408,Oct.2015,doi:10.3233/IFS-151764.
[2]S.Jiang,J.Li,and Z.Mao,“Research on the propagation path of function change in product conceptual design,”Advances in Mechanical Engineering,vol.8,no.10,p.168781401667599,Oct.2016,doi:10.1177/1687814016675998.
[3]S.Ma,Z.Jiang,W.Liu,and C.Huang,“Design Property Network-Based Change Propagation Prediction Approach for Mechanical Product Development,”Chin.J.Mech.Eng.,vol.30,no.3,pp.676–688,May 2017,doi:10.1007/s10033-017-0099-z.
[4]Y.Zheng,Y.Yang,and N.Zhang,“A model for assessment of the impact of configuration changes in complex products,”J Intell Manuf,vol.31,no.2,pp.501–527,Feb.2020,doi:10.1007/s10845-018-01461-w.
[5]S.Ma,Z.Jiang,and W.Liu,“Evaluation of a design property network-based change propagation routing approach for mechanical product development,”Advanced Engineering Informatics,vol.30,no.4,pp.633–642,Oct.2016,doi:10.1016/j.aei.2016.08.002.
[6]Y.Guodong,Y.Yu,Z.Xuefeng,and L.Chi,“Network-Based Analysis of Requirement Change in Customized Complex Product Development,”Int.J.Info.Tech.Dec.Mak.,vol.16,no.04,pp.1125–1149,Jul.2017,doi:10.1142/S0219622017500195.
[7]W.Yang,C.Li,Y.Yu,and B.Li,“Change impact analysis of complex product based on three-parameter interval grey number grey relational model,”In Review,preprint,Mar.2021.doi:10.21203/rs.3.rs-306321/v1.
[8]R.Haibing,L.Ting,L.Yupeng,and H.Jie,“Multi-source design change propagation path optimisation based on the multi-view complex network model,”Journal of Engineering Design,vol.32,no.1,pp.28–60,Jan.2021,doi:10.1080/09544828.2020.1858474.
[9]I.Ullah,D.Tang,Q.Wang,and L.Yin,“Least Risky Change Propagation Path Analysis in Product Design Process:LEAST RISKY CHANGE PROPAGATION PATH ANALYSIS,”Syst Eng,vol.20,no.4,pp.379–391,Jul.2017,doi:10.1002/sys.21400.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: a design method of a product change scheme based on multi-objective particle swarm optimization is provided, the influence of product change effect propagation on product service performance is considered, three indexes of influence degree, change time and change cost of the service performance are comprehensively considered, a multi-objective change node selection model is established, and an integer multi-objective particle swarm optimization algorithm (P-DMOPSO) guided by Problem characteristics is provided to solve the model.
The invention adopts the following technical scheme for solving the technical problems: a product change scheme design method based on multi-objective particle swarm optimization specifically comprises the following steps:
step 1: aiming at a product to be changed, a product multi-network model comprising a service performance layer and a part layer is established, each service performance is determined by the common association of a plurality of related parts, and a coupling relation exists between the service performances;
step 2: each part is used as a node, and a change node multi-target selection model containing three evaluation indexes of service performance influence degree, change cost and change time is generated by decoupling the influence of the associated node on the service performance in change effect propagation;
and step 3: solving the multi-target selection model by using a multi-target particle swarm optimization algorithm to obtain a Pareto optimal solution set, namely obtaining an optimal change scheme;
and 4, step 4: and selecting a solution with the maximum satisfaction degree from the Pareto optimal solution set by using a fuzzy decision method as a final change scheme of the product.
Further, the product multiple network model construction method comprises the following steps:
the model comprises a service performance layer and a part layer; each product in the model contains a number of service capabilities to be considered, denoted as SP1，SP2，…，SPnWhere n is the number of considered service performance indicators;
will the ith partviWith jth service capability SPjIs recorded as w'ij；w'ijLarger, part viWith service capability SPjThe stronger the correlation, the part viFor service performance SPjThe greater the effect of (c); the ith part viWith the kth part vkIs recorded as wik。
Furthermore, three indexes of change cost, change time and service performance influence degree are considered at the same time, and the method for constructing the multi-target selection model of the change node comprises the following steps:
step 2.1: establishing a product node change propagation model, determining the influence range and degree of a change node on a product network through the propagation model, and determining a decision variable of a multi-target selection model of the change node;
step 2.2: evaluating the influence of the change scheme on the service performance of the product, namely calculating the influence degree of the service performance of the change scheme;
step 2.3: evaluating the influence of the change scheme on the change cost and the change time, namely calculating the time cost and the economic cost index of the change scheme;
step 2.4: and generating a multi-target selection model of the change node by combining three indexes of change cost, change time and service performance influence degree.
Further, in step 2.1, a decision variable of the multi-target selection model of the changed node is determined, and the method includes the following steps:
step 2.1.1: establishing a product node change propagation model;
calculating a downstream node v according to the change effect propagation coefficient and the change strength of the upstream nodejThe formula is as follows:
Figure BDA0003359398440000041
in the formula, r (v)j) Is a downstream node vjFor node v, with respect to the change strength ofjAn upstream node v ofi，p(vi,vj) Is v isiFor vjThe alteration effect propagation coefficient of (2); r (v)i) Is node viThe change intensity of (a); us (j) denotes node vjThe set of upstream nodes of (a);
when a node is changed, traversing all neighbor nodes by the formula (1), finding all change propagation paths, and calculating the influence intensity of the change on each related node in the network, namely the change intensity, until the change termination condition is met;
the following two cases are set as termination conditions:
the change effect of the current affected node is propagated until the out-degree is 0, or the change intensity of the current affected node is smaller than the design margin;
based on the steps, all nodes affected by the change and the affected degree of the nodes are determined;
step 2.1.2: coding the selection strategy of the changed node;
when the initial node viAfter the change, a set IV (v) of all nodes affected by the change is determined by the change propagation modeli)＝{iv1,iv2,…,ivmM is the number of affected nodes, ivmIs the m affected node;
the influence strength of all nodes affected by the change, i.e., the change strength, is r (IV (v) respectivelyi))＝{r(iv1),r(iv2),…,r(ivm)}，r(ivm) The influence strength of the mth influenced node;
representing node v by a {0,1} encoded binary vectoriA change scheme or optimization solution to the problem of design changes that arise, then all affected node selection policy encodings are expressed as follows:
CV＝(cv1,cv2,...,cvm),cvj∈{0,1},j＝1,2,…,m. (2)
wherein CV represents a node viResulting design Change, cv j1 represents the pair IV (v)i) Middle node ivjPerforming a change; otherwise, node ivjRemain unchanged.
Further, in step 2.2, the influence degree of the service performance of the change scheme is calculated by the following method:
for node v about initial changeiUsing a node set IV (v)i) Evaluating the influence degree of the scheme on the service performance by the unchanged nodes;
specifically, the overall influence of the change scenario CV on the kth service performance is as follows:
Figure BDA0003359398440000051
wherein, w'ikRepresentative node viAffecting SPI for service performancekThe weight of (c);
Figure BDA0003359398440000052
representing pairs of binary elements cviGet the inverse if node viWithout making a change, then
Figure BDA0003359398440000053
If not, then the mobile terminal can be switched to the normal mode,
Figure BDA0003359398440000054
SP(vi) Is when the node viWhen a change occurs, its overall impact on associated service performance; for kth service capability SPkThe node set of related parts is Vsp (k), the modularity is Q, and (1+ Q) is selected-1Is the decoupling factor.
Further, in step 2.3, CV ═ C (CV) is applied to a modification1,cv2,...,cvm) The corresponding time cost and economic cost are respectively as follows:
Figure BDA0003359398440000055
Figure BDA0003359398440000056
wherein, tiAnd costiRespectively representing the time cost and the economic cost required by the ith part in the change scheme CV to perform the change.
Further, in step 2.4, a multi-target selection model of the changed nodes is generated:
Figure BDA0003359398440000057
meanwhile, corresponding model constraints are added according to actual problems, wherein the constraints comprise that parts cannot be changed due to supply chain problems and the parts must be changed due to design reasons.
Further, in the step 3, a multi-target particle swarm optimization algorithm is used for solving the multi-target selection model, and the steps are as follows:
step 3.1: initialization: randomly initializing binary positions of Nsize particles; setting inertia weight omega and maximum iteration number Iter of algorithmmaxLearning factor c1And c2(ii) a Initializing the individual optimum point of each particle to itself; initializing an external reserve set as an empty set;
step 3.2: calculating objective function values of all particles by adopting an equation (6), selecting solutions which are not dominant from the particle swarm, and storing the solutions in an external reserve set; randomly selecting an initial global optimal point of each particle from the individual optimal points;
step 3.3: updating the position of each particle using the following equation;
Figure BDA0003359398440000061
where k is the iteration number of the algorithm, ω is the inertial weight, c1And c2As a learning factor, rand1、rand2、rand3And rand4Random numbers that obey a uniform distribution of U (0, 1); pbest and Gbest represent the individual and global optima, x, of the particle, respectivelyiAnd uiRespectively representing the position and the speed of the ith particle; sv (u)i) Is composed ofA mapping function;
step 3.4: correcting the position of the particle violating the constraint; calculating the objective function value of each particle by adopting a formula (6), and punishing an infeasible solution by using a penalty function method so as to obtain new adaptive values of the particles violating the constraint;
step 3.5: according to the congestion degree sequence of the particles in the reserve set, updating an external reserve set, and determining a global optimum point Gbest and an individual optimum point Pbest of each particle;
step 3.6: if the end condition is not met, namely the iteration number is more than the maximum iteration number ItermaxReturning to the step 3.3; otherwise, outputting a non-inferior solution set stored in the external reserve set, namely the Pareto optimal solution set.
Further, in the step 4, a fuzzy decision method is used, a fuzzy membership function is adopted to simulate the preference of a decision maker to the task, and a solution with the maximum satisfaction is selected from the Pareto optimal solution set as a final change scheme of the product;
specifically, the optimal solution x is concentrated in a reserve set, i.e. Pareto optimal solutionjIts satisfaction is defined as:
Figure BDA0003359398440000062
Figure BDA0003359398440000063
wherein mu is a fuzzy membership function reflecting the preference of the decision maker; mu.si(xj) Is an optimal solution xjFuzzy membership value of ith index; f. ofi(xj) Represents the optimal solution xjThe ith index value of (1); f. ofi maxAnd fi minThe maximum value and the minimum value of the ith index evaluation result are respectively; weiI is 1,2,3, which represents the bias of the decision maker on three indexes of the change cost, the change time and the service performance influence degree;
the element with the largest μ in the reserve set is the compromise solution, i.e. an optimal variation to meet the user's needs.
Has the advantages that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
1) a complex product multi-layer network model introducing service performance is established. The concept of a service performance layer is introduced for the first time, and compared with the traditional model, the model can more fully reflect the influence of design change on products. And the concept of the influence degree of the service performance is given to evaluate the influence of the change effect propagation on the service performance of the product. The defined degree of influence of service performance can predict more accurately how much a product change affects its service performance.
2) And meanwhile, a multi-target selection model of the change nodes is established by considering the change cost, the change time and the service performance influence degree index. Compared with the conventional common change path planning model, the established change node selection model can comprehensively depict the influence of change on the associated nodes, and the obtained change scheme has stronger operability.
3) An integer multi-objective particle swarm optimization algorithm (P-DMOPSO) guided by problem characteristics is provided for solving the established change node selection model. By introducing an integer particle position updating strategy, the capability of the particle swarm optimization algorithm for processing the model is remarkably improved.
4) The multi-target change node selection model provided by the invention comprehensively considers 3 indexes of service performance, change time and change cost, and can provide multiple groups of alternative change schemes for decision makers at the same time. The decision maker can select the most appropriate change scheme from the alternative scheme set according to the preference of the decision maker. The change selection model provided by the invention comprehensively considers all affected nodes, and the obtained change scheme can simultaneously cover key nodes on a plurality of propagation paths, thereby being more in line with the actual design requirement. By setting appropriate constraint conditions, the feasibility of the change scheme can be ensured to the maximum extent.
Drawings
FIG. 1 is a complex product multiple network model;
figure 2 is a diagram of an improved Sigmod function;
FIG. 3 is a network diagram of creating a network of TV complex products of a certain model;
FIG. 4 is an optimal reroute obtained by comparing the LRCPP algorithm;
fig. 5 is a modified Pareto front end.
Detailed Description
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
The invention relates to a multi-target particle swarm generation method of a complex product change scheme, which comprises the steps of firstly establishing a new complex product multi-network model; then, the influence of the change effect propagation on the service performance of the product is analyzed and evaluated, and the definition of the influence degree of the service performance is given; and then, simultaneously considering a plurality of indexes such as the influence degree of service performance and the like, establishing a multi-target selection model of the change node, and providing an improved particle swarm solution method according to the problem characteristics.
The method specifically comprises the following steps:
step 1: in order to accurately analyze a change propagation mechanism and the influence of the change propagation mechanism on the service performance of a product, a complex product multi-network model introducing the service performance is established for any given product to be changed.
Fig. 1 shows a specific structure of the model, wherein the upper layer is a service performance layer, and the lower layer is a conventional component layer. Each product typically contains multiple service capabilities to be considered, not denoted as SP1，SP2，…，SPnWhere n is the number of considered service performance indicators. The ith part viWith jth service capability SPjIs recorded as w'ij。w'ijLarger, part viWith service capability SPjThe stronger the degree of association; further, the part viFor service capability SPjThe greater the effect of (c). The ith part viWith the kth part vkIs recorded as wik. With v1And v2Two nodes and a 1 st service capability SP1As an example, v1And v2Is connected with an edge of e12With a weight of w12(ii) a Details v2With service capability SP1Degree of association noteIs w'21。
It can be seen that the model has the following characteristics: the network model includes a service performance layer (SPI layer) and a Part layer (Part layer). Compared with the traditional product network model, the model introduces the concept of a service performance layer for the first time. Each service performance is determined by a plurality of related parts, but the degree of association of each related part to the service performance is different. ③ one component may affect multiple service capabilities. Although there is no direct relationship between the service performances, there is also a coupling relationship between the service performances under the influence of the commonly associated parts.
Step 2: and generating a change node multi-target selection model (analyzing a product change influence mechanism and establishing a change node multi-target selection model) containing three evaluation indexes of the influence degree of the service performance, the change cost and the change time by decoupling the influence of the strongly associated node on the service performance in the change effect propagation.
In order to select a more appropriate product change scheme, three indexes of change cost, change time and service performance influence degree are considered at the same time, and a change node selection problem is modeled into a multi-objective combined optimization model. Without loss of generality, parts are subsequently collectively referred to as nodes. Firstly, a product node change propagation model used by the invention is given, and the influence range and the degree of the change node on a product network are determined through the model. Therefore, the decision variables and the value ranges of the multi-target selection model of the change nodes are determined. Then, the definition of the influence degree of the service performance facing to a single node is given, and the integral influence degree of the service performance facing to a change node selection scheme (namely a candidate solution) is further calculated; and finally, considering two indexes of change cost and change time and actual constraints, and establishing a multi-objective optimization model of the problem.
Step 2.1: and determining and coding decision variables of the multi-target selection model of the change node.
Step 2.1.1: and establishing a product node change propagation model.
One node is influenced by the propagation of the change effect of a plurality of upstream nodes according to the change effectThe downstream node v can be calculated according to the propagation coefficient and the product change strengthjThe formula is as follows:
Figure BDA0003359398440000081
in the formula, r (v)j) Is a downstream node vjFor node v, with respect to the change strength ofjAn upstream node v ofi，p(vi,vj) Is viFor vjThe change effect propagation coefficient of (1); r (v)i) Is node viThe change intensity of (a); us (j) denotes node vjThe upstream node set of (2).
When a node is changed, firstly traversing all neighbor nodes by the formula (1), finding all change propagation paths, and calculating the influence strength of the change on each related node in the network, namely the change strength, until the change termination condition is met.
The following two cases are set as termination conditions:
the change effect of the currently affected node is propagated until the out-degree is 0, or the change strength of the currently affected node is smaller than the design margin thereof.
Based on the above steps, all nodes affected by the change and the affected degree thereof can be determined.
Without setting node viAfter the change, the set of affected nodes is IV (v)i)＝{iv1,iv2,…,ivmM is the number of affected nodes, and their influence strength is r (IV (v) respectivelyi))＝{r(iv1),r(iv2),…,r(ivm)}，ivmFor the mth affected node, r (iv)m) The impact strength of the mth affected node.
Step 2.1.2: the change node selection policy is encoded.
When an initial node viAfter a change has occurred, a set IV (v) of all nodes affected by the change can be determined by the change propagation modeli)。
Without loss of the generality of the method,consider a node viProduct design change problem caused by change, and the set of nodes affected by the change is IV (v)i)＝{iv1,iv2,…,ivm}。
Representing node v by a {0,1} encoded binary vectoriAn alternative or optimization solution to the problem of design changes that arise, all affected node selection policy encodings are represented as follows:
CV＝(cv1,cv2,...,cvm),cvj∈{0,1},j＝1,2,…,m. (2)
wherein CV represents a node viResulting design Change, cv j1 represents the pair IV (v)i) Middle node ivjChanging; otherwise, node ivjRemain unchanged.
Step 2.2: and evaluating the influence of the change scheme on the service performance of the product, namely calculating the influence degree of the service performance of the change scheme.
For node v about initial changeiUsing a node set IV (v)i)＝{iv1,iv2,…,ivmAnd evaluating the influence degree of the scheme on the service performance by unchanged nodes.
Specifically, the overall influence of the change scenario CV on the kth service performance is as follows:
Figure BDA0003359398440000091
wherein, w'ikRepresentative node viAffecting SPI for service performancekThe weight of (c);
Figure BDA0003359398440000092
for binary element cviTaking the inverse; if affected node viWithout making a change, then
Figure BDA0003359398440000093
If not, then,
Figure BDA0003359398440000094
SP(vi) Is when the node viWhen a change occurs, its overall impact on associated service performance; for kth service capability SPkSetting the node set of relevant parts as Vsp (k), the modularity as Q, selecting (1+ Q)-1Is the decoupling factor.
Step 2.3: and evaluating the influence of the change scheme on the change cost and time, namely calculating the time cost and the economic cost index of the change scheme.
When a node viWhen the node needs to be changed, the node needs to be redesigned or selected. This not only increases the economic cost of product design, but also extends its construction period. For this reason, the time cost and the economic cost are used as two other indexes to evaluate the merits of one modification.
For one variant CV ═ CV (CV)1,cv2,...,cvm) The corresponding time cost and economic cost are respectively as follows:
Figure BDA0003359398440000095
Figure BDA0003359398440000096
wherein, tiAnd costiRespectively representing the time cost and the economic cost required by the ith part in the change scheme CV to perform the change.
Step 2.4: and simultaneously considering three indexes of change cost, change time and service performance influence degree to generate a change node multi-target selection model:
Figure BDA0003359398440000101
meanwhile, corresponding constraints are added according to practical problems, wherein the constraints comprise that parts cannot be changed due to supply chain problems, and parts must be changed due to design reasons, so that the obtained changing method is feasible.
And step 3: and solving the multi-target model by using an integer multi-target particle swarm optimization algorithm to obtain an optimal change scheme.
In order to solve the model, an improved binary particle swarm optimization algorithm guided by problem characteristics is provided, and the model is solved. The specific execution steps of the improved binary particle swarm optimization algorithm are as follows:
step 3.1: initialization: randomly initializing binary positions of Nsize particles; setting inertia weight omega and maximum iteration number Iter of algorithmmaxLearning factor c1And c2(ii) a Initializing the individual optimum point of each particle to itself; initializing an external reserve set as an empty set;
step 3.2: calculating objective function values of all particles by adopting an equation (6), selecting solutions which are not dominant from the particle swarm, and storing the solutions in an external reserve set; randomly selecting an initial global optimum point of each particle from the individual optimum points;
step 3.3: updating the position of each particle using the following equation;
Figure BDA0003359398440000102
where k is the iteration number of the algorithm, ω is the inertial weight, c1And c2As a learning factor, rand1、rand2、rand3And rand4To obey the random numbers of U (0,1) being uniformly distributed. Pbest and Gbest represent the individual and global optima, x, of the particle, respectivelyiAnd uiRespectively, the position and velocity of the ith particle. In order to prevent the flying speed of the particles from being too fast, an upper limit u and a lower limit u of the particle speed are setmax4 and umin＝-4。sv(ui) For the improved mapping function, the function image is shown in fig. 2.
Step 3.4: correcting the position of the particle violating the constraint; calculating the objective function value of each particle by adopting a formula (6), and punishing an infeasible solution by using a penalty function method so as to obtain new adaptive values of the particles violating the constraint;
step 3.5: according to the congestion degree sequence of the particles in the reserve set, updating an external reserve set, and determining a global optimum point Gbest and an individual optimum point Pbest of each particle;
step 3.6: if the end condition is not met, namely the iteration number is more than the maximum iteration number ItermaxReturning to the step 3.3; otherwise, outputting the non-inferior solution set stored in the external reserve set, namely the Pareto optimal solution set.
And 4, step 4: and obtaining an optimal change scheme meeting the user requirements by using a fuzzy decision method.
And simulating the preference of a decision maker to the task by adopting a fuzzy membership function, and selecting a solution with the maximum satisfaction from the Pareto optimal solution set as a final change scheme of the product.
Centralizing optimal solution x in a reservoirjFor example, its satisfaction is defined as:
Figure BDA0003359398440000111
Figure BDA0003359398440000112
wherein mu is a fuzzy membership function reflecting the preference of the decision maker, and the function is used for simulating the preference of the decision maker to the task; mu.si(xj) Is an optimal solution xjFuzzy membership value, f, of the ith indexi(xj) Represents the optimal solution xjThe ith index value of (f)i maxAnd fi minThe maximum value and the minimum value, we, of the ith index evaluation resultiAnd i is 1,2 and 3, which is used for reflecting the bias of a decision maker on three indexes of change cost, change time and service performance influence degree.
The element with the largest μ in the reserve set is the compromise solution, i.e. an optimal variation to meet the user's needs.
The following is a specific experimental analysis performed on the method of the present invention.
In order to verify the effectiveness of the established model and the proposed algorithm, experimental analysis is carried out by taking a certain model of a TV of the creative RGB electronic limited company as an example. The product comprises 10 modules, 101 parts, such as a Wi-Fi module, a screen module, a power supply module and the like, and a network diagram of a complex product of the product is shown in figure 3. And establishing a DSM matrix and a double-layer product network model according to the physical and functional relation among the parts.
Compared with a classical path optimization algorithm, the efficiency of the method is verified. The selected comparison algorithm is the algorithm described in the document [9] and is marked as LRCPP. After the initial node is changed, the algorithm obtains a continuous path propagation of the change effect. A part 87 (linear voltage stabilizing IC) is selected as an initial change node, and an optimal propagation path is obtained by the LRCPP algorithm shown in FIG. 4. The changed path obtained by the algorithm is shown in a bold way in the figure, and the specific changed path information is as follows:
part 87 (linear regulator IC) → part 61 (patch power inductor 1) → part 52(FPC needle mount) → part 82(DC-DC1) → part 46 (package bipolar transistor +) → part 79 (patch NTC thermistor) → part 66 (patch radio frequency ceramic inductor 2) → part 62 (patch power inductor 2). The → sign indicates the direction of the change path. Further, by calculating 3 index values considered by the present invention using equation (6), it can be obtained that the service performance influence degree of the optimal path is 0.54, the change time is 473 hours, and the change cost is 19.04 ten thousand yuan.
A set of variants can be obtained by operating the method according to the invention, also with the component 87 (linear regulator IC) as initial component. Fig. 5 shows Pareto front ends corresponding to these schemes, where the five-pointed star is the result of the LRCPP algorithm. It can be seen that the results of the LRCPP algorithm are dominated by the multiple solutions obtained by the P-DMOPSO algorithm. Without loss of generality, a variation is randomly selected from these solutions to compare in detail with the results of the LRCPP algorithm. The specific modification is as follows: part 1 (electronic wire), part 12 (packaged LED1), part 13(LENS), part 14 (packaged LED2), part 30 (infrared receiver), part 80 (power amplifier TI), part 82(DC-DC1), part 87 (linear voltage regulator IC), part 88(WIFI chip RTL8188), and part 97(PWM control IC). The service performance influence degree of the modification was 0.24, the modification time was 205 hours, and the modification cost was 14.24 ten thousand yuan.
The modifications obtained by comparing the above two methods are known: (1) the LRCPP algorithm can only find one change propagation path at a time; the P-DMOPSO algorithm provided by the invention can simultaneously obtain a plurality of groups of change schemes for decision makers to select. (2) The index values of a plurality of change schemes obtained by the method are obviously superior to the LRCPP algorithm in consideration of the index of the influence degree of the service performance. This is mainly because the LRCPP algorithm only changes parts on its selected optimal path, ignoring parts on other changed paths that significantly affect service performance. In contrast, the variation obtained by the method of the present invention can cover critical parts on multiple propagation paths simultaneously, as shown in fig. 5. (3) And meanwhile, three indexes of influence degree of service performance, change time and cost are considered, and a plurality of groups of change schemes obtained by the method are superior to the LRCPP algorithm.

Claims (8)

1. A product change scheme design method based on multi-objective particle swarm optimization is characterized in that: the method comprises the following steps:
step 1: aiming at a product to be changed, a product multi-network model comprising a service performance layer and a part layer is established, each service performance is determined by the common association of a plurality of related parts, and a coupling relation exists between the service performances;
step 2: each part is used as a node, and a change node multi-target selection model containing three evaluation indexes of service performance influence degree, change cost and change time is generated by decoupling the influence of the associated node on the service performance in change effect propagation; the model construction method comprises the following steps:
step 2.1: establishing a product node change propagation model, determining the influence range and degree of a change node on a product network through the propagation model, and determining a decision variable of a multi-target selection model of the change node;
step 2.2: evaluating the influence of the change scheme on the service performance of the product, namely calculating the influence degree of the service performance of the change scheme;
step 2.3: evaluating the influence of the change scheme on the change cost and the change time, namely calculating the time cost and the economic cost index of the change scheme;
step 2.4: combining three indexes of change cost, change time and service performance influence degree to generate a change node multi-target selection model;
and step 3: solving the multi-target selection model by using a multi-target particle swarm optimization algorithm to obtain a Pareto optimal solution set, namely obtaining an optimal change scheme;
and 4, step 4: and selecting a solution with the maximum satisfaction degree from the Pareto optimal solution set by using a fuzzy decision method as a final change scheme of the product.
2. The product change plan designing method according to claim 1, characterized in that: the product multi-network model construction method comprises the following steps:
the model comprises a service performance layer and a part layer; each product in the model contains a number of service capabilities to be considered, denoted as SP1，SP2，…，SPnWhere n is the number of considered service performance indicators;
the ith part viWith jth service capability SPjIs recorded as w'ij；w'ijLarger, part viWith service capability SPjThe stronger the correlation, the part viFor service performance SPjThe greater the effect of (c); the ith part viWith the kth part vkIs recorded as wik。
3. The product change plan designing method according to claim 1, characterized in that: in the step 2.1, a decision variable of the multi-target selection model of the change node is determined, and the method comprises the following steps:
step 2.1.1: establishing a product node change propagation model;
calculating a downstream node v according to the change effect propagation coefficient and the change strength of the upstream nodejThe formula is as follows:
Figure FDA0003621141640000011
in the formula, r (v)j) Is a downstream node vjFor node v, with respect to the change strength ofjAn upstream node v ofi，p(vi,vj) Is v isiFor vjThe change effect propagation coefficient of (1); r (v)i) Is node viThe change intensity of (a); us (j) denotes node vjThe upstream node set of (2);
when a node is changed, traversing all neighbor nodes by the formula (1), finding all change propagation paths, and calculating the influence intensity of the change on each related node in the network, namely the change intensity, until the change termination condition is met;
the following two cases are set as termination conditions:
the change effect of the current affected node is propagated until the out-degree is 0, or the change intensity of the current affected node is smaller than the design margin;
based on the steps, all nodes affected by the change and the affected degree of the nodes are determined;
step 2.1.2: coding a change node selection strategy;
when the initial node viAfter the change, a set IV (v) of all nodes affected by the change is determined by the change propagation modeli)＝{iv1,iv2,…,ivmM is the number of affected nodes, ivmIs the m affected node;
the influence strength of all nodes affected by the change, i.e., the change strength, is r (IV (v) respectivelyi))＝{r(iv1),r(iv2),…,r(ivm)}，r(ivm) The influence strength of the mth influenced node;
representing node v by a {0,1} encoded binary vectoriAn alternative or optimization solution to the problem of design changes that arise, all affected node selection policy encodings are represented as follows:
CV＝(cv1,cv2,...,cvm),cvj∈{0,1},j＝1,2,…,m. (2)
wherein CV represents a node viResulting design Change, cvj1 represents the pair IV (v)i) Middle node ivjChanging; otherwise, node ivjRemain unchanged.
4. The product modification design method according to claim 3, wherein: in the step 2.2, the influence degree of the service performance of the change scheme is calculated, and the method includes the following steps:
for node v about initial changeiUsing a node set IV (v)i) The influence degree of the scheme on the service performance is evaluated by unchanged nodes;
specifically, the overall influence of the change scenario CV on the kth service performance is as follows:
Figure FDA0003621141640000021
wherein, wi′kRepresentative node viAffecting SPI for service performancekThe weight of (c);
Figure FDA0003621141640000022
representing pairs of binary elements cviGet the inverse if node viWithout making a change, then
Figure FDA0003621141640000023
If not, then,
Figure FDA0003621141640000024
SP(vi) Is when the node viWhen a change occurs, its overall impact on associated service performance; for kth service capability SPkThe node set of related parts is Vsp (k), the modularity is Q, and (1+ Q) is selected-1Is the decoupling factor.
5. The product modification design method according to claim 3, wherein: in step 2.3, CV ═ CV (CV) is used for a modification1,cv2,...,cvm) The corresponding time cost and economic cost are respectively as follows:
Figure FDA0003621141640000025
Figure FDA0003621141640000026
wherein, tiAnd costiRespectively representing the time cost and the economic cost required by the ith part in the change scheme CV to perform the change.
6. The product modification design method according to claim 3, wherein: in the step 2.4, a multi-target selection model of the change node is generated:
Figure FDA0003621141640000031
meanwhile, corresponding model constraints are added according to actual problems, wherein the constraints comprise that parts cannot be changed due to supply chain problems and the parts must be changed due to design reasons.
7. The product modification plan designing method according to claim 6, wherein: in the step 3, a multi-target particle swarm optimization algorithm is used for solving the multi-target selection model, and the steps are as follows:
step 3.1: initialization: randomly initializing binary positions of Nsize particles; setting inertia weight omega and maximum iteration number Iter of algorithmmaxLearning factor c1And c2(ii) a Initializing the individual optimum point of each particle to itself; initializing an external reserve set as an empty set;
step 3.2: calculating objective function values of all particles by adopting an equation (6), selecting solutions which are not dominant from the particle swarm, and storing the solutions in an external reserve set; randomly selecting an initial global optimum point of each particle from the individual optimum points;
step 3.3: updating the position of each particle using the following equation;
Figure FDA0003621141640000032
where k is the iteration number of the algorithm, ω is the inertial weight, c1And c2As a learning factor, rand1、rand2、rand3And rand4Random numbers that obey a uniform distribution of U (0, 1); pbest and Gbest represent the individual and global optima, x, of the particle, respectivelyiAnd uiRespectively representing the position and the speed of the ith particle; sv (u)i) Is a mapping function;
step 3.4: correcting the position of the particle violating the constraint; calculating the objective function value of each particle by adopting a formula (6), and punishing an infeasible solution by using a penalty function method so as to obtain new adaptive values of the particles violating the constraint;
step 3.5: according to the congestion degree sequence of the particles in the reserve set, updating an external reserve set, and determining a global optimum point Gbest and an individual optimum point Pbest of each particle;
step 3.6: if the end condition is not met, namely the iteration number is more than the maximum iteration number ItermaxReturning to the step 3.3; otherwise, outputting the non-inferior solution set stored in the external reserve set, namely the Pareto optimal solution set.
8. The method of designing a product modification according to any one of claims 1 to 7, wherein: in the step 4, a fuzzy decision method is used, a fuzzy membership function is adopted to simulate the preference of a decision maker to tasks, and a solution with the maximum satisfaction degree is selected from the Pareto optimal solution set to serve as a final change scheme of a product;
in particular, for the reserve set, i.e. Pareto optimal solution setMedium optimal solution xjIts satisfaction is defined as:
Figure FDA0003621141640000041
Figure FDA0003621141640000042
wherein mu is a fuzzy membership function reflecting the preference of the decision maker; mu.si(xj) Is an optimal solution xjFuzzy membership value of ith index; f. ofi(xj) Represents the optimal solution xjThe ith index value of (2); f. ofi maxAnd fi minThe maximum value and the minimum value of the ith index evaluation result are respectively; weiI is 1,2,3, which represents the bias of the decision maker on three indexes of the change cost, the change time and the service performance influence degree;
the element with the largest μ in the reserve set is the compromise solution, i.e. an optimal variation to meet the user's needs.
CN202111361319.3A 2021-11-17 2021-11-17 Product change scheme design method based on multi-objective particle swarm optimization Active CN113987718B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111361319.3A CN113987718B (en) 2021-11-17 2021-11-17 Product change scheme design method based on multi-objective particle swarm optimization

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111361319.3A CN113987718B (en) 2021-11-17 2021-11-17 Product change scheme design method based on multi-objective particle swarm optimization

Publications (2)

Publication Number Publication Date
CN113987718A CN113987718A (en) 2022-01-28
CN113987718B true CN113987718B (en) 2022-06-03

Family

ID=79749046

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111361319.3A Active CN113987718B (en) 2021-11-17 2021-11-17 Product change scheme design method based on multi-objective particle swarm optimization

Country Status (1)

Country Link
CN (1) CN113987718B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116451577B (en) * 2023-04-11 2024-04-02 中国矿业大学 Product change path multi-target optimization method integrating reinforcement learning and differential evolution

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104166873A (en) * 2014-06-26 2014-11-26 中国矿业大学 Multi-objective particle group feature selection method under fuzzy representation of data quality
CN105740984A (en) * 2016-02-01 2016-07-06 北京理工大学 Product concept performance evaluation method based on performance prediction
CN111274757A (en) * 2020-01-19 2020-06-12 南京航空航天大学 Method for realizing automatic layout of multi-branch cable assembly of electromechanical product based on multi-objective optimization
CN113312735A (en) * 2021-05-19 2021-08-27 太原理工大学 DMA partition method for urban water supply pipe network

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160203419A1 (en) * 2013-03-09 2016-07-14 Bigwood Technology, Inc. Metaheuristic-guided trust-tech methods for global unconstrained optimization
CN106446478B (en) * 2016-11-28 2019-04-09 辽宁工业大学 A kind of cutting technology preferred method
CN109631900B (en) * 2018-12-24 2022-08-26 中国矿业大学 Unmanned aerial vehicle three-dimensional flight path multi-target particle swarm global planning method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104166873A (en) * 2014-06-26 2014-11-26 中国矿业大学 Multi-objective particle group feature selection method under fuzzy representation of data quality
CN105740984A (en) * 2016-02-01 2016-07-06 北京理工大学 Product concept performance evaluation method based on performance prediction
CN111274757A (en) * 2020-01-19 2020-06-12 南京航空航天大学 Method for realizing automatic layout of multi-branch cable assembly of electromechanical product based on multi-objective optimization
CN113312735A (en) * 2021-05-19 2021-08-27 太原理工大学 DMA partition method for urban water supply pipe network

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
A PSO-based multi-objective multilabel feature selection method in classification;Yong Zhang et al;《scientific reports》;20170323;第1-12页 *
Evaluation of a design property network-based change propagation routing approach for mechanical product development;Ma Songhua et al;《Advanced Engineering Informatics》;20160822;第633-642页 *
基于广义动态约束满足的产品应用服务构件模型研究;阴艳超等;《计算机集成制造系统》;20090615(第06期);第60-68页 *
基于改进MOPSO的煤炭企业内部供应链优化研究;吉峰等;《中国矿业大学学报》;20161115(第06期);第233-239页 *
面向全局需求适应度的PASC组合优化;阴艳超等;《计算机集成制造系统》;20100915(第09期);第145-154页 *

Also Published As

Publication number Publication date
CN113987718A (en) 2022-01-28

Similar Documents

Publication Publication Date Title
Mendoza et al. NSGA and SPEA applied to multiobjective design of power distribution systems
Guha et al. Machine learning for AC optimal power flow
Zhao et al. A surrogate-ensemble assisted expensive many-objective optimization
CN113361680A (en) Neural network architecture searching method, device, equipment and medium
De Oliveira et al. Efficient hybrid algorithm for transmission expansion planning
CN105631528A (en) NSGA-II and approximate dynamic programming-based multi-objective dynamic optimal power flow solving method
CN113987718B (en) Product change scheme design method based on multi-objective particle swarm optimization
Liu et al. Multi-objective optimization model with AHP decision-making for cloud service composition
Helbig et al. Key challenges and future directions of dynamic multi-objective optimisation
CN116402002A (en) Multi-target layered reinforcement learning method for chip layout problem
Chien et al. An integrated approach for IC design R&D portfolio decision and project scheduling and a case study
Zheng et al. Multi‐objective particle swarm optimisation of complex product change plan considering service performance
Hu et al. Optimal operation of power systems with energy storage under uncertainty: A scenario-based method with strategic sampling
Gosavi et al. Maintenance optimization in a digital twin for Industry 4.0
Zhang et al. DRL-S: Toward safe real-world learning of dynamic thermal management in data center
Massaoudi et al. Navigating the Landscape of Deep Reinforcement Learning for Power System Stability Control: A Review
Fan et al. From specification to topology: Automatic power converter design via reinforcement learning
CN109255389A (en) A kind of equipment evaluation method, device, equipment and readable storage medium storing program for executing
Guo Towards the efficient generation of variant design in product development networks: network nodes importance based product configuration evaluation approach
CN115146455B (en) Complex supply chain multi-objective decision method supported by calculation experiment
Li et al. Implicit stochastic optimization with data mining for reservoir system operation
Divshali et al. Application of bifurcation theory in dynamic security constrained optimal dispatch in deregulated power system
CN111027709B (en) Information recommendation method and device, server and storage medium
Zhang et al. A deep reinforcement learning algorithm for order acceptance decision of individualized product assembling
Liu et al. An improved quantum particle swarm algorithm for solving multi-objective fuzzy flexible job shop scheduling problem

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant