Session: 02-10-01: Variation Simulation and Tolerancing

Paper Number: 68918

Start Time: Friday, 11:45 AM

68918 - Probabilistic Performance Evaluation and Optimization of Medical Plastic Moulded Components Subject to Large Scale Production 

The development of a new medical device consisting of plastic molded components can take years to develop from early concept to product launch. The long development process can be attributed to two essentials complications; 1) the serious consequences for the patient if the device malfunctions, and 2) large-scale production e.g., using multi-cavity tools for molding plastic components, with the inherent variation and material properties (non-linearity, rate-sensitivity, creep, etc.). Mitigation of these complications is often solved by rigorous simulation, physical test programs, and risk assessments of the likelihood and consequences of various possible failure modes. However, these approaches are often not ideally suited to address the two key questions that the design engineer has: 1) How often will my design fail and 2) What should I change to improve the performance of my design. Physical testing does not show an actual distribution of performance, as these are often done on prototype components from the same batch, simulations are typically done on nominal geometry and only provide a safety factor against failure, and risk probabilities are based on educated guesses. In this paper, we present a structured process for providing the design engineer with the needed answers. The process is focused on FEA based variation simulation, virtually creating a full sample space, and simulating the performance of all samples to assess the design sensitivities and probability of failure. 

The present work proposes a novel process and investigates its applicability to evaluate and optimize the probabilistic performance of plastic molded components with multiple conflicting objectives. The final aim of the process is robust and reliability-based optimization but is accomplished by sequential experimentation, which gradually improves the mechanical understanding of the design. Optimization is accomplished by first defining the potential solution space, the relationship between the functional requirements and the sensitive parameters, and then trying to optimize them. This approach promotes a thorough understanding of how design modifications allow for more robust and reliable performance rather than alternative direct optimization approaches.

A live case study in collaboration with a medical design and manufacturing company is testing the proposed virtual design process. The case study includes FEA contact modeling between two plastic molded components with 12 geometrical variables and an augmented fractional factorial design for the creation of second order meta-models to reduce the computational cost. The optimization is based on the meta-models to minimize the failure rate (and to improve design robustness) with respect to three objective functions (contact pressure, strain, torque).

The key novelty is the focus on creating a virtual design process for new products with no or limited legacy knowledge. The technical output of the process results in a sensitivity analysis, response surface meta-models for the conflicting objectives, and a graphical representation of the trade-off between performance, robustness, and reliability of the design. The end result compares the failure rates of the initial design, the design optimized for maximum torque, and the robustness and reliability optimized design. The estimated failure rated is reduced from 78% (nominal design) to less than 1% for the reliability optimized design.

The study concludes that probabilistic evaluation and optimization provide significant benefits compared to the traditional safety factor approach and has the potential of lowering development time and costs by correcting unfortunate designs earlier in the development phases.

Presenting Author: Tim Brix Nerenst Technical University of Denmark

Authors:

Tim Brix Nerenst Technical University of Denmark
Martin Ebro Novo Nordisk, A/S
Morten Nielsen Novo Nordisk, A/S
Kanishk Bhadani Chalmers University of Technology
Gauti Asbjörnsson Chalmers University of Technology
Tobias Eifler Technical University of Denmark
Kim Lau Technical University of Denmark