Use this URL to cite or link to this record in EThOS:
Title: Detection and diagnosis of distributed disturbances in chemical processes
Author: Thornhill, Nina Frances
ISNI:       0000 0001 3532 8266
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2005
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This subject of this thesis is the detection and diagnosis of distributed disturbances in chemical processes. A distributed disturbance affects many variables such as feed, product and recycle flows, column temperature and product composition. It may upset just a single unit for example a distillation column, it may be plant-wide if it affects a complete production process or even site-wide if utilities such as the steam system are involved. Disturbances have an impact on profitability because production and throughput may have to back away from their maximum settings to accommodate process variability. The research has used signal processing, spectral analysis and non-linear time series analysis of measurements from routine process operations and has led to new applications of these methods in chemical process diagnosis. In particular, the use of principal component analysis on the power spectra of process measurements has given a breakthrough in the analysis of non-steady processes because the spectra are invariant to the lags and time delays that can make PCA unreliable in the time domain. The thesis offers novel methods and theoretical insights to support the industrial activity of detection and diagnosis of distributed disturbances. A key insight has been that non- linearity in the time trends of plant measurements is greatest in those measurements closest to the root cause because mechanical filtering by the plant makes the signals more linear as the disturbance propagates away from the source. A non-linearity index derived from process measurements can therefore locate the root cause of a disturbance. A feature of the work has been its focus on industrial implementation. The methods are demonstrated with data from real processes and care was taken to devise robust default settings of parameters in the algorithms to facilitate their application in unseen plants. As demonstrated in a case study, the outcomes of the work will significantly reduce the time spent on analysis and focus attention towards root causes of faults so that maintenance effort is directed effectively.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available