Artificial intelligence as a tool for automatic state estimation and control of bioreactors

摘要

Microbial fermentations in real situations are difficult to monitor on-line or describe by mathematical models. This limits the applicability of model-based control under variable conditions. Methods using artificial intelligence (AI) have been very useful in this context. Their strength lies in their ability to learn from process performance, derive inferences akin to human reasoning and anticipate problems. Artificial neural networks, fuzzy logic and expert systems are common forms of AI for bioreactors. Often two or more of them have to be employed together for optimal performance. Sometimes AI is also used in conjunction with mathematical models and/or operations research techniques. With the right combination of AI and modelling, on-line estimation and control of complex fermentation processes can be done more rapidly, accurately and non-invasively than by instrumental or mathematical techniques and conventional control. © 1997 John Wiley & Sons, Inc. Lab Robotics and Automation 9: 297–304, 1997