Volume weighted probabilistic methods for nitinol lifetime prediction

Nitinol fatigue lifetime prediction typically involves three steps: 1) establish material limits by specimen testing, 2) predict component strains using computational modeling, and 3) compare allowable strains with simulated strains to predict survival or failure.

An improved strain limit diagram format is proposed, including methods to characterize credibility, by representing sample size and survival probability as a function of mean strain and strain amplitude. The importance of controlling specimen type, strain history, and thermal history are also illustrated, including new computational results demonstrating fatigue benefits of pre-strain that are unique to superelastic material

Computational modeling techniques commonly represent results as a “point cloud” of mean strains and strain amplitudes at each integration point in the model. We illustrate some mistakes that are common in current practice, including shakedown effects, and scalar vs. tensor math strain calculations. Finally, we demonstrate the benefits of considering additional results from the model, including strain history, and integration point volume.

In current practice, material limits are subjectively simplified to strain limit diagram threshold line.  A computational point cloud is then compared with this line, to make a binary prediction of “survival” or “failure”. We demonstrate new methods to calculate a fatigue durability hazard probability, based on integrated volume weighted probabilities of survival at each integration point. We then extend this method to consider the influence of inclusion size and distribution for selected materials, using measurements derived using sub-micron computed tomography scanning techniques.