Use Of Survival Models In A Refinery

CATEGORY: RELIABILITY
REVSTAT – Statistical Journal, Volume 11, Number 1, March 2013, 45–65
Use Of Survival Models In A Refinery
S´ılvia Madeira, Phd in Mathematics, ECT of University of ´ Evora, Portugal
silvia.madeira@galpenergia.com
Paulo Infante, CIMA/DMAT, ECT of University of ´ Evora, Portugal
pinfante@uevora.pt
Filipe Didelet, EST, IPS Set´ubal, Portugal
filipe.didelet@estsetubal.ips.pt
Abstract:
Statistical methods are nowadays increasingly useful in industrial engineering. From plant design reliability to equipment analysis, there is much to cover with statistical models in order to improve the efficiency of systems. At Sines refinery authors found it useful to apply a Cox model to a particular critical equipment trying to find process variables that cause its vibration as well as to apply well known distributions to baseline hazard rate.
Sines refinery has been concerned with the shut-downs made by the Turbo- Expander equipment in the FCC unit (Fluid Catalytic Cracking). As it was having problems due to a vibration failure mode causing FCC unit shut-down since year 2000, it was decided to investigate the origin of the vibration. Expander manufacturers and other entities that have this kind of equipment have investigated this problem as it is a big economical issue for the companies as we can see in  and. They highly recommend investigating efficiency in order to detect scaling deposition in expander rotor blades. It is believed that the com- position of some particles resulting from process reaction are the key for scaling, not just erosion. On 2011 turnaround, the procedure of changing the expander’s rotor and shroud was not complete. Only the rotor was changed. This has caused a slight gap between the shroud and the rotor blades due to shroud erosion. This gap is believed to cause less resistance and then, less particles deposition. However, results of Cox Proportional Hazards   demonstrate components in these particles to be linked with high values of expander vibration and with times to failure. In Section 3 authors have a brief description of the equipment and contextualization of the subject. In Section 4 they refer to the goal of this work and variables definition. Then they present some parametric and non-parametric approaches using Kaplan–Meier estimator and Cox Proportional Hazards in Sec- tion 5, and a parametric adjustment to the null model with parametric models. Finally, Section 6 is dedicated to some general considerations of this work.
Free Full Text Source: https://www.ine.pt/revstat/pdf/rs130103.pdf