Application of ridge regression and factor analysis in design and production of alloy wheels

Abstract

This study proposes using statistical approaches to help with both the design and manufacture of wheels. The quality of a wheel is represented by the mechanical properties of spokes. Variation in the mechanical properties of different wheels is attributed to two sources, i.e. between-model variation and within-model variation. The between-model variation is due to different shapes of different wheel models. To model the effect of shapes on the mechanical properties, we first specify eight shape variables potentially critical to the mechanical properties, and then we collect relevant data on 18-wheel models and perform ridge regression to find the effects of these variables on the mechanical properties. These results are linked to the solidification theory of the A356 alloy. The within-model variation is due to natural variability and process abnormality. We extract mechanical data of a particular wheel model from the database. Factor analysis is employed to analyze the data with a view to identifying the latent factors behind the mechanical properties. We then look into the microstructure of the alloy to corroborate the fact that these two latent factors are essentially the Si phase and the Mg2Si phase, respectively. These results can be used to efficiently identify the root cause when the manufacturing process goes wrong.