Hybrid Plant State Prediction for Variable Chemical Recycling Inputs
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 safety concerns in operation assistance, especially when dealing with materials containing impurities.
Innovation Solution
An information processing device that combines a physical model with a machine learning model to provide a hybrid prediction system, using optical spectrum data and plant data to generate complementary prediction results, enhancing operation assistance and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single physical model is used to predict plant state across different raw material variations, then device complexity is reduced, but prediction accuracy and reliability are degraded
Solution Approach 1:
The patent combines a physical model and a machine learning model into a hybrid prediction system. The physical model provides theoretical foundation while the machine learning model captures complex patterns from historical data, achieving both low complexity and high accuracy through model integration
Solution Approach 2:
The prediction system uses a composite modeling approach where multiple models (physical model and machine learning model) are combined to create a hybrid system that leverages the strengths of each individual model, similar to how composite materials combine different materials to achieve superior properties
2Loss of information
If a physical model is used for operation assistance in chemical recycling plants, then interpretability is improved, but adaptability to raw material variations is reduced
Solution Approach 1:
The hybrid model merges the interpretability advantage of physical models with the adaptability advantage of machine learning models, allowing the system to maintain both explainability and flexibility when handling diverse raw materials in chemical recycling plants
Solution Approach 2:
The system applies different modeling approaches to different aspects of the prediction problem: physical models are used where interpretability is critical, while machine learning models are used where adaptability to variations is more important, optimizing the overall system performance
3Measurement precision
If a machine learning model is used to predict plant state with high accuracy, then prediction accuracy is improved, but interpretability and causal understanding are reduced
Solution Approach 1:
The patent integrates machine learning models that provide high prediction accuracy with physical models that provide interpretability, ensuring that both accuracy and causal understanding are maintained through the hybrid architecture
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing device (10) acquires a first prediction result that indicates information concerning a state of a plant that is predicted by a physical model (15). The information processing device (10) acquires a second prediction result that indicates information concerning a state of the plant that is predicted by a machine learning model that is generated by using data concerning the plant. The information processing device (10) displays the first prediction result and the second prediction result.