Abstract
Background
Numerical modelling using Finite Element (FE) Analysis has become ubiquitous in orthopaedic biomechanics, with both commercial and freely available packages widely used. Three FE packages in particular have gained popularity: Abaqus (Simulia); Ansys (ANSYS, Inc.) and FEBio (University of Utah). Although FEBio is now well established, its developers advise that comparisons should be made against more extensively tested software before trusting its results for specific problems. The aim of the study to conduct a comparison of mesh convergence and to provide validated open-source models of the femur for use all three FE packages.
Methods
Three-dimensional FE models of the femur were created in Abaqus. To ensure that all aspects of the models were identical, custom scripts were developed to import the models into other packages. Mesh convergence studies were conducted for each solver using seven mesh densities for linear tetrahedral elements (up to 2 million). Experimental validation used fourth generation Sawbones composite femurs (n=8) with surface strains measured at four locations. The loading applied at the hip was the averaged peak joint reaction force during walking (Bergman et al); experimentally, this loading vector was used for a reduced load of 500N.
Results
The three solvers produced excellent agreement in load-displacement predictions at corresponding locations, being within 0.7% of each other (R2=0.99). The surface strains at matching locations were found to be within 5% for the two commercial software, with FEbio presenting slightly larger differences. For the optimised model, using 450,000 elements, vertical and horizontal displacement predictions for all solvers matched experimental tests closely (R2=0.96), and strains at four locations around the femur were within a median error of approximately 30%.
Conclusions
All three FE packages were able to produce consistent results that matched experimental data. An open source, finite element model of the femur is presented which has been experimentally validated for the three software packages. As far as the authors are aware, this is the first comparison of this kind for the three FE solvers considered.
Level of evidence IIb - Evidence from at least one well designed experimental trial.
