Fault Detection Systems Integrated to Fault-Tolerant Control. Application to Large-Scale Chemical Processes

DAVID A. R. ZUMOFFEN; MARTA BASUALDO

LAP Lambert

Lugar: Saarbrücken; Año: 2012 p. 230

978-3-8484-2794-9

In the last three decades the interest for monitoring systems applied to large-scale chemical processes was increased notably. This is essentially due to the need for more demanding operating conditions of the plant related to security for equipment and personnel, operating costs and environmental restrictions. The increasing complexity in the large-scale process designs and their corresponding control policies need more sophisticated monitoring systems in aspects such as detection rate, robustness, user friendly, modeling and data storage requirements, adaptability, among others. The strong interaction between plant information and control occurs primarily through sensors and actuators. However, these elements are potential sources of common faults in industrial processes. In this context, it is observed that the broad topic of monitoring system designs integrated to fault-tolerant control strategies is still an open problem. Currently, only a few solutions for large chemical plants have been found, most of them applied to academic cases wherethe design framework can not be generalized to others.This book is focused on addressing the fault detection, diagnosis and estimation system (FDDES) design integrated to fault-tolerant control (FTC) for chemical processes. The main objective is automatizing the actions for the abnormal situation management (ASM) through a suitable interaction with the existing control policy. According to the problem dimension dierent tools must be used in the proposed approach. Thus, dierent kind of solutions are considered for single process unit than for large-scale process which have several interacting units. In all cases classical and advanced control policies are considered. The main contribution of this book is to present an improved hybrid FDDES design independent of the plant dimension, complexity, operability, sensors type and control strategies. Tools such as discrete wavelet transform (DWT)and system identication (SI) for lower dimension cases are accounted. Meanwhile, principal components analysis (PCA), fuzzy-logic systems (FLS), and articial neural network (ANN) were integratedsuitably for the FDDES development for large scale plants. Active FTC strategies, either by using the existing control policy or by developing a new one are considered.The book is organized as follows: Chapter 1 introduces the framework of real problems associated with faults commonly found in industrial processes and their consequences. Chapter 2 presents an overview of concepts strongly related with ASM, monitoring system design and FTC from both academic and industrial point of view. Chapter 3 summarizes typical tools, for the two main areas involved here such as process supervision and control eld, used along the book. Chapter 4 addresses the FDDES design problem mainly focused to isolated process units which involve a reduced number of variables. The main results obtained via robust adaptive predictive control (APC) as active FTC strategy are presented in this chapter. Since the results obtained are compared with and without fault detection system, a rigorous assessment is made about the integrated approach FDDES-FTC . Designs based on DWT and SI are discussed here for FDDES development. The final approach, in this chapter, was tested on both academic and industrial systems. The latter is represented by the continuos stirred tank reactor (CSTR) with jacket modeled in Matlab environment. Moreover, new APC alternatives, integration policies and ASM for single control loops are proposed. Typical faults such as oset in sensors, blockage in valves and extra delay in actuators are analyzed by considering their occurrence sequentiallyand insolated. In addition, dierent process operating conditions were tested. Chapter 5 displays a complex hybrid FDDES design for large-scale processes and its integration with FTC for existing control policies. The waste water treatment plant (WWTP) and the pulp mill process (PMP) are proposed here as case studies for testing the approach. The PMP is the largest and most complex benchmark analyzed by the process control community. Moreover, new performance indexes are developed for doing a suitable evaluation about the FDDES behavior as well as the process operating costs with and without FTC. A complete simulation set is presented accounting several operating scenarios which support the nal conclusions presented in Chapter 6. The latter summarizes some future directions in this research eld. In addition, at the end of the book, the appendices are included for give a more deep treatment on specic topics such as recursive identication, predictions with linear models, factorization algorithms, internal model-based control (IMC), and feedforward control (FFC).