A great use of a Weibull analysis, is to estimate warranty cases based on field data and design specifications. This is, to estimate how many failures and/or returns can be expected. Either from the manufacturer point of view or from the field operator. In this case is gearbox component warranty extension.
The following case is an example that I’ve used a couple of times to demonstrate the value of field data, Weibull plots and design specifications in warranty cases.
Six gearboxes from the same OEM have a faulty component. This issue is localized at the planet carrier bearing (see video below as a pure example). If left unattended this fault can cause material loss and damage the gearbox to the point, that the whole gearbox needs to be replaced. To exemplify the monetary value and as a rule of thumb, in the Wind energy business a gearbox exchange onshore can cost approxx. 250k€. Being offshore can raise the cost to 500k€ (depending on weather availability days).
In this case the OEM fixes the problem, retrofitting all the gearboxes at site (11 gearboxes in total), and states that the problem is solved. As a consequence of this fault, the warranty will be extended 4 years from the fix. The characteristic life for this gearbox component is 10 years.
Check out How to Make a Weibull Analysis in 5 Steps – Part 1 to get a short introduction on Weibull plots.
Shall we accept the warranty of 4 years or or do we have base to argue for a longer extension? In other words, are we able to demonstrate that after 4 years of operations this failure mechanism is removed from our gearboxes?
ANSWER (at least one of them):
The warranty should be extended at least to 6 years, if we want to be confident that this failure mechanism will be removed from all our gearboxes.
Looking at the Weibull plot is possible to see that for the same failure mechanism (this is the same Beta as the field failure data) and a characteristic life of 10 years, the B10 is 4 years of operations. This is to the left of our design life line. The B10 represents 1.1 failed gearbox components.
If we would like to be confident that the failure mechanism has been removed, is necessary to move to the right hand side of the design life line. This is for the design life B10, 6.2 years of failure free operations.
In conclusion, 6.2 years is the time needed to prove that the failure mechanism has been removed from our population of 11 gearboxes.
Technical note on B10
B10 is the value where we expect 10% of the population to fail (which is 1.1 gearbox components) given the stated conditions. One exact failure of a gearbox component (this is 9% of the population) would be the correct value. In case of only one failure we will be looking at the B9, which accounts for 6 years.
In this case is critical to know the characteristic life of the component, which might not be known to the operator. The next step can be a Weibayes plot for this component without knowing the design characteristic life. This can be useful, in order to suggest that the failure mode was either eliminated or significantly improved (in this case with 90% confidence).
Weibayes for warranty cases
The characteristic life of the component was quite optimistic, and was not included in the contract. Therefore is not possible to assume anymore the characteristic life of this component to be 10 years.
Will 4 years extension be OK? The field failure shows that almost the whole population will fail between year 1 and 3 with the current beta rate, e.i. within 2 years duration. So, if the failure mode is still there, will not the “fixed” gearboxes fail again within the the four years of extended warranty?
Whenever we do not know the characteristic life of the component and only have field data, is possible to use a Weibayes to check the warranty period is sound. This is when we do not know the characteristic life of the component and only have field data. In this case it translates to, that the beta of the failure mechanism is known ( beta = 4.712 ), but the characteristic life is not.
Weibayes suggests that, after 4 years without any failure we can be confident (is possible to get a % of how confident) the failure mechanism has been removed from the population – the 11 gearboxes). Although will not prove any characteristic design life of the component.
The big advantage in the first case is that the characteristic life (eta = 10 years) is known.
In case you would like to expand on the gearbox bearing design topic for the wind energy busines, you have this great article by Romax, “When is 20 year gearbox life not 20 years?”.
Finally, it is also important to empathize the human factor with this type of analysis. It is preferable not to use it as a negotiation weapon. This should be use as an instrument for the discussions between owner, operator and OEM. Why are we seeing this data? What does it mean? Can we improve it?
Did I miss something important? Do you see flaws in the logic of the analysis? I will be happy to read your opinion in the comments below!