This study reports the development of an artificial neural network computation model to predict the accumulation of crack density and crack length in cancellous bone under a cyclic load. The model was then applied to conduct a parametric investigation into the effects of load level on fatigue crack accumulation in cancellous bone. The method was built in three steps: (1) conducting finite element simulations to predict fatigue growth of different three-dimensional micro-computed tomography cancellous bone specimens considering input combinations based on a factorial experimental design; (2) performing a training stage of an artificial neural network based on the results of step 1; and (3) applying the trained artificial neural network to rapidly predict the crack density and the crack length growth for cancellous bone under a cyclic loading for a given applied apparent strain, cycle frequency, bone volume fraction, bone density and apparent elastic modulus.
Read full abstract