Optimization of Sago Flour Adhesive Concentration in Banana Waste Biocharcoal Briquettes Using Artificial Neural Network-Genetic Algorithm (ANN-GA)

Abstract

This paper presents the work of optimizing the adhesive concentration of sago flour biochar briquettes from banana waste to produce briquettes with the fastest boiling point using Artificial Neural Network with Genetic Algorithm (ANN-GA). Scenarios were carried out on the number of data sets (small and large). ANN is used as a technique for estimating the relationship between the concentration of sago flour adhesive biochar briquettes with the time it reaches the boiling point. The resulting ANN structure is 1:5:3:1:1 with backpropagation algorithm (Levenberg Marquart). The input parameter is the concentration of the biochar briquette adhesive and the output parameter is the time it reaches the boiling point of water. The best ANN performance is in scenario 1 (small data), the number of data sets is 4 with a regression coefficient (R) 0.99222, Mean-Squared Error (MSE) 6.89x 10-5. GA is used as an input parameter optimization technique to find the input parameter set point at what time it reaches the fastest boiling point. From running the GA program, the optimum concentration of sago flour adhesive was 13,9835% and the time to reach the boiling point was 25,017 minutes.