Optical diamond turning of rapidly solidified aluminium alloy grade - 431

The high demand for ultraprecision machining systems is increasing day by day. The technology leads to increased productivity and quality manufactured products, with an excellent surface finish. Therefore, these products are in demand in many industrial fields such as space, national defence, the medical industry and other high-tech industries. Single point diamond turning (SPDT) is the core technology of ultraprecision machining, which makes use of single-point crystalline diamond as a cutting tool. This technique is used for machining an extensive selection of complex optical surfaces and other engineering products with a quality surface finish. SPDT can achieve dimensional tolerances in order of 0.01um and surface roughness in order of 1nm. SPDT is not restricted, but mostly applicable, to non-ferrous alloys; due to their reflective properties and microstructure that discourages tool wear. The focus of this study is the development of predictive optimisation models, used to analyse the influence of machining parameters (speed, feed, and depth of cut) on surface roughness. Moreover, the study aims to obtain the optimal machining parameters that would lead to minimum surface roughness during the diamond turning of Rapidly Solidified Aluminium (RSA) 431. In this study, Precitech Nanoform 250 Ultra grind machine was used to perform two experiments on RSA 431. The first machining process, experiment 1, was carried out using pressurized kerosene mist; while experiment 2 was carried out with water as the cutting fluid. In each experiment, machine parameters were varied at intervals and the surface roughness of the workpiece was measured at each variation. The measurements were taken through a contact method using Taylor Hobson PGI Dimension XL surface Profilometer. Acoustic emission (AE) was employed as a precision sensing technique – to optimize the machining quality process and provide indications of the expected surface roughness. The results obtained revealed that better surface roughness can be generated when RSA 431 is diamond-turned using water as a cutting fluid, rather than kerosene mist. Predictive models for surface roughness were developed for each experiment, using response surface methodology (RSM) and artificial neural networks (ANN). Moreover, RSM was used for optimisation. Time domain features acquired from AE signals, together with the three cutting parameters, were used as input parameters in the ANN design. The results of the predictive models show a close relationship between the predicted values and the experimental values for surface roughness. The developed models have been compared in terms of accuracy and cost of computation - using the mean absolute percentage error (MAPE)