Monitoring metal removal in machining processes has proved to be essential for companies seeking a high level of excellence in the quality of their products and processes, contributing to improved resource allocation and reduced wastage due to nonconforming parts. Multisensory approaches have been employed to monitor these processes, aiming to use signals to train artificial intelligence systems to perform the task of indicating nonconformities in the tools or the product being manufactured. In this study, three artificial intelligence systems were used to estimate diameter of holes produced in sandwich plates—Ti6Al4V alloy was mounted in AA 2024-T3 alloy—and cutting conditions were selected to simulate a common aircraft fuselage manufacturing process. A multilayer perceptron artificial neural network (MLP ANN), an adaptive neuro-fuzzy inference system (ANFIS) and a radial basis function (RBF) neural network were trained to estimate the diameter of machined holes. The multisensory approach includes an acoustic emission sensor, accelerometer, dynamometer and an electric power sensor. The optimum configuration for each artificial intelligence system was determined based on algorithms designed to examine the influence of each system’s signals and specific parameters on the final result of the estimate. The results indicated the MLP ANN was more robust in withstanding data variation. The ANFIS system and RBF network showed markedly varying results in response to variations in the obtained data during training, suggesting these systems should always be trained with the dataset presented in the same order. A satisfactory response between the multisensory approach and MLP network was observed. The vertical component of force, along the z axis, was the only parameter able to present valid results for all the artificial intelligence systems analysed.
Read full abstract