Damage patterns recognition in dressing tools using PZT-based SHM and MLP networks

In order to promoting the optimization of the theme: "grinding-dressing", this study intends to contribute to the fill the gap of works completed with the damage diagnostic systems in dressing tools. For this purpose, this work aims to use neural models based on multilayer Perceptron networks (MLP) to improve the damage pattern recognition in diamond dressing tools based on electromechanical impedance (EMI). Thus, experimental dressing tests were performed with a single-point diamond-dressing tool and a low-cost lead zirconate titanate (PZT) transducer to acquire the impedance signatures at different dressing passes. The proposed approach was able to select the optimal frequency range in impedance signatures to determine the dressing tool condition. To achieve this, representative damage indices in several frequency bands were considered as input to the proposed intelligent system. This new approach open the door to effective implementation of future works for a broader situation in grinding process.