Open Collections

An analysis of three model-based estimation methods for diesel engine condition monitoring

University of British Columbia

In the context of engine fault detection and isolation, we focus on three main aspects: engine modeling and validation, engine parameter identification, and cylinder pressure waveform reconstruction. The problem of diesel engine modeling is solved using Euler's equation, under the assumption that the crankshaft is perfectly rigid. The model inputs are represented by the cylinder pressures and the output is the flywheel angular velocity fluctuation. Simulation results obtained with MATLAB show a root mean square (RMS) error of 0.0891 rad/sec between the estimated and the actual crankshaft angular velocity fluctuation for the normal operating condition. The elimination of a strong sinusoidal trend for the faulty condition results in a RMS error range of 0.0973 rad/sec to 0.1836 rad/sec. The identification methods involved the off-line standard least-squares technique, the recursive gradient estimator and the on-line least-squares estimators with exponential forgetting. The parameters of interest are engine inertia, and torque fluctuation. The RMS velocity error for the normal operation has a value of 0.0559 rad/sec, which represents approximately 30% improvement over the initial result, before identification. The issue of cylinder pressure waveform reconstruction is addressed. The inverse dynamics are solved by redefining the system input as the torque due to gas pressure. The cylinder pressure waveform is approximated by an impulse-like periodic function. We considered the problem of fault detection and isolation. The procedure uses 6 pressure templates. The estimated pressure variations are obtained using a standard least-squares technique. An under-fueling fault in the i-th firing cylinder can be determined exactly by the minimum value of the estimated pressure variation. Using the pressure correction we are able to improve the estimation of the gas pressure torque. The RMS torque error for the normal operation reduces to 99.24 Nm. The case of an under-fueling fault is characterized by a reduced RMS torque error range of 85.7 Nm to 198.1 Nm. The pressure waveform reconstruction is characterized by a RMS pressure error range of 0.155 MPa to 0.277 MPa for the normal operating condition. For each of the six under-fueling faults, the pressure waveform corresponding to the faulty cylinder is reconstructed. The RMS error range is of 0.155 MPa to 0.386 MPa.

Thesis/Dissertation

application/pdf

2009-01-18

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/3771

Electrical and Computer Engineering

University of British Columbia

UBCV

DSpace