Probabilistic analysis of manufacturing uncertainties for an automotive turbocharger centrifugal compressor using numerical and experimental methods

Abstract

The compressor impellers for mass-produced turbochargers are generally die-casted and machined to their final configuration. Manufacturing uncertainties are inherently introduced as stochastic dimensional deviations in the impeller geometry. These deviations eventually propagate into the compressor functionality as variability in output performance. This paper presents a probabilistic evaluation of the impact of manufacturing uncertainties on the compressor performance using the non-intrusive Monte Carlo method coupled with metamodels prepared for compressor performance modeling using computational fluid dynamics (CFD) and comprehensive experimentation. The probabilistic evaluation or the uncertainty quantification (UQ) of manufacturing uncertainties has identified the most sensitive and uncertain of the impeller geometric parameters contributing the most to variability in compressor performance. Consequently, the tolerances of different impeller dimensions can be reviewed, thereby influencing the production quality and costs.