The full text of this item is not available at this time because the student has placed this item under an embargo for a period of time. The Libraries are not authorized to provide a copy of this work during the embargo period, even for Texas A&M users with NetID.
Data-Based Methods and Algorithms for Fault Detection and Diagnosis of Chemical and Bio-Chemical Processes
MetadataShow full item record
Fault detection and isolation techniques are important aspects in the chemical industry to ensure safe and proper operation. Some industrial processes are complex, and it is difficult to have an accurate mathematical model describing these processes. In such cases, data-based methods are useful tools for fault detection and diagnosis. Process data in chemical processes are usually contaminated with noise and are highly nonlinear in nature with complex correlated variables. The objective of this work is to develop methods and algorithms for early and accurate detection and isolation of faults within a process, allowing corrective actions to be taken quickly. First, a novel optimized diagnostic method for the detection of an abnormal event is developed by combining nonlinear fault detection methods with a composite hypothesis statistical test to accurately and reliably detect faults in the system. The multiscale kernel partial least square (MSKPLS)- based generalized likelihood ratio test (GLRT) algorithm has been developed to handle process noise and correlated data sets to increase reliability when determining faults within a chemical system. Once the fault has been detected in the process data, it is important to isolate and identify the faulty equipment. A fault isolation and identification approach is developed by combining the GLRT-based contribution plot to isolate the faulty variable in the process data, and the principle component analysis (PCA)-based Euclidian distance classifier to accurately identify faults. A novel hybrid observer and fault detection model is also developed for a batch bioreactor system to monitor and estimate the states of the bioreactor. The hybrid observer model was developed by a combination of the neural network partial least square (NNPLS) regression (to estimate a subset of process states) and an unknown input nonlinear observer or state-estimation technique to accurately determine a complete set of state estimates. In addition, a standalone software incorporating all the developed fault detection, diagnosis and prediction algorithms has been developed. The standalone software is designed with GUI (graphical user interface) capable of being applied both in an online and offline mode. The developed GUI software will aid in the implementation of the data-based algorithms to industrial processes.
Botre, Chiranjivi Chandrakant (2019). Data-Based Methods and Algorithms for Fault Detection and Diagnosis of Chemical and Bio-Chemical Processes. Doctoral dissertation, Texas A&M University. Available electronically from