Uncertainty and Sensitivity Analysis of Blunt Impact Tissue Response Using a Swine Finite Element Model
Abstract
This report describes the uncertainty and sensitivity analysis of tissue response from nonlethal relevant impacts using a swine finite element model. A case study on projectile impact over the lung region was conducted. We used response surface methodology (RSM) and Latin hypercube sampling (LHS) for selecting the input parameters of the mechanical properties of muscle and skin, rib, and lung. The output responses were the peak energy density (PED), contact force, rib stress, and the lung contusion percentage for lung impact simulations. In RSM, we used a face-centered central composite design (CCD) to select extreme cases for the FEM simulations and the quadratic regression fitting function in MATLAB to fit the response surface models to the output of FEM. We statistically analyzed the goodness of fit and the significance of the fitting coefficients with their p-values. We quantified the uncertainty with the cumulative distribution function (CDF), boxplot, histogram, mean and standard deviation for both methods. For sensitivity analysis, we used p-values and line plots drawn with the response surface models to identify which parameters had the greatest effect on tissue response, including higher order parameters and parameters that interacted with each other. In LHS, 50 samples of input parameters were selected based on a truncated Gaussian distribution. We quantified the uncertainty with the CDF, boxplot, histogram, mean and standard deviation. We used scatter plots and rank correlation to analyze the sensitivity and calculated the coefficient of variation. The LHS method provides more accurate uncertainty quantification than RSM, as it provides a better representation of the statistical spread of material properties. Uncertainty analysis showed that for lung impacts at 40 m/s, the contact force, rib stress, and contusion had relatively low normalized coefficient of variation values of tilde 0.08. The PED has a higher coefficient of variation of tilde 0.24.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 12, 2018
- Accession Number
- AD1067098
Entities
People
- Jianxia Cui
- Laurel Ng