Research On Product Model Evolutionary Design Based On Feature Functions

Creative design is one of the most important ploys which enterprises can choose faced on the intense competition in the market. The creative design method can provide for designers a broader line of thougnt, relaxed design environment, moreover, the design data which it often produces in the stage of concept design should be reused in the follow-up design phases. In response to these requests of the creative design method, the product model evolutionary design based on feature functions was proposed.The evolutionary design system based on genome, population, fitness function, evolutionary operation was established by the thoughts of backward design, case-based design, etc., and with the traits that the evolutionary mechanism does not rely on environment knowledge and logic thinking. It can stimulate designers'image thinking and inspiration thinking to widen their creative thoughts, and it can reduce design burden by automatically completing the design process, in addition, the model image features based on the geometry API (Application Program Interface) in CAD can be reused in the detail design stage.After introducing the background, objects and tasks, activities and thoughts of modern design technology, the product model-oriented evolutionary design method was researched.1) The concepts and methods such as product structure, product function, product genome, population, fitness function, plans evaluation, evolutionary operation, etc. were firstly discussed. In fact, the genome as another form of product model is composed of image components; the population is a set of similar product instances and a knowledge repository containing design logic laws; the fitness function is a kind of quantitative model of design plans evaluation; the evolutionary operation is a concrete reasoning mechanism which the evolutionary design adopted. Then the recent research development on the four aspects of modern design, intilligent design, evolutionary design were summarized.2) The product genome expressed by API as product model image feature function codes including the information of structure and function was proposed at the basis of product structure modelling and function modelling. The geometry API and topology API were respectively defined as construction genes and coordination or topology genes in the genome, so the gene granularity is beneficial to the diversity of evolutionary design plans, and furthermore is not losing geometry and engineering semantics.The product function is modelling by functional behavior. The construction genes'functional behavior expressed by functional surfaces and placed in their attribute table, is used for fitness calculation. The genome is a symbol string of construction genes and topology genes, which is ranked according to the generated sequences and spatial topological relations. The construction genes are classified into primary genes and secondary genes in the light of their roles in the product modelling. The genome is a kind of tree structure, in which the primary genes are the root nodes of each branch, the topology genes are son nodes, and the secondary genes are grandson nodes, therefore, the genome can be converted into a binary tree structure which is easier to ergodicity of genes and their function information or similarity calculation.3) The population is the source of evolutionary creative design, the method building up the population with similar individual genomes was proposed. The similarity between two genomes is related with the number of the equal gene blocks and the average similarity of the equal genes. The gene block corresponding with the shape features, which is a sequence of gene codes with the same sort, can display the product characteristics under no-split condition, which was picked out by the string pattern matching algorithm.By comparing with the target genome, the individual similarity is obtained, on which the different genome groups are divided. It is the content-based 3D model search technology by comparing and matching symbol strings to get the similar genome, which is different from other search technology.The population database as a C/S structure can provide the best individual space for the evolutionary design module.4) The 3-level fitness function is used to evaluate the creative plans in the process of evolutionary design, which fits into automatic, high-speed, diversity evolutionary design. The limit indexes at the first level such as weight and dimension are the basic constraint conditions. The basic functional indexes at the second level the new qualified genomes must possess are some functional surfaces which form the benchmark function model as the reference. The parameter vectors on matching functional surfaces of the genome to be evaluated by comparing with the benchmark function model are input into the RBF neural network and then the genome fitness is output. The subjective indexes at the third level are scored by human, their average scores is the fitness. The third level is a widely distributed, large evaluating system to keep pace with the high-speed evolutionary design. The sum of the three classes of fitness is the fitness of the evaluated genome for selection operation.5) The multi-subgroup immune evolutionary mechanism is adopted to meet a large-scale creative design. The initial population is at first divided into several subgroups, the evolutionary design will have been running independently in every subgroup for some generations, and then the excellent individuals from these subgroups are regrouped to continue their evolution until some relatively optimal plans emerge. The three operators which are immune selection, crossover covered by gene block, mutation of parameter in genes, are introduced in the evolutionary design.The evolutionary creative design platform in the ACIS-HOOPS environment adopts parallel computing and message-passing paradigm to conform all modules. The evolutionary test of some teapot models demonstrated the effectiveness and feasibility of the model-oriented evolutionary creative design process from genome formation, similarity calculation, benchmark function model and fitness function to evolutionary operations.To sum up, some novel concepts and methods from the perspective of genome, population, fitness function, evolutionary operators in the product model-oriented evolutionary design have been studied, and a kind of evolutionary design thinking and design process model have been put forward for the product model innovation in the concept design stage. The evolutionary design method has the great practical value to the creative design of product structures and models in the field of auto parts, home appliances and supplies, crafts, and so on.