Visiting Lecturer Program (1)
Speaker: Ali Zuashkiani
Ph.D. candidate
Department of Mechanical and Industrial Engineering,
University of Toronto
Local Organizer : Mohammad Baharnemati
General Talk:
Title: A survey on different tools applied in Condition Based Maintenance
Time: Sunday, Sep 19, 2004 at 4 pm
Location: Department of Industrial Engineering, Sharif University of Technology, Tehran
Specialized talk:
Title : Incorporating expert knowledge in the estimation of parameters of a Condition Based Maintenance ModelTime : Sunday, Sep 19, 2004 at 5 pm
Location : Department of Industrial Engineering, Sharif University of Technology, Tehran
Abstract:
General talk:
The ever-growing field of Condition Based Maintenance (CBM) has caused reliability engineers to wander which technique can fit their needs and conditions in the most appropriate way. In this talk emphasize will be on the most popular and applicable tools applied in CBM. Both statistical tools such as regression models, trending, PHM and non-statistical models including Neural Networks, Expert systems will be introduced. In addition to their definition some of their advantages and disadvantages accompanied by their applications will be presented in this talk.
Specialized talk:
In recent years, the increasing need for competitiveness and therefore productivity has forced industries to delegate more manual tasks to industrial machines, which means more automation. Thus, currently, there is a greater need for keeping the machines working, which has attracted both industries and academia to develop and apply more accurate reliability and maintenance techniques. As a result, models, which can use all related data to predict the failure time of a system in the best way, are in the center of attention. These models generally fall under the Condition Based Maintenance (CBM) category, which is a recently developed field of maintenance able to make use of all the data related to a system for predicting the system's health conditions.
CBM not only measures the effect of the age of the system, but also takes into account other influential factors on the lifetime of a system, such as metal particles in the engine circulating oil, vibrations intensity, temperature etc. To date, two major categories of techniques have been applied in CBM:
1. Non-statistical tools including neural networks, fuzzy logic, and expert systems
2. Statistical tools including Proportional Hazards Model, log linear regression model, linear regression model, and Proportional Intensities Model [1].
Statistical methods are among the most widely used tools in CBM and have been applied for many years. Among them the Proportional Hazards Model (PHM) is unique because of its structure that allows covariates or explanatory variables to affect the hazard rate of a system. This is a more sensible assumption of the real world in comparison with other tools. Another advantage of PHM that has made it more practical is its ability for handling time dependent covariates. Moreover, unlike most conventional models, PHM is capable of handling right censored and tied survival data.
A practical difficulty in developing statistical models is their need to large sets of data in order to produce reliable outputs. In the absence of such data, researchers have been probing for methodologies that can make use of other types of information such as experts' knowledge. This search was paid off in the late 1970 and as a result Bayesian statistic was applied in practice. The structure of Bayesian methods allows incorporation of both experts' knowledge and statistical data in model building.
The aim of current research is creating a methodology to incorporate experts' knowledge as well as statistical data in estimating the parameters of PHM. The results of this research are believed to be beneficial not only to industries but also to different fields of study that currently use PHM such as biomedical, finance, organization demography etc.