Warranty Analysis Bug

[Yakura]

I'm using Weibull++ ver. 7 to perform an analysis in a Warranty Analysis Folio using a 'Nevada' Format. I have pasted the data for both the shipments and returns into the forms from Excel.

The problem is that the number of failures that Weibull++ reports is 712 failures less than the sum of failures on the original spreadsheet. The number of sales I entered is 3,414 greater than the sum of failures and suspensions that Weibull++ reports.

I have 42 months of sales and 42 months of returns. Returns began five months after shipments.

When I exported the Weibull++ sales and returns forms back into Excel, the sums are the same as what I intended to enter.

The examples in the Life Data Analysis Reference manual show the sum of failures plus suspensions being equal to the total sales.

Version 6 software produces the same results as version 7.

If I cannot trust the counts, the results are also suspect. Can you please tell me what is wrong?

[Tarik El-Azzouzi]

Hi,
Do you have returns during the same sales period, ie returns whith time-to-failure = 0?
For assistance, contact customer support and send them your file.

[Yakura]

Thanks Tarik,

That explains what happened (along with an Excel problem).

Best regards
Yakura

[HankC]

I observed the same thing. Not counting 0 time in service does create a confusion. Now my question is when I calculate predicted # of units failed at certain time-in-service by using the total unit and multiply the unreliability, I should use the true total units or the Weibull++ adjusted total units (not counting the unit in 0 TIS)? My guess is I need to use the adjusted number since that is how Weibull++ calculate the reliability. Correct?

Hank

[Tarik El-Azzouzi]

Hi,
you should use all the units you started with (sales data) unless you have suspensions at T=0, then these units are treated as if they never existed (they were take out of the population before they accumulated any hours, it's like they were never used).

Weibull++ can provide the forecasts for you, you don't need to do this by hand, it's quite tedious. You need to calculate the number of units at risk and estimate conditional reliabilities, as described in http://www.weibull.com/LifeDataWeb/warranty_prediction.htm.

[Temanag]

I have an another question---> "Users Guide" page 238.

Can anybody tell me why there are only few (one) failures in the first period of the warranty forecast and forty in the second period ....?

I entered for the forecast start date 2. Mai 2040. The result 0 failures 2.Mai 2040 and 13 failures in Juni 2040. ?..?..


I don`t understand why there are always few failures in the first observed period in the forecast and why the number of failures icrease significantly in the next periods .

Thanks

[David]

Hi Temang,

The end time for the analysis is specified as 5/1/2006. The first column on the Forecast sheet is 5/2/2006 which represents the expected number of failures for one more day given that they have made it to 5/1/2006. Since it is only for a single day, the expected failures is small. The next column is for a full month so therefore the values increase.

I hope this helps.

[Temanag]

Thanks you for your answer :)

No matter what I enter for the start date of the forecasting period, the number of failures in the first forecasting period is always very small and in the second, third and so on periods considerably greater. There is no logic... :(
I can`t explain myself why in the first period no or only one falirure happend and than in the second period suddenly a good deal more.

[David]

Hi Temanag,

Well, first let me say that you are not going crazy. :) There is an issue with the calculations for the first month's forecasting results. The interval from the end date of the analysis to the first month projection is not being handled properly. I apologize for the confusion. This will be addressed within the next service release of Weibull++ 7 (7.1.4). Unfortunately, I do not have a schedule as to when it will be released. If you would like, I can let you know when it is available.

I would like to let you know that you can also email questions to Support@ReliaSoft.com. Thanks again for your patience and for bringing this to our attention.