A data-driven prognostic model using time series prediction techniques in maintenance decision making

Siti Azirah, Amai (2014) A data-driven prognostic model using time series prediction techniques in maintenance decision making. Doctoral thesis, Universiti Teknikal Malaysia Melaka.

	Text (24 Pages) A Data-Driven Prognostic Model Using Time Series Prediction Techniques In Maintenance Decision Making 24Pages.pdf - Submitted Version Download (262kB)
	Text (Full Text) A data-driven prognostic model using time series prediction techniques in maintenance decision making.pdf - Submitted Version Restricted to Registered users only Download (6MB)

Abstract

In recent years, current maintenance strategies have extensively evolved in condition-based maintenance solution in order to achieve a near-zero downtime of equipment function. One of these support elements is the use of prognostic. Prognostic has progressed as a specific function over for the last few years. It provides failure prediction and remaining useful lifetime (RUL) estimation of a targeted equipment or component. This estimation is beneficial for production or maintenance people as it allows them to focus on proactive rather than reactive action. While some prognostic models are created based on the historical failure data, others remain as simulation models serving as a pre-exposure effect analysis. Although the concept of a data-driven prognostics model using condition monitoring information has been widely proposed, the validation in predicting the target value continues to be a challenge. In addition, the prognostics have not been applied directly within the maintenance decision making. Hence, the aim of this study is to design a data driven prognostics model that predicts the series of future equipment condition iteratively and allows the process of maintenance decision making to be carried out. The initial phase of this research deals with a conceptual design of data-driven prognostics model. This conceptual design leads to the formulation of a generic data acquisition and time series prediction techniques, which are the key elements to predictive prognostic solution. In this case, there are four techniques have been used and formulated to have better prognostic results namely: Double Exponential Smoothing (DES), Neural Network (NN), Hybrid DES-NN and Enhanced Double Exponential Smoothing (EDES). The intermediate phase of this research involves the development of a computational tool based on the proposed conceptual model. This tool is used for model implementation that uses the experimental data to test the ability of the prognostics model for failure prediction and RUL estimation. It also demonstrates the integration of prognostics model in maintenance decision making. The final phase of this research demonstrates the implementation of the model using industry data. In this phase, the industrial implementation takes into account the performance accuracy to verify the operational framework. The results from the model implementations have shown that the proposed prognostic model can generate the degradation index from the data acquisition, and the formulated EDES can predict RUL estimation consistently. By integrating it with the maintenance cost model, the proposed prognostic model also can perform time–to-maintenance decision. However, the accuracy of the prognostic and maintenance results can be increased with a huge and quality data. In conclusion, this research contributes to the development of data-driven prognostics model based on condition monitoring information using time series prediction techniques to support maintenance decision.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Machining
Subjects:	T Technology > T Technology (General) T Technology > TJ Mechanical engineering and machinery
Divisions:	Library > Tesis > FTMK
Depositing User:	Norziyana Hanipah
Date Deposited:	29 Jul 2015 06:39
Last Modified:	20 Apr 2022 09:48
URI:	http://eprints.utem.edu.my/id/eprint/14793
Statistic Details:	View Download Statistic

Actions (login required)

View Item