基于BP神经网络和GA研究啤酒糟不溶性膳食纤维的酶法脱脂工艺

采用脂肪酶对酶碱法制备啤酒糟不溶性膳食纤维(IDF)的脱脂工艺进行研究，并对制备得到的IDF 成分和功能特性进行分析。在正交试验的基础上，基于BP 神经网络建立脂肪酶脱脂模型，利用遗传算法优化工艺条件。BP 神经网络建立的脱脂模型误差为0.0001，具有较强的逼近能力。优化得到的最佳工艺条件是加酶量0.7g、酶解温度39.6℃、酶解时间5.6h。在此条件下，脂肪去除率达74.1%。制备得到IDF 的溶胀性达6.05mL/g，持水力达318.2%，具有较好的生理活性。

Combining BP Neural Networks and Genetic Algorithms for Optimizing the Lipase Degreasing of Brewer s Spent Grains for Insoluble Dietary Fiber Production

Lipase degreasing was carried out as the first step in the production of insoluble dietary fiber from brewer's spent grains. Orthogonal array design was employed to provide a basis for the establishment of a mathematical model predicting the degreasing efficiency as a function of lipase hydrolysis conditions, namely enzyme amount, hydrolysis temperature and length of hydrolysis and the subsequent genetic algorithm optimization of these conditions. The established model displayed an error of 0.0001, indicating high reliability. The optimum values of enzyme amount, hydrolysis temperature and length of hydrolysis were found to be 0.7 g (on the basis of 10 g brewer s spent grains), 39.6 ℃ and 5.6 h, respectively, and the resultant degreasing efficiency reached up to 74.1%. The swelling capacity and water retention capacity of prepared IDF were 6.05 mL/g and 318.2%, respectively.