A MECHANISTICALLY BASED APPROACH TO PROBABILITY MODELING FOR CORROSION FATIGUE CRACK-GROWTH

An approach and methodology for developing a probability model for lif e prediction and its utility are demonstrated. The goal of this probab ilistic approach is to make stochastically tight estimates of life for conditions that are beyond the range used in typical supporting data. Probability models, versus statistically based models, are used to de scribe the influences of fundamental variables through mechanistically based models of the failure processes. The efficacy of a probabilisti c model, therefore, depends crucially upon the availability and qualit y of a mechanistically based model. The approach and methodology is il lustrated through the use of a simplified model for reaction controlle d corrosion fatigue crack growth for ferrous alloys in aqueous environ ments. The stochastic contributions from material properties and from the effects of environmental (electrochemical and thermal) and loading variables are assessed.