Articles
Artificial intelligence model development for ultrasound-assisted extraction of bioactive compounds from rooibos processing waste for postharvest preservation
Article number
1466_57
Pages
471 – 480
Language
English
Abstract
Rooibos (Aspalathus linearis) is a major South African herbal tea crop that generates substantial processing waste (spent leaves, stems, and dust) rich in bioactive compounds, including aspalathin, nothofagin, and other polyphenols.
Valorizing this waste through efficient green extraction provides a sustainable source of natural antioxidants and antimicrobials for food preservation, reducing reliance on synthetic additives and addressing environmental concerns associated with landfill disposal.
The present study aimed to optimize ultrasound-assisted extraction (UAE) of bioactive compounds from rooibos processing waste using response surface methodology (RSM) with central composite design and artificial neural network (ANN) modeling.
The independent variables investigated were extraction temperature, sonication time, power level, and concentration, with responses being yield, total phenolic content (TPC), total flavonoid content (TFC), and antioxidant activity (2,2-diphenyl-1-picrylhydrazyl (DPPH) and 2,2’-azino-bis(3-ethylbenzoline-6-sulfonic acid (ABTS)) assays.
The combination of all independent variables significantly influenced the responses (p≤0.05), with temperature exerting the strongest effects across most responses.
Compared with the quadratic RSM model, the ANN model consistently outperformed it.
Specifically, ANN achieved higher coefficient of determination (R2) values (up to 0.98 for TPC) compared to RSM (0.84-0.92), along with lower root mean square error (RMSE) and mean absolute error (MAE) across all responses.
Additionally, ANN predictions showed stronger linear agreement with experimental data and reduced residual dispersion, confirming enhanced predictive precision and generalization.
Therefore, the developed predictive model can guide industrial-scale valorization, supporting circular economy principles and the development of natural food preservatives.
Valorizing this waste through efficient green extraction provides a sustainable source of natural antioxidants and antimicrobials for food preservation, reducing reliance on synthetic additives and addressing environmental concerns associated with landfill disposal.
The present study aimed to optimize ultrasound-assisted extraction (UAE) of bioactive compounds from rooibos processing waste using response surface methodology (RSM) with central composite design and artificial neural network (ANN) modeling.
The independent variables investigated were extraction temperature, sonication time, power level, and concentration, with responses being yield, total phenolic content (TPC), total flavonoid content (TFC), and antioxidant activity (2,2-diphenyl-1-picrylhydrazyl (DPPH) and 2,2’-azino-bis(3-ethylbenzoline-6-sulfonic acid (ABTS)) assays.
The combination of all independent variables significantly influenced the responses (p≤0.05), with temperature exerting the strongest effects across most responses.
Compared with the quadratic RSM model, the ANN model consistently outperformed it.
Specifically, ANN achieved higher coefficient of determination (R2) values (up to 0.98 for TPC) compared to RSM (0.84-0.92), along with lower root mean square error (RMSE) and mean absolute error (MAE) across all responses.
Additionally, ANN predictions showed stronger linear agreement with experimental data and reduced residual dispersion, confirming enhanced predictive precision and generalization.
Therefore, the developed predictive model can guide industrial-scale valorization, supporting circular economy principles and the development of natural food preservatives.
Publication
Authors
M.A. Muhammad, T. Kaseke, K.A. Jimoh, O.A. Fawole
Keywords
artificial neural network, Aspalathus linearis, bioactive compounds, response surface methodology, rooibos, ultrasound-assisted extraction
Online Articles (74)