Initial age of equipment for analysis

[AromaticsThailand]

I don't understand about initial age of equipment that I would like to know for calculate time to failure.

Suppose, equipment installation in year 1997 and use until now (2008), so, when I calculate TTF, I must track failure data since 1997 - 2008, right? or can I track from desire period e.g. since 2003 - 2008?

I'm very confuse. Please let me clear and understand.

[Arai.M]

Per your prior posts/replies, it seems like you will end up building some kind of repairable systems model (via rda/rga). In that case, you are looking at failures over time and a way of modeling that. The model will try to mimic the failure intensity of the system over time. If that failure intensity behaves similarly over time (e.g. increasing failure intensity at a certain rate), then the more data you have (going back to 1997), the better. You have more data to capture the behavior. However, if 2003 represents some change (e.g. better data collection system, change in the repair efficiency, etc.) you might as well just use data that is more representative of what is happening today. Remember these models will be used to predict behavior so if there is anything from that data that does not reflect current conditions, then the model might not do too good of a job at predicting them.
In theory, all other things remaining equal, the only difference between using 2003 vs. 1997 as your starting point is that you have to be carefull on how you get results out of your model (and confidence intervals will be wider with a smaller sample size). You are calling t=0 a different point so the initial age will be different depending on what model you are quering...

[AromaticsThailand]

That's clear for me about my question! Thanks you.

I would like to summary in my understood and please be correct me.

1. Repariable system is component or system can be repair/replace when it fail. (I still confuse about non-repairable system)
2. If I have more failure data to analyze, I will getting the exactly result.
3. If 2003 I have improved the system, So I would initial age of the system in 2003 to be analyzed.

[Arai.M]

A general definition for a repairable system is a system that can be restored to an operating condition following a failure. Now following a failure, if the system is replaced by a new one (or the repair is so good it is equivalent to a replacement), LDA can be used and you can fit a Weibull distribution for example. This kind of analysis is done with non-repairable components/systems.

If a small portion of the system is replaced/repaired (for example, the failed component is replaced), then what happens after a failure (i.e. or when that next failure will occur) depends on the past history of the system. You need to use a model that can capture such dependencies. A life distribution cannot do this as it is taking every event as identical to and independent from the previous one.