|
|
|
|
|
 |
 |
|
| Reliability HotWire | |
| Tool Tips | |
|
This simply indicates that you are missing or do not know the failure information for the current system from time equal to zero up until the specified start time. This may be due to improper data collection or the data may have not been collected in the first place. The start time relates only to the current system and does not have any relation to the start times for any of the other systems, if any, in the analysis.
It depends. In general, when not using the Crow Extended model, the non-cumulative times-to-failure should not be transferred to Weibull++. Transferring the time between failures of a system to Weibull++ and fitting a distribution would imply that the system’s failure events are independent. This would mean that no corrective action is taken after a failure event and that the repair brings the system to a brand new condition (As Good As New). However, the failure events for the systems in reliability growth analysis are dependent. When a failure occurs, a repair or corrective action may take place. If a repair is performed, it generally does not make the system As Good As New, rather it brings the system to its previous condition before the failure (As Bad As Old). If a corrective action is taken in order to improve the system’s reliability, then this intervention changes the system’s failure behavior. Therefore, the failure events are dependent and are being recorded with regard to the system's accumulated time. From this, you can see that the data for reliability growth and life data analysis are collected under different assumptions and processes. The non-cumulative failure times in reliability growth analysis do not follow a life distribution, such as Weibull, and should not be treated in this manner. However, when analyzing failure times data in RGA using the Crow Extended model, the non-cumulative times-to-failure for certain failure modes could be transferred and analyzed using Weibull++ 6, if it is installed on your computer. Consider the data in the following figure (only the first 20 rows of data are shown).
The A and BD modes could be transferred to Weibull++ and a distribution could be fitted to the time between failures for each unique mode (unique Mode ID). This is because the distribution analysis in this case is done for each unique failure mode and no assumption is made about the system’s condition. In addition, A and BD modes are modes where no corrective action is taken during the test, therefore their failure behavior is not altered. Notice that the A modes should also have a mode classification so as to distinguish between the different A modes. On the other hand, time between failures for BC modes, if any, should not be analyzed using Weibull++ since these times-to-failure are dependent due to the corrective actions taken. (The example data set does not contain BC modes, which are modes where a fix was put in place during the test in order to reduce its occurrence.) To transfer the data to Weibull++, select Weibull++ 6 from the Tools menu then select the options displayed next.
Click OK and the data will be transferred to Weibull++ 6 as shown in the next figure. Once again, only the first 20 rows are displayed.
From here, you can separate the data based on the Subset ID so that a new data sheet will be created for each individual failure mode. Once this has been completed, you can analyze the data and determine if the failure mode has a decreasing, constant or increasing failure rate. |
|
 |
Copyright © 2004 ReliaSoft Corporation, ALL RIGHTS RESERVED |
|
Comments or Questions: Webmaster@ReliaSoft.comweibull.com is a service of ReliaSoft Corporation Copyright © 1992-2008 ReliaSoft Corporation, All Rights Reserved Last updated: 10/29/2008 04:49:18 PM |
SITE
[Home] [Site
Contents] |
What is the meaning of a start time greater than zero when entering
data for fielded systems in 
