A comparison of networked approximators in parallel mode identification of a bioreactor

This paper presents a simulation based comparison of Multilayer Perceptron (MLP), Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and Least Squares Support Vector Machines (LS-SVM) in parallel mode identification of a chemical process displaying several challenges. The paper provides a graphical analysis of the nonlinear behavior for the system under investigation, a case study of purely parallel identification scheme, the effects of noise in the training data on the prediction performance and the performance comparison of the standard approaches under limited amount of numerical data. The results have shown that the emulators utilizing the MLP structure are superior to the others in terms of predicting the system trajectories, locating the limit cycle, noise driven response and predicting the steady state conditions given only 582 pairs of training data. Furthermore, as opposed to others, with the MLP structure, these qualities disappear smoothly as the noise level is increased gradually.