Scholarlink

<p>The sintering conditions have a significant impact on the quality of the sintered ore. In order to enhance the control efficiency, this paper conducts a comprehensive evaluation, prediction, and optimization study of the sintering state. Collect and process data from the entire sintering process. Based on metallurgical theory and production experience, the burn through point position, burn through point temperature, and bed thickness are selected as the characterization indicators for the sintering state. Transform the data direction of the three indicators. Using the entropy weight method, principal component analysis, and expert judgment to derive weights for calculating the comprehensive evaluation score of the sintering state. The number of clusters is determined to be three using the elbow method. Utilizing the K-means algorithm, the evaluation scores are classified into three levels. Feature selection is conducted using the Boruta algorithm. Using recurrent neural networks, long short-term memory, and multilayer perceptron as foundational learners to build a meta-learning model for predicting the comprehensive sintering grade. The predictive model achieves an accuracy of 91.21&#xa0;pct, precision of 90.10&#xa0;pct, recall of 92.24&#xa0;pct, and an F1 score of 0.91. When the comprehensive grade is low, a set of operational parameter adjustment plans is generated. Select the plan with the smallest Euclidean distance from the original operational parameters as the final optimization solution, thus achieving an overall improvement in the sintering state.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comprehensive Evaluation, Prediction, and Optimization of Sintering State

  • 期刊论文

  • 发表日期：

    2025年2月19日

查看全文
  • Ming-yu Wang,
  • Man-sheng Chu,
  • Jue Tang,
  • Quan Shi,
  • Zhen Zhang,
  • Zhi-feng Zhang

摘要

The sintering conditions have a significant impact on the quality of the sintered ore. In order to enhance the control efficiency, this paper conducts a comprehensive evaluation, prediction, and optimization study of the sintering state. Collect and process data from the entire sintering process. Based on metallurgical theory and production experience, the burn through point position, burn through point temperature, and bed thickness are selected as the characterization indicators for the sintering state. Transform the data direction of the three indicators. Using the entropy weight method, principal component analysis, and expert judgment to derive weights for calculating the comprehensive evaluation score of the sintering state. The number of clusters is determined to be three using the elbow method. Utilizing the K-means algorithm, the evaluation scores are classified into three levels. Feature selection is conducted using the Boruta algorithm. Using recurrent neural networks, long short-term memory, and multilayer perceptron as foundational learners to build a meta-learning model for predicting the comprehensive sintering grade. The predictive model achieves an accuracy of 91.21 pct, precision of 90.10 pct, recall of 92.24 pct, and an F1 score of 0.91. When the comprehensive grade is low, a set of operational parameter adjustment plans is generated. Select the plan with the smallest Euclidean distance from the original operational parameters as the final optimization solution, thus achieving an overall improvement in the sintering state.