| Summary: | 博士 === 國防大學理工學院 === 國防科學研究所 === 98 === The purpose of this dissertation is devoted to improving the performance of Proton Exchange Membrane Fuel Cells (PEMFC) by using Metamodel-Based Design Optimization (MBDO). The benefits of MBDO are an efficient and systematic design and analysis in engineering problems. There are numerous operating parameters to affect the performance of PEMFC, such as the operating temperature of fuel cell, the anode and cathode humidification temperature and the flow rate of reactant. It is necessary to adjust effectively the conditions of different control factors for meeting the need and change in the environment. Moreover, the fuel cell would be expected to get an optimal design on the robust performance parameters.
The MBDO procedure involves three strategies. In order to reduce the cost of experiments and get useful information to system, the first strategy is to choose an experimental design for establishing the database of metamodel. On the primary stage, a fractional factorial design was employed as the screening experiment to identify the significance of each factor and the interaction between two factors each other. Then, the follow-up design would be more efficient by using Taguchi orthogonal arrays. In the present work, metamideling was to adopt the configuration of neural network based on the back-propagation network and the radial basis network. Owing to the ability of establishing a nonlinear model of a complex system, metamodeling can be used to complement the lack of the Taguchi experiments that maybe only find a local solution. Hence, the available expression can be proposed for the non-linear function between the input-control factors and the output-response quality. Finally, on the method of optimization, the genetic algorithm with a high capability of the global optimization has been used to seek for the best combination of the parameters in order to meet the requirement of the quality characteristics.
The research of PEMFC by using MBDO is divided into Z type and 4U type for parallel flow geometries in this study. The most important strategy in MBDO is to validate the correctness of metamodeling in the analysis of process. The reliability of the metamodel can be verified through comparing it with the initial result of the Taguchi analysis; it is helpful to realize the effects of experimental parameters on PEMFC system. By way of the confirmation experiments, the interactions have been verified to exist and be of significance between the temperature of fuel cell and the temperature of reaction gas humidification. So, it is noticed how to decide adequately the combination of parameters between the significant interactions in order to get the best performance of fuel cell. In this study, the flow passage of type-4U is provided with the less variance of the output performance under the best combination of parameters; the performance of output target can meet the requirements of robust design.
|
|---|
