Performance evaluation of ANFIS and RSM in modeling biodiesel synthesis from soybean oil

• Published in: Biosensors and bioelectronics. X Vol. 15; p. 100408
• Main Authors: Kumar, Sunil, Bansal, Sumit
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2023; Elsevier
• Subjects: Adaptive neuro-fuzzy inference system; Biodiesel; Prediction; Response surface methodology; Soybean; Response surface methodology; Soybean; Adaptive neuro-fuzzy inference system; Biodiesel; Prediction
• ISSN: 2590-1370; 2590-1370
• DOI: 10.1016/j.biosx.2023.100408

The parameters of biodiesel generation from Soybean oil having the molar ratio, reaction duration, and catalyst concentration for constant temperature, were modelled Using response surface methodology. a molar ratio of 6–12, NaOH of 1–2% w/w, time of 30–60 min, and temperature of 35–55 °C. an adaptive neuro-fuzzy inference system (ANFIS) and the response surface methodology-based Box–Behnken experimental design. a significant regression model with an R2 value of 0.9411 was obtained. The ANFIS model was used to link each of the four input factors with the outcome variable (biodiesel yield). In the training, an R2 value of 0.9918 was attained. The findings showed that the constructed models accurately depicted the processes they described. •In this study, developed RSM and ANFIS models to estimate the conversion of Soybean oil to fatty acid methyl ester (FAME) using NaOH and methanol within a mechanically stirred reactor. To anticipate experimental outcomes, we employed RSM Box-Behnken design and ANFIS techniques to construct these predictive models. Following the design and testing phase, we compared the predictive performance of both models using statistical metrics, namely R2 and RMSE values. The RSM model achieved R2 and RMSE values of 0.9411 and 1.4548, respectively, while the ANFIS model demonstrated values of 0.9918 and 0.5586. These results indicate that the ANFIS model exhibits greater resilience in predicting FAME conversion values compared to the RSM model.