Study Of Firefly Optimization Method Based On Surrogate Model And Its Isight Application

Recently considerable disciplines require a higher efficiency for production design with the evolvement of technology. However, the traditional CAE analysis is time costly and computing expensively for designing the sophisticated system or product, meanwhile it is perhaps difficult to obtain the engineering optimal solution in brief time, even cannot be open, customized and modular. Therefore, computer assistant optimization (CAO) for handling the complex product design becomes a new research orientation in view of this situation. Practical instances manifests that CAO could be employed to solve a wide range of engineering design optimization problems which involves multiple design strategies to promote the properties and qualities of products and decline the lead time of products, and thus satisfy the optimization design requirements of users. Nevertheless, how to build a complete design process? How to choose the main technical points of the design? How to customize or integrate the design process of the main technical points according to different requirements? For these problems, this paper studied three kinds of surrogate models and the involved key technologies in building the surrogate model in support of central college funding and military project of one unit of ordnance industrial group. We also provided a crucial chaotic firefly algorithm based on Gaussian mutation. The secondary development is involved based on the Isight platform of CAO software to apply the research outcome systematically. We herein proposed a firefly algorithm optimization based on surrogate model. At last, we obtained acceptable results by experiments of three types of mechanical design problems. In this paper, the main work can be roughly divided into the following several parts:(1) For surrogate model, this paper studied and analyzed three surrogate models embedded in Isight. Comparing with the involved diverse methods for selecting sample points, the usual process of predictive parsing based on surrogate model is proposed. Adopting the CEC2014 test functions as examples, we selected cross validation method for error analysis, discussed the applicability issue of three kinds of surrogate model problems and different performance indexes under the different conditions, and thus provided the gist for the selection of surrogate model.(2) For optimization, a chaotic firefly algorithm (GHFA) based on Gaussian mutation is presented aiming at the early-maturing and the poor capability of local search of standard firefly algorithm. We compared the convergence speed and calculation accuracy of the algorithm with genetic algorithm (GA) and the standard firefly algorithm (FA) by four benchmark functions, the results indicated that GHFA has quick convergence ability and prevents premature phenomenon. Then analyzed the optimizing capability by example verification of planet wheel design.(3) For the secondary development of Isight, we presented a general flow for the firefly optimization design based on the different surrogate models with GHFA proposed in this paper. Three mechanical design examples based on the three surrogate models in this paper are studied and analyzed and we obtained a well result.This paper proposed the secondary development method based on the Isight to apply the research fruit on surrogate models and optimization methods systematically, as well as the user function customization, the gradual design flow combining the surrogate model and firefly algorithm is proposed, and three different mechanical design problems are addressed with an apparent effect, providing assistant for diversified CAO methods, which is beneficial to engineering application.