Reliability of repairable systems

[Willie Lottering]

[Originally Posted: 11/15/99-- Transferred by ReliaSoft Moderator]

How will you: 1. Utilize the Weibull 5 software to estimate the reliability 0f say 15 systems on test for 120 hours -- but where systems are repair after failure and return to testing. 2. Utilize the Weibull 5 softare to determine the burn-in period for the above systems before subjecting them to qualification testing 3. Can i utilize you software for maintainability analysis? 4. Are you going to introduce the ability to do Availability analysis in your software?

[Adamantios Mettas]

[Originally Posted: 11/15/99-- Transferred by ReliaSoft Moderator]

ReliaSoft's BlockSim 1.0 is specifically designed for performing Reliability, Maintainability and Availability analyses. If you are interested you can download an evaluation copy from the Web.

[David J. Groebel]

[Originally Posted: 11/15/99-- Transferred by ReliaSoft Moderator]

There are a few different ways you can go about in analyzing this type of data. Actually, your problem is not as simple as it sounds, since it is a repairable system data. The data types available in Weibull++ refer to whether the data is exact, censored, or interval. Repairable system data can also be exact, censored, or interval. One method of analyzing this data using Weibull++ is by separating them into different failure modes (or subsystems). If you assume that each time a subsystem is repaired it is returned to its original operational condition, you have in a sense a sample from this subsystem. This assumption is called As-Good-As-New. You can then fit a distribution to the data for this subsystem, as well as for all the subsystems you have data. For example, consider a system in testing. The first time it failed was due to Subsystem1 at 100hr. It was repaired and a new Subsystem1 was installed. The system was put back into test. The next failure was due to Subsystem2 at 200hr. Subsystem2 was repaired (as good as new), and the system was put back in the test. The third failure was again due to Subsystem2 at 300hr. Upon repair, Subsystem1 was also replaced preventively without failing. What we have so far is:

Subsystem1: 1 failure at 100hr, 1 suspension at 200hr (because it was replaced preventively at 300hr, thus it operated for 200hr). So far these data are Non-Grouped Times-to-Failure with Suspensions.

Subsystem2: 1 failure at 200hr, and 1 failure at 100hr. So far these data are Non-Grouped Times-to-Failure.

Continuing this logic, data for Subsystem1 and Subsystem2 can be collected. Using Weibull++, separate distributions can be fitted to each of these subsystems. The system can then be treated as a series system. Assuming independence, the reliability of the system is the reliability Subsystem1 times the reliability of Subsystem2. You can then create a table of Time and the corresponding System Reliability (the product of the two subsystem reliabilities). You can then use a Free Form data sheet in Weibull++, and fit a line through these data and plot them.

In addition, you could use our BlockSim software package to model the entire system using a reliability block diagram (RBD). You could then associate a failure and/or repair distribution with each subystem. From the RBD you can then calculate the reliability of the system.

Another approach would be to use the NHPP (or AMSAA) model. This model is used in Reliability Growth analysis. It is not available in Weibull++, but in our RG 1.0 package.