| Summary: | 碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 98 === Process parameters are critical to flat plate collector performance, and the key process parameters for designing and manufacturing a flat plate collector include collector materials, absorber materials, number of collectors, collector tube diameter, absorber film type, and understructure insulation material thickness. The quality characteristics are the efficiency coefficient and the heat loss coefficient. Therefore, this study examined the effect of various levels of key process parameters on flat plate collector quality. The Taguchi orthogonal array table was used to design the experiment. The main effect analysis and analysis of variance were conducted on quality data obtained from the experiment in order to determine the optimum parameters for single quality. Quality data from experiment were preprocessed by a grey relational generating operation, and the grey relational theory, coupled with entropy measurement, was employed to determine the optimum process parameter-level combination. Finally, Taguchi verification was carried out to verify experimental and computational confidence intervals and experimental results.
In addition, this study applied a back-propagation neural network and Levenberg-Marquardt algorithm to build the flat plate collector process parameter prediction system. It also set control factors as network input and quality characteristics as output, and conducted network learning training. The prediction error rate was within 5%, proving that the prediction system, as established in this study, has excellent prediction capability.
|
|---|
