Stent finite element analysis for crimping, deployment and pulsatile fatigue

Stent finite element analysis predicts how a stent or stent-graft deforms when it is crimped, released into a vessel and loaded by every heartbeat, and where the strain at the struts is highest. We build it around your existing solver, so a strut layout, heat treatment or oversizing range is compared on mean and alternating strain, radial force and a fatigue margin before samples are cut and a pulsatile rig slot is booked.

The design freeze for the next size matrix is close, and a longer lesion has forced a change to the strut layout. The durability rig reports fracture or survival only at the end. Which size governs, which crown breaks first, and what does the change cost in radial force? A wrong guess becomes a failed run and a moved submission date. FDA's guidance on non-clinical engineering tests for intravascular stents recommends a fatigue analysis and says accelerated durability testing validates it: the analysis is the plan, the rig the check.

What does stent finite element analysis have to capture?

One model, three load steps: crimping packs the stent into its delivery system, release lets it expand until the vessel wall stops it, and pulsatile loading cycles the vessel diameter with every heartbeat. ASTM F2514-21, the guide to finite element analysis of metallic vascular stents, lists the same sequence and says such analysis is frequently used to find the worst-case size for fatigue testing and to tell designs apart.

A balloon-expandable stent is deformed plastically, so recoil and the fatigue strength of the steel or cobalt-chromium decide it. A self-expanding stent is superelastic nitinol: the review by Robertson, Pelton and Ritchie (2012) notes a strain excursion of up to 10% as it is constrained into the delivery system, then slight oversizing that holds a constant mean strain while heartbeat and body motion add the amplitude. FDA recommends measuring the radial force such a stent exerts on the wall, so the model reports it beside the strain.

For each crown the strain history gives a mean ε_m = (ε_max + ε_min)/2 and an amplitude ε_a = (ε_max − ε_min)/2, the values FDA recommends taking from the stress and strain analysis.

How do you model nitinol and the vessel wall?

Nitinol is the hard part. Its stress barely changes along the transformation plateau, the plateau moves with temperature (a Clausius-Clapeyron relation), and FDA notes that its properties vary widely with processing, so it recommends physical experiments on actual post-processing samples. We calibrate the superelastic law to your test files at body temperature. The review is plain about limits: fatigue studies have not produced a universally applicable constitutive relationship, and the multiaxial response is uncertain because the transformation depends on the loading path.

The vessel is a boundary condition with consequences: a rigid tube over-constrains the stent, while taper, curvature and an eccentric or calcified lesion change where it touches. FDA recommends validating with bench testing, for example by comparing the force-displacement result of a radial loading test with a simulation of it, and ASTM F2477-24 deploys samples into a mock, elastically simulated vessel. So we validate on the bench vessel first, then swap in anatomy.

How is a stent fatigue safety factor calculated?

FDA recommends a Goodman analysis or another fatigue life method, a Goodman diagram or similar graphic, and safety factors in a table that explains how they were calculated. For a conventional metal the modified Goodman relation is σ_a/S_e + σ_m/S_u = 1/n, with S_e the fatigue strength at the target life, S_u the tensile strength and n the safety factor.

Superelastic nitinol does not follow it. The review concludes from published coupon tests that, up to the end of the stress plateau, life is driven by strain amplitude rather than mean strain, and that Goodman and Soderberg analyses are not appropriate. In the stent-cell coupon data it summarises, raising the mean strain from 1.5% to 4% raised the tolerated amplitude at 10⁷ cycles from 0.4% to 0.8%: one alloy, process and test, not an allowable for your device.

What stays outside a stent model: surface, corrosion and scatter

The model computes strain on a nominal surface; fatigue is decided at the real one. The review reports that electropolishing slightly but significantly raised nitinol life at low cycle counts and made no significant difference at high ones. FDA recommends examining fretting corrosion within the durability test, inspecting samples afterwards for corrosion, and names ASTM F2129 for corrosion potential. A strain model sees none of it.

That is why the test exists: FDA says it evaluates failure modes such as fretting, abrasion, wear and fracture. The ASTM F2477-24 scope covers pulsatile diameter change only, so peripheral stents need justified bending, axial and torsional boundary conditions. It includes endovascular prostheses with metal stents; FDA's stent guidance addresses stents, not stent-grafts.

How much of the anatomy and tolerance range does the analysis cover?

Diameter, taper, curvature and lesion shape vary between vessels, and so do the stent's own dimensions within tolerance. FDA recommends addressing dimensional variation within allowable tolerances and explaining why the modelled size is the worst case. We turn that into a sweep over parametric geometry, an anatomy set from your data and your size matrix, reporting the strain distribution per crown and the combinations not covered. It describes loads, not one patient.

How credible does a stent model have to be?

ASME V&V 40-2018, Assessing Credibility of Computational Modeling through Verification and Validation: Application to Medical Devices, says credibility should be commensurate with the degree to which the model is relied on as evidence and the consequences of an incorrect decision. FDA's November 2023 guidance, Assessing the Credibility of Computational Modeling and Simulation in Medical Device Submissions, applies that risk-informed idea to physics-based models, and ASTM F2514-21 builds on V&V 40 for stent models.

Our context of use: the model chooses the size and crown for the durability samples and ranks strut variants. The pulsatile test still reports survival or fracture, so the model's influence on that decision is limited and the credibility goals follow. If the model should carry more, say by dropping a size from the matrix, the goals rise and the regulator or notified body is asked first. Verification is mesh convergence at the crowns; validation is your radial force and diameter data; uncertainty is plateau stress, friction, tolerance and vessel stiffness, swept to a range.

Where does AI help stent design, and where does it not?

• Design-space exploration. Full runs at designed points train a Kriging surrogate of crown strain, radial force and crimp strain against the strut geometry; an expected-improvement rule picks the next run. Li and colleagues (2016) did this for a balloon-expandable stent: seven geometric variables, an ANSYS model per design and 27 iterations to the optimum for dogboning and recoil; the authors say it still needs experimental validation.
• Anatomy and tolerance sweeps. A surrogate returns crown strain for far more combinations than full runs allow; anything near the boundary is re-run in full.
• Calibration. An inverse fit of nitinol constants to your tests returns a range.

It does not replace the fatigue test, predict corrosion or surface effects, or extrapolate. Crown strain is not smooth where a hinge crosses onto the transformation plateau, so we train with the plateau state, test on held-out full runs and gate anything outside the range to the full model. FDA's credibility guidance covers physics-based, mechanistic or other first-principles models, so the physics model stays the evidence and the surrogate chooses the runs. See surrogate models for CFD and FEA, AI surrogate models you can sign off and our uncertainty quantification page.

From one stent study to a design workflow you keep

The return comes when your engineers can put the next strut layout or size matrix through the same chain.

• A parametric stent model in your existing solver: strut geometry and tolerances as parameters, vessel and lesion as inputs, crimp, release, contact and pulse in one load-step template.
• A calibration script from your plateau and tensile test files to the material law, with fit and range on record.
• A post-processor writing mean and alternating strain per crown, the safety-factor table and the diagram from each run.
• A validation record against your radial force test and a diameter or strain record, tolerances set before the run.
• A surrogate with a stated range and a gate to the full model outside it.
• Handover and training, so your team runs and extends it. See our training and software development services, and the cardiovascular implants page.

Questions engineers ask about stent finite element analysis

If a size matrix or strut change is waiting on a fatigue answer, send us the strut layout, the alloy and heat-treatment data with your plateau and austenite-finish curves, the vessel and oversizing range, and the radial force and durability tests you have run or planned. We reply with how we would build the model, the outputs it would report, what the validation record should contain and what a surrogate could cover. Start with a conversation, or read the capability page.