An investigation of data-based fault detection methods in petroleum refineries için kapak resmi
An investigation of data-based fault detection methods in petroleum refineries
Başlık:
An investigation of data-based fault detection methods in petroleum refineries
Yazar:
Yasmal, Aslı, author.
Yazar Ek Girişi:
Fiziksel Tanımlama:
x, 91 leaves: charts;+ 1 computer laser optical disc.
Özet:
The petroleum refineries are complex systems vital for energy and production sectors. During production, these complex systems might experience various faults, including fluid leaks in unit operations. The detection of leaks is important for a reliable, safe, and efficient operation. Among the possible leak detection mechanisms, data-based leak detection methods are promising in terms of low investment cost, less human intervention, ability to detect small leaks in advance and direct integration capability to distributed control systems. The aim of this study is to investigate data-based leak detection methods in a heat exchanger in a petroleum refinery. To that end, possible leaking problems in petroleum refineries are assessed, multiple leak cases from a real heat exchanger in a petroleum refinery are determined, literature studies are searched for appropriate data-based leak detection methods, applicability of a set of data-based leak detection methods is studied with a literature benchmark data set, and the real cases of heat exchanger leaks are studied with the determined leak detection methods. Data sets for multiple leak cases of a heat exchanger are obtained from a TUPRAS refinery. The benchmark data set is obtained from Tennessee Eastman Process (TEP). Discrete Wavelet Transform (DWT), Auto Encoder (AE), and Exponentially Weighted Moving Average (EWMA) are selected as the data-based leak detection methods. The selected data-based methods are first studied with TEP data set, and good fault detection capability is observed. Then, the real leak cases are studied. All three data-based methods are found successful in detecting the actual leak cases. For some of the cases, leaks are detected with data-based methods in advance of the operation engineers noticing the leak.
Yazar Ek Girişi:
Tek Biçim Eser Adı:
Thesis (Master)--İzmir Institute of Technology: Chemical Engineering.

İzmir Institute of Technology:Chemical Engineering--Thesis (Master).
Elektronik Erişim:
Access to Electronic Versiyon.
Ayırtma: Copies: