This book is about the design procedure of soft sensors and their applications for solving a number of problems in industrial environments.
Industrial plants are being increasingly required to improve their production efficiency while respecting government laws that enforce tight limits on product specifications and on pollutant emissions, thus leading to ever more efficient measurement and control policies. In this context, the importance of monitoring a large set of process variables using adequate measuring devices is clear. However, a key obstacle to the implementation of large-scale plant monitoring and control policies is the high cost of on-line measurement devices.
Mathematical models of processes, designed on the basis of experimental data, via system identification procedures, can greatly help, both to reduce the need for measuring devices and to develop tight control policies. Mathematical models, designed with the objectives mentioned above, are known either as virtual sensors, soft sensors, or inferential models.
In the present book, design procedures for virtual sensors based on data-driven approaches are described from a theoretical point of view, and relevant case studies referring to real industrial applications, are described. The purpose of the book is to provide undergraduate and graduate students, researchers, and process technologists from industry, a monograph with basic information on the topic, suggesting step-by-step solutions to problems arising during the design phase. A set of industrial applications of soft sensors implemented in the real plants they were designed for, is introduced to highlight their potential.
Theoretical issues regarding soft sensor design are illustrated in the framework of specific industrial applications. This is one of the valuable aspects of the book; in fact, it allows the reader to observe the results of applying different strategies in practical cases. Also, the strategies adopted can be adapted to cope with a large number of real industrial problems.
The book is self-contained and is structured in order to guide the interested reader, even those not closely involved in inferential model design, in the development of their own soft sensors.
Moreover, a structured bibliography reporting the state of the art of the research into, and the applications of, soft sensors is given.
All the case studies reported in the book are the result of collaboration between the authors and a number of industrial partners. Some of the soft sensors developed are implemented on-line at industrial plants.
The book is structured in chapters that reflect the typical steps the designer should follow when developing his own applications. The reader can refer to the following scheme as a guide with which to search the book for solutions to particular aspects of a typical soft sensor design. Also, soft sensor design procedure is not straightforward and the designer sometimes needs to reconsider part of the design procedure. For this reason, in the scheme, a path represented by grey lines overlaps the book structure to represent possible soft sensor design evolution.
The state of the art on research into, and industrial applications of, soft sensors is reported in Chapter 1. Chapters 2 and 3 give some definitions and a short description of theoretical issues concerning soft sensor design procedures. Chapter 9 deals with the related topic of model-based fault detection and sensor validation, giving both the state of the art and two applications of sensor validation. Technical details of plants used as case studies are reported in the Appendix A.
As a complement to the bibliography section, where works cited in the book are listed, a structured bibliography is provided, in Appendix B, with the aim of guiding the reader in his or her search for contributions on specific aspects of soft sensor design.
Readers wishing to apply the techniques for soft sensor design described in the book will find data taken from real industrial applications in the book web site: www.springer.com/1-84628-479-1.