Novel Data-Based and Model-Based Algorithms for Process Monitoring and Equipment Degradation Tracking

Process monitoring is a critical component of many industries, required in order to maintain product quality and enhance process safety, thereby increasing economic benefits. Process monitoring methods provide a means of determining if a process is operating as expected, or if it is experiencing faulty or abhorrent conditions, e.g., process drifts or disturbances that disrupt the operation, which can result in plant shutdowns and economic losses due to down time and maintenance. Process monitoring methods can be broadly categorized into qualitative based models, quantitative based models, and data-based models. A primary objective of this work is to enhance the performance of monitoring algorithms by integrating the advantages of various data-driven and model-based methods. Data-based fault detection methods such as principal component analysis (PCA) and its extensions, will be integrated with composite hypothesis tests, such as the generalized likelihood ratio (GLR) charts in order to obtain superior fault detection performance when compared to conventional methods. The applicability of the developed fault detection algorithms will be examined using different illustrative examples, such as the Tennessee Eastman (TE) process. Monitoring process drifts and equipment degradation is another area of concern in process industries. Therefore, a second objective of this work is to develop an algorithm capable of detecting drifts in processes and equipment degradation, even when operating under control, by utilizing state estimation methods that are able to determine when a process is operating under sub-par conditions. The developed algorithm will be applied on an illustrative example of a heat exchanger, using both simulated synthetic and experimental data, to demonstrate its simplicity and practical applicability. This should enable the process engineer to make better executive decisions regarding the running of the plant. Pipeline flow and leak detection, specifically in subsea pipelines is another important issue that needs to be addressed, and therefore a third objective of this work is to design and develop an experimental setup to collect different sensor measurements, and utilize different fault detection and classification algorithms in order to study pipeline flow behavior