Surphysis

Interface qualification

Will this joint, coating or barrier survive its duty cycle?

A stack, a duty cycle and your own characterisation files in; fracture margin, breakthrough time and survival probability out, with the measurement that would change the answer.

Case C-0412, sapphire to IN718mock
Cross-section of a ceramic-to-metal joint with a five-step graded interlayer; a temperature front crosses the stack and the fracture margin at the critical interface counts up G / Gc at the critical interface 0.00

Three cases

How it works

four steps, one audit trail
  1. 01

    Describe the stack

    Choose materials from the passport library or type them in. Set the duty cycle: temperature history, pressure, environment, number of cycles.

  2. 02

    Attach what you measured

    Accepted: XRD as .xy or .xrdml, EBSD as .ctf or .ang, SEM as .tif with a scale bar, permeation as .csv. Each file tightens a prior.

  3. 03

    Watch the engine reason

    Stress, minimax grading, fracture, grain-boundary kinetics and Sobol sensitivity run as numbered rules R00 to R24. Every step is logged with its rule reference.

  4. 04

    Take the verdict to the bench

    Fracture margin, breakthrough time, survival probability, a ranked measure-next list, and the recipe as editable code with an acceptance test plan.

Public scorecard

all blind predictions
42
mock
Blind predictions logged
since 2026-06-01
36 / 42
mock
Within stated error band
86 % hit rate
0.18
mock
Mean absolute log error
dex, life metrics
11
mock
Published failure datasets
4 mechanisms

Every prediction is logged before the published outcome is looked up. Misses stay on the board with the rule that caused them. The tiles above are placeholders until the first blind predictions are scored.

Origin

Surphysis began as the modelling stack behind a series of papers on ceramic-to-metal graded interlayers for additive manufacturing. In that work a chromium diffusion barrier that lattice-diffusion theory called safe for a hundred laser pulses was consumed in a single pulse once the measured grain size of 15 nm to 25 nm entered the model.

The platform exists so that every verdict is computed from measured microstructure and a stated set of assumptions rather than from handbook values. Interface toughness by mode mixity, grain-boundary diffusion coefficients and thin-film residual stresses are curated with provenance and uncertainty, and the engine is scored in public against published failure data.