Statistical modeling and optimization of the bleachability of regenerated spent bleaching earth using response surface methodology and artificial neural networks with genetic algorithm

Abstract

In this study, the statistical modeling and optimization of the regeneration of spent bleaching earth (SBE) for re-use in the bleaching of crude palm oil (CPO) oil was examined. Having a good model will assist with the successful optimal regeneration of SBE and hence minimize the environmental pollution associated with its current disposal method which is based on dumping as landfills. The SBE samples were de-oiled with the Soxhlet extraction method, using n-hexane for 1 h at 60 °C; treated at temperatures ranging from 300–500 °C; at carbonization time between 30 and 45 min; and with hydrochloric acid concentrations between 1 and 2 M, at a constant stirring time of 30 min, respectively. The operating conditions for the experiment were according to the Central Composite Design (CCD) experimental design using the Design Expert software version 13. The modeling and optimization of the SBE regeneration process was carried out with the Response Surface Methodology (RSM) and Artificial Neural Network (ANN) techniques. Five regression models were developed from the RSM approach and the best one selected based on model selection parameters recommended in the literature. Similarly, ten ANN models with the number of neurons in the hidden layer that varied from 2 to 16 were considered and the best one selected using the mean square error (MSE) and correlation coefficients (R) for the training, validation and testing performances. Results showed that the ANN technique led to a model with a better predictive ability than the RSM one. The optimum experimental bleachability of 71.5% for the regenerated de-oiled SBE was obtained at carbonization temperature of 500 °C, hydrochloric acid concentration of 2M and carbonization time of 45min. Using the Genetic Algorithm (GA), the ANN model resulted in an optimum bleachability of 70.87% with corresponding optimum factors at 468.19 °C, 2 M and 45 min, while the RSM approach gave an optimum bleachability of 73.52% at the corresponding factors of 498.99 °C, 1.57 M and 41.14 min for the carbonization temperature, acid concentration and carbonization time, respectively. The optimum experimental bleachability of the regenerated SBE achieved was 12.5% higher than that of virgin bleaching earth (VBE).

Profiling of Secondary Metabolites of Optimized Ripe Ajwa Date Pulp (Phoenix dactylifera L.) Using Response Surface Methodology and Artificial Neural Network

“…Contrarily, the most widely used metrics among the numerous model selection techniques are the predicted residual sum of squares (PRESS), the Akaike information criterion (AIC), and the Bayesian information criterion (BIC) [ 24 ]. The PRESS measures a model’s propensity to predict; hence, the lower the PRESS score, the greater the model’s prediction [ 25 ]. The PRESS value range in the current study is between 0.0464 and 0.5369, which supports the claim.…”