Time series prediction techniques for estimating remaining useful lifetime of cutting tool failure

• Main Authors: Asmai, S. A., Burairah, Hussin, Mohd. Yusof, Mokhtar, Shibghatullah, Abdul Samad
• Format: Article
• Language: English
• Published: Praise Worthy Prize 2014
• Subjects: TJ Mechanical engineering and machinery; QA75 Electronic computers. Computer science; QA76 Computer software
• Online Access: http://eprints.utem.edu.my/id/eprint/14089/1/azirah_Irecos_V2_Accepted.doc; http://eprints.utem.edu.my/id/eprint/14089/

Prognostic have progressed over in the last few years as a specific function. It provides remaining useful lifetime (RUL) estimation of the targeted equipment or component in which able to be beneficially used by production or maintenance people to be readily advanced through preventive maintenance actions. In order to get accurate RUL for predicting future failure, RUL estimation is depending on the current condition of equipment. However, existing prognostic works use historical run-to-failure data and simulation-based model which is difficult to predict the future failure occurrence from the current certain level of degradation equipment. Therefore, this paper reported the use of time series prediction techniques in estimating RUL from established degradation index. Artificial Neural Network (ANN) with time series and Double Exponential Smoothing(DES) approaches with some modification is used to carry out the prediction steps. The modification obtained two variants of multi-step time series name predictions namely hybrid ANN-DES and Enhanced Double Exponential Smoothing (EDES). All the techniques are compared and evaluated to investigate the performance accuracy based on RMSE. The results shows that the EDES has a better solution in RUL estimation compare than other techniques