在对铸坯质量缺陷类型及其主要影响因素分析总结的基础上，确定以特殊钢大方坯常见的中间裂纹、中心裂纹和中心偏析为研究对象，利用BP神经网络建立了该3种典型缺陷的预测模型。基于冶金理论和连铸生产大量历史数据的统计分析，提炼出影响以上3种内部缺陷的20个主要工艺参数，进而提出20153的预测模型网络拓扑结构。采用生产现场数据制做了预测模型的训练样本集和测试样本集。利用训练样本集将该神经网络训练至设定预报误差以内，再用测试样本集对所构建的网络进行了测试。基于训练成熟的神经网络模型，进一步编制在线预报系统，实现铸坯质量在线实时预报。
Based on the summaries and analyses of various types of strand defects and their causes, the halfway crack, central crack and central segregation of special steel bloom were adopted as the samples for the establishment of a quality prediction model using BP neural network. Based on the metallurgical theory related to the defects introduction and the analyses to a great deal of statistical quality data of the bloom castings, 20 defectcausing parameters with the continuous casting process were selected as the input of the model, a 20153 network topology structure for the artificial neural network prediction model was developed with back propagation algorithm. The training sample sets and test sample sets for the ANN model were prepared from the real production of the continuous casting production. The BP network was trained by training sample sets and tested by test sample sets in sequence. Based on the well trained BP network, the online quality prediction system was developed, and the realtime quality of bloom was predicted on line.
参考文献
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%