How multivariable predictive controllers work, how to design and step-test them, and why most of them end up switched off
The flyer below is the printable A4 version. Use the buttons above to download it or the request form.
APC and control engineers, process engineers supporting APC applications, console supervisors and operations engineers who own the benefits
An MPC application is a set of models, limits and economics that must be maintained like any other piece of equipment; the industry's experience is that half of them are off within three years. The course explains what a predictive controller actually does with its models and objective function, then works through the lifecycle that keeps it on: benefits estimation, base-layer preparation, step-test planning, model identification, commissioning, and the monitoring that detects model mismatch before the operators lose confidence. Exercises use a distillation column case in ProcessVision.
Why APC pays: variability reduction and constraint pushing · MPC anatomy: models, prediction horizon, control horizon, move suppression · steady-state optimiser and economics · benefits estimation and functional design · base-layer audit with CeraControl · inferential and soft-sensor requirements · step-test planning: signals, amplitude, duration, PRBS
Model identification and model-quality judgement · controller tuning weights and limit ranges · operator interface, alarms and ISA-101 considerations · commissioning and benefit verification · monitoring KPIs: service factor, limits active, prediction error, model mismatch · sustaining: MOC for APC, re-identification triggers, roles · workshop on a column application in ProcessVision
ISA-95 / IEC 62264ISA-101ISA-18.2API RP 554IEC 61131-3Send a short brief — plant, service of interest and what you need to achieve — and a Cerasus engineer will respond with a proposed scope and next steps.
Prefer email or phone? sales@cerasus.ai · +974 5506 2743