High-Performance and Distributed Computing in a Probabilistic Finite Element Comparison Study of the Human Lower Leg Model with Total Knee Replacement

TL;DR

This study compares passive and active flexion in human lower leg finite element models with knee replacements, using probabilistic simulations and high-performance computing to analyze sensitivity and key parameters affecting contact pressures.

Contribution

It introduces a comprehensive probabilistic analysis of passive flexion models, extending previous active flexion studies, utilizing high-performance distributed computing for extensive simulations.

Findings

Sensitivity of passive flexion to input parameters identified

Probabilistic models highlight key factors influencing contact pressures

High-performance computing enables large-scale simulation analysis

Abstract

Reliability theory is used to assess the sensitivity of a passive flexion and active flexion of the human lower leg Finite Element (FE) models with Total Knee Replacement (TKR) to the variability in the input parameters of the respective FE models. The sensitivity of the active flexion simulating the stair ascent of the human lower leg FE model with TKR was presented before in [1,2] whereas now in this paper a comparison is made with the passive flexion of the human lower leg FE model with TKR. First, with the Monte Carlo Simulation Technique (MCST), a number of randomly generated input data of the FE model(s) are obtained based on the normal standard deviations of the respective input parameters. Then a series of FE simulations are done and the output kinematics and peak contact pressures are obtained for the respective FE models (passive flexion and/or active flexion models). Seven output performance measures are reported for the passive flexion model and one more parameter was reported for the active flexion FE model (patello-femoral peak contact pressure) in [1]. A sensitivity study will be performed based on the Response Surface Method (RSM) to identify the key parameters that influence the kinematics and peak contact pressures of the passive flexion FE model. Another two MCST and RSM-based probabilistic FE analyses will be performed based on a reduced list of 19 key input parameters. In total 4 probabilistic FE analyses will be performed: 2 probabilistic FE analyses (MCST and RSM) based on an extended set of 78 input variables and another 2 probabilistic FE analyses (MCST and RSM) based on a reduced set of 19 input variables. Due to the likely computation cost in order to make hundreds of FE simulations with MCST, a high-performance and distributed computing system will be used for the passive flexion FE model the same as it was used for the active flexion FE model in [1].