Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
DEVELOPMENT OF ARTIFICIAL NEURAL NETWORKS (ANN) MODELS TO PREDICT THE PRODUCTION OF CUMULATIVE BIOGAS FROM FOOD WASTE (FW), FRUITS AND VEGETABLES WASTE (FVW) AND THEIR CODIGESTION (CD)
Submission Author:
Florian Alain Yannick Pradelle , RJ
Co-Authors:
Michel Carvalho, Florian Alain Yannick Pradelle, Brunno F. Santos
Presenter: Florian Alain Yannick Pradelle
doi://10.26678/ABCM.COBEM2023.COB2023-1040
Abstract
Neural Network (ANN) models for estimating the cumulative volume of biogas produced by a diverse range of biodigester configurations. A comprehensive database was created, incorporating information from 47 literature references, resulting in a total of 2098 conditions. The database included eight variables: biomass type, reactor type, volatile solids (VS), hydraulic retention time (HRT), organic load rate (OLR), temperature, pH, and reactor volume. Data wrangling analysis was conducted to prepare the database, including the removal of outliers and missing data using histograms. The ANNs, developed in Matlab software, were evaluated using various topologies, with the number of neurons in the hidden layer ranging from 8 to 11 and different activation functions, including the output layer. The models' performance was assessed using the coefficient of determination (R2) and the sum of squared errors (SSE). Additionally, response surface assessments were conducted to evaluate the models' applicability across a range of operational conditions. The final database consisted of 2,098 conditions, and the investigations demonstrated the feasibility of developing a predictive model for cumulative biogas production with acceptable performance indexes. The response surfaces identified regions with enhanced production performance, characterized by a combination of process variables consistent with previous literature.
Keywords
Biofuel production, Biomass, Database analysis, Response surface assessment
