Issues regarding artificial neural network modeling for reactors and fermenters

In recent years researchers in many areas have used artificial neural netwo rks (ANNs) to model a variety of physical relationships. While in many case s this selection appears sound and reasonable, one must remember than ANN m odeling is an empirical modeling technique (based on data) and is subject t o the limitations of such techniques. Poor prediction occurs when the train ing data set does not contain adequate "information" to model a dynamic pro cess. Using data from a simulated continuous-stirred tank reactor, this pap er illustrates four scenarios: (1) steady state, (2) large process time con stant, (3) infrequent sampling, and (4) variable sampling rate. The first s cenario is typical of simulation studies while the other three incorporate attributes found in real plant data. For the cases in which ANNs predicted well, linear regression (LR), one of the oldest empirical modeling techniqu es, predicted equally well, and when LR failed to accurately model/predict the data, ANNs predicted poorly. Since real plant data would resemble a com bination of situations (2), (3), and (4), it is important to understand tha t empirical models are not necessarily appropriate for predictively modelin g dynamic processes in practice.