Monte Carlo Simulation of Sudden Death Bearing Testing

Abstract

Monte Carlo simulations combined with sudden death testing were used to compare resultant bearing lives to the calculated bearing life and the cumulative test time and calendar time relative to sequential and censored sequential testing. A total of 30,960 virtual 50-mm bore deep-groove ball bearings were evaluated in 33 different sudden death test configurations comprising 36, 72, and 144 bearings each. Variations in both life and Weibull slope were a function of the number of bearings failed independent of the test method used and not the total number of bearings tested. Variations in L ₁₀ life as a function of number of bearings failed were similar to variations in life obtained from sequentially failed real bearings and from Monte Carlo (virtual) testing of entire populations. Reductions up to 40% in bearing test time and calendar time can be achieved by testing to failure or the L ₅₀ life and terminating all testing when the last of the predetermined bearing failures has occurred. Sudden death testing is not a more efficient method to reduce bearing test time or calendar time when compared to censored sequential testing.

Keywords:

• Rolling-Element Bearings
• Life Prediction
• Weibull Analysis
• Sudden Death Testing
• Computer Simulation
• Monte Carlo Analysis

ACKNOWLEDGMENT

The authors would like to acknowledge the contribution of Gregory Hickman of Georgia Southern University for generating many of the spreadsheets used in the Monte Carlo analysis in this article.

Presented at the STLE/ASME Tribology Conference in Ponte Vedra Beach, Florida October 26-29, 2003

Final manuscript approved November 11, 2003

Review led by Ted Bailey

Notes

^a Life based on Zaretsky's rule and lubricant life factor (15).

^b Life based on Lundberg-Palmgren equation (9) and lubricant life factor (Zaretsky (15)).

^a Life based on Zaretsky's rule and lubricant life factor (15).

^a Based upon calculated L ₁₀ life of 6912 hours (from Table 1).

^b Compare to assumed Weibull slope of 1.11.

^c Maximum and minimum values of life do not necessarily correlate with maximum and minimum values of Weibull slope.