NEURAL NETWORK MODELING OF THE EFFECT OF COOLING SLOPE CASTING PARAMETERS ON PARTICLE SIZE OF PRIMARY SILICON CRYSTALS OF SEMISOLID CAST INGOTS OF Al-20Si (wt%)

Abstract: Cooling slope-casting processing is a relatively new technique to produce semisolid cast feedstock for the thixoforming process. Simple equipment, ease of operation, and low processing costs are the main advantages of this process in comparison with existing processes such as mechanical stirring, electromagnetic stirring, etc. The processing parameters of cooling slope casting are length, angle and the material of the inclined plate and their combinations, which usually affect the micro structural evolutions of the primary solid phase. In order to clarify the effect of the processing parameters on the evolution of the particle size, based on experimental investigation, Artificial Neural Network (ANN) was applied to predict the primary silicon crystals (PSCs) size of semisolid cast ingot via a cooling slope casting process of Al-20%(wt.%) Si alloy. The results demonstrated that the ANN, with 2 hidden layers and topology (4, 3), could predict the primary particle size with a high accuracy of 94%. The sensitivity analysis also revealed that material of the cooling slope had the largest effect on particle size.