Hybrid Plant State Prediction Using Physical and ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In chemical recycling plants, the variation in raw materials makes it difficult to prepare a unified physical model, leading to degraded prediction accuracy and reduced validity and safety of operation assistance, especially when dealing with materials containing impurities.
Innovation Solution
An information processing device that uses a hybrid model combining a physical model and machine learning models to predict the state of a chemical plant, incorporating optical spectrum data and plant data to provide accurate operation assistance by complementing the physical model's output with machine learning predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single physical model is used to match respective variations, then the model can be simplified and easier to operate, but the prediction accuracy is degraded
Solution Approach 1:
The patent divides the prediction system into multiple specialized machine learning models, each trained to handle specific variations in raw materials. Instead of using one general physical model, the system segments the prediction task into multiple models that can be selectively applied based on the type of variation encountered, thereby maintaining high accuracy for each specific case while keeping individual models relatively simple.
Solution Approach 2:
The system dynamically selects which machine learning model to use based on the characteristics of the input data and raw material variations. This dynamic selection allows the system to adapt to different conditions in real-time, maintaining high prediction accuracy without requiring a single complex model that covers all possible scenarios.
2Device complexity
If a single physical model is used to match respective variations, then the model structure can be simplified, but the validity and safety of operation assistance are decreased
Solution Approach 1:
The patent segments the prediction system into multiple specialized machine learning models, each optimized for specific raw material variations. This segmentation allows each model to be relatively simple in structure while collectively providing comprehensive and reliable predictions across all variations, thereby maintaining both low complexity and high reliability.
Solution Approach 2:
The system incorporates feedback mechanisms that monitor prediction results and select the appropriate model based on current conditions. This feedback loop ensures that the most suitable model is used for each specific case, enhancing the validity and safety of operation assistance while keeping individual models simple.
3Measurement precision
If multiple machine learning models are used to handle raw material variations, then the prediction accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent divides the prediction task into multiple specialized machine learning models, each handling specific types of raw material variations. This segmentation improves prediction accuracy for each specific case while keeping individual models relatively simple, thereby managing overall system complexity through modular design.
Solution Approach 2:
The system employs a universal framework that can selectively apply different machine learning models based on the input characteristics. This multi-functional approach allows the system to handle various raw material variations with high accuracy while maintaining a unified and manageable system architecture through centralized model selection and coordination.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing device (10) acquires a prediction result that indicates a state of a plant by using a physical model (15). The information processing device (10) acquires a plurality of prediction results that indicate a state of the plant, by using a plurality of machine learning models (1~N) that are generated by using data concerning the plant. The information processing device (10) outputs information concerning a state of the plant by using the prediction result and the plurality of prediction results.