High gravity and high productivity ethanol fermentation with self-flocculating yeast cells (original) (raw)

An artificial neural network (ANN) model was established based on the prototype experiment of up flow anaerobic sludge bed and anaerobic filter (UASBAF) reactor treating industrial wastewater. It was adopted a back propagation algorithm with momentum and adaptive learning rate. Because of dynamic model was too rough to be used practicability (Sinha et al., 2002), mathematical model of UASBAF was established by the Matlab neural network toolbox (YI Sai-li and LUO Wen-shen, 2004). Then, each investigate influence of parameters to the performance of the reactor was compared using method of partitioning connection weights (PCW) which was able to reflect the direct and indirect influences of all the factors on the reactor performances (Gevrey et al., 2003; TIAN Qi-chuan, 2005). The comparative influences are: pH values > influent of chemical oxygen demand (COD) > hydraulic retention time (HRT) > alkalinity. In addition, by fractionally fixing affecting factors three-dimensional figures were plotted to analyze ways and processes of input parameters affecting the reactor vividly and qualitatively. And ultimately a series of strategies was proposed to optimize the working condition of the system, which provided an effective way of controlling anaerobic reactor simply and transparently.