Introduction:
Along with component failure analysis, the engineer would like to determine how long before the next failure of duplicate components might occur and also determine an appropriate replacement frequency. Weibull Analysis is a valuable tool for developing answers to those questions. Weibull Analysis can also indicate if failures are occurring with infant mortality, randomly or simply wearing out. Batch problems (defective components) can be quickly identified by Weibull Analysis. An additional benefit to the Weibull Analysis is the ability to analyze issues with very few failures. In general, the Weibull method involves:

- Plotting and interpreting data 
- Failure forecasting and prediction
- Evaluating corrective action plans
- Engineering change substantiation
- Maintenance planning and cost-effective replacement strategies
- Spare parts forecasting
- Warranty analysis and support cost predictions
- Recommendations to management in response to service problems

Background: An example of a Weibull Method Case History

A new mode of failure was observed with engine failures caused by a bonded drive failure. Below are two theories on why the bonded drives were failing.

Theory 1: The engines were being started during cold weather. If the water froze in the water pump housing, the bonded drive would fail due to torsional overload. 

Theory 2: The polymeric material in the bonded drive was worn out after a certain amount of usage.  Due to capacity constraints at the rebuild shops, some of the engines were being subjected to service lines extending beyond the OEM’s recommended engine overhaul limits.

Procedure:
The data used for analysis with the Weibullsmith^tm software included the ‘miles to date’ or the ‘miles to failure’, whichever was applicable. The plot shown in Figure 1 was developed using the mileage data.  The plot shown in Figure 2 was developed using the cost of changing out the bonded drives on a scheduled maintenance (planned costs) versus letting the drives operate to failure (unplanned costs). Figure 2 shows the optimum change out time for bonded drives. 

Discussion:
The graphed data in Figures 1 and 2 provides the following relevant information:
- The regression is good indicating that the model is applicable. 
- The slope of the line, or beta, is very steep indicating that “rapid wear out” is the primary failure mode.
- Had the failures been caused by frozen water pumps during cold winter starts the failures would have been random in nature with a slope of one. 
- Optimum replacement for the bonded drives occurs at the lowest point in the curve or at 493,333 miles.

Conclusion:
The final results indicated that the primary failure mode was rapid wear out. The recommendations made to management were to replace the bonded drive at a mileage of 500,000.  This recommendation was adopted and only one documented failure occurred after implementing the changeout criteria.

Figure 1 - Weibull plot of bonded drive failures.

Figure 2 - Plot indicating optimum bonded drive replacement mileage.