响应面法与基于神经网络的遗传算法在优化棉粕固态发酵中外源添加物用量上的应用

采用响应面法和基于神经网络的遗传算法,对提高棉粕固态发酵中游离棉酚降解的外源添加物水平(尿素添加量、碳酸钠添加量、菜粕添加量等因素)进行优化,并对这2种方法的优化效果进行比较.结果表明:采用响应面法优化,当尿素、碳酸钠和菜粕的添加量分别为0.97%、2.47%和24.32%时,预测的游离棉酚最大降解率为77.71%,实际降解率为79.10%;3因素中,添加碳酸钠对棉酚降解的影响最显著;而采用神经网络协同遗传算法优化,当尿素、碳酸钠和菜粕的添加量分别为0.98%、2.45%和23.66%时,预测值和实际值分别为81.36%和80.09%;采用神经网络模拟结合遗传算法的优化方法拟合度为99.91%,高于响应面法的91.91%,且均方差(RMSE)较低,为0.13,表明在棉粕固态发酵优化中,采用基于神经网络的遗传算法比响应面法具有更好的结果拟合度和准确性.

Optimization of additive contents in cottonseed meals during the solid-state fermentation using response surface methodology and artificial neural network-based genetic algorithm

The response surface methodology(RSM) and an artificial neural network-based genetic algorithm(ANN-GA) were carried out to investigate the effects of urea content,Na2CO3 content and rapeseed meal content on free gossypol detoxification from cottonseed meals by solid-state fermentation.The modeling and optimizing abilities of the two methods were compared.The results showed that according to RSM,the optimal additive contents for free gossypol detoxification were 0.97% urea,2.47% Na2CO3 and 24.32% rapeseed me...