Type
Text
Type
Dissertation
Advisor
Tang, Wendy | Robertazzi, Thomas G | Hong, Sangjin | Brown, Kevin A | Tsoupas, Nicholaos.
Date
2015-12-01
Keywords
Electrical engineering | Accelerator, Bayesian, Machine Protection, Reliability, Superconducting magnets, System identification
Department
Department of Electrical Engineering.
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/77457
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
Relativistic Heavy Ion Collider (RHIC) at Brookhaven National Laboratory is used to study primordial form of matter that existed in the universe shortly after the Big Bang. Enormous energy (72 MJ) is stored inside RHIC in the form of ion beams and superconducting magnet currents during operation. The accelerator Machine Protection System (MPS) is used to safeguard against undesirable energy leakage due to the faults developing in the collider, and needs to be highly reliable. The most crucial parts of MPS are the Beam Permit System (BPS) and the Quench Detection System (QDS). The BPS monitors the health of RHIC subsystems and takes active decisions regarding safe disposal of the stored energy. The first segment of this dissertation aims towards Bayesian reliability analysis and quantitative estimation of system level catastrophic events of BPS which can lead to significant downtime of RHIC, and to identify the weak links in the system. A dynamic Monte Carlo failure model is developed, with modules having exponential lifetime distribution with competing risks. The module failures are calculated by Fault Tree Analysis, which traces down system level failures to component failures. This model is verified by an equivalent mathematical probabilistic model. A Bayesian reliability model is then employed to integrate the failure model and the historical failure data of BPS. It is based on a two-parameter Weibull distribution with unknown scale and shape parameters, and implemented using Markov Chain Monte Carlo algorithm. The QDS is responsible for detecting the superconducting magnet quenches and initiates the magnet energy dump. The second segment of this dissertation aims towards the accurate determination of developing quench failures, through remodeling the superconducting magnet behavior using Nonlinear System Identification. This reduces the false failures in the system, thereby enhancing the availability of the system. A mathematical memory model is conceptualized to define the highly nonlinear behavior of magnets undergoing saturation and hysteresis. This model shows good compliance with the data. It eliminates the manual calibration of hundreds of magnet lookup tables every year. More importantly, this work generates design recommendations for reliable protection systems of upcoming eRHIC project at Brookhaven National Laboratory, first of its kind in the world. | 136 pages
Recommended Citation
Chitnis, Prachi, "Bayesian Reliability Analysis and Nonlinear System Identification for Complex Particle Accelerator Protection Systems" (2015). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 3269.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/3269