Aix Marseille Univ, CNRS, ISM, Marseille, France.
Aix Marseille Univ, CNRS, ISM, Marseille, France.
J Biomech. 2020 May 7;104:109724. doi: 10.1016/j.jbiomech.2020.109724. Epub 2020 Mar 2.
In musculoskeletal modelling, adjusting model parameters is challenging. This paper proposes a multivariate statistical methodology to adjust muscle force-generating parameters optimally. Dynamic residuals are minimized as muscle force-generating parameters are varied (maximal isometric force, optimal fiber length, tendon slack length and pennation angle).First, a sensitivity and a Pareto analyses are carried out in order to sort out and screen the set of parameters having the greatest influence regarding the dynamic residuals. These parameters are then used to create a response surface following a Design of Experiments (DoE) approach. Finally, this surface is used to determine the optimum levels of the design variables (muscle force-generating parameters). The proposed methodology is illustrated by the adjustment of a three-dimensional musculoskeletal model of a sheep forelimb. After adjustment, the reserve actuator values of the elbow and wrist joints were reduced, on average, by 18%, and 16%, respectively. These results demonstrate that the use of multivariate statistical strategies is an effective way to adjust model parameters optimally while reducing dynamic inconsistencies. This study constitutes a step towards a more robust methodology in musculoskeletal modelling, focusing on muscular parameter tuning.
在肌肉骨骼建模中，调整模型参数是具有挑战性的。本文提出了一种多变量统计方法，以最优的方式调整肌肉产生力的参数。在调整肌肉产生力的参数（最大等长力、最佳纤维长度、肌腱松弛长度和肌纤维倾斜角）时，最小化动态残差。首先，进行灵敏度和帕累托分析，以整理和筛选对动态残差影响最大的参数集。然后，使用这些参数根据实验设计（DoE）方法创建响应面。最后，使用该表面确定设计变量（肌肉产生力的参数）的最佳水平。该方法通过调整绵羊前肢的三维肌肉骨骼模型进行说明。调整后，肘部和腕关节的备用执行器值平均分别降低了 18%和 16%。这些结果表明，使用多变量统计策略是一种有效的方法，可以在减少动态不一致性的同时优化调整模型参数。本研究是朝着肌肉骨骼建模中更稳健的方法发展的一步，重点是肌肉参数调整。
