Prediction of the Inelastic Behaviour of Radius Segments: Damage-based Nonlinear Micro Finite Element Simulation vs Pistoia Criterion

The Pistoia criterion (PC) is widely used to estimate the failure load of distal radius segments based on linear micro Finite Element (mu FE) analyses. The advantage of the PC is that a simple strain-threshold and a tissue volume fraction can be used to predict failure properties. In this study, the PC is compared to materially nonlinear mu FE analyses, where the bone tissue is modelled as an elastic, damageable material. The goal was to investigate for which outcomes the PC is sufficient and when a nonlinear (NL) simulation is required. Three types of simulation results were compared: (1) prediction of the failure load, (2) load sharing of cortical and trabecular regions, and (3) distribution of local damaged/overstrained tissue at the maximum sustainable load. The failure load obtained experimentally could be predicted well with both the PC and the NL simulations using linear regression. Although the PC strongly overestimated the failure load, it was sufficient to predict adequately normalized apparent results. An optimised PC (oPC) was proposed which uses experimental data to calibrate the individual volume of overstrained tissue. The main areas of local over-straining predicted by the oPC were the same as estimated by the NL simulation, although the oPC predicted more diffuse regions. However, the oPC relied on an individual calibration requiring the experimental failure load while the NL simulation required no a priori knowledge of the experimental failure load.

Automated summary

The study compared the Pistoia criterion (PC) and nonlinear finite element analyses in predicting the failure load of distal radius segments. The results showed that both PC and nonlinear simulations could predict the experimental failure load well, with the PC tending to overestimate the failure load. An optimized PC (oPC) was proposed, which calibrated the volume of overstressed tissue using experimental data.