Description
Summary:This thesis presents a comparison between methods for tool life prediction. The main objective of the thesis is to have an accurate prediction of the RUL and select the best method for prediction. An experiment has been conducted using Kistler dynamometer and Olympus metallurgical microscope on a HASS VF-6 milling machine to acquire the sensor force signals and actual tool wear respectively. The force signal gives the significant statistical features of the data. The features are extracted using statistical measure and reduced using a stepwise regression model. The prediction methods are Support Vector Regression and Neural Network. Both the models are trained using the MATLAB software. The results of the models are compared against each other to select the best method. Moreover, the methods are also applied on data taken from PHM Society. This data serves as a preliminary result and fundamental knowledge for my own experiment. The models trained in this project are compared with the existing models. These results show that the proposed methods are suitable for predicting the remaining useful life.