Investigating the microwave parameters correlating effects on total recovery of bioactive alkaloids from sesame leaves using orthogonal matrix and artificial neural network integration

Abstract

In this study, the microwave‐assisted processing of Sesamum indicum leaves was intensified to predict the total alkaloid content via Taguchi orthogonal design and neural network model. Under the optimum microwave condition, the maximized alkaloid content was estimated to be 15.40 (mg/g) % using irradiation time (A) of 60 sec, microwave power (B) of 300 W, oven temperature (C) of 60°C, the mass ratio (D) of 12.5 g/ml, and solvent concentration (E) of 80%. The significant effects of each microwave extraction parameter declined in the order listed: C > B > A > D > E which indicated a 31 and 29% contribution of microwave power and oven temperature, respectively. Moreover, higher coefficients of determination (R2 = 0.9999), and a decreased mean square error (MSE) of 2.69E‐07 suggest good accuracy in predicting total bioactive alkaloid recovery, as proven by cross‐validation trials. Under these microwave conditions, a total of 115 alkaloids were tentatively identified with lindelofine (m/z = 308.18) and supinin (m/z = 306.17) being the most abundant pyrrolizidine alkaloids. Novelty impact statement Although many researchers reported the total alkaloids recoveries from different parts of Sesame plant; the process intensification of microwave processing from leafy parts using Taguchi design and ANN approach have not been investigated. This study established an optimum condition for the total recovery of alkaloid content from S. indicum using orthogonal optimization methodology and artificial neural network

Heat-reflux processing of black peppercorn into bioactive antioxidant oleoresins: a three-functioned Taguchi-based grey relational grading

“…Furthermore, when the number of iterations in the training process was increased, the mean square error produced from the training process showed a minor decrease in magnitude. The lower RMSE values further supported the validity of a solid training programme and improved prediction [ 17 ]. As indicated in the graph, the best fit point occurs at epoch number five.…”

“…Artificial neural network (ANN) is an intelligent modelling tool that deals with non-linear behaviour of inputs and desired outputs data [ 16 ]. The relationship between several input and output variables can be easily reached by optimization [ 17 ]. It has been reported that Grey-Taguchi combined with ANN enhances process performance [ 11 ].…”