engleski

Computer integrated production of baker"s yeast

In the work is presented computer integrated production CIP of baker"s yeast in company " Podravka", Koprivnica, Croatia. The production automation is based on the ABB Master computer system. Under computer control are the following production phases: storage tank management molasses preparation washing sterilisation ferementations product separation ( filtration ) product storage The product drying and packing are at the moment only production phases outside integrated computer control. The computer integrated production CIP is based on 1100 on-line signals, of which are 120 analogue inputs and outputs, and with 80 PID control loops. The plant is monitored with sophisticated process instrumentation. The fermentation control is primarily based on feedback loop for regulation of ethanol partial pressure. Developed are neural network models for on-line prediction of ethanol pressure and inverse model for molasses feed rate. The output variables are modelled with time horizons from 1 min. up to 20 min. The models are developed for internal model control IMC structure of fermentations. Applied are feedforward neural networks with static neurones and sigmoid activation transfer functions. Process dynamics is approximated with nonlinear autoregression moving averages ( sequences ) ARMA. Input patterns to neural networks are decomposed into linear and nonlinear principal components PCA. Neural network modelling is focused on development of models with maximum predictivity accuracy.

computer integrated production; process automation; neural networks

nije evidentirano