Complex product design change method under supplier interruption risk

Abstract

The invention relates to a design change method of a high-speed train bogie, and provides a design change model combining a complex network and a Bayesian network and a solving method of the design change model aiming at the problem of complex product design change in a supplier interruption scene. A part function and structure relation matrix is determined based on the interval intuitionistic fuzzy set, the weight of the matrix is determined by adopting the network difference degree, and a product part complex network is constructed; defining a supplier risk quantitative index, and constructing a supplier interruption risk assessment model based on the AHP and the Bayesian network; constructing a supplier-combined product module reconfiguration master-slave optimization model, and solving by adopting a nested genetic algorithm to obtain a product module reconfiguration optimal scheme; according to the method, the supplier side of the product design change reason is considered, the design change caused by supplier interruption is identified, the research and development speed of the product is improved, the quality of the product is ensured, and the competitive capacity of an enterprise is enhanced.

Description

Technical Field

[0001] The present application relates to the technical field of complex product design, and in particular to a method for changing the design of a high-speed train bogie. Background Art

[0002] Design changes are inevitable during product development. In the current process of market globalization, companies are increasingly inclined to take supplier factors into consideration in the early stages of product design changes in order to better control costs, ensure quality, and shorten time to market. In response to complex product design changes in supplier disruption scenarios, it is required to analyze product design changes from the supply side. However, traditional product design changes focus on changes in demand, but insufficient research on supply-side risks leads to insufficient response strategies for companies when problems arise on the supply chain side. Therefore, accurately identifying and controlling the spread of design changes can speed up product development, ensure product quality, and enhance the competitiveness of companies. Summary of the invention

[0003] In order to solve the problems existing in the prior art, the purpose of the present invention is to propose a method for dealing with complex design changes under the risk of supplier interruption, thereby reducing the cost losses of enterprises under supplier interruption, and quickly restoring the production and supply of their products, thereby improving the competitiveness of enterprises in the market.

[0004] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0005] A method for dealing with complex design changes under supplier disruption risk, the method comprising the following steps:

[0006] Step S1: Based on the complex network component association relationship representation, the relationship network between components is constructed by the functional relationship and structural relationship of the product;

[0007] Step S2: Using the ant colony algorithm combined with the capacity of complex network nodes to search for the change propagation path, with the goal of minimizing the change propagation intensity;

[0008] Step S3: Based on the search for the change propagation path, the modules that need to be changed can be obtained, and the modules that need to be changed are used to solve the product design change plan and the supplier selection.

[0009] The method of constructing a relationship network between parts through the functional relationship and structural relationship of the product includes: determining the functional relationship matrix and structural relationship matrix of the parts based on interval intuitionistic fuzzy sets, determining the weights of the functions and structures respectively by using the network difference method, and determining the comprehensive relationship matrix.

[0010] The method of using the ant colony algorithm in combination with the capacity of complex network nodes to search for the change propagation path with the goal of minimizing the change propagation intensity includes the following steps:

[0011] Step S21: The change cost is defined as the additional cost of changing a node in the network. Therefore, the additional cost of changing a node is expressed as:

[0012]

[0013] In the formula, n represents the total number of nodes, C(V i ) is the node V i The cost of the change;

[0014] Change time is the time required from the initiation of the change request to the completion of all change designs; it is expressed as:

[0015]

[0016] Where n represents the total number of nodes, T(V i ) is the node V i The time required for the change;

[0017] The change propagation probability refers to the probability of node i propagating to node j. When there is no connecting edge between nodes i and j, the propagation probability is zero. The larger the propagation probability, the more likely it is that the propagation will pass through the edge. The propagation probability is calculated based on the weights between the edges and is expressed as:

[0018]

[0019] In the formula, w(V i ) is the node V i The sum of the weights of the connected edges, W ij is the weight between node i and node j;

[0020] The change propagation index is used to determine the properties of the node, which is classified as absorber / carrier / diffuser. During the propagation process, the diffusion node should be avoided as much as possible. The change propagation index is expressed as:

[0021]

[0022] Where D out (i) represents the out-degree of the node, D in (i) represents the in-degree of the node;

[0023] The calculation of change propagation intensity is based on the above four indicators. The smaller the change cost, change time and change propagation index, the better, while the larger the change propagation probability, the better. In summary, the change propagation intensity is defined as:

[0024] CPIij =(w1C c +w2T c +w3D c )(1 - P ij )

[0025] Wherein, W1 + W2 + W3 = 1, which respectively represent the weights of each index and can be calculated by the entropy weight method;

[0026] Step S22: During the design process, designers will reserve a certain degree of design tolerance. Define the design tolerance as the node capacity of its complex network. The larger the capacity of a node, the stronger its risk resistance ability. The node capacity is defined as follows:

[0027] ρ i =(1 + β)L0(v i )

[0028] Wherein, β > 0 is the tolerance parameter of the network, indicating the tolerance size of the design parameter. The larger β is, the stronger its risk resistance ability, but when it is too large, it will lead to a decline in product performance. Therefore, the parameter should be within the specified range; L0(V i ) represents the initial load of the node and can be obtained by calculating the degree of the node. The calculation formula is as follows:

[0029]

[0030] Wherein, K i , K j represent the degrees of the component nodes i and j, q is a proportionality constant, which takes the value of 0.5 in this paper, a(0 < a < 1) is the control parameter of the initial load of the node, and the larger a is, the greater the difference in the initial load between nodes; S i represents the set of neighbor nodes of node i;

[0031] Based on the above definitions of the change propagation intensity and the node capacity, the objective function and constraints for searching the optimal change propagation path are defined as follows:

[0032]

[0033] Wherein, k is the number of modules for change propagation, ρ0 represents the change impact of the initial change node and can be given by domain experts; the formula represents the objective of minimizing the change propagation intensity, and the formula represents that when the absorption capacity of the node is greater than the change impact of the initial node, the propagation stops.

[0034] Based on the search for the change propagation path, the modules that need to be changed can be obtained. Solving the product design change plan and the selection of suppliers for the modules that need to be changed includes the following steps:

[0035] Step S31: using the master-slave optimization method to solve the change plan, the upper layer optimization target is the established comprehensive utility function, and the target is to maximize the utility value;

[0036] Convert customer requirements into configuration parameter values ​​for each module to be configured; this can be calculated by the weighted Euclidean distance from the target vector;

[0037] After normalizing the configuration item parameters, the weighted Euclidean distance method is used to calculate the module instance M ij The similarity with the target vector is as follows:

[0038]

[0039] In the formula, n represents the number of modules, m represents the number of instances, and CP ii Configure the parameter item for the target, CP iil Represents module M i The lth configuration parameter item of the jth instance of is the configuration parameter item after normalization, Rmd ij represents the requirement matching degree of the jth instance of the i-th module, W il Indicates the weight of the lth configuration parameter item of the i-th module;

[0040] The upper utility function is represented by U,

[0041]

[0042] Where C is the total cost of the supplier, R s is the supplier risk value, rmd represents the customer demand matching degree, and sim is the similarity of product modules;

[0043] The objective function of the upper layer is expressed as follows:

[0044]

[0045] x ij ∈{0,1},i=1,2,...n;j=1,2..,m,s=1,2...,S

[0046]

[0047] In the formula, χ ij Represents the selected module instance, where i represents the module number, j represents the jth module instance of the i-th module, and similarly y ijs represents the supplier of the selected module instance, i and j are defined as above, s represents the sth supplier of the selected module instance, and x ij ,y ijsAll are binary decision variables that can only take the value of 0 or 1, indicating the selection or non-selection of the current module instance or supplier. The three parameters n, m, and S are the number of modules, module instances, and supplier selections, respectively. il Represents the weight of the lth configuration item parameter of the i-th module;

[0048] Step S32: By establishing a Bayesian network of supplier risk indicators, the risk level of the supplier needs to be evaluated based on Bayesian formula reasoning;

[0049] Step S33: using a nested genetic algorithm to reconfigure and solve the changed parts.

[0050] Beneficial technical effects:

[0051] 1. The inventors have found in practice that the existing complex product design change methods do not adequately analyze the design changes caused by the supply side and lack a quantitative method for suppliers. The present invention considers the supplier interruption risk and defines its quantitative indicators, and constructs a supplier interruption risk assessment model based on AHP and Bayesian networks to actively identify and avoid risks during product design changes.

[0052] 2. The inventor has found in practice that in the existing solution for reconfiguration of product modules in combination with suppliers, product configuration and suppliers are considered in the same dimension. The present invention proposes a master-slave optimization model and a nested genetic algorithm for reconfiguration of product modules in combination with suppliers, which realizes the optimal solution for reconfiguration of product modules in combination with suppliers under the optimal path of design changes.

[0053] The method for dealing with complex design changes under the risk of supplier interruption takes the supplier side of the product design change into account, identifies the design changes caused by supplier interruption, and is conducive to improving the product development speed, ensuring product quality, and enhancing the competitiveness of enterprises. It also provides a method for the rapid recovery of product production interruptions caused by supplier interruptions. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a complex network diagram of component relationships.

[0055] Figure 2 This is the flow chart of the ant colony algorithm.

[0056] Figure 3 This is the flow chart of the ant colony algorithm.

[0057] Figure 4 A schematic diagram of a Bayesian network.

[0058] Figure 5 This is the Bayesian network diagram of supplier risk.

[0059] Figure 6Design a flow chart for the algorithm.

[0060] Figure 7 This is a schematic diagram of the upper layer encoding.

[0061] Figure 8 This is a schematic diagram of the lower layer encoding.

[0062] Fig. 9 This is the algorithm convergence diagram.

[0063] Fig.10 This is a comparison chart of the methods. DETAILED DESCRIPTION

[0064] To make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0065] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the invention claimed for protection, but merely represents some embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features and technical solutions in the embodiments may be combined with each other.

[0067] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0068] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the invention product is usually placed when in use, or the orientation or positional relationship commonly understood by those skilled in the art. Such terms are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0069] The purpose of this invention is to propose a method for dealing with complex design changes under the risk of supplier interruption to guide the customized design of complex products. Specifically, it includes the following steps:

[0070] Step S1: Based on the complex network representation of component association relationships, a relationship network between components is constructed by analyzing the functional and structural relationships of the product.

[0071] S1 includes the following steps:

[0072] Step S11: Determine the functional relationship matrix and structural relationship matrix of the parts based on interval-valued intuitionistic fuzzy sets, use the network difference method to determine the weights of the functions and structures, and determine the comprehensive relationship matrix.

[0073] The interval intuitionistic fuzzy set sets the decision-making subject weights in the form of interval fuzzy numbers. Compared with the traditional method of using precise numerical weighting, this method increases the objectivity and authenticity of the decision. The defined component association strength is shown in the following table.

[0074] Table 1 Functional relationship

[0075] Correlation Description Association strength Interval value Together they complete a major function, neither of them can be missing Very strong [0.9,0.95],[0.02,0.05] Working together on a main function Strong [0.70,0.75],[0.20,0.35] Complete a certain auxiliary function together generally [0.50,0.55],[0.40,0.45] Working together on an auxiliary function Weaker [0.20,0.25],[0.70,0.75] Interact with different functions Very weak [0.02,0.05],[0.90,0.95] Completely independent of each other Does not exist [0,0],[1,1]

[0076] Table 2 Structural relationship

[0077] Correlation Description Association strength Interval value Cannot be disassembled, such as welding, bonding, casting, etc. Very strong [0.9,0.95],[0.02,0.05] Difficult to disassemble, such as connecting by heat pressing, cold shrinking, etc. Strong [0.70,0.75],[0.20,0.35] Difficult to disassemble, such as connected by rivets, keys, pins, etc. generally [0.50,0.55],[0.40,0.45] Easier to disassemble, such as connecting by threads Weaker [0.20,0.25],[0.70,0.75] Easy to disassemble, such as connecting by clamping, magnetic suction, etc. Very weak [0.02,0.05],[0.90,0.95] No connection Does not exist [0,0],[1,1]

[0078] The corresponding interval value is determined according to the relationship strength between the two components. Let A = ([a, b], [c, d]) be an interval fuzzy number, ([a, b], [c, d]) correspond to the position values ​​of the interval values ​​in the above table, and S (α) be the score function value. Based on the score function formula (1), the corresponding association strength weight can be calculated.

[0079]

[0080] The functional relationship matrix and the structural relationship matrix can be calculated from the above formulas respectively, and the comprehensive relationship matrix can be obtained by merging these two matrices.

[0081] Step S12: Taking the high-speed train bogie as the application object, modeling analysis is carried out, the structure of the bogie is appropriately simplified, and its main functions are divided into five major functions: load-bearing, traction, buffering, guiding, and braking. According to the main components corresponding to the main functions, a total of 32 parts of the bogie are determined, as shown in Table 3.

[0082] Table 3 Part Number

[0083] Components Node Number Components Node Number Gearbox <![CDATA[M1]]> Bearings <![CDATA[M 17 ]]> Axle box positioning node <![CDATA[M2]]> Axle box <![CDATA[M 18 ]]> Traction rod <![CDATA[M3]]> Side beam <![CDATA[M 19 ]]> Axle box spring <![CDATA[M4]]> beam <![CDATA[M 20 ]]> Coupling <![CDATA[M5]]> Longitudinal connecting beam <![CDATA[M 21 ]]> A series of vertical shock absorbers <![CDATA[M6]]> Motor mount <![CDATA[M 22 ]]> Air Springs <![CDATA[M7]]> Gearbox hanger <![CDATA[M 23 ]]> Anti-roll torsion bar <![CDATA[M8]]> Anti-snaking shock absorber mount <![CDATA[M 24 ]]> Anti-snaking shock absorbers <![CDATA[M9]]> Axle box positioning node mounting seat <![CDATA[M 25 ]]> Brake cylinder <![CDATA[M 10 ]]> Vertical shock absorber mounting seat <![CDATA[M 26 ]]> Secondary lateral shock absorber <![CDATA[M 11 ]]> Empty spring mount <![CDATA[M 27 ]]> Traction motor <![CDATA[M 12 ]]> Foundation brake mount <![CDATA[M 28 ]]> wheel <![CDATA[M 13 ]]> Secondary lateral shock absorber mounting seat <![CDATA[M 29 ]]> Axles <![CDATA[M 14 ]]> Traction rod node <![CDATA[M 30 ]]> Brake force amplifier <![CDATA[M 15 ]]> Brake pad composition <![CDATA[M 31 ]]> Shaft mounted brake disc <![CDATA[M 16 ]]> Horizontal stop <![CDATA[M 32 ]]>

[0084] The method in step S11 is used to output the obtained comprehensive relationship matrix to the complex network visualization software gephi to obtain the relationship network between the components as follows: Figure 1 As shown, Figure 1The thickness of the middle edge represents its weight.

[0085] Step S2: Based on the high-speed train bogie complex network established in the previous step, the ant colony algorithm is used in combination with the complex network node capacity to search for the change propagation path, with the goal of minimizing the change propagation intensity.

[0086] S2 includes the following steps:

[0087] Step S21: The change cost is defined as the additional cost of changing a node in the network. Therefore, the additional cost of changing a node can be expressed as:

[0088]

[0089] In the formula, n represents the total number of nodes, C(V i ) is the node V i The cost of the change;

[0090] Change time is the time required from the initiation of the change request to the completion of all change designs. It can be expressed as:

[0091]

[0092] In the formula, n represents the total number of nodes, T(v i ) is the node v i Time required for the change.

[0093] The change propagation probability refers to the probability of node i propagating to node j. When there is no connecting edge between nodes i and j, the propagation probability is zero. The larger the propagation probability, the more likely it is that the propagation will pass through the edge. This paper calculates the propagation probability based on the weights between edges, which can be expressed as:

[0094]

[0095] In the formula, w(v i ) is the node v i The sum of the weights of the connected edges, w ij is the weight between node i and node j.

[0096] The change propagation index is used to determine the properties of the node, which can be classified as absorber / carrier / diffuser. During the propagation process, the diffusion node should be avoided as much as possible. The change propagation index can be expressed as:

[0097]

[0098] Where D out (i) represents the out-degree of the node, D in (i) represents the in-degree of the node.

[0099] The calculation of the change propagation intensity is based on the above four indicators. The change cost, change time, and change propagation index are preferably as small as possible, while the change propagation probability is preferably as large as possible. In summary, the change propagation intensity is defined as:

[0100] CPI ij =(w1C c +w2T c +w3D i )(1 - p ij )(6)

[0101] In the formula, w1 + w2 + w3 = 1, which respectively represent the weights of each indicator and can be calculated by the entropy weight method.

[0102] Step S22: During the design process, designers will reserve a certain degree of design tolerance. The design tolerance is defined as the node capacity of its complex network. The larger the capacity of a node, the stronger its risk resistance ability. The node capacity can be defined as follows:

[0103] ρ i =(1 + β)L0(v i )(7)

[0104] In the formula, β > 0 is the tolerance parameter of the network, indicating the tolerance size of the design parameter. The larger β is, the stronger its risk resistance ability, but when it is too large, it will cause the product performance to decline. Therefore, the parameter should be within the specified range. L0(v i ) represents the initial load of the node and can be obtained by calculating the degree of the node. The calculation formula is as follows:

[0105]

[0106] In the formula, k i 、k j represent the degrees of the component nodes i and j, q is a proportional constant, and in this paper, q takes the value of 0.5. a(0 < a < 1) is the control parameter of the initial load of the node. The larger a is, the greater the difference in the initial load between nodes. S i represents the set of neighbor nodes of node i.

[0107] Based on the above definitions of the change propagation intensity and node capacity, the objective function and constraints for searching the optimal change propagation path are defined as follows:

[0108]

[0109] In the formula, k is the number of modules for change propagation, and ρ0 represents the change impact of the initial change node, which can be given by domain experts. Equation (9) represents the objective of minimizing the change propagation intensity, and equation (10) represents that when the absorbable capacity of the node is greater than the change impact of the initial node, the propagation stops.

[0110] The ant colony algorithm is used to search for the change propagation path based on the above-defined goals and constraints. The flowchart of the algorithm is as follows: Figure 2 shown.

[0111] It is known that the original supplier of the wheel at node 13 cannot normally supply the original configuration module due to some external uncertainty. Since there is no other alternative supplier for the original configuration module of the wheel, the company considers making design changes to the wheel and then purchasing from an optional supplier. The propagation path obtained by solving the product design change path with node 13 as the initial change node is as follows: Figure 3 shown.

[0112] Step S3: Based on the above search for the change propagation path, the module that needs to be changed can be obtained. The module that needs to be changed is then used to solve the optimal product design change plan and supplier selection.

[0113] S3 includes the following steps:

[0114] Step S31: The change plan is solved by using the master-slave optimization method. The optimization target of the upper layer is the established comprehensive utility function, and the target is to maximize the utility value. The lower layer is the product of the supplier's total cost and risk value, and the target is to minimize the product value.

[0115] Customer demand matching refers to the process when there is a new customer order requirement. Professional designers need to combine mapping rules, configuration rules, and design experience to convert customer requirements into configuration parameter values ​​for each module to be configured. It can be calculated by the weighted Euclidean distance from the target vector.

[0116] After normalizing the configuration item parameters, the weighted Euclidean distance method is used to calculate the module instance M ij The similarity with the target vector (i.e., the degree of demand matching) is shown in the following formula.

[0117]

[0118] In the formula, n represents the number of modules, m represents the number of instances, and CP ij Configure the parameter item for the target, CP ijl Represents module M i The lth configuration parameter item of the jth instance, CP * ijl is the configuration parameter item after normalization, Rmd ij represents the requirement matching degree of the jth instance of the i-th module, w il Indicates the weight of the lth configuration parameter item of the i-th module.

[0119] Module similarity refers to the fact that the greater the similarity between the updated product configuration and the original configuration after the product change occurs, the more favorable it is for the enterprise to change the product. The similarity of the product configuration scheme is calculated based on the similarity between the optional configuration modules and the original configuration modules.

[0120] Customer demand matching and module similarity are part of the optimization objectives of the upper-level product configuration, which are included in the product utility function. The ultimate goal of the upper layer is to maximize the utility function, which is the comprehensive value of the customer demand matching and module similarity of the upper-level product configuration, the lower-level supplier cost, product completion time, and supplier risk.

[0121] The upper utility function is represented by U, as shown in (11), where C is the total cost of the supplier, Rs is the supplier risk value, rmd represents the customer demand matching degree, and sim is the similarity of the product module.

[0122]

[0123] The objective function of the upper layer can be expressed as follows:

[0124]

[0125] In the formula, x ij Represents the selected module instance, where i represents the module number, j represents the jth module instance of the i-th module, and similarly y ijs Indicates the supplier of the selected module instance. The definitions of i and j are the same as above. rmd ij Indicates the customer demand matching degree of the selected module instance, sim ij is the similarity of the product module of the selected module instance, s represents the sth supplier of the selected module instance, and x ij ,y ijs are binary decision variables that can only take the value of 0 or 1, indicating the selection or non-selection of the current module instance or supplier. The three parameters n, m, and S are the number of modules, module instances, and suppliers that can be selected, respectively. k is the number of modules to propagate changes, and w is the number of modules to propagate changes. il represents the weight of the lth configuration item parameter of the i-th module. Formula (13) defines the binary decision variable x ij can only be 0 or 1; Formulas (14)-(15) indicate that the module to be configured M i Only one module instance can be selected from the configurable space; formula (16) indicates that the sum of the weights between module configuration parameter items is 1.

[0126] The optimization model of the lower layer consists of three components: product completion time, supplier total cost, and supplier risk. The parameters of the lower layer are fed back to the upper layer to calculate the utility function of the upper layer.

[0127] This application defines the product completion time as the time required from the manufacturer sending an order request to the supplier to the completion of product production and assembly when a product change occurs. Product completion time includes development time and logistics time. Development time is defined as the time required for a product from design, production, to final assembly, which can be calculated from the supplier's historical experience data and expressed as T P Indicates. Logistics time: T T =D / v, D represents the transportation distance, v represents the transportation speed. In summary, the product completion time can be expressed as:

[0128] T(x ij ,y ijs )=T P +T T (17)

[0129]

[0130] In the formula, n, m, and S represent the number of modules, the number of module instances, and the number of suppliers, respectively. represents the development time of the supplier selected for the jth module instance of the i-th module, Represents the logistics time of the supplier selected by the j-th module instance of the i-th module. The other parameter definitions are the same as those in the utility function above.

[0131] The total cost of the supplier can be calculated based on the completion time of the product. The total cost includes research and development cost, procurement cost, and logistics cost. The research and development cost mainly includes product development and management costs. The expression of the supplier cost of the lower-level optimization target is:

[0132] C s (x ij ,y ijs )=C P +C M +C T (twenty two)

[0133] C P =C UP ·T P (twenty three)

[0134]

[0135] C T =C UT ·T T (26)

[0136]

[0137] In the formula, χ ijRepresents the selected module instance, where i represents the module number, j represents the jth module instance of the i-th module, and similarly y ijs Indicates the supplier of the selected module instance. i and j are defined as above. n, m, and S represent the number of modules, the number of module instances, and the number of suppliers, respectively. C s is the total cost, C P is the development cost, C M is the purchase cost, C T For logistics costs, represents the procurement cost of the supplier selected by the jth instance of the i-th module, C UP and C UT Represent the supplier's unit time development cost and unit time logistics cost respectively. and They respectively represent the unit time development cost and unit time logistics cost of the supplier selected for the j-th instance of the i-th module.

[0138] Step S32: By establishing a Bayesian network of supplier risk indicators, the risk level of the supplier needs to be evaluated based on Bayesian formula reasoning.

[0139] The determination of supplier risk indicators is based on supplier risk assessment in relevant literature and is determined as technology, quality, delivery, cost, and service, a total of five first-level indicators. Each first-level indicator has a corresponding second-level indicator, and a total of 11 second-level indicators are obtained, as shown in Table 4.

[0140] Table 4 Supplier risk indicators

[0141]

[0142] The Bayesian network is formed on the basis of traditional Bayesian theory combined with graph theory, which can express complex systems in an intuitive model form. The graph model consists of network nodes and directed edges, such as Figure 4 The figure shows a schematic diagram of a Bayesian network, in which D1 and D2 are the prior probabilities of the root node, D3 is the conditional probability of the intermediate node, and D4 is the conditional probability of the leaf node.

[0143] Based on the prior probabilities of nodes D1 and D2 and the conditional probabilities of D3 and D4, the Bayesian formula can be used to infer the posterior probabilities of nodes D3 and D4. If node D4 is the final supplier risk, the obtained posterior probability is the risk value of the evaluated supplier.

[0144] The method of determining the conditional probability of the node is based on the expert scoring method, which divides the risk status of the node into high (H), medium (M), and low (L). The conditional probability is calculated by combining the AHP method to compare the scores.

[0145] The expert can get the score table shown in Table 5 by answering the question "Ignoring the influence of other indicators, when the D1 node is in the H state, how much is the probability of the D3 node being in the H state compared to the probability of being in the M and L states?"

[0146] Table 5 Conditional probability comparison matrix

[0147]

[0148] Table 6 Conditional probability matrix

[0149]

[0150] The AHP method is used to calculate the probability of node D3 being in the H, M and L states when D1 is in the H state, as shown in Table 6. Similarly, the conditional probability of D3 when D1 is in the M and L states can be calculated in turn. The same is true for node D2.

[0151] The above method can be used to calculate the influence of individual nodes. According to Kim and Pearl

[18] , when a node R in a Bayesian network has two parent nodes S and T, its probability conditional on S and T can be approximated by the following equation.

[0152] P(R|S,T)=μ×P(R|S)×P(R|T) (28)

[0153] Where μ is a normalization factor used to ensure the following:

[0154]

[0155] In the formula, r represents the state of node R.

[0156] The above formula can be used to calculate the comprehensive conditional probability of node D3, which can be expressed as Table 7.

[0157] Tab.6 conditional probability combination

[0158]

[0159]

[0160] by For example, this probability can be calculated using the following formula:

[0161]

[0162] In the formula, and It can be calculated based on the conditional probability comparison matrix in the previous article.

[0163] The above steps can be used to calculate the conditional probabilities of child nodes with multiple parent nodes. The comparison matrix scoring method reduces the complexity of the problem compared to the traditional scoring method based on node state combinations, and has advantages in calculating the conditional probabilities of child nodes with multiple states and multiple parent nodes.

[0164] For the prior probability determination of the root node, some secondary indicators such as the proportion of key equipment and the fluctuation of supply costs can be obtained through statistical data. Qualitative indicators such as technical level can be obtained through expert scoring.

[0165] Based on the above process, we can get the prior probability of all root nodes and the conditional probability of child nodes. Based on the Bayesian formula, we can calculate the final supplier risk. The calculation formula is as follows:

[0166]

[0167] In the formula, P(A|B) represents the possibility of event A occurring under the condition that event B occurs, that is, conditional probability, P(B) represents the prior probability, and P(B|A) represents the posterior probability.

[0168] The GeNle software is used to implement the modeling of the supplier risk Bayesian network. The final supplier risk can be calculated by inputting the prior probability and conditional probability of the node. The supplier risk Bayesian network established by taking technical indicators as an example is as follows Figure 5 shown.

[0169] From the perspective of risk-averse enterprises, such enterprises are more inclined to choose suppliers with low risk. Therefore, in the subsequent master-slave optimization model, the risk value of the supplier is the degree of high risk.

[0170] Step S33: using a nested genetic algorithm to reconfigure and solve the changed parts.

[0171] The master-slave optimization model established is actually a 0-1 nonlinear programming problem, and in this master-slave optimization model, all decision variables are discontinuous. The genetic algorithm can better handle such problems with large search space and discontinuous decision variables by using random search technology. Therefore, the genetic algorithm is used to solve the product design change plan combined with supplier selection. Product configuration and supplier selection are different entities, which is a problem with a hierarchical structure. Therefore, the master-slave optimization problem is solved by nesting genetic algorithms. The design process of the nested genetic algorithm is as follows: Figure 6 Shown

[0172] The detailed steps are as follows:

[0173] (1) Set the initial population, number of iterations, crossover probability, and mutation probability parameters for the upper and lower layers.

[0174] (2) Initialize the upper population, Figure 7 The upper chromosome encoding schematic shown adopts an integer encoding strategy. For example, the module to be configured M1 contains 3 optional physical configuration instances, then a configuration instance is randomly selected from the three, and so on, the corresponding physical configuration instances are selected from the remaining configurable modules to form a configuration scheme.

[0175] (3) Determine whether the individuals in the upper population meet the constraints. If so, the upper population is transferred to the supplier decision model in the lower layer; otherwise, the individual fitness value is set to 0.

[0176] (4) Initialize the lower population, such as Figure 8 As shown, each instance of the selected configuration solution includes one or more optional suppliers, such as the physical configuration instance M 13 It contains three optional suppliers, then one supplier is randomly selected from these three, and so on, to form a selection plan for the lower-level suppliers.

[0177] (5) Determine whether the population in the lower layer meets the constraints. If so, calculate the fitness value of the individuals in the population; otherwise, set the individual fitness value to 0, and then perform selection, crossover, and mutation operations.

[0178] (6) Determine whether the lower population has reached the maximum number of iterations. If so, the algorithm stops the operation and feeds back the optimal individual and optimal value to the individuals in the upper population to calculate the corresponding fitness value of the upper population; if not, continue to step 5.

[0179] (7) Determine whether the upper population has reached the maximum number of iterations. If the set number of iterations is reached, the algorithm stops the operation and outputs the optimal solution and value. Otherwise, perform selection, crossover, and mutation operations and execute step 3.

[0180] The nested genetic algorithm is coded using Python programming language. The specific parameters of the algorithm are set as follows: the maximum number of iterations of the upper and lower layers are 150 and 50 respectively, and the other parameters are set the same, the population size is 50, the crossover probability is 0.8, and the mutation probability is 0.03. The convergence effect diagram of the final algorithm is shown in Figure 9. The algorithm reached the convergence value at the 86th time.

[0181] In order to analyze the joint impact of the master-slave optimization method on product configuration and suppliers, the optimization results of this paper are compared with the normalized optimization method. The normalized method considers product configuration and suppliers in the same dimension and does not optimize the supplier layer separately. Fig.10The comparison of the results shows that the master-slave optimization method is superior to the normalized method in terms of utility value, cost and time.

[0182] In the context of globalization, complex product design changes and supply chain disruption risks are unavoidable, and manufacturing companies hope to consider supplier disruption risks in the early stages of design changes. Based on this, the present invention proposes a design change model and a solution method that combines complex networks and Bayesian networks. The overall idea is to first construct a complex network of product structure to determine the optimal change path, then evaluate and predict supplier risk indicators, and finally reconfigure the product design solution based on the optimal design change path. Taking the design change problem of a certain type of high-speed train bogie as an example, the case analysis results show that the solution results that consider supplier risks and use the master-slave optimization method are better than the solution results that only consider cost or normalization methods.

[0183] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described in the present invention. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the above specific implementation methods. Therefore, any modification or equivalent replacement of the present invention; and all technical solutions and improvements thereof that do not depart from the spirit and scope of the invention are included in the scope of the claims of the present invention.

Claims

1. A method for dealing with complex design changes under supplier interruption risk, the method comprising the following steps: Step S1: Based on the complex network component association relationship representation, the relationship network between components is constructed by the functional relationship and structural relationship of the product; Step S2: Using the ant colony algorithm combined with the capacity of complex network nodes to search for the change propagation path, with the goal of minimizing the change propagation intensity; Step S3: Based on the search for the change propagation path, the modules that need to be changed are obtained, and the product design change plan and supplier selection are solved using the modules that need to be changed.

2. The method for coping with complex design changes under supplier interruption risk according to claim 1, characterized in that: The method of constructing a relationship network between parts through the functional relationship and structural relationship of the product includes: determining the functional relationship matrix and structural relationship matrix of the parts based on interval intuitionistic fuzzy sets, determining the weights of the functions and structures respectively by using the network difference method, and determining the comprehensive relationship matrix.

3. The method for coping with complex design changes under supplier interruption risk according to claim 1, characterized in that: Step S2 specifically includes the following steps: Step S21: The change cost is defined as the additional cost of changing a node in the network. The additional cost of changing a node is expressed as: In the formula, n represents the total number of nodes, C(v i ) is the node V i The cost of the change; Change time is the time required from the initiation of the change request to the completion of all change designs; it is expressed as: In the formula, n represents the total number of nodes, T(v i ) is the node V i The time required for the change; The change propagation probability refers to the probability of node i propagating to node j. When there is no connecting edge between nodes i and j, the propagation probability is zero. The propagation probability is calculated based on the weights between the edges and is expressed as: In the formula, w(v i ) is the node V i The sum of the weights of the connected edges, W ij is the weight between node i and node j; The change propagation index is used to determine the properties of nodes, which are classified as absorbers / carriers / diffusers. The change propagation index is expressed as: Where D out (i) represents the out-degree of the node, D in (i) represents the in-degree of the node; The calculation of change propagation intensity is based on the above four indicators. The smaller the change cost, change time and change propagation index, the better, while the larger the change propagation probability, the better. Change propagation intensity is defined as: <h2 style=";text-align:left;direction:ltr">CPI<h2 style=";text-align:left;direction:ltr"> ij <h2 style=";text-align:left;direction:ltr"> (w1C<h2 style=";text-align:left;direction:ltr"> c <h2 style=";text-align:left;direction:ltr"> +w2T<h2 style=";text-align:left;direction:ltr"> c <h2 style=";text-align:left;direction:ltr"> +w3D<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> )(1-p<h2 style=";text-align:left;direction:ltr"> ij <h2 style=";text-align:left;direction:ltr"> ) In the formula, W1+W2+W3=1, which represents the weight of each indicator, calculated by the entropy weight method; Step S22: During the design process, the designer will reserve a certain degree of design tolerance, which is defined as the node capacity of the complex network. The larger the capacity of a node, the stronger its ability to resist risks. The node capacity is defined as follows: r i =(1+β)L0(v i ) In the formula, β>0 is the tolerance parameter of the network, which indicates the tolerance of the design parameters. The larger β is, the stronger its ability to resist risks is. i ) represents the initial load of the node, which is obtained by calculating the degree of the node. The calculation formula is as follows: where K i , K j represent the degrees of component nodes i and j, q is a proportionality constant, taken as 0.5 in this paper, a (0 < a < 1) is a control parameter for the initial load of nodes, and the larger a is, the greater the difference in the initial load between nodes; S i denotes the set of neighbor nodes of node i; Based on the above definitions of change propagation intensity and node capacity, the objective function and constraints of change propagation path search are defined as follows: Where k is the number of modules for change propagation, and ρ0 represents the change impact of the initial change node; Formula Indicates that the goal is to minimize the change propagation intensity. The formula It means that when the absorption capacity of a node is greater than the change impact of the initial node, the propagation stops.

4. The method for coping with complex design changes under supplier interruption risk according to claim 1, characterized in that: Step S3 specifically includes the following steps: Step S31: using the master-slave optimization method to solve the change plan, the upper layer optimization target is the established comprehensive utility function, and the target is to maximize the utility value; Convert customer requirements into configuration parameter values ​​for each module to be configured; calculated by weighted Euclidean distance from the target vector; After normalizing the configuration item parameters, the weighted Euclidean distance method is used to calculate the module instance M ij The similarity with the target vector is as follows: In the formula, n represents the number of modules, m represents the number of instances, and CP ii Configure the parameter item for the target, CP iil Represents module M i The lth configuration parameter item of the jth instance of is the configuration parameter item after normalization, Rmd ij represents the requirement matching degree of the jth instance of the i-th module, W il Indicates the weight of the lth configuration parameter item of the i-th module; The upper utility function is represented by U, Where C is the total cost of the supplier, R s is the supplier risk value, rmd represents the customer demand matching degree, and sim is the similarity of product modules; The objective function of the upper layer is expressed as follows: s.tx ij ∈{0,1},i=1,2,...n;j=1,2...,m,s=1,2...,S In the formula, χ ij Represents the selected module instance, where i represents the module number, j represents the jth module instance of the i-th module, and similarly y ijs Indicates the supplier of the selected module instance. The definitions of i and j are the same as above. rmd ij Indicates the customer demand matching degree of the selected module instance, sim ij is the similarity of the product module of the selected module instance, S represents the Sth supplier of the selected module instance, and χ ij ,y ijs are binary decision variables that can only take the value of 0 or 1, indicating the selection or non-selection of the current module instance or supplier. The three parameters n, m, and S are the number of modules, module instances, and supplier selections, respectively. k is the number of modules to be propagated, and w is the number of modules to be propagated. il Represents the weight of the lth configuration item parameter of the i-th module; Step S32: By establishing a Bayesian network of supplier risk indicators, the risk level of the supplier needs to be evaluated based on Bayesian formula reasoning; Step S33: using a nested genetic algorithm to reconfigure and solve the changed parts.

5. The method for coping with complex design changes under supplier interruption risk according to claim 4, characterized in that: The optimization model of the lower layer consists of three components: product completion time, supplier total cost, and supplier risk. The parameters of the lower layer are fed back to the upper layer to calculate the utility function of the upper layer.

6. The method for coping with complex design changes under supplier interruption risk according to claim 5, characterized in that: Product completion time includes development time and logistics time. Development time is defined as the time required for a product to go from design, production, to final assembly. It is calculated based on the supplier's historical experience data and is expressed in T P Indicates that logistics time: T T =D / v, D represents the transportation distance, v represents the transportation speed; the product completion time is expressed as: T(x ij ,y ijs )=T P +T T In the formula, χ ij Represents the selected module instance, where i represents the module number, j represents the jth module instance of the i-th module, and similarly y ijs Indicates the supplier of the selected module instance. i and j are defined as above. n, m, and S represent the number of modules, the number of module instances, and the number of suppliers, respectively. represents the development time of the supplier selected for the jth module instance of the i-th module, Represents the logistics time of the supplier selected by the j-th module instance of the i-th module.

7. The method for coping with complex design changes under supplier interruption risk according to claim 6, characterized in that: The total cost of the supplier is calculated based on the completion time of the product. The total cost includes research and development cost, procurement cost, and logistics cost. The research and development cost mainly includes product research and development, management and other costs. The expression of the lower-level optimization target supplier cost is: C s (x ij ,y ijs )=C P +C M +C T C P =C UP .T P C T =C UT .T T In the formula, χ ij Represents the selected module instance, where i represents the module number, j represents the jth module instance of the i-th module, and similarly y ijs Indicates the supplier of the selected module instance. i and j are defined as above. n, m, and S represent the number of modules, the number of module instances, and the number of suppliers, respectively. C s is the total cost, C P is the development cost, C M is the purchase cost, C T For logistics costs, represents the procurement cost of the supplier selected by the jth instance of the i-th module, C UP and C UT Represent the supplier's unit time development cost and unit time logistics cost respectively. and They respectively represent the unit time development cost and unit time logistics cost of the supplier selected for the j-th instance of the i-th module.

8. A computer storage medium storing a program executable by a processor, wherein the program executable by the processor is used to implement the method for coping with complex design changes under the risk of supplier interruption as described in any one of claims 1 to 7 when executed by the processor.

Images