Expected Downtime Model

After studying the operating practice, the existing maintenance and failure data, the system can be modelled using the following assumptions (Pillay (2(X)1), Pillay et al. (2001a, b))  

It could be argued that some of the above assumptions may not be practical. For example, inspections/repairs could never be carried out perfectly. However, such assumptions which have been widely used by many safety/reliability researchers are made mainly for demonstrating the described method with ease. 

As a consequence of the above assumptions, the model of b(T) given in Equation (8.1) can be simplified as  

Consequently, the expected downtime per unit time function D(T) is given by Equation (8.3) below  

The product k/T will give the expected number of defects within the time horizon considered. This is normally based on the historic data gathered for the equipment or system. 

---

We base the determination of the long term expected downtime costs on the assumption that different waiting times are mutually independent. This assumption is only approximative, because the waiting time for a spare part for one repair gives an indication of the on hand inventory, and thus of the waiting time for the next repair. The approximation is similar to the one used in the analysis of the METRIC model (Sherbrooke 1968). 

A linear regression model is developed for the expected smck-pipe downtime based on physical performance measures. For a specific test set-up i, the model output is the expected downtime Oi- It is... 

Following this, there can be estimated e.g. the expected cost over each cycle C(T), or expected downtime Ej (T) of an inspection cycle of length T (see e.g. (Christer 1999). The examples of such the model implementation are given e.g. in (Redmond 1997, Wang 2008). 

In one hand, a drop in downtime due to preventive maintenance results of the increase in frequencies. In the other hand, the corrective maintenance, and thus the number of failures, increases. These two observations also make it possible to validate the model that restores the expected behaviour of the system. 

---