基于新型模糊神经网络连铸漏钢诊断预报的研究

Study on Breakout Diagnosis Prediction Based on a Novel Fuzzy Neural Network in Continuous Casting Process

基于某钢厂连铸现场采集的历史数据,通过对AFLC模糊神经网络学习算法和网络结构的改进,建立了融合模糊模式识别和模糊聚类的新型竞争型模糊神经网络,并将模型应用于连铸漏钢预报的过程中。结果表明,模型能够有效地识别连铸粘结漏钢过程中两种典型的温度模式和预报拉漏事故的发生。在警戒参数为0.88的条件下,该模型对两种典型温度模式的预报率分别达到95.6%和97.8%,报出率都达到100%。

Based on the history data acquired in a steel plant and the modification of the learning algorithm and structure of AFLC fuzzy neural network, a novel competed fuzzy neural network model with fuzzy pattern recognition and fuzzy clustering was established, and was applied to the breakout prediction of continuous casting process. The results s.show that the model can effectively indentify two typical temperature patterns of sticking breakout and predict possible leakages of liquid steel. When the vigilance Parameter is 0. 88, the prediction rates of the model for these two typical temperature patterns can reach 95.6 % and 97.8 % respectively, and both of the quote rate can reach 100 %.