A study on the enzymatic hydrolysis of steam exploded napiergrass with alkaline treatment using artificial neural networks and regression analysis

To facilitate the enzymatic saccharification of widely available lignocellulosic biomass, napiergrass was subjected to a two-stage pretreatment process consisting of steam explosion (SE) followed by alkaline delignification. SE was performed under various reaction temperatures, reaction times, and particle sizes. The experimental results show that the two-stage pretreatment process was effective at removing xylan and lignin and significantly enhanced the enzymatic digestibility of napiergrass. Up to 85% of the lignin in the original material was removed, leaving a cellulose-rich residue that was highly susceptible to enzymatic hydrolysis; the enzymatic digestibility reached 96.1%. SE alone was not sufficient, because considerable amount of lignin still remained in the steam-exploded solids. In addition to experimental work, models predicting digestibility as a function of SE conditions were developed using artificial neural network and regression techniques, and their performance was compared. Three different methods were used, i.e., the back-propagation neural network (BPNN), the multiple linear regression (MLR), and the partial least-square regression (PLS). The input of the model was the three parameters of steam explosion (temperature, time, and particle size), while the enzymatic digestibility was the output. The results show that the BPNN model provided reasonable predictive performance. In addition, the SE temperature was found to be the most significant factor among the three parameters studied, followed by the SE time.