This tool demonstrates a two-step uncertainty quantification (UQ) methodology that can be used in concert with verification and validation (V&V) to support credibility assessment of computer modeling techniques used for medical devices. To illustrate the methodology, this tool provides a Jupyter/Python notebook with a minimal working example of a nitinol beam in bending simulated in ABAQUS software.
The UQ method is composed of two parts:
- A sensitivity analysis which determines which simulation input parameters are most influential.
- Uncertainty analysis based on Latin hypercube sampling of probability density functions representing the remaining input parameters.
The output of this tool is a statistical distribution of the output quantity of interest which can be used to understand the uncertainty of the model prediction. The user of the tool will need to characterize their simulation inputs with probability density functions. This can be done through repeated measurements on multiple samples, known measurement uncertainty in insturments, or numerous other ways to determine input uncertainty.
This tool draws from methods outlined in ASME V&V 20 Section 3 which discusses local and global UQ techniques [1]. We also direct the reader to ASME V&V 10 Section 4.5 & 4.6 for more discussion of UQ [2].
This methodology is intended to support credibility assessments performed under the ASME V&V 40-2018 framework, which establishes risk-informed credibility requirements for computational modeling of medical devices [3].
The methodology and Jupyter notebook/Python code presented in this tool are intended to facilitate the use of uncertainty quantification techniques in numerical simulations of medical devices. Uncertainty quantification can be a complex process but is necessary when the results of the numerical simulation will significantly influence decision making in the design or testing of the device.
The methodology presented is applicable to deterministic numerical simulations whose input parameter uncertainties can be characterized by probability density functions, but the Jupyter notebook/Python code example is specific to finite element analysis of a nitinol beam in bending simulated in ABAQUS software. Intended users include simulation code developers, device designers, and simulation analysts. These methods could be applied during device design for internal design decision making or for software devices in which the simulation output is the device output. Effective use assumes proficiency in finite element analysis, basic probability and statistics (PDF fitting, Monte Carlo sampling), and familiarity with ASME V&V verification and validation concepts.
This tool was developed using methods recommended in ASME V&V 10-2019 Sections 4.5 and 4.6 and ASME V&V 20-2009 Section 3.
This tool has been demonstrated in a peer-reviewed publication describing ABAQUS computational solid mechanics simulations of a mock medical device:
[4] Carr, I.A., Aycock K.I., Paranjape, H., Bonsignore C., Weaver J.D., Craven, B.A., Uncertainty of Finite Element Strain Predictions for a Nitinol Medical Device: Influence of Input Parameter Probability Distribution on Output Uncertainty. Cardiovascular Engineering and Technology (2026).
The ABAQUS solver and superelastic constitutive model used in the worked example were verified in prior peer-reviewed work via the method of manufactured solutions [Aycock et al. 2020] and the method of exact solutions [Aycock et al. 2024]. Solution verification of the worked example was performed via mesh refinement and grid convergence index calculation, yielding a total estimated numerical uncertainty of 2.51% in the peak strain amplitude. This tool has not been qualified through the Medical Device Development Tools (MDDT) program.
Using this tool requires an estimate of the uncertainty of all of the relevant input parameters. Gathering these data can present difficulties depending on the materials used and the overall device design. Typically, this will require the user to evaluate a simplified version of their full device to reduce the complexity of the parameterization.
The supporting documentation is provided as a Jupyter notebook linked here with fully executable code in this repository.
A static version is provided in PDF form at this link.
Use of the tool requires installation of free and open-source software including Python 3, NumPy, pandas, SymPy, Matplotlib, and Jupyter, which are available through installation of the Anaconda Distribution.
[1] American Society of Mechanical Engineers. ASME V&V 20-2009 (R2016) Standard for Verification and Validation in Computational Fluid Dynamics and Heat Transfer. (2016) [2] American Society of Mechanical Engineers. ASME V&V 10-2019 Standard for Verification and Validation in Computational Solid Mechanics. (2020). [3] American Society of Mechanical Engineers. ASME V&V 40-2018 Assessing Credibility of Computational Modeling Through Verification and Validation: Application to Medical Devices. (2018). [4] Carr, I.A., Aycock K.I., Paranjape, H., Bonsignore C., Weaver J.D., Craven, B.A., Uncertainty of Finite Element Strain Predictions for a Nitinol Medical Device: Influence of Input Parameter Probability Distribution on Output Uncertainty. Cardiovascular Engineering and Technology (2026).
• RST_CDRH@fda.hhs.gov Tool Reference [DOI]
• Catalog of Regulatory Science Tools to Help Assess New Medical Devices
