What ROI can you realistically expect from APC and optimization?
Three published reference implementations with their sources, where the value usually comes from — and what actually changes outcomes inside a real project.
Book a 30-min technical fit callPublished cases: ETH Zurich OptiControl-II · Holcim Lägerdorf (ABB) · Dillinger (TU Wien) — sources linked
The results are real — and they are conditional
These references are peer-reviewed papers and vendor technical articles about real plants, each linked to its source. They are not marketing extrapolations — one of them states openly that the business case did not close for that building. Every process is different, instrumentation varies, the team and operations matter, and the gap between a published case and your asset is exactly what the Remote Diagnostic Package is built to estimate.
Three published reference cases
ETH Zurich — MPC in an occupied office building
OptiControl-II ran MPC for seven months in a fully occupied Swiss office building. It controlled the building reliably with good comfort; in simulation, MPC used 17% less non-renewable primary energy than the previously installed rule-based control. The authors also found that, with the tools then available, the initial investment was too high to justify on operating savings alone — the engineering effort decides the business case. Sturzenegger et al., IEEE Transactions on Control Systems Technology, 2016.
Source (DOI) →Holcim Lägerdorf / ABB — cement precalciner
Model predictive precalciner temperature control (ABB Expert Optimizer), installed in August 2006 while fossil fuels were being replaced by alternative fuels. Precalciner temperature variability fell by more than 50%, and from June 2007 the calciner ran coal-free on five alternative fuels, with coal kept on standby. Kiener & Gallestey (ABB), World Cement, March 2008; Marx & Gallestey (ABB), 2008.
Source (PDF) →Dillinger / TU Wien — slab reheating furnace
Nonlinear MPC of a slab reheating furnace, commissioned in February 2011. Over more than 7,000 slabs, the mean final temperature error fell from 13.0 K (systematic overheating) to 0.9 K, and the share of slabs outside the target band from 59% to 12% — the authors note the band was set conservatively, so these are not reject rates. The system was then installed on two more furnaces in the same plant. Steinboeck, Wild, Kugi, Control Engineering Practice, 2013.
Source (DOI) →Where the value usually comes from
High-energy continuous processes
Cement, lime, steel reheating, paper drying, large compressors, cryogenics: energy and yield. Payback tends to be fastest where energy is the dominant cost and the existing loops can be tuned.
Quality-driven semi-continuous processes
Reheating furnaces, glass, ceramics, batch reactors: less off-spec product and more stable downstream operation.
Mechanical and machine systems
Wind turbines, large machines, robotic cells: load reduction, damping and shorter commissioning time.
Soft sensors and estimators
When a missing measurement is the actual bottleneck, a model-based estimator unblocks tighter control or better quality decisions.
What actually changes the outcome
Where the bottleneck really sits
Most projects under- or over-invest because the bottleneck was misidentified. The biggest single ROI lever is naming the right problem before committing to a solution architecture.
Instrumentation and data quality
Advanced control without sound measurement infrastructure underperforms. A short instrumentation review often pays back faster than any controller upgrade.
Operations and acceptance
Operator trust, alarm hygiene, and a clear handover plan determine whether the gain survives the first six months in production.
Model fidelity vs. complexity
The right model is as simple as the control problem allows. Over-modelled systems are slow to commission and brittle to maintain; under-modelled systems leave money on the table.
How the Remote Diagnostic Package estimates your range
The Remote Diagnostic Package takes your existing documentation, control loops, data samples, and a one-hour technical interview, and returns a 1-page executive summary plus a detailed technical report. The report names the dominant bottleneck, the realistic ROI range for your specific case, the data quality risks, and the recommended next phase — in writing, with deliverables fixed.
Want a calibrated ROI estimate for your specific asset?
A short call is enough to decide whether the Remote Diagnostic Package can give you a realistic answer in five working days.
Book a 30-min technical fit call →