AI Control Setpoint Validation Using Historical Process Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing process control systems face challenges in optimizing control setpoints due to the complexity of relationships between setpoints, material inflows, and throughput and quality of output in manufacturing and industrial processes.
Innovation Solution
A computer-implemented method that generates an artificial intelligence recommended control setpoint for a process, compares it to historical control setpoints, and updates the control system setpoint for optimal control, using evidence-based processing to refine or verify the AI recommendation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data-driven predictive AI models are used to generate control setpoint recommendations, then adaptability to changing process operating conditions is improved, but the risk of generating unstable or suboptimal setpoint recommendations increases
Solution Approach 1:
The system implements feedback by comparing AI-generated setpoint recommendations against historical process data and performance outcomes. The historical data store contains past process states, control setpoints, and resulting outcomes. The system retrieves relevant historical instances and uses them to validate whether AI recommendations align with proven successful patterns, creating a feedback loop that enhances reliability while maintaining adaptability.
Solution Approach 2:
The system performs preliminary action by pre-storing and organizing historical process data, state instances, and control outcomes before they are needed for validation. The historical data is structured and indexed in advance, allowing rapid retrieval and comparison when AI generates setpoint recommendations. This preliminary preparation enables quick verification without delaying the control decision process.
2Productivity
If AI models perform local search optimization to maximize objective functions, then process optimization capability is improved, but computational complexity and processing time increase
Solution Approach 1:
The system performs preliminary action by pre-computing and storing historical process data, state instances, and their associated outcomes in an organized historical data store. This advance preparation allows the system to quickly retrieve relevant historical patterns during optimization without performing computationally intensive analysis in real-time, thus reducing processing time while maintaining optimization capability.
Solution Approach 2:
The system uses copying by creating simplified representations of historical process states and outcomes that can be quickly compared against current AI recommendations. Instead of re-analyzing raw historical data, the system works with pre-processed copies of historical instances that capture essential patterns, enabling fast validation and comparison operations.
3Reliability
If the system validates AI recommendations against historical process data, then reliability of control setpoints is improved, but system complexity and data processing requirements increase
Solution Approach 1:
The system applies segmentation by dividing the validation process into distinct modular components: (1) AI setpoint generation module, (2) historical data retrieval module, (3) comparison and validation module, and (4) recommendation acceptance/rejection module. Each component performs a specific function independently, making the overall complex system manageable and maintainable while achieving reliable validation.
Solution Approach 2:
The system uses an intermediary approach by introducing a historical data store and comparison mechanism that mediates between AI-generated recommendations and final control decisions. This intermediary layer provides objective validation based on historical evidence, reducing the complexity of direct AI decision-making while improving reliability through evidence-based verification.
Data Source
AI summary
Setpoint control processing is provided which includes generating, for a current state of a process, an artificial intelligence recommended control setpoint for the process, and obtaining, by one or more processors, historical process state instances, and associated control setpoints, related to the current state of the process. The artificial intelligence recommended control setpoint for the process is compared, by the one or more processors, to the associated control setpoints of the historical process state instances related to current state of the process. Based on a result of the comparing, a control system setpoint for the process is updated for optimal control of the process for a current process optimization objective, and the process is operated using the updated control system setpoint to optimally control the process for the current process optimization objective.