Are Neural Networks the Right Tool for Process Modeling and Control of Batch and Batch-like Processes?

The prevalence of batch and batch-like operations, in conjunction with the continued resurgence artificial intelligence techniques for clustering classification applications, has increasingly motivated exploration applicability deep learning modeling feedback control processes. To this end, present study seeks to evaluate viability general, neural networks particular, toward process via a case study. Nonlinear autoregressive exogeneous input (NARX) are evaluated comparison subspace models within framework model-based control. A polymethyl methacrylate (PMMA) polymerization is chosen as simulation test-bed. Subspace-based state-space NARX identified first compared their predictive power. then implemented model (MPC) compare performance both approaches. comparative analysis reveals that performed better than power performance. Moreover, were found be less versatile adapting new operation. results indicate further research needed before may become readily applicable