Monitoring diesel particulate filters
Abstract
Knowledge of the mass of particulate matter stored in a diesel particulate filter is imperative for making sound filter regeneration decisions. Currently, the mean-value pressure drop is used to estimate filter particulate load. However, this approach is fraught with accuracy and repeatability problems. In this research, dynamic pressure signal features are examined for this purpose. Specifically, the normalized firing frequency components are extracted from the dynamic pressures upstream, downstream, and across the filter using the complex Fourier Transform for periodic signals. Experimental data show that the firing frequency features appear to have smaller variability and better repeatability than the mean-value pressure drop. In back-to-back filter loading experiments, the mean-value pressure drop varied by around 25% while variations in the firing frequency features varied by less than about 7%. The firing frequency features can be used alone or in conjunction with the mean-value pressure drop to improve particulate load estimation. For a loading difference of 0.25 g/L in the diesel particulate filter, discrimination errors are as high as 3.4% using only the mean-value pressure drop, while 0% error is achieved using the firing frequency features. The discernible particulate load changes can be reduced to less than 0.25 g/L when the mean-value and firing frequency features are combined. A 1-D compressible unsteady-flow model was constructed for the flow in the channels of wall-flow diesel particulate filters to study the impact of parameter variations on the firing frequency features. Using this model, the permeability of the porous wall and particulate cake layer demonstrate a nonlinear relationship with the mean-value pressure drop and the firing frequency features. On the other hand, particulate cake layer density has an approximately linear relationship. The simulated firing frequency features were not normalized and showed similar percentage changes with parameter variations. This suggests that the normalization applied to the experimental data is important in suppressing the variations in the firing frequency features. Another important contribution of this research is the construction and validation of a particulate matter measuring system capable of approximately ±1 g accuracy during steady-state engine operation.
Degree
Ph.D.
Advisors
Meckl, Purdue University.
Subject Area
Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.