● Beta OpSimNav — WASP & decarbonization decision intelligence — is now live. Visit OpSimNav →
OpSimTech / Method & Toolkit
Approach · Method Selection & Simulation Toolkit

The Right Physics-AI Method For Every Problem

No single algorithm fits every application. The art is matching the method to the physics, the data, and the deployment constraint. Here's how we decide — and the full toolkit we decide between.

/ 01 — Method Selection

How we pick the Physics-AI method.

Four cells — each is a class of problem we see often, paired with the method we use for it and the reason.

Application type
Method we use
Why
/ 01
Continuous flow fieldsaerodynamics · hydrodynamics
PINNs
Governing equations fully encodable; no labeled data needed. The PDE residual is the loss.
/ 02
Complex manufacturingelectrode mixing · coating · calendering
Physics-Guided Surrogates · ROM / ML
PINNs intractable at 3D + time; physics constraints embedded in architecture or loss instead.
/ 03
Sensor-rich operationsrunning fleets · instrumented lines
Data-Driven + Physics Validation
High-frequency sensor data dominates; CFD/SPH serve as offline validators, not runtime solvers.
/ 04
All applicationscross-cutting fidelity strategy
Multifidelity Coupling
Low-fidelity online; high-fidelity offline to train, anchor, and validate the twin.
We don't force every problem into a PINN.
We engineer the right solution for your physics.
/ Honesty PINNs remain an active research area. We apply them where they are proven, and use physics-guided ML where they are not. Acknowledging model limits is part of our quality standard.
/ 02 — Simulation Toolkit

Our simulation toolkit.

Three coherent layers across the OpSimTech stack — from first-principles solvers to system-level twin enablers.

A · High-Fidelity Physics Solvers

/ Foundation assets
/ Cap 01

CFD

Fluid dynamics, turbulence, heat transfer. RANS, LES, DES for steady & transient flows.

/ Cap 02

SPH

Smoothed particle hydrodynamics. Free-surface, slurry, mixing, multiphase.

/ Cap 03

DEM

Discrete element method. Granular mechanics & particle-level manufacturing.

B · Physics-AI Acceleration Layer

/ OpSimTech IP
/ Cap 04

PINNs

Physics-residual-trained neural solvers. PDEs encoded directly in loss.

/ Cap 05

Physics-Guided Surrogates

ROM, Gaussian processes, ML with physics constraints in architecture or loss.

/ Cap 06

Simulation-Driven Optimization

Sweep thousands of designs against multi-objective constraints.

C · System-Level Twin Enablers

/ The twin in production
/ Cap 07

Multifidelity Modeling

Couple low- and high-fidelity intelligently — speed where you can, accuracy where you must.

/ Cap 08

Real-Time Data Assimilation

Sensor streams fused into simulation states. The twin stays current with the real asset.

/ Cap 09

Decision Support Engine

Optimization & scenario analysis on top of the twin. Outputs are decisions, not numbers.

/ Get in touch

Got a problem
we should look at?
Let's talk method.

Request a Demo Email an engineer