The Research Of The RBF Network In The Evaluation Of The Product Conceptual Design

In today's rapidly evolving global and fiercely competitive economy, companies need to innovate continuously. It is also more difficult to the designing of the product. Conceptual designing is very important in the designing stage; it determined the basic feature and the main frame of the product. The key that how to select the best designing consequence according with kinds of requirement from many designing consequence is gaining the accurate multiple evaluation of the evaluating target. Product model evaluation is to make final decision solve complex problems that have more solutions. Multi-Target evaluation has already become an important step to improve design quality for product model. Therefore, the problem of the product model designing becomes a hot spot in the field of the product conceptual design, and more and more experts and students pay close attention to it.Industrial product model design is the creative action that satisfied people's physiology and psychology requirement, coordinated the relation of the product and the environment, and raised the additional value of the product. Its evaluating content contained: the product model accorded with the requirement of the human project; the model's pattern can express the function of the product clearly; the coordination of the model's color; the moderation of the model surface's decoration; the concordant requirement between the model and the environment; the economical requirement of the model and so on.Now neural network model is extensively used to solve the problem of the calculation and the evaluation because that the neural network has very good self-organization, self-study and the ability of the association and memory, and it can gain the steelyard weight of the importance of the evaluative target by studying the sample, and it is good to weaken the human factor and pledge the effect of the steelyard weight. To solve these problems, it is mainly by studying and memory the real data or evaluating consequence and then attains the purpose of the calculation and the evaluation. Radial Basis Function neural network is a kind of feed-forward neural network. It has good ability to classify the vector and high-speed computing, and it can press on towards willfully non-linear function by willfully accuracy, so it is named the neural network which is the best suitable for the pattern identification task.Because of the above research current situation, a corresponding evaluation system is formed in this thesis in the light of the characteristic of the mobile phone innovative and conceptual design. Mainly including work:1,The factors existed in the mobile phone creative designing evaluation is presented based on the research of the creative conceptual design, ergonomic and the fuzzy conception.This thesis summarized the factors which affect the industrial product model evaluation from the factors of substance and people and the fuzzy feature existed in the product model evaluation.2,The two level synthetical evaluating model is raised based on the Radial Basis Function neural network.Radial Basis Function neural network, a neural network model which is usually used in the evaluation nowadays, is researched in the thesis. In the light of the question existed in the RBF network model, the Island migrating Neural network ensemble Model based on an island migrating model and the thought of several neural networks ensemble is presented. Combined with the characteristics of the mobile phone the two level synthetical evaluating model is raised.3,The evaluating system is realized based on the mobile phone.Based on research of the creative conceptual design, ergonomic and Radial Basis Function neural network the thesis realized the evaluating system used vc++6.0 and matlab6.5 and the technology of data base. And it gained a good result from evaluating. The result has higher precision and quite good dependability. It not only provides the train of thought and the improving foundation for the designer, ensures the designing quality but also raises the designing efficiency. And the Radial Basis Function neural network has good extensive ability and takes on wide application foreground.