In an earlier article, the central composite design was applied to the determination of geometrical features of casts in a two-phase transformation process to produce the wheel covers of automobiles whereby the A356 alloy is reinforced with organic substances for composite property enhancement. This article reexamines the assumptions in that circumstance to revise and expand the optimisation through the response surface methodology to a new method, Box-Behnken design (BBD), to facilitate a comprehensive treatment of the sand casting product parameters. Casting geometrical optimisation can be modelled to involve lengths, breadths, widths, heights, densities of casts and weight loss, varied at three discrete levels. The parameters are translated into codes (–1,0,1) with specified actual, minimum and maximum values. The framework, validated by published literature data, indicates its feasibility in a real-life circumstance. This article assessed the effects of the casting geometry parameters on the responses. Besides, it examined the accuracy of the parameters to predict in the regression models deployed. It was concluded that the BBD and the regression models are adequate and predict correctly. The BBD can be applied by composite developers to improve casting dimensional accuracy and economics.
Large sample theory for semiparametric regression models with two-phase, outcome dependent sampling
