Chatter Identification on a Mobile Milling Machine: Experimental Insights Using Low-frequency Internal Sensors

On-site milling with mobile machine tools is a popular solution for the modification of large or immobile metal parts. For such operations the transportation and setup of the machine constitutes a delicate process. Hence, mobile machine designs must find a good compromise between power, structural rigidity, and adaptability. However, this compromise can lead to flexible structures which are susceptible to chatter.

The present article is dedicated to chatter identification in a numerically controlled mobile milling machine using low-frequency internal sensors. This solution is low-cost and removes the need for further sensor installation during on-site operations. Milling operations are carried out for a wide range of parameters and chatter is diagnosed using a high-frequency accelerometer. Then, time domain statistical features on the internal sensor signals are computed to evaluate their potential for chatter identification. Two strategies for chatter classification are tested: firstly, using an empirically determined threshold on the features, and, secondly, using Support Vector Machine (SVM).

Using the linear thresholding technique, maximum agreement between the accelerometer classification and the internal sensor classification is 91 % for chatter and non chatter data. However, using the C-SVM method, classification was improved up to a 98 % agreement for the chatter cases and a 95 % agreement for the non-chatter cases.

Thus the authors conclude that using a C-SVM classification method allows accurate chatter detection in the mobile milling machine using internal sensor data.