Aspen Nonlinear Controller
Non-linear model predictive control
Aspen Nonlinear Controller is a multivariable model predictive controller capable of full automation and dynamic optimization of product grade transitions. It uses a combination of simple first-principle models (general equations) in conjunction with nonlinear approximation technology that can be safely used directly within a controller. Aspen Nonlinear Controller is a core element of AspenTech’s aspenONE® Advanced Process Control applications.
Features
- Full non-linear model predictive control: Fully nonlinear (dynamic and steady state, interacting nonlinearities).
- Reduced modeling complexity: Ability to capture all relationships in a single model per controlled variable.
- All parts of the solution reference the same model. Models are gain, time constant, and delay type models.
- State-space model formulation with Extended Kalman Filter for unmeasured disturbance handling.
- Little or no step testing required.
- Less reliance on lab feedback during transitions: Highly accurate models, reliable process gain profiles.
- Constraint ranking: Targets and limits for different variables can be ranked so that under conditions of infeasibility, the optimizer will give up on lower ranked constraints first.
- Dead time and dynamics: Independent dead time alignment for each pair of relationships. General state space models and parametric dynamics supported.
- Guaranteed gain and extrapolation: Bounded Derivative networks guarantee gains will be within specified bounds. Models extrapolate sensibly outside data in existing operating regions.
- Variable dynamics support: Variable dynamics and deadtimes are supported. Model dynamics can be adjusted online in real-time.
- Gain constrained dynamics: Accurate modeling of non-linear gains across the entire operating space.
- Ease of use: Operator interfaces designed by operators for operators.
- Interactive graphical simulation environment: Simulate open loop and closed loop performance. Validate controller tuning including move plan, optimization, and unmeasured disturbance rejection.
Benefits
- Full automation and optimization of even complex-grade transition strategies
- Easily model multiple catalysts and donors in a single model
- Faster grade transitions
- Reduced maintenance compared to multiple model per CV designs
- Robust solution (a single model per CV eliminates model conflict problems associated with controllers where the inferential model is different to the control model)