FRAMEWORK FOR ENHANCING FAULT-DIAGNOSIS CAPABILITIES OF ARTIFICIAL NEURAL NETWORKS

Neural networks have demonstrated excellent performance in facilitatin g automatic fault detection and diagnosis in many engineering applicat ions. Their primary advantage over model-based and knowledge-based exp ert systems is that they require very little development time or exper tise. However, neural networks perform only as robustly as the data fr om which they are trained. Therefore, understanding the content and li mitations of process data used to train the network is crucial. This p aper presents neural networks as part of a fault recognition framework for diagnosing process inefficiencies. In this framework, incorporati ng a small amount of process knowledge helps minimize data limitations and maximize the neural network's performance. Data preprocessing and filtering lead to significant improvement in recognition performance and markedly reduced training time. In addition, the framework allows detection of the ''unknown'' class. The fault recognition framework wi ll be demonstrated via a simulated continuous stirred tank reactor sys tem which operates under realistic disturbances and noisy measurements .