Complex product customization design method based on demand uncertainty

Abstract

The invention relates to the field of product customization design, in particular to a complex product customization design method based on demand uncertainty. The method comprises the following steps: modeling a customization design process by utilizing a system engineering theory to obtain a customization design model; mining demand information and demand preferences of the customer from the first data of the complex product; using a cloud model to represent the uncertainty of the demand preference as a data interval, and converting the uncertainty into a demand weight based on the data interval; obtaining first modules of the complex product, constructing a mapping matrix based on a coupling relationship between the demand information and the first modules, and embedding the demand weight into the mapping matrix to obtain a relative weight between the first modules; based on a reinforcement learning method, constructing a customized design mathematical model in combination with the customized design model; constructing a problem space based on the demand information, and constructing a solution space based on the first module and the relative weight; and under the constraint of the solving space, solving the problem space through the customized design mathematical model to obtain a customized design scheme.

Description

Technical Field

[0001] The present invention relates to the field of product customized design, and particularly to a method for customized design of complex products based on demand uncertainty. Background Art

[0002] The manufacturing industry is transforming towards mass customization and personalized production models. Such models emphasize product customized design according to specific customer needs to meet diverse customer demands. At the same time, the rapid development of artificial technologies has introduced revolutionary innovations into product design, improving design efficiency and accuracy. However, despite the theoretical advantages of mass customization and personalized production models, in actual operation, there are still challenges in efficiently integrating existing component resources of enterprises and quickly completing product design.

[0003] In the process of customized design of complex products, main attention is paid to aspects such as data consistency, demand preference acquisition methods, management and utilization of complex relationships within the product, etc., mainly including three stages: the requirement analysis stage, the conceptual design stage, and the detailed design stage. The requirement analysis stage and the conceptual design stage are key links in customized design. Existing technologies have made certain progress in dealing with customer demand preferences and complex relationships within the product architecture. However, problems such as dynamic acquisition of customer demand preferences and dynamic changes in the intensity of complex relationships within the product architecture still exist, and more perfect methods are needed to improve design quality and customer satisfaction. Therefore, how to handle rapidly changing solution scenarios and uncertainties of design factors in the process of complex product customized design remains an urgent problem to be solved. Summary of the Invention

[0004] The purpose of the present invention is: aiming at the uncertainty of rapidly changing customized requirements of customers in the current process of complex product customized design, through the embodiments of the present application, a method for customized design of complex products based on demand uncertainty is provided, which realizes quantitative analysis of the uncertainty of customized requirements and is beneficial to solving the best customized design scheme.

[0005] In order to achieve the above-mentioned invention purpose, the present invention provides the following technical solutions: A method for customized design of complex products based on demand uncertainty, comprising: Decomposing and abstracting the customized design process of complex products by using model-based systems engineering theory, and modeling the customized design process by using a graphical modeling language to obtain a customized design model; Obtaining first data of the complex product, and mining customer demand information and demand preferences from the first data; using a cloud model to represent the uncertainty of the demand preferences as a data interval, and converting the uncertainty into a demand weight based on the data interval; the demand information and the demand weight are correlated with each other; Obtain the first module in the complex product, analyze the coupling relationship between the requirement information and the first module, construct a first mapping matrix based on the coupling relationship, embed the requirement weights into the first mapping matrix, and obtain the relative weights between the first modules. Based on the reinforcement learning method, construct a customized design mathematical model in combination with the customized design model; construct a problem space based on the requirement information, and construct a solution space based on the first module and the relative weights; under the constraints of the solution space, solve the problem space through the customized design mathematical model to obtain a customized design solution for the complex product.

[0006] As a preferred technical solution of the present application, the requirement information includes objective requirement data and subjective requirement data: the method for mining the customer's requirement information and requirement preferences from the first data includes: obtaining the index values in the objective requirement data through natural language processing technology; setting evaluation grades for the subjective requirement data through expert evaluation.

[0007] As a preferred technical solution of the present application, the method for using the cloud model to represent the uncertainty of the requirement preferences as a data interval includes: using the reverse cloud algorithm to convert the data of the requirement information into a cloud model, and extracting the characteristic parameters of the cloud model; establishing the upper bound calculation formula and the lower bound calculation formula of the data interval through the 3En rule of the cloud model, and substituting the characteristic parameters for calculation to obtain the data interval.

[0008] As a preferred technical solution of the present application, the method for using the reverse cloud algorithm to convert the data of the requirement information into a cloud model includes: using the reverse cloud algorithm to convert the index values into an objective cloud model, and using the reverse cloud algorithm to convert the evaluation grades into a subjective cloud model; fusing the objective cloud model and the subjective cloud model to obtain a multi-level evaluation cloud model.

[0009] As a preferred technical solution of the present application, before fusing the objective cloud model and the subjective cloud model, it further includes: selecting experts in the field of complex product design, calculating the expert weights between experts according to the evaluation criteria, and embedding the expert weights into the subjective cloud model.

[0010] As a preferred technical solution of the present application, the method for converting the uncertainty into requirement weights based on the data interval includes: calculating the subjective requirement weights according to the median of the data interval; calculating the objective requirement weights according to the width of the data interval; fusing the subjective requirement weights and the objective requirement weights into requirement weights.

[0011] As a preferred technical solution of the present application, before constructing the first mapping matrix based on the coupling relationship, it further includes screening the first module, and the method includes: Record the coupling relationship between the requirement information and the first module as the first coupling relationship, and construct a requirement-module association matrix based on the first coupling relationship; set a first strength threshold, and record the process of screening the requirement-module association matrix with the first strength threshold as the first screening rule; Record the coupling relationship between the first modules as the second coupling relationship, and construct a module association matrix based on the second coupling relationship; set a second strength threshold; record the process of screening the module association matrix with the second strength threshold as the second screening rule; Record the first modules that simultaneously satisfy the first screening rule and the second screening rule as the modules to be configured.

[0012] As a preferred technical solution of the present application, the method for obtaining the relative weights between the first modules includes: constructing a first mapping matrix for the coupling relationship between the requirement information and the modules to be configured based on the DSM structure design matrix; embedding the requirement weights into the first mapping matrix to realize the mapping of customer requirement preferences into the relative weights between the modules to be configured.

[0013] As a preferred technical solution of the present application, the method for constructing a requirement-module association matrix based on the first coupling relationship includes: using an association rule mining algorithm to perform data mining on the existing customized solution instance library, obtaining the requirement information and the first modules that are frequently associated in the existing customized solutions, obtaining a frequent item set, and deriving the first coupling relationship into a requirement-module association matrix based on the frequent item set.

[0014] As a preferred technical solution of the present application, the constraints of the solution space further include customized design rules; the method for obtaining the customized design rules includes: constructing a customized rule model according to the customized design rules between the first modules; inputting the frequent item set into the customized rule model, and outputting the customized design rules corresponding to the customized design solution.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. The present invention combines the MBSE theory and the SysML modeling language, analyzes the customized requirements of complex products, constructs a customized design requirement model using requirement diagrams and use case diagrams, and defines the problem space of complex product customized design; analyzes the functions and modules of the product system, constructs a customized design product family model using activity diagrams, use case diagrams and module definition diagrams, and defines the design space and solution space of complex product customized design; analyzes the associations between modules, the associations between requirements and modules, and the customized design rules, constructs a customized design knowledge model using internal module diagrams and parameter diagrams to support designers in completing corresponding design tasks, and transforms the customized design problem of complex products into a more understandable and processable graphical method, which can help designers better understand the requirements and constraints of the system during the system design stage.

[0016] 2. Aiming at the problem that it is difficult to ensure the information integrity, consistency and traceability in the traditional document-based design process, the key characteristics and behaviors of the customized design process are decomposed and abstracted through the MBSE theory and the SysML modeling language, improving the adaptability and data consistency of the design model. At the same time, through the analysis of multiple design models, the problem space and solution space are defined, so that the best design solution can be selected through the solution model.

[0017] 3. With the continuous improvement of the customization level of complex products, the dominant position of customer customized requirements in the design process becomes more and more prominent. Under the influence of various factors, customer requirements show characteristics such as fuzziness and dynamics. In addition, different users' demand preferences for products at different times also show differences, further increasing the complexity and variability of customer requirement information. The present invention conducts an analysis of customized design requirement preferences based on the cloud model, combines the subjective requirement weight and the objective requirement weight of all requirement information into the final customer requirement weight, associates the customer requirement weight with the corresponding requirement information, and completes the analysis of customized design requirement preferences for complex products under the influence of the uncertainty of requirement information, which can capture and analyze customer requirement preferences more accurately and improve customer satisfaction and design quality.

[0018] 4. The conceptual design of complex products is a process in which the design thinking diverges and then converges, and the design solution becomes clear from fuzzy. The research focus of conceptual design lies in the product architecture. A variety of ideas and methods are integrated to construct a product function architecture and a logical architecture that meet the quality function requirements, and attention is paid to the nodes representing functions or structures in the product architecture. However, for problems such as a large number of modules and complex coupling relationships in the complex product architecture, it is still difficult to solve. This application proposes a method of mining the coupling relationship strength between requirement information and the first module through the FP-Growth algorithm, constructs a requirement-module association matrix, optimizes the decision-making process in the conceptual design stage, and improves the design quality.

[0019] 5. The customization solution for current complex products is mostly based on algorithm-based configuration solutions, mainly evolutionary algorithms and their improved algorithms. It incorporates comprehensive considerations of design elements in the solution objectives and processes of the algorithms, and also includes design variables and their possible impacts on the final solutions in the decision-making process of the algorithms, making the solution process faster. However, with the rapid changes in the solution scenarios and the influence of many difficult-to-quantify and potential design factors, the solution process of the design scheme must have the ability of dynamic adjustment. The present invention proposes a method for custom design solution using the Deep Q-Network (DQN) algorithm, which can quickly solve the optimal custom design scheme and improve the automation level and response speed of the design. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic flow diagram of the custom design process; Figure 2 It is a schematic structural diagram of the custom design model; Figure 3 It is a schematic flow diagram of the requirement preference analysis based on the cloud model; Figure 4 It is a schematic flow diagram of the requirement-module mapping based on FP-Growth; Figure 5 It is a schematic flow diagram of the custom design solution based on DQN. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0022] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely represents some embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of the present invention.

[0023] It should be noted that, without conflict, the embodiments in the present invention and the features and technical solutions in the embodiments may be combined with each other.

[0024] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0025] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by terms such as "upper" and "lower" is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the inventive product is customarily placed during 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, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the present invention. In addition, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0026] Embodiment 1: Refer to Figures 1 - 5 As shown, the embodiment of the present application provides a complex product customization design method based on demand uncertainty to solve the problem that it is difficult to quantify the customer customization demand preferences. By using a multi-level evaluation cloud model to analyze the customization design requirements, the uncertainty of the customer customization requirements is quantified into the weights of the demand information.

[0027] The technical solution in the embodiment of the present application to solve the above problem that it is difficult to quantify the customer customization demand preferences is as follows: Use the model-based systems engineering theory to decompose and abstract the customization design process of complex products, and use a graphical modeling language to model the customization design process to obtain a customization design model.

[0028] In this embodiment, based on the model-based systems engineering theory MBSE, the top-down and object-oriented design ideas are adopted to gradually decompose and abstract the customization design process of complex products, extract key characteristics and behaviors, and obtain customization design objects such as customer customization requirements, module functions, behaviors, structures, and parameters. Then, an object-oriented modeling language is used to graphically express the entire customization design process of complex products and the customization design objects, and a customization design model is constructed. The customization design model includes a customization design process model, a customization design multi-view model, and a customization design solution model.

[0029] The model-based systems engineering MBSE includes nine types of diagram models: package diagram, requirement diagram, activity diagram, sequence diagram, state machine diagram, use case diagram, parameter diagram, module definition diagram, and internal module diagram. The object-oriented modeling language includes, but is not limited to, the Systems Modeling Language SysML and the Unified Modeling Language UML. The Systems Modeling Language SysML is a reuse and extension of the Unified Modeling Language UML, providing additional modeling constructs and extensions to meet the specific needs of the systems engineering field, emphasizing the modeling and analysis of the structure, behavior, and interaction of systems.

[0030] For example Figure 2As shown in the figure, in this embodiment, the system modeling language SysML, which is compatible with MBSE, is used to model the customization design process of the complex product. The steps include: S11: Establish a customization design process model based on the activity diagram. The customization design process model is a model that provides overall guidance for the entire customization design process, covering the dynamic knowledge in customization design, and is used to deeply associate and model the customization design process and related customization design data of complex products. The customization design process model represents the dynamic characteristics of the activity units executed sequentially in the customization design process through the control flow and decision-making process of the activity diagram, such as business processes or workflows. The customization design process model represents the data transfer process between activity units through the object flow of the activity diagram. The object flow represents the data input into the activity unit or the output data generated by the activity unit. Through the object flow, the activity diagram can clearly show the data dependency relationship and transfer process of data in each activity unit in the system.

[0031] S12: Establish a customization design requirement model based on the use case diagram and the requirement diagram. The customization design requirement model expresses information such as requirement objects, requirement attributes, and relationships between requirements required for the customization design of complex products, is used to express customer personalized customization requirements, and defines the problem space in the customization design process of complex products. The use case diagram can realize the modeling and management of stakeholders, and the requirement diagram can accurately define the requirement information of system stakeholders.

[0032] S13: Establish a customization design product family model based on the use case diagram and the module definition diagram. The customization design product family model is the core of the complex product customization design model, which defines both the customization design space of complex products and the solution space of customization design schemes. The use case diagram is used to realize the modeling of product functional requirements, and the module definition diagram is used to define the product structure hierarchy relationship, module information, and the association relationship between modules.

[0033] S14: Establish a customized design knowledge model based on the internal module diagram and the parameter diagram. For the requirement information, the complex coupling relationships between the first modules, and the association relationships formed by the complex structures among the first modules, they can be efficiently expressed and associated and managed through the parameter diagram and the internal module diagram. The customized design knowledge model is used to represent the requirement information - the first modules, the association relationships between the first modules, and the association strength, providing an important knowledge source for the subsequent customized design solution process. The customized design knowledge model includes an association relationship model and a customization rule model. The association relationship model is constructed based on the internal module diagram; the customization rule model is constructed based on the parameter diagram. Among them, the internal module diagram is used to clarify the association relationships between the first modules and the association relationships between the requirement information and the first modules. The parameter diagram is used to graphically describe the customization design rules between the first modules. The customization design rules include the constraint relationships of module attributes and / or module connections, usually in the form of production rules or functions.

[0034] S15: Establish a solution model for customized design based on the parameter diagram. Since the parameter diagram can efficiently integrate analysis models such as external solution algorithms, under the guidance of the customized design process model, the parameter diagram is used to construct a customized design solution model to ensure the accuracy and efficiency of the solution. Combining the data and knowledge in the customized design multi-view model, the process of solving the customized design solution is realized by introducing a reinforcement learning algorithm. The customized design multi-view model includes the customized design requirement model, the customized design product family model, and the customized design knowledge model.

[0035] The present invention combines the MBSE theory and the SysML modeling language, analyzes the customized requirements of complex products, constructs a customized design requirement model using a requirement diagram and a use case diagram, and defines the problem space of the customized design of complex products; analyzes the functions and modules of the product system, constructs a customized design product family model using an activity diagram, a use case diagram, and a module definition diagram, and defines the design space and solution space of the customized design of complex products; analyzes the associations between modules, the associations between requirements - modules, and the customized design rules, constructs a customized design knowledge model using an internal module diagram and a parameter diagram, which is used to support designers to complete the corresponding design tasks, and transforms the customized design problem of complex products into a more understandable and processable graphical way, which can help designers better understand the requirements and constraints of the system in the system design stage.

[0036] Aiming at the problem that it is difficult to ensure information integrity, consistency, and traceability in the traditional document-based design process, the key characteristics and behaviors of the customized design process are decomposed and abstracted through the MBSE theory and SysML modeling language, improving the adaptability and data consistency of the design model. At the same time, by analyzing multiple design models, the problem space and solution space are defined, so that the best design solution can be selected through the solution model.

[0037] After establishing the basic customized design model, it is further necessary to analyze the uncertainty of different customer demand information and demand preferences in the customized design process using the customized design model. The complex product customized design process under the influence of uncertainty in this application includes three parts: the customized design requirement analysis stage, the customized design conceptual design stage, and the customized design detailed design stage.

[0038] First is the intelligent customized design requirement analysis stage. In this embodiment, the uncertainty of demand preferences is analyzed by establishing a cloud model, and the uncertainty is quantified as the weight of demand information.

[0039] Obtain the first data related to the complex product, mine the customer's demand information and demand preferences from the first data; use the cloud model to represent the uncertainty of the demand preferences as a data interval, and convert the uncertainty into the demand weight of the demand information based on the data interval; the demand information and the demand weight are mutually associated.

[0040] Complex products have multiple stakeholders, and there are also contradictions and priorities among multiple stakeholders. For example, in the high-speed rail design process, one of the concerns of the operator is the construction cost, and one of the concerns of the passengers is the carriage environment. The better the carriage environment is decorated, the higher the construction cost. At the same time, the key points of different sources of customer requirements for product attributes vary. For example, due to cost pressure, the operator hopes that the number of seats is as many as possible under reasonable circumstances, while the passengers hope that the number of seats is as few as possible for a comfortable riding environment.

[0041] Aiming at the uncertainty of the demand information of complex products in the customized design process, this application uses the interval model in the convex model to describe it, converts the uncertainty into a data interval, and completes the analysis of the demand preferences of complex product customized design with the support of the customized design requirement model. For example Figure 3 as shown, the steps include: S21: Uncertainty analysis of the requirement information in the complex product customization design process. With the continuous change of customer requirement information, as well as the continuous progress and wide application of technologies such as big data and artificial intelligence, the customer's requirement preferences are also changing rapidly. At the same time, it has become easier to obtain and mine the customer's requirement information and requirement preferences. In this embodiment, the first data related to the complex product is obtained through two methods, online and offline. The first data includes online data and documents such as complex product design documents, design tender invitations, and online evaluations. The requirement information of the customer is mined from the first data through natural language processing technologies such as keyword matching. The requirement preference is an evaluation term related to the requirement information, such as an adverb of degree, a negative word, etc.

[0042] In this application, the customer's requirement information is divided into objective requirement data and subjective requirement data. The objective requirement data includes requirement information of technical indicators. In the prior art, there are a series of mature methods, such as natural language processing technologies like keyword recognition and matching, which can effectively extract the index values of the objective requirement data from the design tender invitation and / or design document, and then guide the parameter design of the first module.

[0043] The subjective requirement data includes the fuzzy requirement preferences of the customer for the complex product, which reflects the difference in the degree of attention of the customer to the product attributes and is the key factor determining customer satisfaction. For example, the comfort preference of passengers for high-speed rail seats. In order to accurately obtain the subjective requirement data, it can be obtained through individual interviews with customers and / or mined from a large amount of online requirement data. However, due to personal knowledge limitations, the ambiguity of language, and the complex internal correlation relationships in the product system, customers often face difficulties in describing their expectations for product attributes, resulting in the ambiguity and uncertainty of the obtained subjective requirement data, which hinders the final solution of the customization design scheme.

[0044] Since the objective requirement data has certainty and can be directly used to guide the customization design of complex products; while the subjective requirement data has uncertainty and is difficult to be directly used for solving the mathematical model of the customization design scheme in the customization design process; therefore, in this embodiment, different cloud models are used to mine the objective requirement data and the subjective requirement data respectively.

[0045] S22: Obtain customer personalized customization requirement data in two parts: objective and subjective. The objective requirement data is mined from online data and / or documents and stored in the enterprise PLM system. The subjective requirement data is mined through expert evaluation. The steps include: selecting experts in the field of complex products, calculating the expert weights among experts according to the evaluation criteria, and then having the experts give evaluation grades for each requirement information. Subsequently, both the obtained objective requirement data and the subjective requirement data are input into the customized design requirement model for data storage and management. The customized design requirement model transmits the objective requirement data and the subjective requirement data to the cloud model for further analysis.

[0046] The evaluation criteria set multiple dimensions, such as expert position, work experience, project experience, service time, and education level, to comprehensively and objectively evaluate the professional capabilities and experience of experts. Corresponding scores are set for each dimension, and then the comprehensive scores of experts are calculated based on these scores. The comprehensive scores are normalized to obtain the expert weights of each expert.

[0047] The setting of the evaluation grades is based on the concept of a multi-level cloud model, a method for dealing with uncertainty and ambiguity. First, determine the evaluation term set and the corresponding evaluation grades. For example, define a five-level evaluation term set, including "extremely poor", "poor", "average", "good", and "excellent", corresponding to the equivalent "0", "1", "2", "3", "4" respectively. Use the forward cloud generator algorithm to create multiple forward cloud models for the defined evaluation term set. Each forward cloud model corresponds to an evaluation term, and three characteristic parameters of the forward cloud model are defined: expectation Ex, entropy En, and hyperentropy He. Specific threshold ranges are set for each evaluation grade, and evaluation data of multiple experts for requirement information are obtained, and the specific evaluation data is mapped to the corresponding evaluation grades. For example, define the range of the expectation Ex of the forward cloud model, compare the expected value calculated based on the evaluation data of multiple experts with the range of the expectation Ex, and determine different evaluation grades.

[0048] S23: Use the reverse cloud algorithm to establish an objective cloud model for the objective requirement data and a subjective cloud model for the subjective requirement data respectively. The reverse cloud algorithm extracts the characteristic parameters required by the cloud model, including: expected value Ex, entropy En, and hyper-entropy He, through statistical analysis of the requirement data. In this embodiment, the multi-step reverse cloud transformation algorithm MBCT-SR is used to convert the index values of the obtained objective requirement data into an objective evaluation cloud model, and convert the evaluation grades obtained through expert evaluation into a subjective evaluation cloud model. After embedding the expert weights into the subjective evaluation cloud model, the objective evaluation cloud model and the subjective evaluation cloud model are fused to generate a final multi-level evaluation cloud model. Based on the expected value Ex, entropy En, and hyper-entropy He of the multi-level evaluation cloud model, upper and lower calculation formulas for the data interval are established through the 3En rule, and the uncertainty of each requirement information is expressed as a data interval.

[0049] S24: According to the characteristic parameters of the cloud model itself, obtain the data interval through the multi-level evaluation cloud model. Calculate the subjective requirement weight and the objective requirement weight of the customer requirement information respectively through the data values in the data interval. For example, for the subjective requirement weight, the data interval corresponding to the requirement information can be obtained, the first median of the data interval can be calculated, and the ratio of the first median to the sum of the medians of all data intervals can be calculated to obtain the subjective requirement weight; for the objective requirement weight, the data interval corresponding to the requirement information can be obtained, the difference between the upper bound and the lower bound of the data interval can be calculated to obtain the first width of the data interval, and the ratio of the first width to the sum of the widths of all data intervals can be calculated to obtain the objective requirement weight.

[0050] With the continuous improvement of the customization level of complex products, the dominant position of customer customization requirements in the design process has become increasingly prominent. Under the influence of various factors, customer requirements exhibit characteristics such as fuzziness and dynamics. In addition, different users' demand preferences for products at different times also show differences, further increasing the complexity and variability of customer requirement information. Through the analysis of customization design requirement preferences based on the cloud model, the present invention fuses the subjective requirement weight and the objective requirement weight of all requirement information into the final customer requirement weight, associates the customer requirement weight with the corresponding requirement information, and completes the analysis of customization design requirement preferences for complex products under the influence of the uncertainty of requirement information, can capture and analyze customer requirement preferences more accurately, and improve customer satisfaction and design quality.

[0051] Secondly is the customized design concept design stage. There are complex coupling relationships among multiple modules and different structures in complex product systems; the same requirement information may be associated with multiple modules, and the same module may also be associated with multiple requirement information. These intricate coupling relationships will increase the difficulty of customized design, increase the risk of unpredictable changes in the system, and thus lead to the instability of the product system.

[0052] For example Figure 4 As shown, obtain the first module in the complex product, analyze the coupling relationship between the requirement information and the first module, construct a first mapping matrix based on the coupling relationship, and embed the requirement weights into the first mapping matrix to obtain the relative weights between the first modules.

[0053] S31: Analyze the coupling relationship between the complex product requirements and the first module, and the coupling relationship between the first modules. There are close connections among multiple first modules and in-depth integration among multi-disciplinary knowledge within the complex product system. For example, there are interactive and associated coupling relationships among thousands of components in the structure of the complex product; among the first modules, there are cross-overs in multiple disciplinary fields such as mechanical, hydraulic, electrical, and control, forming intricate coupling relationships; during the R & D process, the design tasks among multiple departments and disciplines are intertwined, making the coupling relationship among the design tasks of the first module extremely complex; at the same time, the design and simulation links are carried out alternately, and the transfer process between design data and simulation verification data also presents intricate coupling relationships. Therefore, in the process of customized design of complex products, there will be multi-domain and multi-granularity complex coupling relationships between the structure and knowledge within the product.

[0054] In this embodiment, these coupling relationships are extracted from existing requirement information instances and module instances, and are converted into corresponding matrices based on the strength of the coupling relationship to achieve the quantification of the coupling relationship.

[0055] S32: Denote the coupling relationship between the requirement information and the first module as the first coupling relationship, and construct a requirement-module association matrix based on the first coupling relationship. Use the association rule mining algorithm to perform data mining on the existing customized solution instance library, obtain the requirement information and the first module that frequently occur in the existing customized solutions, get the frequent item set, and export the first coupling relationship as a requirement-module association matrix based on the frequent item set. For example, in this embodiment, use the FP-Growth algorithm to analyze the customized solution instance library, traverse all customized solutions, count the occurrence times of the requirement information and the first module, sort them according to the times, establish an FP-Tree based on the sorted results, obtain the frequent item set through the FP-Tree and calculate the support degree and confidence degree, and then construct a requirement-module association matrix based on the frequent item set, where the matrix elements are the support degree or the confidence degree.

[0056] Input the frequent item set into the customized rule model, and output the customized design rule. For example, during the high-speed rail design process, the noise control inside the carriage, the application of sound insulation materials, and the application of shock absorption technology are a frequent item set. Therefore, the customized design rule may include: if it is necessary to control the noise level inside the carriage, then a sound insulation module and a shock absorption module are required.

[0057] Denote the coupling relationship between the first modules as the second coupling relationship, and construct a module association matrix based on the second coupling relationship. The module association matrix includes a module structure association matrix and a module function association matrix. Based on the analysis of complex products, the structural relationship and functional relationship between modules can be directly derived from the customized design knowledge model, and the module structure association matrix and the module function association matrix can be constructed correspondingly. For example, using the first structure of the module as the row label and column label of the matrix, the element of the matrix is the number of other structures coupled with the first structure; or the frequency of data interaction between other structures and the first structure; or 1 or 0, where 0 indicates that the data interaction direction between other structures and the first structure is unidirectional interaction, and 1 indicates bidirectional interaction. In this way, the corresponding module structure association matrix or module function association matrix can be obtained.

[0058] S33: Set a first strength threshold for the demand module association matrix, and set a second strength threshold for the module association matrix. The first strength threshold can be the minimum support threshold, minimum confidence threshold, lift, and leverage in the association rule mining algorithm, which are used to evaluate the strength and effectiveness of association rules and are set by experts according to specific application scenarios and dataset characteristics. The second strength threshold can be the minimum number of couplings of the smallest structure or module set by experts or the minimum data interaction frequency. Use the first strength threshold to screen the demand module association matrix, use the second strength threshold to screen the module association matrix, and obtain the first modules that meet both screening rules, denoted as modules to be configured. Derive non-customized modules from the customized design product family model, and combine them with the modules to be configured to form a solution space.

[0059] S34: Based on the DSM structure design matrix, construct a first mapping matrix for the coupling relationship between the requirement information and the to-be-configured modules. For example, use the requirement information and the to-be-configured modules as the rows and columns of the first mapping matrix respectively. The matrix elements can be 1 or 0, where 1 indicates that the requirement information is related to the to-be-configured module, and 0 indicates no relation; or they can be the number of data interactions between the requirement information and the to-be-configured modules. Embed the requirement weights obtained in the customized design requirement analysis stage into the first mapping matrix to realize the mapping of customer requirement preferences into the relative weights between the to-be-configured modules. Obtain the set of to-be-configured modules whose relative weights are greater than the preset weight threshold, denoted as the first set. Pass the relative weights and the first set into the customized design product family model for completing the customized design of complex products under the influence of uncertainty in the detailed design stage of customized design.

[0060] For the dynamic fuzzy uncertainty existing in the customized design of complex products, use the interval model in the convex model to describe it, and with the support of the customized design product family model and the customized design knowledge model, complete the mapping between the complex product requirement information and the first modules, and map the customer requirement preferences into the relative weights between the first modules.

[0061] The conceptual design of complex products is a process in which the design thinking diverges and then converges, and the design scheme becomes clear from being fuzzy. The research focus of conceptual design lies in the product architecture, integrating various ideas and methods to construct a product function architecture and a logical architecture that meet the quality function requirements, and paying attention to the nodes representing functions or structures in the product architecture. However, for problems such as a large number of modules and complex coupling relationships in the complex product architecture, it is still difficult to solve. This application proposes a method for mining the coupling relationship strength between the requirement information and the first modules through the FP-Growth algorithm, constructing a requirement-module association matrix, optimizing the decision-making process in the conceptual design stage, and improving the design quality.

[0062] Finally, it is the detailed design stage of customized design. For example Figure 5 As shown, based on the reinforcement learning method, construct a customized design mathematical model in combination with the customized design model; construct a problem space based on the requirement information, and construct a solution space based on the first modules and the relative weights; under the constraints of the solution space, solve the problem space through the customized design mathematical model to obtain the customized design scheme of the complex product.

[0063] The customized design mathematical model is established based on the DQN model in the reinforcement learning method. Using the parameter graph, a deep learning network is established for the complex constraint relationships between modules in the customized design of complex products to obtain the solution model of the scheme, which is used as an agent in reinforcement learning. The first set is derived from the customized design product family model, and the corresponding requirement information and instance data of the to-be-configured modules are used to construct a training set. Referring to the problem space defined by the customized design requirement model, customer customized requirements are randomly generated and the solution model of the scheme is trained. The relative weights in the customized design product family model, the first set, and the customized design rules in the customized design knowledge model are imported into the solution model of the scheme to solve the optimal customized design scheme, and the customized design of complex products under the influence of uncertainties is completed.

[0064] Most of the current customized solutions for complex products are based on algorithmic configuration solutions, mainly evolutionary algorithms and their improved algorithms. The comprehensive consideration of design elements is incorporated into the solution objective and process of the algorithm, and the design variables and their possible impacts on the final solution are also incorporated into the decision-making process of the algorithm, making the solution process faster. However, with the rapid change of the solution scenario and the influence of many difficult-to-quantify and potential design factors, the solution process of the design scheme must have the ability of dynamic adjustment. The present invention proposes a method for customized design solution using the deep Q-network (DQN) algorithm, which can quickly solve the optimal customized design scheme and improve the automation level and response speed of the design.

[0065] Embodiment 2: This embodiment provides a medium and a device for storing and executing the complex product customized design method.

[0066] A computer storage medium stores a program executable by a processor, and the program executable by the processor is used to implement the complex product customized design method when executed by the processor.

[0067] A complex product customized design device, the device comprising: a sensor for collecting various data related to the complex product customized design; at least one processor capable of processing the various data related to the complex product customized design collected by the sensor; at least one memory for storing the various data related to the complex product customized design collected by the sensor and at least one program; when the at least one program is executed by the at least one processor, enabling the at least one processor to implement the complex product customized design method, the processor and the memory can be connected through a bus or other means for storing at least one program; the memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely disposed relative to the processor, and these remote memories can be connected to the device through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and combinations thereof. When the at least one program is executed by the at least one processor, enabling the at least one processor to implement the complex product customized design method.

[0068] Those of ordinary skill in the art can understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or can be implemented as hardware, or can be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, and the computer-readable medium can include a computer storage medium (or a non-transitory medium) and a communication medium (or a transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules or other data. The computer storage medium includes but is not limited to RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, the communication medium generally includes computer-readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0069] The above embodiments are only used to illustrate the present invention and do not limit the technical solutions described in the present invention. Although the present specification has described the present invention in detail with reference to the above respective embodiments, the present invention is not limited to the above specific embodiments. Therefore, any modification or equivalent replacement of the present invention; and all technical solutions and their improvements that do not depart from the spirit and scope of the invention are covered within the scope of the claims of the present invention.

Claims

1. A complex product customization design method based on demand uncertainty, comprising the following steps: Decomposing and abstracting the customized design process of complex products by using the model-based system engineering theory, modeling the customized design process by using a graphical modeling language, and obtaining a customized design model; Acquire first data of the complex product, and mine customer demand information and demand preferences from the first data; Using a cloud model, the uncertainty of the demand preference is expressed as a data interval, and the uncertainty is converted into a demand weight based on the data interval; the demand information and the demand weight are associated with each other; Acquire a first module in the complex product, analyze the coupling relationship between the demand information and the first module, construct a first mapping matrix based on the coupling relationship, embed the demand weight into the first mapping matrix, and obtain the relative weights between the first modules; Based on the reinforcement learning method, a custom design mathematical model is constructed in combination with the custom design model; Constructing a problem space based on the demand information, and constructing a solution space based on the first module and the relative weight; Under the constraints of the solution space, the problem space is solved by the customized design mathematical model to obtain a customized design solution for a complex product.

2. A complex product customization design method based on demand uncertainty according to claim 1, characterized in that: The demand information includes objective demand data and subjective demand data: The method for mining the customer's demand information and demand preferences from the first data includes: obtaining the indicator values ​​in the objective demand data through natural language processing technology; and setting evaluation levels for the subjective demand data through expert evaluation.

3. A complex product customization design method based on demand uncertainty according to claim 2, characterized in that: The cloud model is used to express the uncertainty of the demand preference as a data interval, and the method includes: The data of the demand information is converted into a cloud model using a reverse cloud algorithm, and the characteristic parameters of the cloud model are extracted; the upper and lower calculation formulas of the data interval are established through the 3En rule of the cloud model, and after substituting the characteristic parameters into the calculation, the data interval is obtained.

4. A complex product customization design method based on demand uncertainty according to claim 3, characterized in that: The data of the demand information is converted into a cloud model using a reverse cloud algorithm, the method comprising: The index value is converted into an objective cloud model using the reverse cloud algorithm, and the evaluation level is converted into a subjective cloud model using the reverse cloud algorithm; the objective cloud model and the subjective cloud model are fused to obtain a multi-level evaluation cloud model.

5. A complex product customization design method based on demand uncertainty according to claim 4, characterized in that: Before fusing the objective cloud model with the subjective cloud model, the method further includes: Experts in the field of complex product design are selected, expert weights between experts are calculated according to evaluation criteria, and the expert weights are embedded in the subjective cloud model.

6. The method for complex product customization design based on demand uncertainty according to claim 1 is characterized in that: The method of converting the uncertainty into a demand weight based on the data interval includes: The subjective demand weight is calculated according to the median of the data interval; the objective demand weight is calculated according to the width of the data interval; and the subjective demand weight and the objective demand weight are merged into the demand weight.

7. The method for complex product customization design based on demand uncertainty according to claim 1 is characterized in that: Before constructing the first mapping matrix based on the coupling relationship, the first module is screened, and the method includes: Recording the coupling relationship between the demand information and the first module as a first coupling relationship, and constructing a demand module association matrix based on the first coupling relationship; setting a first strength threshold, and recording the process of screening the demand module association matrix by the first strength threshold as a first screening rule; Recording the coupling relationship between the first modules as a second coupling relationship, constructing a module association matrix based on the second coupling relationship; setting a second strength threshold; and recording the process of screening the module association matrix by the second strength threshold as a second screening rule; The first module that satisfies both the first screening rule and the second screening rule is recorded as a module to be configured.

8. A complex product customization design method based on demand uncertainty according to claim 7, characterized in that: The method for obtaining the relative weights between the first modules includes: Based on the DSM structure design matrix, a first mapping matrix is ​​constructed for the coupling relationship between the demand information and the modules to be configured; the demand weights are embedded in the first mapping matrix to realize the mapping of customer demand preferences into relative weights between the modules to be configured.

9. The method for complex product customization design based on demand uncertainty according to claim 7, characterized in that: The method of constructing a demand module association matrix based on the first coupling relationship includes: An association rule mining algorithm is used to perform data mining on an existing customization solution instance library to obtain the demand information and the first module that are frequently associated in the existing customization solutions, obtain frequent item sets, and derive the first coupling relationship as a demand module association matrix based on the frequent item sets.

10. A complex product customization design method based on demand uncertainty according to claim 9, characterized in that: The constraints of the solution space also include customized design rules; and the method for obtaining the customized design rules includes: A customized rule model is constructed according to the customized design rules between the first modules; the frequent item sets are input into the customized rule model, and the customized design rules corresponding to the customized design solution are output.

Images