Optimization of Sugarcane Bagasse Conversion Technologies Using Process Network Synthesis Coupled with Machine Learning

Abstract

Sugarcane bagasse is a commonly generated item from the food industry in the world—the amount of sugarcane bagasse production is increasing yearly. In 2017, the reported sugarcane production in Malaysia was 30,000 kg, which resulted in 9,800 kg of sugarcane bagasse. Sugarcane bagasse produces steam as waste management in Malaysia or simply in landfills. This study aims to optimize sugarcane bagasse conversion technologies using process network synthesis. A superstructure of sugarcane bagasse was created via P-Graph, with multiple pathways or processes being considered. Data needed for the sustainability assessment of each pathway was acquired from various journal sources, including conversion fraction, operating and capital cost, greenhouse gas emission, and the selling price of products were implemented into the superstructure. Then, the data from the feasible structure generated would be analyzed using machine learning via Waikato Environment for Knowledge Analysis software. The data sets were analyzed using this software using the selected algorithm as P-graph developed 17 feasible solution structures. All 17 generated solution structures were analyzed using six different classifier algorithms. The multilayer perceptron algorithm had the best and the least error in classifying the data. Hence, the multilayer perceptron algorithm proved that the correlation between products produced from sugarcane bagasse and the profitability of the process was significant. Therefore, the model can be a basis for determining the best process for sugarcane bagasse conversion technologies.