FFNN-GA-Assisted Process Modeling and Optimization of Banana Fiber Pretreatment for Retrieval of Cellulose fiber and Other Value-Added Byproducts

Abstract Present work deals with the alkaline pretreatment of pseudo stem fibers of banana and critical process parameter optimization through artificial neural network (FFNN) followed by genetic algorithm (GA) to maximize lignin retrieval, simultaneous reduction of sugar loss and derivatization into inhibitors, keeping cellulose fiber intact for pharmaceutical and textile industries. Critical process parameters like NaOH concentration, pretreatment time & temperature were considered as input variables and reducing sugar and lignin concentrations as response variables. A total of 32 experimental observations were obtained using a one-variable-at-a-time (OVAT) approach. Quadratic regression and a feed-forward artificial neural network (FFNN) were used to model the relationship between the pretreatment variables and responses. The final FFNN employed a 3-15-2 architecture and was trained using the Levenberg-Marquardt back-propagation algorithm. The model showed strong agreement between target and predicted responses, with overall correlation coefficients (R) of 0.99137 for lignin and 0.99864 for reducing sugar, with overall R² values of 98.28% and 99.73%. Based on the trained model, 43 additional combinations were generated, expanding the dataset to 75 observations for computational optimization. The GA identified an optimum condition of 3.42% NaOH, 11.10 h, and 71.15°C, with predicted reducing sugar and lignin concentrations of -0.240 and 1.037 mg/mL. The Optimized condition was experimentally validated, which resulted in 0.15 mg/mL reducing sugar and 1.05 mg/mL lignin. These concentrations corresponded to 9.40 mg/g reducing sugar and 65.80 mg/g lignin released per gram of initial fiber, while the released sugar and lignin were equivalent to 0.94% and 6.58% of the initial fiber mass. The treatment also resulted in 17.33% weight loss, indicating the removal of non-cellulosic components while retaining the cellulose rich fiber structure. FE-SEM, XRD and FTIR and showed a homogenous structure with desirable fiber characteristics. Hence, the FFNN-GA approach was found to be an efficient tool for optimizing the removal of non-cellulosic components from banana pseudostem with the released lignin, reducing sugar, furfural and phenolic as value added products for various industry oriented biorefinery processes such as pharmaceuticals, textile, biofuel, biomedical and filtration industries etc.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11589518/v1
Primary Topic
Biofuel production and bioconversion
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

FFNN-GA-Assisted Process Modeling and Optimization of Banana Fiber Pretreatment for Retrieval of Cellulose fiber and Other Value-Added Byproducts

Sangeeta Negi, Priya Kaithvar
Research Square
Biofuel production and bioconversion
preprint

FFNN-GA-Assisted Process Modeling and Optimization of Banana Fiber Pretreatment for Retrieval of Cellulose fiber and Other Value-Added Byproducts

Sangeeta Negi, Priya Kaithvar
preprint en

Abstract

Abstract Present work deals with the alkaline pretreatment of pseudo stem fibers of banana and critical process parameter optimization through artificial neural network (FFNN) followed by genetic algorithm (GA) to maximize lignin retrieval, simultaneous reduction of sugar loss and derivatization into inhibitors, keeping cellulose fiber intact for pharmaceutical and textile industries. Critical process parameters like NaOH concentration, pretreatment time & temperature were considered as input variables and reducing sugar and lignin concentrations as response variables. A total of 32 experimental observations were obtained using a one-variable-at-a-time (OVAT) approach. Quadratic regression and a feed-forward artificial neural network (FFNN) were used to model the relationship between the pretreatment variables and responses. The final FFNN employed a 3-15-2 architecture and was trained using the Levenberg-Marquardt back-propagation algorithm. The model showed strong agreement between target and predicted responses, with overall correlation coefficients (R) of 0.99137 for lignin and 0.99864 for reducing sugar, with overall R² values of 98.28% and 99.73%. Based on the trained model, 43 additional combinations were generated, expanding the dataset to 75 observations for computational optimization. The GA identified an optimum condition of 3.42% NaOH, 11.10 h, and 71.15°C, with predicted reducing sugar and lignin concentrations of -0.240 and 1.037 mg/mL. The Optimized condition was experimentally validated, which resulted in 0.15 mg/mL reducing sugar and 1.05 mg/mL lignin. These concentrations corresponded to 9.40 mg/g reducing sugar and 65.80 mg/g lignin released per gram of initial fiber, while the released sugar and lignin were equivalent to 0.94% and 6.58% of the initial fiber mass. The treatment also resulted in 17.33% weight loss, indicating the removal of non-cellulosic components while retaining the cellulose rich fiber structure. FE-SEM, XRD and FTIR and showed a homogenous structure with desirable fiber characteristics. Hence, the FFNN-GA approach was found to be an efficient tool for optimizing the removal of non-cellulosic components from banana pseudostem with the released lignin, reducing sugar, furfural and phenolic as value added products for various industry oriented biorefinery processes such as pharmaceuticals, textile, biofuel, biomedical and filtration industries etc.

Research Square
Motilal Nehru National Institute of Technology (IN)
Council of Scientific and Industrial Research, India, Motilal Nehru National Institute of Technology Allahabad
Responsible consumption and production
Biofuel production and bioconversion
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon