Tom Mikolyuk - 3D-Printed Mechanical Heart Valve

The University of Washington's Interdisciplinary Honors Program offers its students opportunities to pursue extra projects in their major classes to double-count credit towards their graduation from the program. In Spring 2022 M E 356 (Machine Design Analysis), I was advised by Prof. Nathan Sniadecki in a personal exploration of the aspects of machine design and solid mechanics as applied to medical devices, an emerging area of interest of mine.

Initial research

In normal operation, heart valves accept blood flow in one direction, preventing blood from flowing back into the origin of the flow. Valve failure often results in backflow (regurgitation of blood back into ventricles/atria), which has potentially life-threatening consequences. Metallic mechanical heart valves are installed to remedy valve failure, but they are generally prohibitively expensive to acquire, install, and maintain, and are prone to causing complications in the patient and to fracture through fatigue cycling.

In the initial stage of research, my goal was to familiarize myself with the problem by learning about the history of heart valves, new advances in replacement heart valve technology, and the design constraints involved in the making of replacement heart valves, ultimately adopting those constraints into guidelines appropriate for a quarter-long project using student makerspace resources.

The slide deck that I presented to Prof. Sniadecki early in the project, summarizing key market research findings and first prototyping steps.

Prototyping, iteration 1

I chose to do the prototyping for this project using 3D printing because it's a relatively accessible, cheap, and efficient way to produce many iterations of prototypes, which enables experimentation with forms of solutions that may not be considered "good" from the get-go. While aortic valves are generally ovular in shape, the process of creating a triangular bileaflet heart valve served as a good way to experiment with printing practices on Dremel fused-deposition modeling printers, including finer resolutions and printing joints in place. It also served as a good creative warmup and a great place to start refining once the problems with this prototype (stress concentrators, bad fits into aortae, bad print quality, potential for blood clotting on deposited layers) became clear.

Isometric view of a CAD model of iteration 1, with dashed lines showing invisible bileaflet and hinge contours (modeled in Fusion 360).

Close up view of narrow hinge of a CAD model of iteration 1, with dashed lines showing invisible bileaflet and hinge contours

Close up view of wide hinge of a CAD model of iteration 1, with dashed lines showing invisible bileaflet and hinge contours

Top side of 3D-printed iteration 1 (2x scale), showing filled-in gaps where there should be openings in model

Bottom side of 3D-printed iteration 1 (2x scale), showing bileaflet scaffolding present in model almost entirely missing from print

Prototyping, iteration 2

As a result of Prof. Sniadecki and I setting up short biweekly meetings to check in on project progress throughout the 11-week-long academic quarter, the project was split up into five neat two-week periods which functioned like design sprints. Keeping up with the numerous ideas thrown back and forth in these 30-minute meetings and producing sketches of higher quality from them proved to be a challenge, functioning more as back-of-the-napkin notes rather than primetime-ready sketches.

Spurred by the need to create precise features on a small length scale, I pivoted to resin 3D printing (stereolithography) and created a elliptic frame to house two thin semielliptic leaflets resting on an internal lip. This design also allowed me to experiment with not only printing hinges in place with a more precise method, but also with press-fitting components with more precise tolerances.

Notes from the meeting preceding the development of iteration 2, exploring potential alternatives to printing hinges in place.

A screenshot of CHITUBOX's graphical user interface with 1x and 2x scale cut-open models of iteration 1, intended to be press-fit together, put onto a virtual 3D printer buildplate. The CHITUBOX software package automatically adds supports onto 3D printer beds, making prints higher quality and much easier to detach from the buildplate before beginning post-processing.

Cured and processed prints of iteration 2 (1x, 1.2x, and 2x scale) done near the halfway mark of the project, with all components still attached to supports.

A close-up of the printed-in-place hinges on the 2x scale model of iteration 2, showing that although the hinges are visually distinct inside the frame, the cavities are still too small so the hinge fused to the frame during printing.

A close-up of a hinge on a 1x scale model of a leaflet in iteration 2, showing its offset from the central axis of the overall design. This offset resulted in a rotational moment being engaged onto the hinge when the leaflet rotated out of position, making it more susceptible to fracturing earlier than the rest of the part.

Prototyping, iteration 3

One of my personal goals for this project was to gain a better understanding of moments, torques, and pressures on shafts in contexts extending beyond gear combinations covered in class. Applying elementary mechanics of materials relationships onto a part with well-defined geometry would yield a good theoretical estimate for the maximum loads applied during blood flow through the valve, enabling an analysis of the materials in use and a comparison to FEA results. A decrease in frontal area not covered by the leaflets, as well as the intagration of two straight shafts connected to the leaflets and running parallel to the central axis of the design, reduces stress concentrations in the leaflets and reduces the area of the design over which backflow may occur.

This iteration was marked by print failures, including poor adhesion of parts to the supports, the buildplate, and to each other, underscoring the importance of factoring in maintenance and quality control when estimating processing times for products. Instead of playing around with prototypes and refining them further, a lot of my time spent on this project late in the quarter involved pulling together many resources in the on-campus makerspaces to troubleshoot and repair the available SLA 3D printers.

Notes from the meeting preceding the development of iteration 3, outlining options for manual force analysis through linear-elastic assumptions.

A 3D print that failed from bad adhesion, as shown by parts missing from the tops of support structures on the left and uneven thickness of parts on the right, after post-processing.

A 3D print that failed from bad adhesion, as shown by parts missing from the tops of support structures on the right and uneven thickness of parts on the left, immediately before post-processing.

A 2x scale model of iteration 3 (left) compared to a printed-in-place 1x scale model of iteration 2 (middle) and a dime (right), showing the small sizes of features bearing forces resulting from blood flow.

Renderings of iteration 3

Isometric view of the top of a CAD model of iteration 3, with the leaflets shown in the open position.

Diagonal view of the bottom of a CAD model of iteration 3, with the leaflets shown in the open position.

Short animation of press-fit iteration 3 assembly.

Close-up view of the parallel hinges in iteration 3, from above.

Force analysis

One of the important constraints for a heart valve design is that it shouldn't break under peak loads from blood flow. 1D linear-elastic mechanics, combined with some reverse-engineering of material properties, provides for a useful approximation of the entire leaflet as a thin cylindrical shaft. The safety factor of over 100 likely indicates that even if this is an unsafe assumption to make, the cured resin is strong enough that a single blood flow cycle won't cause it to fracture. However, this analysis does not include fatigue cycling, which is how historical heart valve designs have failed.

Finite element analysis

One of the skills I wanted to get practice with through this project is using ANSYS to run linear-elastic simulations, a topic touched on in class through simulations on springs and holes in plates. Because the theoretical force analysis above condensed a leaflet to a slender cylindrical shaft and assumed 1D linear-elastic mechanics were a good approximation, comparing FEA results to results from hand calculations was a valuable learning opportunity. A test applying the same force to the same location on the same size beam but with fixed supports instead of simple supports yielded a maximum deflection value over 2000% larger than the hand-calculated maximum deflection value.

Even though I knew this before starting this project, I now have first hand experience with the concept of "garbage in, garbage out" — there are so many variables in setting up even the simplest of finite element simulations, and it's important to consider each of them carefully with real-life/hand-calculated analogs.

Screenshot of ANSYS Mechanical graphical user interface, showing a solution to a Static Structural deformation test on a shaft with the same dimensions as in the theoretical force analysis

Next steps

If I had more time...

• Replacing press-fit tabs with hook connectors to snap parts together

• Fatigue analysis using theory and FEA simulations

• Exploration of material interactions in vivo — thrombogenesis, leeching, calcification, for example — and effects on material properties (like fracture), which may elucidate use of different manufacturing methods or biocompatible materials

• Apply adhesives or modify design to reduce backflow through cracks in press-fit parts

• Recreate CAD model parametrically to allow valves of differing radii and closer fits to individual aortae

It's one thing to look at a CAD model, it's another thing to have a scale prototype of it, and it's yet another thing to translate it into a mass-produced product

Acknowledgments

Many thanks to Dr. Nathan Sniadecki, Professor in Mechanical Engineering at the University of Washington, for invaluable feedback, insights, and mentorship throughout the project.

Thanks to Evan O'Neill, Jason Wu, and McCarty Innovation and Learning Laboratory staff for equipment access and troubleshooting support.