Industrial Process Control Algorithm Switching for Stable Operation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The industry lacks an efficient orchestration methodology for seamlessly switching among various control algorithms to manage industrial processes effectively, particularly in conditions of uncertainty and complexity, where traditional methods like Model Predictive Control, reinforcement learning, and rule-based systems have limitations in scalability, adaptability, and stability.
Innovation Solution
A method and controller that dynamically select the most suitable control algorithm based on monitored process variables, incorporating a penalty for algorithm switching to ensure stability and optimal performance, using an orchestrator with a classifier, optimizer, and transition manager to facilitate smooth transitions between algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple control algorithms are used to handle diverse process conditions, then adaptability and robustness improve, but system complexity and difficulty of orchestration increase
Solution Approach 1:
The patent combines multiple control algorithms (MPC, RL, rule-based control) into a unified orchestration system that selects and coordinates them based on process conditions. This merging approach allows the system to leverage the strengths of each algorithm while managing complexity through centralized coordination.
Solution Approach 2:
The orchestration system serves as a universal controller that can handle diverse process conditions by selecting from multiple control algorithms. It provides multi-functionality by adapting to different scenarios (normal operation, disturbances, critical conditions) using the appropriate algorithm, making the system versatile across various operating conditions.
2Adaptability or versatility
If control algorithms are switched frequently to optimize performance, then adaptability improves, but system stability deteriorates due to high-frequency switching
Solution Approach 1:
The switching penalty parameter is dynamic rather than fixed. It adapts based on process conditions, being higher during stable operation to prevent unnecessary switching and lower during critical transitions to enable timely algorithm changes. This dynamic adjustment balances adaptability and stability.
Solution Approach 2:
The system continuously monitors process variables and performance metrics to dynamically adjust the switching penalty. This feedback mechanism ensures that switching decisions are based on actual process needs rather than arbitrary thresholds, maintaining stability while enabling necessary adaptability.
3Productivity
If advanced control algorithms like MPC and RL are used, then control performance improves, but reliability decreases due to sensitivity to uncertainties and unmodeled dynamics
Solution Approach 1:
The system changes the operational parameters of control algorithms based on process conditions. For example, it adjusts the horizon parameters, weighting matrices, and exploration-exploitation balances dynamically. This allows advanced algorithms to maintain high performance while adapting to uncertainties by modifying their behavior rather than relying on fixed parameters.
Solution Approach 2:
The orchestration system acts as an intermediary layer between the process and control algorithms. It handles uncertainties and unmodeled dynamics by selecting appropriate algorithms and adjusting their parameters, shielding the advanced algorithms from direct exposure to uncertainties while maintaining their performance benefits.
4Reliability
If rule-based control systems are used to handle uncertainties, then reliability improves, but adaptability and control performance deteriorate
Solution Approach 1:
The control system is segmented into different functional layers: rule-based systems handle critical safety and reliability functions, while advanced algorithms (MPC, RL) handle performance optimization. This segmentation allows each component to excel at its designated function without compromising the other.
Solution Approach 2:
The system dynamically transitions between rule-based and advanced control modes based on process conditions. During normal operation, advanced algorithms provide high performance; during critical conditions, the system switches to rule-based control for reliability. This dynamic switching enables both adaptability and reliability.
Data Source
Figure 1~2
Figure 3~4
AI summary
The present disclosure relates to a method for controlling an industrial process (1). The method comprises monitoring, over time, at least one process variable of the industrial process. The method also comprises, at each of a plurality of consecutive time points, selecting a control algorithm, from a group of predefined control algorithms, for controlling the industrial process based on said monitored at least one process variable. The method also comprises using the latest selected control algorithm to control the industrial process. The selecting comprises selecting the control algorithm which optimizes a predefined function including value(s) of the monitored at least one process variable at said time point and a preset penalty for switching from a currently used control algorithm, of the group of control algorithms, to another one of said control algorithms.