摘要
精确有效的发酵过程模型不仅能够定量揭示过程信息间的关联,实现对难以实时监测变量的预测,而且是进一步控制和优化的前提;基于数据驱动的发酵过程建模方法得到了广泛研究与应用,然而其仅考虑发酵过程的非线性特征和数据具有多采样率的特点,忽略了过程数据中测量噪声对模型的影响;为此,提出基于栈式降噪自编码器的发酵过程回归建模方法,该方法不仅具有较强的非线性拟合能力,半监督的学习策略也能够充分挖掘发酵过程中的所有数据信息,同时可以从含噪声的过程数据中提取出鲁棒性的特征,使模型具有噪声适应性;通过青霉素仿真对比实验结果表明,该模型的预测性能更好。
An accurate and effective fermentation process model can not only quantitatively reveal the correlation between process information and realize the prediction of variables that are difficult to monitor in real time,but also is a prerequisite for further control and optimization.The data-driven modeling method has been widely researched and applied.However,it only considers the nonlinear characteristics of the fermentation process and the characteristics of the data with multiple sampling rates,and ignores the influence of measurement noise in the process data on the model.For this reason,a regression modeling method for fermentation process based on stacked denoising autoencoder is proposed.This method not only has strong nonlinear fitting ability,but also semi-supervised learning strategy can fully mine all data information in the fermentation process.At the same time,robust features can be extracted from the noisy process data,so that the model has certain noise adaptability.The results of penicillin simulation and comparison experiments show that the prediction performance of this model is better.
作者
岳向阳
赵忠盖
刘飞
YUE Xiangyang;ZHAO Zhonggai;LIU Fei(Ministerial Key Laboratory of Advanced Process Control for Light Industry(Ministry of Education),Jiangnan University,Wuxi 214122,China;Institute of Automation,Jiangnan University,Wuxi 214122,China)
出处
《计算机测量与控制》
2021年第7期136-139,155,共5页
Computer Measurement &Control
关键词
发酵过程
建模
栈式降噪自编码器
半监督学习
噪声
fermentation process
stacked denoising autoencoder
semi-supervised
modeling
noise