| Summary: | 碩士 === 建國科技大學 === 自動化工程系暨機電光系統研究所 === 96 === The following article is a discussion of prediction of temperature distribution in the copper-beads packed bed with a confined air jet using artificial neural networks (ANN). Using experimental results, an optimum ANN model was built. This experiment investigated under a confined rectangular air jet. The packed bed was made of copper-beads and its size was 120 mm in length (L), 60 mm in width (W) and 30 mm in height (H). The input and output parameters include the relative packed bed height, the diameter of the copper bead, the relative jet nozzle width, the jet nozzle speed, the room temperature, the thermocouple position and its wall temperature. The Back Propagation (BP) algorithm was used to built ANN model. Moreover, the best performance simulation is by using Taguchi method due to calculate error parameter of BP, and in result we found ANN can be apply to the temperature distribute analysis. The absolute value of inaccuracy can be controlled below 5%. The outlet average speed was used to study parameter sensitivity, results shown very good agreement with experimental data. Therefore, we can use this network to predict any position’s temperature in the flow channel, and to reduce the time and the cost of experiment.
|
|---|
