Robust Engine Design Using Engine Simulations 2003-01-0371
During the design stage, certification testing, or in field problem solving, t is of value for engine designers and engineers to have an understanding of how robust the engine design is to variation in the manufacturing process, in-use wear, controller and the testing processes. In this paper, a sensitivity analysis is performed on a parametric GTpower diesel engine model and using Robust Design methods NOx defects are reduced.
Sensitivity analysis is conducted using a Plackett-Burman DOE. The DOE is performed on a 6 cylinder, direct-injection, turbocharged diesel engine model in GTpower, while Minitab is used for the experimental design and the factorial sensitivity analysis.
It was found that the NOx population distribution was unacceptably high, yielding a 7.4% defect rate relative to an upper control limit of 5 (g/kw-hr). The sensitivity analysis provided a parameter ranking which allowed the identification of the vital few factors which were controlling the NOx and SFC populations.
While timing retard of 1 degree could reduce the NOx defect rate to an acceptable “three sigma level” of 0.135% or only 135 NOx defects in 100,000 engines, it simultaneously caused an unacceptable 1% increase in fuel consumption.
Alternatively, the Response Surface Optimization (RSM) method was used within AutoDOE to re-optimize the key parameter settings to not only reduce the NOx defect rate to “three sigma”, but do it without loss of fuel consumption.
This work demonstrates that sensitivity analysis and Robust Design & Optimization methods allow engine designers an efficient way to understand the impact of variability on engine performance for the entire production population and to control the engine performance to meet emissions requirements with less loss of fuel consumption than with one-at-a-time modifications to nominal design settings alone.
