Experimental Validation of an Unscented Kalman Filter for Estimating Transient Engine Exhaust Composition with Fourier Transform Infrared Spectroscopy

摘要

Fourier transform infrared (FTIR) spectroscopy is a prevalent technique for measuring the comprehensive chemical composition of engine emissions. However, its applicability to transient emissions is limited due to recirculation of exhaust from past engine cycles within a FTIR gas cell and nonstationarity of the infrared beam intensity. An unscented Kalman filter is developed to overcome these limitations and obtain accurate, time-resolved estimations of engine exhaust composition from FTIR measurements. Residence time distribution within the FTIR gas cell is modeled using the well-mixed assumption, while the Fourier transform of an interferogram generated from a linearly evolving, uniformly broadened absorption line is used to deduce transient gas cell composition values from FTIR measurements. The filter utilizes both models, as well as measurement noise statistics, to infer the composition of sample entering the FTIR gas cell during a measurement period. To validate the filter, FTIR measurements of air with transient, trace amounts of acetylene are conducted. A variety of composition profiles are explored with different combinations of composition standard deviation (15 and 45 ppm) and duration between set points (0.4 and 1 s). The results demonstrate that measurement noise becomes less impactful as the magnitudes of composition fluctuations increase, while residence time effects become less significant as the duration of fluctuations increase. Improvements in estimated composition are achieved by the filter in every case, with an average improvement of 32% over unfiltered FTIR measurements. Experiments are conducted using sample flow rates of 12 and 25 standard liters per minute. More accurate measurements and estimations are attained at higher sample flow rates, highlighting the importance of maximizing flow rate to reduce residence time effects for transient measurements.