On-Line Roughness Fault Detection Using Current Profile Measurement

With the advent of new technologies, customers are becoming more demanding of inexpensive and higher quality products. As a result, manufacturers are becoming keener to early fault detection during and post manufacturing processes. Fault detection has many benefits such as failure cost reduction, unplanned machine downtime reduction, and safety enhancement. In this paper, a hybrid fault detection method is proposed to differentiate between high and low roughness surfaces during turning using motor current measurement. The proposed method is based on modeling spindle motor current profile and using profile parameters to detect any major deviations from the average healthy state. Since several parameters are monitored simultaneously, multivariate statistical process control is utilized to flag deviations. The effectiveness of the proposed method is demonstrated, using two data sets of two different materials, Al 6061 and Al-SiC metal matrix composite. The proposed method successfully detected all faulty profiles in the two cases, when using sum of sine functions. However, one healthy profile was identified by proposed method as faulty in the case of Al-SiC. The proposed method is scalable and can be further refined and to actual manufacturing settings