Device model explorer

Two architectures, one set of reliability questions

Switch between our own design and the HeartMate 3 reference device. The geometry, the operating parameters and the derived load spectra all change with the selection — so you can see exactly which conclusions belong to the architecture and which belong to us.

Model

WebGL is unavailable here, so the model cannot render. Every number it drives is in the tables below.

drag to orbit · scroll to zoom

What each part does

Point at any part to isolate it in the model. Selecting one pulls the assembly apart.

Architecture

Blood flow

Build

View

Operating point

3,000 rpm

0°

Artificial pulse

Impeller blades

6

Motion

Rotation is slowed about 14× so the blades stay readable. The artificial-pulse surge keeps its true two-second period.

The circulation our model was missing

Two circuits, one device

A total artificial heart replaces both ventricles, so it has to drive two loops in series — the same blood, the same flow, against pressures that differ by a factor of six. Until 5 August 2026 our reliability model contained only the left one.

LUNGS gas exchange BODY systemic capillary bed RIGHT pulmonary side 15 mmHg ≈ 5 L/min LEFT systemic side 100 mmHg ≈ 5 L/min one levitated rotor · an impeller on each face Vena cavae superior + inferior deoxygenated · anastomosis 1 Pulmonary artery the one artery carrying deoxygenated blood anastomosis 4 of 4 Pulmonary veins the one vein carrying oxygenated blood anastomosis 2 of 4 Aorta oxygenated anastomosis 3 of 4 CARDIANOVA — REPLACES BOTH VENTRICLES
The pressure ratio is the design problem. Both sides must move the same 5 L/min, but the right side works against roughly 15 mmHg and the left against roughly 100. For a centrifugal stage head scales with (speed × diameter)², so on a shared shaft the right impeller would need about 0.39 of the left's diameter at equal speed — it cannot be a mirror image. The remaining trim comes from vane height and, in service, from the axial position of the levitated rotor. Colour follows oxygenation, not vessel type: the pulmonary artery carries deoxygenated blood and the pulmonary veins carry oxygenated blood — the two exceptions in the body.

What this cost the reliability model. Re-running the existing pipeline with both circulations, four anastomoses and a suspension that now also trims left-right balance: R(5 yr) falls from 72.4% to 62.5% for the shared-rotor topology, and the per-subsystem requirement to reach the 90% target tightens from 0.9869 to 0.9913 because there are twelve subsystems in series rather than eight. Two independent pumps do slightly worse still, at 60.8% across sixteen subsystems — fewer parts in series beats hydraulic independence. The four anastomoses alone account for about a quarter of predicted failures at five years.

Parameters

Parameters

ParameterValueSourceReliability consequence

The comparison that matters

Clearance, drawn to one scale

Our contact-bearing architecture and the reference device differ by a factor no material choice can close. This figure does not change with the selector — it is the reason the selector exists.

DRAWN TO ONE COMMON SCALE — 1 px = 12 µm CardiaNova Architecture A — contact journal bearing 30 µm — the entire failure budget wear depth consumes this; predicted life 2.4–242 yr journal ⌀ 6 mm · 148,700 km of sliding over five years HeartMate 3 — rotor to housing, µCT of an explanted pump 1,000 µm top gap — 33× the clearance above levitated rotor — no contact, no wear surface 1,750 µm bottom gap — 58×
On one scale, the contact bearing’s whole failure budget — the 30 µm the wear depth has to eat through — is the thin red line. The reference device runs its rotor in a blood gap 33 to 58 times larger. That is why the Archard wear coefficient, the single dominant uncertainty in our study, does not appear in its reliability model at all. Gap values from µCT reconstruction of an explanted pump; 0.5 mm radial and 1.0 mm axial are also reported by the design team but were not verified from the primary paper.