摘要
利用静态多层前馈神经网络模型对腊样芽孢杆菌(BaeilluscereusDM423)(简称DM423)分批培养过程的生物量进行软测量.构建了以培养时间、温度、pH、溶氧和葡萄糖浓度作为网络的输入变量,生物量浓度为输出变量的5-5-1型的3层前馈网络,网络对DM423的生物量浓度估算精确度较好.分别对葡萄糖浓度、pH和溶氧的实际测量值施加干扰,对网络的鲁棒性进行检测,结果表明,所建立的神经网络具有一定的抗干扰能力.
The biomass concentration of Baeillus cereus DM423 (DM423) during the batch cultivation was measured by feedforward multilayer neural network based on software sensor. A 5-5-1 feedforward neural network was constructed with five inputs of culture time, temperature, pH, dissolved oxygen and glucose concentration and one output of DM423 biomass concentration. The biomass concentration was fitted well with the experimental value by the neural network. The robustness of the neural network was tested applying perturbation to the experimental values of glucose concentration, pH and dissolved oxygen. The results showed that the constructed neural network offered a good ability of resisting interference.
基金
国家自然科学基金青年基金(20306007)
国家自然科学基金重点项目(20436020)
