Effect of the AWJM Method on the Machined Surface Layer of AZ91D Magnesium Alloy and Simulation of Roughness Parameters Using Neural Networks.

This paper investigates the effect of change of the abrasive flow rate and the jet feed on the effectiveness of machining of AZ91D casting magnesium alloy. The evaluation of the state of the workpiece surface was based on surface and area roughness parameters (2D and 3D), which provided data on: irregularities formed on the workpiece edge surface (water jet exit), the surface quality after cutting, the workpiece surface chamfering, microhardness of the machined surface, and of specimen cross-sections (along the water jet impact). The process was tested for two parameter settings: abrasive flow rate 50 at cutting speed vf = 5⁻140 mm/min, and abrasive flow rate 100% (0.5 kg/min) at vf = 5⁻180 mm/min. The results demonstrate a significant effect of the abrasive flow rate and the jet feed velocity on the quality of machined surface (surface roughness and irregularities). In addition, selected 2D surface roughness parameters were modelled using artificial neural networks (radial basis function and multi-layered perceptron). It has been shown that neural networks are a suitable tool for prediction of surface roughness parameters in abrasive water jet machining (AWJM).