Steelworks Operation Support Using Learning Models for Total Cost
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for supporting operations in steelworks face challenges in accurately tuning calculation models due to deviations between theoretical and actual performance data, leading to suboptimal solutions.
Innovation Solution
A method utilizing a learning model trained with performance data to estimate total costs in steelworks operations, incorporating blast furnace, steelmaking, and energy division data, and outputting optimized conditions to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional calculation models based on theoretical values are used for optimization, then the optimization processing can be performed, but the accuracy of the optimal solution deteriorates due to deviations from actual performance data
Solution Approach 1:
The patent transforms the static theoretical calculation model into a dynamic learning model that continuously adapts its parameters based on actual performance data. The learning model updates its internal parameters through machine learning algorithms, allowing it to reflect real-world deviations and provide more accurate optimization results that align with actual operational conditions.
2Measurement precision
If calculation models are tuned using recent performance data, then the accuracy improves, but the difficulty of tuning increases during genetic algorithm processing
Solution Approach 1:
The patent replaces the manual or complex iterative tuning process with an automated learning model that performs parameter adjustment automatically. The learning model uses machine learning algorithms to absorb performance data and adjust model parameters without requiring complex manual tuning interventions, thereby reducing tuning complexity while maintaining or improving accuracy.
3Productivity
If theoretical values are used for optimization search, then the processing can be completed, but the suitability for actual operation deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the learning model continuously receives actual performance data from operations and uses this feedback to adjust its predictions and recommendations. This closed-loop feedback system ensures that the optimization results remain suitable for actual operations by constantly aligning the model's understanding with real-world outcomes, while maintaining processing efficiency through automated learning.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of supporting operation is to be executed by an information processing apparatus and to be used in a steelworks including a blast furnace division, a steelmaking division, and an energy division. The method includes inputting data on blast furnace operation specifications and a pig iron ratio that have an effect on the steelmaking division and the energy division into a learning model, estimating a total cost using the learning model, based on the inputted data on the blast furnace operation specifications and the pig iron ratio, and outputting the estimated total cost. The learning model is a model trained using performance data including performance values of blast furnace operation specifications and total cost.