A neural based modeling approach for predicting effective thermal conductivity of brewer’s spent grain

Abstract

Purpose

Brewer's spent grain (BSG) is the main by-product of the brewing industry, holding significant potential for biomass applications. The purpose of this paper was to determine the effective thermal conductivity (k_eff) of BSG and to develop an Artificial Neural Network (ANN) to predict k_eff, since this property is fundamental in the design and optimization of the thermochemical conversion processes toward the feasibility of bioenergy production.

Design/methodology/approach

The experimental determination of k_eff as a function of BSG particle diameter and heating rate was performed using the line heat source method. The resulting values were used as a database for training the ANN and testing five multiple linear regression models to predict k_eff under different conditions.

Findings

Experimental values of k_eff were in the range of 0.090–0.127 W m⁻¹ K⁻¹, typical for biomasses. The results showed that the reduction of the BSG particle diameter increases k_eff, and that the increase in the heating rate does not statistically affect this property. The developed neural model presented superior performance to the multiple linear regression models, accurately predicting the experimental values and new patterns not addressed in the training procedure.

Originality/value

The empirical correlations and the developed ANN can be utilized in future work. This research conducted a discussion on the practical implications of the results for biomass valorization. This subject is very scarce in the literature, and no studies related to k_eff of BSG were found.