# Knowledge base: Warsaw University of Technology

Back

## Metoda doboru zbioru sensorów dla diagnostyki procesów przemysłowych na podstawie grafu przyczynowo-skutkowego

### Anna Sztyber

#### Abstract

The aim of this dissertation is to develop a method for selecting a set of sensors for diagnostics of industrial processes. Diagnosed process is described by a Graph of a Process (GP) (Ostasz, 2006), which is a causal graph with additional knowledge regarding faults. Graph vertices represents process variables, control signals, measurements, and faults. Directed edges shows causal influences between variables. The sensor placement problem for fault diagnosis is formulated as follows: find the set of measuring instruments providing fault detectability and isolability. There is often lack of detailed description of industrial processes, therefore qualitative models are widely used. In this thesis comprehensive survey of qualitative models in fault diagnosis is presented including critical analysis of their features. Graph of a Process is used for finding model structures for fault detection. Model structure is described by an output variable and a set of input variables. Each structure can be used for finding neural, fuzzy or parametric model. New method is presented for finding all models structures given the set of sensors. On this basis, fault-symptom relation can be found and fault detectability and isolability can be obtained. Necessary and sufficient conditions for the set of measuring instruments are formulated, that diagnostics specification is met. Next, algorithms are presented suitable for searching the space of solutions. Last part of the dissertation contains comparison of obtained solutions with results of structural analysis.
Record ID
Diploma type
Doctor of Philosophy
Author
Title in Polish
Metoda doboru zbioru sensorów dla diagnostyki procesów przemysłowych na podstawie grafu przyczynowo-skutkowego
Language
(pl) Polish
Certifying Unit
Faculty of Mechatronics (FM)
Discipline
automation and robotics / (technology domain) / (technological sciences)
Status
Finished
Defense Date
04-03-2015
Supervisor
External reviewers
Antoni Ligęza, prof. AGH Kraków Antoni Ligęza, prof. AGH Kraków,, Undefined Affiliation
Marcin Witczak, prof. Uniw. Zielonogórski Marcin Witczak, prof. Uniw. Zielonogórski,, Undefined Affiliation
Pages
151
Keywords in English
Graph vertices represents process variables, control signals, measurements, and faults.
Abstract in English
The aim of this dissertation is to develop a method for selecting a set of sensors for diagnostics of industrial processes. Diagnosed process is described by a Graph of a Process (GP) (Ostasz, 2006), which is a causal graph with additional knowledge regarding faults. Graph vertices represents process variables, control signals, measurements, and faults. Directed edges shows causal influences between variables. The sensor placement problem for fault diagnosis is formulated as follows: find the set of measuring instruments providing fault detectability and isolability. There is often lack of detailed description of industrial processes, therefore qualitative models are widely used. In this thesis comprehensive survey of qualitative models in fault diagnosis is presented including critical analysis of their features. Graph of a Process is used for finding model structures for fault detection. Model structure is described by an output variable and a set of input variables. Each structure can be used for finding neural, fuzzy or parametric model. New method is presented for finding all models structures given the set of sensors. On this basis, fault-symptom relation can be found and fault detectability and isolability can be obtained. Necessary and sufficient conditions for the set of measuring instruments are formulated, that diagnostics specification is met. Next, algorithms are presented suitable for searching the space of solutions. Last part of the dissertation contains comparison of obtained solutions with results of structural analysis.
Thesis file
• File: 1
PhD_all.pdf
Request has been registered
Citation count
2

Uniform Resource Identifier
urn:pw-repo:WUTadeb592ff2c64eab8f4151976a64950b