An Expert System for Real-time Fault Diagnosis and Its Application in PTA Process

Increasing complexity and automation of the chemical process industries requires more reliable and efficient real-time fault diagnosis systems. Here, a real-time expert system based on wavelet transform and fuzzy ART neural network is introduced for fault diagnosis, providing fault prediction to help operators before abnormal situations occur. Data preprocessing, knowledge base structure, representation of knowledge, generalized inference engine and graphic user interface are technically considered. Industrial applications in pure terephthalic acid (PTA) process indicate that the real-time expert system diagnoses abnormal events efficiently and promptly and it has many specialties such as friendly interface, easy to train and maintain and also reliable under changing process conditions