Regenerative thermal oxidizers (RTO) are commonly used in the pharmaceutical and chemical industries to treat emissions of volatile organic compounds but are prone to overheating risks stemming from variations in the flow and concentration of exhausted gases from diverse workshops. This paper proposes a dynamic analysis method based on correlation between various sensor-point data to determine the cause of the RTO overheating in such a multi-source waste gas treatment system. This method continuously adjusts the length of the data in the correlation analysis to comprehensively capture the dynamic changing characteristics of the data. And a strategy for the accumulation of association degrees is introduced to reveal long-standing potential causal relationships. The effectiveness of proposed method was validated in an industrial exhaust gas treatment system of a pharmaceutical company with over 1000 sensor points. The fault diagnosis results are consistent with those of the actual on-site investigation in 5 real cases. Our proposed method can provide relatively accurate results within 2–3 h, much faster than the manual investigation of the workers (2 days on average). Moreover, compared to traditional static correlation analysis methods, our method can narrow down the investigation scope by more than 5 times. This study provides a practical and effective method for tracing the source of anomalies in complex dynamic systems, and also contributes to the safe and stable operation of chemical engineering systems.
Read full abstract