Tool breakage monitoring based on sequential hypothesis test in ultrasonic vibration-assisted drilling of CFRP.

Tool condition is highly relative to the productivity, quality, and safety of ultrasonic vibration-assisted drilling (UVAD) of carbon fiber-reinforced polymer (CFRP). Tool breakage can cause the degradation of drilling quality and maybe even lead to unexpected machine downtime. Therefore, tool breakage monitoring is the key technique of ensuring drilling quality and realizing fully automated drilling. However, existing tool breakage monitoring methods based on machine learning need training model previously, which is impractical for the actual drilling process. In this work, a novel tool breakage monitoring method based on a sequential probability ratio test (SPRT) in UVAD of CFRP is proposed. Three different damage levels are introduced to simulate the tool breakage in the drilling process. The vibration signals collected under different tool damage levels in the experiment are preprocessed by low-pass filtering to remove the disturbance frequency generated by the ultrasonic spindle system. To reduce data redundancy, the signals are downsampled according to the useful frequency band and the feature parameter extracted from test signals is finally fed into the SPRT model as a test sequence to recognize tool damage levels. Root mean square error (RMSE) between the same conditions and between different types of conditions was selected as the criteria to evaluate the reliability of the method. The test results and error analysis show that the method is effective and reliable to classify different tool breakage conditions during UVAD of CFRP. [ABSTRACT FROM AUTHOR]