Machine Learning Based Predictive Modeling of Ball Nose End Milling using Exogeneous Autoregressive Moving Average Approach

Ball nose end milling is widely used for achieving high precision surface finish of free form aerospace components. Accurate predictive modeling is essential for effective control and automation of such high precision machining processes. In the present work, an LM6 aluminum alloy was ball nose end milled under varying machining conditions and the resultant surface roughness was recorded. This multi input single output machining system was modeled using the ARMAX (autoregressive moving average with exogenous inputs) structures as per the system identification methodology. Machine learning was employed for optimum parameter estimation of the ARMAX models. All derived models were validated on the basis of FIT%, FPE (final prediction errors), MSE (mean squared errors) and the number of model parameters. The ARMAX 6520 model emerged as the best performing structure with a FIT of 84.83%, FPE 1.61, MSE 0.02729 and 23 parameters. This predictive model can be utilised for closed loop control of ball nose end milling for precision aluminum alloy aerospace components.