AI Control Model Evaluation Using Plant Simulation Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face difficulties in easily verifying the effect of AI control introduction to a system, such as a plant, due to the need for skilled engineers to manually generate and adjust plant simulator and AI control models, increasing cost and time.
Innovation Solution
An information providing apparatus and method that automatically generates and evaluates multiple simulation and AI control models using operational data, determining the best models based on evaluation values, and generates an AI control introduction effect report.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If skilled engineers manually generate and adjust plant simulator models and AI control models, then the reliability and accuracy of the AI control verification is improved, but the complexity of the process and the time required increase significantly
Solution Approach 1:
The system enables automated model generation and evaluation where the computer automatically generates multiple candidate models, evaluates them using evaluation functions, and selects optimal models without requiring skilled engineer intervention. This self-service approach maintains verification reliability while eliminating manual complexity.
Solution Approach 2:
The system automatically adjusts and optimizes model parameters through computational evaluation functions that assess multiple candidate models with different parameter configurations. This automated parameter optimization replaces manual expert adjustment, reducing process complexity while maintaining or improving verification accuracy.
2Reliability
If skilled engineers manually generate and adjust plant simulator models and AI control models, then the reliability of the AI control verification is improved, but the time required for the process increases
Solution Approach 1:
The system performs preliminary automated model generation and evaluation before actual AI control implementation. By pre-generating multiple candidate models and evaluating them in advance using automated processes, the system reduces the time required for verification while maintaining reliability through comprehensive automated assessment.
Solution Approach 2:
The automated evaluation process continuously assesses multiple candidate models using evaluation functions, maintaining continuous useful action throughout the model selection process. This continuous automated evaluation eliminates idle time between manual steps while ensuring thorough verification, reducing total verification time without sacrificing reliability.
3Ease of operation
If automated model generation and evaluation is implemented, then the ease of operation and cost are improved, but the manufacturing precision of the models may decrease
Solution Approach 1:
The automated process segments the model generation and evaluation into distinct automated stages: candidate model generation, evaluation function assessment, and optimal model selection. This segmentation enables ease of operation through automated workflows while maintaining precision through systematic evaluation of multiple candidate models with different parameter configurations.
Solution Approach 2:
The system implements automated feedback loops where evaluation functions assess candidate models and feed results back into the selection process. This feedback mechanism ensures that only models meeting precision criteria are selected, maintaining manufacturing precision while enabling ease of operation through automated iterative improvement and selection.
Data Source
AI summary
A server device generates N simulation models on the basis of operational data collected from a system, determines a simulation model in which a simulation model evaluation value for evaluating the simulation model is a maximum value on the basis of the operational data, generates M AI control models on the basis of the operational data and the determined simulation model, and determines the AI control model in which an AI control model evaluation value for evaluating the AI control model is a maximum value on the basis of the operational data.