Design of a water-cooling condenser based on the integration of Taguchi's method and MANFIS

The design of water-cooling condensers is a typical nonlinear and Multi-Input Multi-Output (MIMO) problem. In order to find out the manufacturing parameters to realize the optimal performance of this condenser, Taguchi's optimization and inverse-model technology are integrated in this paper. First, the Taguchi method is used to determine the minimum required times of the experiment and the important factors which affect the out performance of most of this condenser. Next, the Adaptive Neuro-Fuzzy Inference System (ANFIS), being widely used to model different kinds of nonlinear systems, is further extended to a MIMO-ANFIS (MANFIS) architecture to train the inverse model for this condenser. Since the well-trained model has been characterized with the unique ability of figuring out the inverse relationship of each input-output set, the manufacturing parameters can be found. A case study involving statistical characterization and multiple criteria optimization on the design of a water-cooling condenser is successfully presented to confirm the effectiveness of the proposed method.