Methodology for Establishing Damage Criteria Using Probability Distribution Function on Component Level Tests - a Case Study

摘要

Automobile components are usually subjected to complex varying loads. Thus, fatigue failure is a common mode of failure in automobile components. Accurately predicting the fatigue life is the key point for light weight and also reliability design of automobile components. Various life prediction theories are being used in the automotive industry for damage analysis using material S-N curves. However, due to variability in manufacturing, material spec etc. it is difficult to predict the experimental lives using conventional theories. Probability based statistical modeling is prevalent in the industry for life prediction. Probabilistic plots of cycles to failure to constant amplitude loads are plotted and used for prediction purpose. As the component is subjected to varying loads in real world, defining a single parameter i.e. damage would be more relevant compared to loads. The paper combines the damage theory and probability distribution function to determine damage criteria for a component with variable cyclic loads that may represent a real customer usage pattern. The methodology is illustrated by testing an engine mount bracket made of Aluminum alloy in component level test. A number of samples were tested under different constant amplitude loads to simulate a field failure. Test sample size was increased to include uncertainty. Stress was measured at these loads and damage was calculated using conventional theories using miners rule. The parameters of the distribution were computed as a function of damage from the resultant samples and damage criteria for the component was established with high confidence levels (90%). The applicability of the distribution function was confirmed using Hollander Proschan Goodness of fit tests. This methodology can be easily adapted for different components of any material (with S-N curve) and can be helpful in defining criteria to take quick durability related decisions based on stress/damage data observed in field/track.