An intelligent approach for improved predictive control of spray drying process

Abstract

A flexible meta modelling approach is presented to predictive control of a drying process using Adaptive Neuro-Fuzzy Inference System (ANFIS), Artificial Neural Network (ANN) and Partial Least Squares (PLS) analysis. In the proposed approach, the PLS analysis is used to pre-process actual data and to provide the necessary background to apply ANN and ANFIS approaches. A reasonable section of this study is assigned to the modelling with aim at predicting the granule particle size and executing by ANFIS and ANN. ANN hold the promise of being capable of producing non-linear models, being able to work under noise conditions and being fault tolerant to the loss of neurons or connections. Also, the ANFIS approach combines the advantages of fuzzy system and artificial neural network to design architecture and is capable of dealing with both limitation and complexity in the data set. The efficiencies of ANFIS and ANN approaches in prediction are compared and the superior approach is selected. Finally, by deploying the preferred approach, several scenarios are presented to estimate the predictive control of spray drying as an accurate, fast running and inexpensive tool. This is the first study that presents a flexible intelligent approach for predictive control of drying process by ANN, ANFIS and PLS. The approach of this study may be easily applied to other drying process.