Online detection of run out in microdrilling of tungsten and titanium alloys

Run out is one of the major problems in microdrilling processes, causing unexpectedly short tool life, sudden breakage, and dimensional inaccuracy. In this study, a two-step monitoring system for run out detection is proposed. The first step uses the fast Fourier transform for extracting features from the online measured force signals. In the second step, a neural network-based model predicts the process condition from the previously obtained features. The model was trained and tested by using force signals obtained from tungsten and titanium alloys, which are widely applied in electronic and aerospace industries. A 0.1-mm-diameter microdrill was used in the experimental study, and three different feed rates were applied for each material. The trained model was validated with data that was not used in the training process. In this validation, the system was able to detect more than 70 % of the run out conditions with less than 10 % of false detections. For microdrills, detecting and reducing run out can yield considerable gains in tool life and productivity.