engleski

Principal component analysis of integrated chemical plants

Modelling of dynamical behaviour of an integrated chemical plant by the use of the principal composition algorithm ( PCA ) is analysed. In the work equivalence between singular value decomposition ( SVD ) and PCA is shown. The plant is a serie of two chemical nonisothermal continuous stirred tank reactors with a recycle loop. The plant dynamics is simulated by a computer model of the mass balances under stochastic variations of feed composition and the reactor temperatures. The process variables are sampled for loading and score variables determination. The PCA analysis reveals that due to interaction of variables in the integrated plant a significant reduction in state dimension is achieved. From the loading and score plots possible on-line monitoring and control structures are discussed.

Principal components; singular value decomposition; partial least squares; state reduction; neural networks

nije evidentirano