Evaluation Model Selection by Raw Material Property Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In industrial facilities like refineries, generating a new evaluation model for each change in crude oil production region or time is resource-intensive and time-consuming, increasing processing load and causing operational inefficiencies.
Innovation Solution
A model selection apparatus that stores multiple evaluation models associated with raw materials and selects the appropriate model based on property data, using clustering and threshold adjustments to determine similarity in raw material properties, thereby reducing the need for frequent new model generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a new evaluation model is generated for each change in crude oil production region or time, then the evaluation accuracy is improved, but the processing load increases and time loss occurs
Solution Approach 1:
The system performs preliminary clustering of crude oil data by production region and time, pre-identifying groups of crude oils with similar properties. When a new evaluation is needed, the system retrieves the pre-clustered model corresponding to the crude oil's group rather than generating a new model from scratch, significantly reducing processing time while maintaining evaluation accuracy.
Solution Approach 2:
The system creates universal evaluation models that can serve multiple crude oil types within the same cluster. Instead of generating dedicated models for each individual crude oil sample, a single model trained on clustered data can evaluate all crude oils in that cluster, reducing the total number of models needed and the time required to generate them.
2Measurement precision
If a new evaluation model is generated for each change in crude oil production region or time, then the evaluation accuracy is improved, but the processing load increases
Solution Approach 1:
The system performs preliminary clustering of crude oil data by production region and time, pre-identifying groups of crude oils with similar properties. When a new evaluation is needed, the system retrieves the pre-clustered model corresponding to the crude oil's group rather than generating a new model from scratch, significantly reducing processing time while maintaining evaluation accuracy.
Solution Approach 2:
The system creates copyable evaluation models that can be reused across different crude oil evaluations. Once a model is trained for a specific cluster of crude oils, it can be copied and applied to any crude oil within that cluster, eliminating the need to regenerate the model and reducing computational processing load.
3Productivity
If clustering is used to select evaluation models based on raw material properties, then the need for new model generation is reduced, but the device complexity increases
Solution Approach 1:
The system segments the continuous space of crude oil properties into discrete clusters based on production region and time characteristics. Each cluster corresponds to a specific evaluation model, transforming a complex continuous selection problem into a simpler discrete classification task. This segmentation reduces the complexity of model selection while maintaining efficiency.
Data Source
AI summary
There is provided a model selection apparatus including: an evaluation model storage unit configured to store each of a plurality of evaluation models capable of outputting an index for evaluating a state of a facility that is configured to manufacture a product from a raw material, in association with the raw material; a property data acquisition unit configured to acquire property data indicating a property of the raw material which is used in the facility; a model selection unit configured to select a target model for evaluating the state of the facility based on the property data, from among the plurality of evaluation models, when a target raw material in the facility is used; and a target model output unit configured to output the target model.