An ensemble model to predict the minimum spouting velocity for two types of spouted beds

Abstract

Spouted beds are extensively used in different industries. The minimum spouting velocity is a critical hydrodynamic parameter needed for the design and scale-up of these systems. This study deals with an ensemble method involving Multi-Layer Perceptron (MLP) network, the Adaptive Neuro-fuzzy Inference System (ANFIS), the Gaussian Process Regression (GPR), and a Regression Tree (RT) model for predicting minimum spouting velocity for coarse particles in conical spouted beds equipped with nonporous draft tubes (CSBDTs) and conical-cylindrical spouted beds. Some dimensionless variables related to the physical properties of the conical bed and particles’ material were constructed and used as model inputs. The proposed method was validated against the measured data. The results revealed that the ensemble method was capable of predicting the minimum spouting velocity with an average absolute relative error (AARE) of 10.52% and 10.35% for the testing data points in the case of CSBDTs and conventional spouted beds, respectively. It was also shown that the proposed ensemble method provided the best results in comparison with the other methods. Therefore, the ensemble method is a useful and powerful technique for predicting the minimum spouting velocity in spouted beds with and without the draft tube.

Keywords:

• Spouted bed
• draft tube
• minimum spouting velocity
• ensemble system
• neural network
• GPR

Disclosure statement

No potential conflict of interest was reported by the author(s).