Diagnosis of Process Nonlinearities and Valve Stiction: Data Driven Approaches (Advances in Industrial Control)

were published in the series as the contributed volume, Process Control Performance Assessment: From Theory to Implementation with Andrzej Ordys, Damian Uduehi, and Michael Johnson as Editors (ISBN 978-1-84628-623-0, 2007). Along with this good progress in process controller assessment methods, researchers have also been investigating techniques to diagnose what is causing the process or control loop degradation. This requires the use of on-line data to identify faults via new diagnostic indicators of typical process problems. A significant focus of some of this research has been the issue of valve problems; a research direction that has been motivated by some industrial statistics that show up to 40% of control loops having performance degradation attributable to valve problems. Shoukat Choudhury, Sirish Shah, and Nina Thornhill have been very active in this research field for a number of years and have written a coherent and consistent presentation of their many research results as this monograph, Diagnosis of Process Nonlinearities and Valve Stiction. The Advances in Industrial Control series is pleased to welcome this new and substantial contribution to the process diagnostic literature. The reader will find the exploitation of the extensive process data archives created by today’s process computer systems one theme in the monograph. From another viewpoint, the use of higher-order statistics could be considered to provide a continuing link to the earlier methods of the statistical process control paradigm. In this book, Higher Order Statistical (HOS) theory is used to develop indices for detecting and quantifying signal non-Gaussianity and nonlinearity. These indices, together with specific patterns in the mapping of process output and controller output are used to diagnose the causes of poor control loop performance. Often valve stiction is the main cause of poor control performance. A generalized definition of valve stiction based on the investigation of real plant data is proposed. A simple data-driven model of valve stiction is developed. The model is simple, yet powerful enough to properly simulate the complex valve stiction phenomena. Both open and closed loop results have been presented and validated to show the capability of the model. Conventional invasive methods such as the valve travel test can detect stiction easily. However, they are expensive, time consuming and tedious to use for examining thousands of valves in a typical process industry. A non-invasive method that can simultaneously detect and quantify control valve stiction is presented. The method requires only routine operating data from the process. Over a dozen industrial case studies have demonstrated the wide applicability and practicality of this method. In chemical industrial practice, data are often compressed for archival purposes, using various techniques. Compression degrades data quality and induces nonlinearity in the data. The issues of data quality degradation and nonlinearity induction due to compression are investigated in this book. An automatic method for detection and quantification of the compression present in the archived data is discussed. Compelling and quantitative analyses have been recommended to end the practice of process data compression. Process Control Performance Assessment is a practical guide to the application of control benchmarking to real, complex, industrial processes. It presents commercial solutions as well as current and future techniques still under development and contains real full-scale-implementation industrial case studies from the oil and gas, power and chemical industries providing a hierarchical perspective on benchmarks. The effect of benchmarking and control tuning on process performance and revenue optimisation is demonstrated. The most frequently occurring practical issues associated with implementing benchmarking algorithms in industrial processes are discussed in detailed and concise terms and guidelines given for avoiding common problems. The impact of problems like valve stiction and plant-wide disturbances on process performance is reviewed and methods for diagnosing causative fault conditions suggested. The benefit of non-intrusive and intrusive benchmarking techniques for optimising control systems using performance assessment and diagnostic tools is the potential to enhance company operational economics through increased throughput and better product quality. The book is appropriate for industrial and academic control and process engineers who wish to learn about methods of benchmarking through directed study or self study. Advances in Industrial Control aims to report and encourage the transfer of technology in control engineering. The rapid development of control technology has an impact on all areas of the control discipline. The series offers an opportunity for researchers to present an extended exposition of new work in all aspects of industrial control 'Process Control Performance Assessment'is a guide to the application of control benchmarking to industrial processes. It presents commercial solutions and techniques still under development and contains real industrial case studies from the oil and gas, power and chemical industries. The effect of benchmarking and control tuning on process performance and revenue optimisation is demonstrated. Practical issues to do with implementing benchmarking algorithms in industrial processes are covered in detail and guidelines for avoiding common problems given. The impact of problems like valve stiction and plant-wide disturbances on process performance is reviewed and methods for diagnosing causative fault conditions suggested. The benefits of using benchmarking techniques for optimising control systems are increased throughput and product quality. The book is appropriate for industrial and academic control and process engineers wishing to learn about methods of benchmarking through directed study or self study.

The subject matter of the book is concerned with the detection and diagnosis of process nonlinearities from routine process data. In general, processes can be treated as locally linear and measures of overall process performance can be monitored from routine operating data. However when process performance is not satisfactory then it is imperative that the cause of poor performance be diagnosed. Poor performance can be due to several reasons. Statistics abound on the cause of poor control performance. It has been documented that as many as 40% of the control loops in industry perform unsatisfactorily because of valve problems, a majority of them due to valve stiction, causing the closed loop system to become nonlinear. The development of signal processing methods to detect and quantify process nonlinearity from routine process data is the main subject matter of this book.