Reduce maintenance resources and increase plant availability by utilising web-based condition monitoring systems and Markovian modeling techniques

Abstract

The purpose of this research is two-fold. Firstly, to decrease maintenance resources and increase plant availability and secondly, to investigate the feasibility of using Web-based condition-monitoring techniques as a preventative maintenance tool. This is achieved by using Markovian modelling and its associated mathematical operations, among other techniques, which in turn leads to the manipulation of the Stochastic Reliability equation to determine the key drivers of poor plant performance. In addition, the findings are elaborated on by applying chaos theory principles and logic. In this exercise, a Web-based condition monitoring device (Motornostix® Canary) was installed on a critical overhead crane in a pilot plant (ESM) to test the feasibility of the system in the steel production environment. This research also aims to elaborate on the outlook of the current global steel market as well as present the author's views on the topic. To achieve the outcomes of this research, proper methodology and hypotheses were applied to process the information collected and data generated. The following results of the literature study were amongst the most important: The global steel industry is increasingly competitive and the market has changed radically from its previous model'. Traditional "Third-World/Iron Curtain" countries are becoming major players in the global steel industry and the world economic playing field as a whole 2. Markovian models are memory-less, discrete and not dependant on the route followed to achieve the current state of the system3. Markovian models are lacking as an application in a chaotic environment as they can only simulate linear systems. Linear systems exist more in theory than in practice. Living systems cannot be equated via linear methods4. The Newtonian paradigm has to be exchanged for a fresh way of approaching maintenance issues5. The study has been approached from the perspective that Markovian models do work, if only with a limited degree of predictability over time. However, as has been proven by subsequent findings, Markovian models alone will not suffice in increasing the pilot plant's availability. Intuitive and practical decisions must be applied, in addition, for the outcomes to be both accurate and to impact on business. It was also noted that failure modes have more than one driver, thus distorting failure and repair rate data into distributions that are not Poison or exponential forms. The inconsistency of the failure rate compounded the difficulty of applying the Markov modeling techniques to this system. To date, there has been no outside research done on the immediate benefits of implementing Web-based condition monitoring systems. All available papers on the subject have been published by manufacturers of this equipment. Therefore, this research delivers a "third-party" perspective on the effectiveness of these devices as implemented on a pilot plant when used on overhead cranes, whilst quantifying their impact on safety, cost and availability. The results have proven formidable. Failure rate at the pilot plant was drastically decreased, with no instances of failure (that could have been prevented by this system) occurring during 2006. Not only was the trial a conclusive success but the application of this technology in other areas of the steel production industry has since begun in earnest.