Development of Adaptive Neuro Fuzzy Inference System Model for CNC Milling of AA5052 Alloy with Minimum Quantity Lubrication by Natural Cutting Fluid

In view of the improvements in manufacturing sectors, which are an important component of any economy’s growth, there is a significant need for new and advanced materials, particularly alloy materials need to be analysed and investigated so that new technologies may be effectively utilized. Materials with low weight and high strength, such as aluminium alloys, are recommended for a variety of applications that require both strength and corrosion resistance, such as marine applications and high-temperature applications. Aluminium alloy Al 5052, a nonferrous material with outstanding properties is an Al-Mg alloy with high thermal conductivity and corrosion products that are non-toxic. Minimum Quantity Lubrication (MQL) is a cost-effective and environmentally friendly method of lubrication employed in a variety of machining processes. The investigation of CNC milling of AA5052 alloy with standard Tungsten Carbide (WC) tool inserts with MQL settings are detailed in this paper. Coconut oil has been used as a natural cutting fluid in this research. The experiments have been planned using Taguchi’s experimental design approach, which considers speed, feed, and depth of cut as distinct process variables. An L27 orthogonal array has been selected for performing the experiments based on the input variables and their levels. Performance indicators include Material Removal Rate and Surface Roughness. Taguchi’s single response analysis was used to obtain the optimal process parameters. Grey system concept has been used for determining multiple performance index. ANOVA analysis is employed to examine the significant process parameters. Appropriate implementation of artificial intelligent decision-making technologies aids the manufacturer in achieving advantages in a variety of engineering applications. ANFIS is a predictive model for forecasting the necessary performance metrics that are constructed using this perspective. The resulting model’s performance is also examined and the predicted values were compared to the experimental values. Correlation coefficient of the prediction model is 0.99975. Error analysis has been conducted and the convergence of these values demonstrates the usefulness of the constructed model and demonstrates that it is capable of accurately anticipating the intended performance indicators.