Fault detection using spectral methods: Wavelets and correlation techniques
Abstract
Oxides of nitrogen (NOx), particulate matter (PM) and hydrocarbon (HC) emissions are closely regulated by the Environmental Protection Agency (EPA) and California Air Resources Board (ARB). Exhaust gas recirculation (EGR) is one of the in-cylinder NOx control strategies commonly used in diesel engines. EGR valve is an important component in the EGR loop and is used to precisely meter the EGR flow. EGR low-flow and high-flow result in increased PM, HC and NOx emissions respectively. Control valve stiction is known to cause limit cycling behavior because of the stick-slip effect. Such cycling behavior results in both low-flow and high-flow conditions. Besides stick-slip, control valve stiction also introduces a deadband into the system. The following work describes the stiction detection method. The proposed scheme consists of detecting each of the two stiction symptoms individually. A wavelet-based approach that checks for local regularity of the valve displacement signal is used to detect stick-slip. A time-delay estimation scheme based on estimates of the cross-correlation function between the valve's open loop input and output is used to detect the deadband. Besides this two-method detection scheme, an approach to detecting stiction using the estimates of the bispectrum of the displacement signal is also described. The data used in this exercise was generated using a simulation model of a control valve following a reference signal in a closed loop.
Degree
M.S.M.E.
Advisors
King, Purdue University.
Subject Area
Electrical engineering|Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.