theses_web.jpg

Thesis for the Degree of Master of Science

 

Year 2006
George Win Ainiveettil

Simulation case study of causal based reasoning and inference methods in paper making grade change process


Fault detection and isolation (FDI) is important in industrial processes considering the dynamic nature of the process variables. Fault detection is an indication of fault or disturbance in the system. Fault isolation is the method of finding the exact location of the fault .Heavy losses are incurred in the process industry as a result of faults such as process disturbances and sensor drifts. It is necessary to establish relations between the causes and the consequences of the fault in order to develop an effective fault detection and isolation system.

The literature review concentrates on general methods of fault detection and isolation based on causal based techniques. The emphasis is given on reasoning methods and assigning causality to process variables in industrial environment. For this purpose, case studies were investigated in different process industries. A general idea about the causal representation of the variables in the process, information flow, and algorithms for fault detection and isolation were investigated in industrial processes.

The experimental part of the thesis focused on applying a suitable method of fault detection and isolation. To carry out this task, a real time papermaking grade change process was selected. A dynamic simulator was used to carry out the necessary simulation runs. The process was analysed for normal conditions and faulty conditions. Based on the causal relations between the quality variables and simulation runs of the system, a model has been developed. The model approach relied on the operator knowledge and causal relations between the variables.

The causal model developed worked well under process disturbances. The fault diagnosis algorithm searched the paper machine sections based on the causal relations to identify the faults. The results showed the importances of a causal model in industrial applications. Flows of information between different sections of the process are important to identify the disturbances. The operator will be able to identify the cause of fault quickly and necessary corrective actions can be done.


This info last modified 24 Sep 2017 by Jerri Kämpe-Hellenius