Hybrid MPC and Expert Control for Industrial Process Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced Process Control (APC) technologies in industrial processes face limitations, particularly with Model Predictive Control (MPC) struggling to accurately predict variable trajectories in all operating regions and rule-based approaches being non-optimal due to reliance on human expertise, leading to inefficient operation and potential equipment breakdowns.
Innovation Solution
A hybrid control system that combines Model Predictive Control (MPC) with an expert component, utilizing double-entry control variables where one instance is set by an engineer/operator and the other by an expert system, allowing for dynamic adjustment to prevent instability and optimize process operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If Model Predictive Control (MPC) is used to control industrial processes, then the system can handle large systems through scalable algorithms, but it performs poorly when its models cannot accurately predict variable trajectories in all operating regions
Solution Approach 1:
The patent combines Model Predictive Control (MPC) with rule-based control systems into a unified hybrid control architecture. The rule-based component compensates for MPC's weaknesses in regions where the mathematical model cannot accurately predict process behavior, while MPC handles the scalable control of large systems. This merging allows the system to leverage the strengths of both approaches simultaneously.
2Adaptability or versatility
If rule-based control approaches are used with many inputs and outputs, then the system can address process interactions, but it becomes unwieldy with thousands of rules that must be written
Solution Approach 1:
The hybrid control system merges rule-based control with MPC, where the rule-based component provides high-level guidance and constraints, while MPC handles the detailed control calculations. This division reduces the number of individual rules needed compared to a pure rule-based system, as MPC's mathematical model captures general process behavior without requiring explicit rules for every scenario.
Solution Approach 2:
The MPC component serves multiple functions simultaneously: it optimizes process variables, handles constraints, manages interactions between multiple inputs and outputs, and provides a framework for incorporating rules. This multi-functionality reduces the need for separate rule-based logic for each control objective.
3Productivity
If operators push processes to their limits to operate at optimum points, then productivity increases, but equipment or process breakdowns occur leading to downtime and lost profits
Solution Approach 1:
The hybrid control system continuously monitors process variables and uses feedback from both the MPC model predictions and rule-based evaluations to adjust control actions. This real-time feedback allows the system to operate near optimal points while detecting early signs of instability or approaching critical limits, enabling preventive adjustments before breakdowns occur.
Solution Approach 2:
The rule-based component of the hybrid system includes predefined rules that detect incipient overload conditions and trigger corrective actions before actual breakdowns occur. This preliminary detection and intervention allows the system to operate closer to limits safely by anticipating and preventing failure modes before they manifest.
4Reliability
If operators keep processes well within limits with a cushion, then equipment breakdowns are prevented, but lost profits occur due to suboptimal operation
Solution Approach 1:
The continuous feedback mechanism in the hybrid control system allows real-time optimization by adjusting operating points based on current process conditions, equipment state, and predicted future behavior. This enables the system to safely operate closer to optimal points dynamically rather than maintaining static conservative margins, thereby improving profitability while preserving reliability.
Data Source
AI summary
Controlling and optimizing industrial processes by integrating MPC-based approaches and expert system approaches. At least two different control variables with identical models are used. An expert system adjusts at least one of the control variables to change a setpoint or range or the like while standard MPC techniques change another control variable to address appropriate classes of control problems.