Modeling and analysis of engine cold -test cells for optimizing driveline design for structural reliability and engine assembly defect diagnostics

Kamran Ahmed Gul, Purdue University

Abstract

Engine cold-testing techniques are used to determine engine assembly defects not easily characterized by the traditional hot tests. By running an unfired cold engine as a load using an electric motor and a driveline, steady state torque measurements can be used to determine cylinder compression quality, combustion yield energy for injector integrity, and reciprocating and rotating force abnormalities at low engine speeds. However, to diagnose transmission defects using gear noise measurements, transient tests must be run at significantly higher engine speeds. During these tests, cold engine test-cell drivelines experience large torsional amplitudes due to the excitation of the system resonances by various engine harmonics. These excessive torsional vibrations result in structural degradation of the test systems and also make the fault detection process difficult in two ways: by compromising the quality of torque waveforms and by preventing the accurate measurement of gear noise. To address the undesirable vibration and diagnostic signal degradation problems torsional vibration models of two different engine cold-test cells are first developed. These test-cell models are validated experimentally by comparing the model predictions of driveline vibratory torque and quality of waveforms with measured responses and are also used to generate appropriate design modifications in the drivelines. The driveline design optimization problem is challenging from several aspects: the coupling between the design variables, the discontinuous nature of variables and the conflicting objectives of minimizing torsional amplitudes as well as waveform distortion. An optimization scheme is proposed that utilizes embedded sensitivity functions to identify model parameters that help to suppress the torsional resonances and addresses the issue of coupling between the driveline design variables by treating them as discontinuous in the overall physically realizable range but continuous in local small ranges of variation where one parameter can be modified without affecting the other. It is shown that by using only the most influential model parameters as design variables, the optimization problem can be reduced to a lower dimensional search space which reduces the computation time thereby eliminating the need of advanced search methods. It also reduces the number of components to be modified to a minimum to produce cost-effective design modifications in the existing driveline designs. Based on this approach, it is found that by modifying the inertia and stiffness properties of the rubber coupling of the driveline, the resonant vibration problem of interest can be mitigated. The design modifications are implemented in both test cells resulting in a significant reduction in torsional amplitudes and waveform distortion levels with a corresponding increase in the sensitivity to faults. The modeling and analysis methodology developed in this work can be applied in designing future cold-test cells and other systems where vibration phenomena raise concerns of structural integrity and diagnostic signal degradation.

Degree

Ph.D.

Advisors

Adams, Purdue University.

Subject Area

Automotive engineering|Mechanical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS