Economic re-optimization, fast enough to actually run every cycle.
A plant-wide multistart Ipopt NLP re-solves the economic optimum as feed costs and product prices drift, then pushes the new targets straight to the APC layer holding your process there — measured on this server, not a projected number.
How the three tiers close the loop
RTO sets the economic target, APC holds it every control tick, and the reconciled digital twin keeps the model RTO optimizes against honest — each tier is independently real, and now linked.
What actually runs each cycle
A real multistart nonlinear economic optimization — not a spreadsheet LP wrapped in a scheduler.
Multistart Ipopt, every cycle
Each re-optimization cycle solves the plant-wide nonlinear program from several random starting points and keeps the best feasible plan — the same robustness a legacy RTO suite claims, without the separate license.
Sub-second per cycle, measured
Every multistart count on this page is timed live on this server against a stated cycle budget, not asserted — see the benchmark below.
Pushes straight to your APC layer
A closed-loop RTO link re-solves on a schedule and writes the new economic targets to a linked MIMO DMC controller automatically — no manual hand-off between the economic layer and the control layer.
Reconciled against live plant data
The steady-state gate checks the plant is actually settled before RTO re-optimizes against it, and the digital twin's OPC-UA/MQTT reconciliation keeps the model it solves against honest.
No fragile equation-oriented crash
The multistart NLP degrades to its best feasible plan instead of raising — a bad cycle reports status, it doesn't take the loop down.
Web-native, no desktop install
Runs in the browser against the same solver every simulation uses — no separate RTO workstation, no client software to patch and maintain.
Live RTO cycle-solve benchmark
The real plant-wide multistart Ipopt NLP (flowsim/solver/rto.py), timed on this server at increasing multistart-restart counts — not example figures.
| Multistart restarts | Solve time | Cycle budget | Status | Status | Run live |
|---|---|---|---|---|---|
| 1 multistart | 9.9 ms | 1000 ms | solved | Passed | |
| 4 multistarts | 38.8 ms | 1000 ms | solved | Passed | |
| 8 multistarts | 62.7 ms | 1000 ms | solved | Passed | |
| 16 multistarts | 109.7 ms | 1000 ms | solved | Passed |
Median of 3 repeated cycles of the real olefins-furnace RTO benchmark, against a 1 s cycle budget — measured on this server, not fabricated.
Why plants switch off legacy RTO suites
A structural contrast, not a feature checkbox — the specifics of any vendor's current product are theirs to publish.
| Traditional RTO suites | MaximaLabs RTO | |
|---|---|---|
| Licensing | A separate RTO module, priced and licensed apart from the base simulator | The same web app and account as every other simulation |
| Deployment | A dedicated engineering project — commissioning, step-testing, on-site tuning | Model the plant network in the browser and schedule cycles the same day |
| Execution | Desktop client software, patched and maintained per site | Web-native, same CasADi/Ipopt solver core every simulation runs |
| Failure mode | Equation-oriented solves that can crash on a bad iteration | Multistart NLP degrades to its best feasible plan and reports status — never raises |
| APC hand-off | Often a separate integration project between the RTO and APC vendors | A closed-loop RTO link pushes targets to a MIMO DMC controller directly |
Scripted from Python
The same plant-wide NLP the RTO panel and the live benchmark above actually solve.
from flowsim.solver.rto import (
RtoCycle, default_olefins_furnace_network, run_rto_schedule,
)
network = default_olefins_furnace_network()
cycles = [
RtoCycle(label="naphtha up 15%", feed_cost={"naphtha": 6.9}),
RtoCycle(label="ethylene demand cut", product_demand_max={"ethylene": 80.0}),
]
trajectory = run_rto_schedule(network, cycles, n_starts=8)
for point in trajectory:
print(point.label, point.result.status, point.result.objective, point.solve_seconds)Deploy without an 18-month project
A traditional RTO rollout is a separately licensed module, a dedicated engineering project, and a multi-month commissioning window before the first economic cycle runs.
Model the plant network
Feeds, process units, pooling/blending, and products with their costs, prices, and demand caps — the same plant-wide NLP the benchmark below actually solves.
Schedule the market drift
Define the feed-cost, product-price, and demand-cap changes each RTO cycle should react to, or link it to live data so cycles run against actual conditions.
Link it to APC and go closed-loop
Point the RTO link at a MIMO DMC controller and set a re-solve schedule — new economic targets flow to the control layer automatically, gated on the plant actually being settled.
See it on your own process
Every number on this page comes from a real, converged flowsheet — open the workspace and run one yourself.