摘要
针对板带轧机液压AGC系统在线故障诊断问题,建立了一种基于非线性自回归滑动平均模型(NARMA的递归神经网络,通过AIC定阶法确定模型阶次。运用生产实际数据,通过动态学习算法完成对网络的训练,使网络映射系统的动力学特性。该网络模型避免了故障的自学习,能够很好地实现故障检测。试验研究证明了该神经网络方法进行轧机液压AGC系统在线故障诊断的可行性和有效性。
For on-line fault diagnosis of hydraulic AGC system on strip rolling mill, a recursive neural network model based on NARMA was established. The model order is determined by AIC method. By training with dynamic learning algorithm and actual production data, neural networks can map system dynamic characteristics. This network model can avoid fault self-learning and has better diagnosis capability. Feasibility and efficiency of this method was verified by experiment.
出处
《钢铁》
CAS
CSCD
北大核心
2005年第5期45-48,共4页
Iron and Steel
基金
国家自然科学基金项目(50375135)
河北省自然科学基金资助项目(E2005000323)
