Journal of Engineering Research
Innovation and Scientific Development

Softcomputing Model and Optimization of Soursop (Annona muricata) Seed Pyrolysis in a Fixed Bed Reactor

Document Type : Research Paper

Authors
1Chijioke Boniface Ugwuodo,  2Henry Ukochukwu Itiri,  3Iheonu Nduka Emmanuel,  4Kelechi Noble Akataobi,  5Ngozi Benedict Sunday, 
  1. 1  Department of Chemical Engineering, Michael Okpara University of Agriculture, Umudike
  2. 2  Department of Chemical Engineering, Michael Okpara University of Agriculture, Umudike
  3. 3  Department of Chemical Engineering, Nazarbayev University, Astana, Kazakstan
  4. 4  Department of Chemical Engineering, Michael Okpara University of Agriculture, Umudike
  5. 5  Department of Chemical Engineering, Michael Okpara University of Agriculture, Umudike
Abstract

The introduction of machine learning in predicting yield for the pyrolysis process stimulates the wide usage of first-generation biomass, especially in producing bio-oil and biochar. This study focused on soft computing optimization of soursop seed pyrolysis using ANN and ANFIS. Soursop (Annona muricata) seeds, often discarded as agricultural waste, contain valuable organic components suitable for thermochemical conversion. This study explores the pyrolysis of soursop seeds in a fixed-bed reactor. It models the process using soft computing techniques, Artificial Neural Networks (ANN), and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) to optimize bio-oil yield. Samples of the soursop seed were washed, oven-dried, and ground the soursop seed using a mill, screening it to various particle sizes before valorizing it. Seventeen (17) pyrolysis experiments were conducted using varying temperatures, particle sizes, and inert gas flow rates. The resulting bio-oil yields were used to train both ANN and ANFIS models. Model performance was evaluated using statistical metrics such as mean square error (MSE) and coefficient of determination (R²). ANFIS achieved superior predictive accuracy (R² = 0.999, MSE = 0.0005) compared to ANN (R² = 0.997, MSE = 0.7235), confirming its effectiveness in modeling the pyrolysis process. Optimal conditions for bio-oil yield (27.8%) were identified at 500°C, 3.5 mm particle size, and 1.25 L/min inert gas flow rate. Physicochemical characterization of the bio-oil revealed favorable fuel properties, including a high heating value (41.5 MJ/kg), acceptable viscosity, flash point (195.8°C), and pour point (−25°C). GC-MS and FTIR analyses further confirmed the presence of essential fatty acids and functional groups consistent with fuel-grade bio-oils. This study demonstrates that soft computing models, particularly ANFIS, are effective tools for optimizing bio-oil production from biomass such as soursop seeds. The findings support the valorization of agricultural waste as a viable pathway for sustainable energy production and provide a framework for scaling bio-oil production processes in bioenergy industries.

Graphic Abstract
JERISD PUBLICATION LOGO
Vol 3, Number 1
March 2025
Pages 84-90
Files
Download: PDF
History
  • Received: 16/02/2025

  • Revised: 20/03/2025

  • Accepted: 24/03/2025

  • Published: 26/03/2025
Statistics