|title:||Lifecycle simulation using Baysian calculation|
|author:||Ferdinand H. Utomo|
|published in:||January 2001|
Master of Science thesis
Delft University of Technology
|PDF (2.753 KB)|
This thesis covers the Life Cycle Simulation in the Process Industry. It
covers two main areas: 1) apply the Bayesian approach for assessing the degradation
of equipment of process plants and derive applicable knowledge from the results; and
2) develop a data model that facilitates the intake of diverse data from different
sources: the process applications with which the chemical processes are designed and
projected as a prior expectation and the real time process management and ERP-applications
that provide the details with which prior estimates can be checked, thus
providing the outcome of the Bayesian assessment. As a result, an application has
been designed, that incorporates the different aspects of the thesis.
The thesis provides feedback on a chemical process that is analyzed. It shows the logic of the modeling that is needed, when applying Bayesian calculation. As well, the Bayes’ approach and how it can be applied in a plant environment is discussed. It is concluded that this approach seems to be fit for assessing degradation patterns of equipment, to express that degradation in lost revenue over the years to come and to derive meaningful advice to the organization in terms of: change maintenance patterns or replace equipment items due to excessive degradation.
To apply this Bayes’ approach in a practical setting, the data streams of different applications have to be imported and prepared for the simulation. In this thesis, the data model that has to support the simulation and the flexibility that is needed to be able to integrate new data, are discussed. It has been illustrated that a hierarchical object oriented set-up provides a logical data model and that from the model, data can be retrieved in a flexible way, so the Bayes’ approach can be changed and fed with new variables at will.
The thesis then highlights the way the interface is used to derive helpful conclusions to the user of the simulation. The principles on how data should be provided are discussed and it is explained how this guidance is applied for this particular interface. In this part, it is as well illustrated that the usefulness of the application is enhanced by abstracting knowledge from the Bayesian calculations; and it is visualized how to provide this as a consult to the end-user.
In the final part of the thesis, the effectiveness of the Bayes approach in the specific plant setting, a process that is run at the University of Delft as a test-bed, is analyzed