Modelling the Electrostatic Fluidised Bed (EFB) coating process using Support Vector Machines (SVMs)

Abstract

An Electrostatic Fluidised Bed (EFB) coating process is used as an eco-friendly alternative to an electrostatic spraying process to coat components of particularly complex shapes with powder paints. Although fluidised beds are well known systems and are widespread throughout several industrial domains, the implementation of appropriate process control procedures is still extremely difficult. Fluidised bed processes are governed by the hydrodynamic behaviour of the suspended powders. The solution of the hydrodynamic laws in closed form is often not realisable because they are complicated or require large amounts of computational time. In contrast, empirical or simplified analytical models as well as learning machine techniques are often used for the control and automation of fluidised bed processes. Therefore, the current study proposes modelling an EFB coating process using Support Vector Machines (SVMs). SVMs were determined to appropriately match the experimental coating thicknesses and demonstrate good prediction capability. The SVMs were compared with both empirical and Artificial Neural Network (ANN) models to demonstrate how an SVM could be a particularly interesting alternative for modelling “in service” and high-duty equipment.