Application of 1-D Convolutional Neural Network for Cutting Tool Condition Monitoring: A Classification Approach

In any machining activity, the role of cutting tool is to remove excessive material while producing desirable surface finish of workpiece. Tool Condition Monitoring (TCM) assists early detection of wear/faults, which in turn increase productivity and reduce downtime. In this paper, application of 1-D Convolutional Neural Network (CNN) is presented that classifies cutting tool state as healthy or faulty. Total five distinct tool conditions (1 normal and 5 faulty) were collected in terms of tool post vibrations evolved during machining. Deep Neural Networks require large amount of data to yield better results, data augmentation was performed with the help of up-sampling. Detailed design and training of 1-D CNN is described along with hyper parameters tuning followed by testing of trained classifier considering train/test data and independent blind data. This model performed 88% of classification accuracy for blind data.