基于DBN和EEMD方法的精轧测温传感器故障诊断与分类

Fault Diagnosis and Classification of Temperature Sensor for Finishing Rolling Based on DBN and EEMD

为了减少轧制过程温度测量不准引发的钢产品质量问题,该文采用DBN网络和EEMD模态分析相结合的方法,建立精轧后测温传感器故障诊断与分类模型。首先建立DBN的温度预报模型,对精轧后温度进行预测;其次对预报温度与实际温度的差值进行EEMD特征提取;最后,构建DBN故障诊断与分类模型,用方差、方差百分比、能量、能量百分比方法分别或者联合的方法构建特征向量集,并将这些向量集作为故障模型的输入,对正常情况、漂移、精度下降、冲击、固定偏差等故障进行诊断与分类。以六架精轧机组的带钢生产为例进行仿真实验,结果表明,建立的模型具有很好的故障诊断与分类能力。

In order to reduce the quality problems of steel products caused by inaccurate temperature measurement in rolling process,the fault diagnosis and classification model of temperature sensor after finishing rolling was established by using deep belief networks(DBN)method and ensemble empirical mode decomposition(EEMD)method.The temperature prediction model of DBN and fault classification model of DBN were constructed respectively.Firstly,the temperature prediction model of DBN was established to predict the temperature after finishing rolling.Secondly,the feature extraction was performed for the difference between the predicted temperature and the actual temperature by using EEMD.Finally,the DBN fault diagnosis and classification model was constructed,and the feature vector sets were constructed by variance,variance percentage,energy and energy percentage methods respectively or jointly,these vector sets were used as the input of the fault model to diagnose and classify the faults such as normal condition,drift,precision degradation,impact and fixed deviation.Taking the strip production of six finishing mills as an example,the simulation results show the model is effective for fault diagnosis and classification.