| Summary: | 碩士 === 東海大學 === 化學工程學系 === 92 === This study utilizes neural network computation and response surface methodology (RSM) to search for the optimal operating conditions of dioxin emissions from an incineration system.
Dioxin emissions and operating variables data from a commercial fluidized-bed incinerator are employed in this study. The published data show that dioxin emission is related to operating variables, namely, total air supply, secondary/primary air ratio, furnace bed temperature, furnace top temperature, boiler out temperature, oxygen composition at bag filter outlet, carbon monoxide composition at bag filter outlet, hydrochloric composition at bag filter outlet, nitrogen oxides composition at bag filter outlet, calcium hydroxide absorbent addition, and ammonia absorbent addition. A neural network model is built based on these emissions and their corresponding operating variables. After that, a statistical screening is carried out to obtain major operating variables. Modified neural network models based on the major operating variables are built and coupled with a RSM search to reveal the operating conditions for the lowest dioxin emissions. Neural network (NeuralWare), SAS (SAS Institute Inc.), and Design Expert (Stat-Ease, Inc.) are employed to complete neural model, to screen major variables, and to locate optimal conditions, respectively.
Results indicate that major operating variables are total air supply, oxygen composition at bag filter outlet and furnace top temperature. Based on the first two major operating variables, i.e., total air supply and oxygen composition at bag filter outlet, the lowest dioxin emission is 2.54 ng-TE/Nm3 at a total air supply of 30393. Nm3/h and an oxygen composition at bag filter outlet of 11.8 %. Also, considering three operating variables all together, the lowest dioxin emission is about 2.33 ng-TE/Nm3 as being operated under total air supply (32237. Nm3/h), oxygen composition at bag filter outlet (12.98 %) and furnace top temperature (884.℃).
|
|---|
