基于BP神经网络的棉秆酶解糖化的模拟与优化

Simulation and optimization of enzymatic hydrolysis of cotton stalk based on back propagation neural network

新疆棉秆资源丰富,将棉秆酶解生成可发酵糖,有利于其高值化利用。本研究以棉秆为原料,将其经过Na OH处理后,以纤维素酶用量、酶解温度、水解时间和固含量为输入参数,以酶解得到的还原糖产率为目标输出,在Box-Behnken设计实验的基础上,采用BP神经网络对经Na OH处理后的棉秆在纤维素酶中酶解的过程进行模拟与优化,建立了棉秆在纤维素酶中酶解糖化的神经网络模型,优化了棉秆在纤维素酶中酶解的工艺。模型分析结果表明:优化的神经网络模型均方误差小,回归值为0.9571,网络性能稳定,预测结果准确。通过模型优化获得最优的酶解条件为:酶用量为70 FPU/g,温度为46.31℃,酶解时间为72 h和固含量为7.5%,获得还原糖产率最高为62.14%;XRD、SEM和FT-IR的分析结果表明:Na OH预处理能有效除去木质素,使纤维素的含量相对升高,与纤维素酶的接触位点增多,从而提高棉秆的水解产率。

Cotton stalk resource is abundant in Xinjiang province, and transformation of it into fermentable sugars is one of the most promising way to make the high value utilization of cotton stalk. In this study, hydrolysis of NaOH treated cotton stalk in the presence of cellulase solution was investigated. Cellulase loading, reaction temperature, time, and solid content were regarded asinput parameters, reducing sugar yield acts as output parameters. Back propagation neural network(BPNN) was developed to predict and optimize the process of hydrolysis of NaOH treated cotton stalk in cellulase solution based on the experimental sets designed by Box-Behnken. The analysis results showed that the optimized BPNN model had a small mean square error and a regression value of 0.9571, indicating that BPNN model could provide reasonable predictive performance. The optimal hydrolysis condition was obtained when the enzyme concentration was 70 FPU/g, hydrolysis temperature was 46.31℃,hydrolysis time was72 h and the substrate content was 7.5%. The maximum reducing sugar yield of 62.14% was achieved. SEM, XRDand FT-IR analysis results showed that lignin could be effectively removed after pretreated with NaOH, which made the content of cellulose increase relatively, furthermore, more available sites were exposed to cellulase during the pretreatment process, which improved the hydrolysis efficiency of cotton stalk in cellulase solution.