AI Copilot for Process Simulation — Backed by Deterministic Solvers
Describe a flowsheet in plain English, debug a failed convergence instantly, and optimize a parameter. The AI proposes — the thermodynamic engine confirms every single number.
No hallucinated results · every number traces to a solver residual · runs in your browser
Try it yourself
Pick a prompt shape below — the copilot handles all three the same way: propose, then let the solver confirm.
The AI proposes; the solver still computes every number.
Why it doesn't hallucinate
The LLM never touches a physical number. It only ever proposes a flowsheet or a parameter change; the deterministic engine computes every result.
Three prompts engineers run every day
Not abstract — these are the shapes of request the copilot handles today.
Natural-language flowsheet generation
12 unit ops auto-placed (absorber, stripper, lean/rich exchanger, reboiler), property package set to electrolyte NRTL, initial conditions estimated — ready to run.
Automated convergence debugging
“Your reflux ratio (5.0) drives a distillate flow above the feed — the material balance is infeasible. Lowering it to 2.5 resolves it.” The solver re-runs and confirms.
Multi-variable optimization
The copilot proposes the search space; SciPy/Pyomo runs the actual sweep and returns the exact tray count, feed stage, and reboiler duty — every point solver-verified.
Free academic and individual tier available · zero desktop setup