Doris Payne
The wastewater from the pulping and papermaking industry is one of the main industrial pollution sources and hard to be treated because of complex composition. Once the effluent does not meet the discharge standard, it will cause great implications for the environment. Thus, some important indicators which reflect the effluent quality of papermaking Wastewater Treatment Processes (WWTPs), such as Chemical Oxygen Demand (COD), have caught much attention of researchers and operators. . Generally, the information acquirement of these indicators just can rely on traditional manual measurements and assays, which makes a great challenge for process monitoring. In order to know the operating state of a treatment system, the sensors with high accuracy and short response time are particularly important. However, hostile operating conditions and some inevitable disturbances make the measurement accuracy of sensors decrease over time. Therefore, designing an advanced monitoring scheme to improve the detection ability of sensor faults is important for industrial processes. To this end, many traditional Multivariate Statistical Process Monitoring (MSPM) methods, such as Principal Component Analysis (PCA) and Partial Least Squares (PLS), have been successfully applied to develop an effective process model proposed a multi-scale PCA method extracting cross-correlation and auto-correlation characteristics of sensors in detecting and identifying industrial boiler process faults. Applied PCA to discover the fixed and drifting biases of sensors, while the joint angle analysis technique was used to isolate these faults in variable air volume systems. These PCA-based methods are usually based on an assumption that data follow a gaussian distribution. In practice, however, some collected variables have non-Gaussian behavior. Thus, Independent Component Analysis (ICA) and its extension methods are proposed to solve this problem.
