Stratified operation of a gasoline engine is one of the most efficient technologies for fuel economy improvement. This operation requires detailed knowledge and governance of component tolerances (fuel injector and spark plug) in order to ensure robust and smooth engine operation without unacceptable torque fluctuations. The coefficient of variation (COV) is a metric in engine development and calibration for fluctuation of indicated mean effective pressure (IMEP), resp. torque (critical to quality), which means for the customer that the engine is running smoothly (critical to satisfaction). It denotes the relation of standard deviation of IMEP over 300 combustion cycles to the average IMEP over these cycles. COV performance must be below the specified levels, as a function of operating point, which can be translated into limit states at chosen engine speeds.
In this study a meta-model is fitted to a large DoE data set and used for optimizing and verifying COV capability in a most probable limit state scenario, leading rather to test design than reliability prediction. The parameter strategy is simplified by eliminating the independence of two key input variables, in order to reduce the dimensionality of the noise space. By comparing the sensitivities at selected operating points, the most important responses are used to verify a most critical point with respect to assembly situation, in order to optimize the capability of the system against specified failure modes.