刚性罐道是立井采矿中煤炭运输以及设备运输环节的重要组成机械,对罐道做出精确的故障诊断具有保证和提高立井采矿生产效率的重要意义.面对罐道难以实现精确故障诊断的问题,在Improved Convolution Neural Networks with Training Interference(ITICNN)的基础上添加了 Inception_v4 与 Long Short Term Memory(LSTM),提出 了一种高精度卷积神经网络 Improved Convolution Neural Network Based on ITICNN(IBICNN),并采用了一种方向振动传感器收集罐道振动信号,用IBICNN的卷积层提取振动信号数字特征,并用LSTM提取多方向振动信号之间的相关性信息的刚性罐道故障诊断方法.通过搭建实验平台对罐道实验模型进行了故障诊断研究,取得了 99.4％的诊断率.为了解决罐道在使用过程中伴有大量噪声从而难以诊断故障的问题,在Adaptive Batch Normalization(AdaBN)算法的基础上进行了改进,并采取自迁移学习的方法,提高了 IBICNN的抗噪声能力,在噪声含量为100％的情况下取得了 90.11％的诊断率.
Fault Diagnosis of Cage Way Based on Noise Data Driven CNN and LSTM
Rigid cage way is an important part of coal transportation and equipment transportation in shaft mining.Accurate fault diagnosis of cage way is of great significance to ensure and improve the production efficiency of shaft mining.To solve the problem of difficulty in achieving accurate fault diagnosis for cage way,Inception_v4 and Long Short Term Memory(LSTM)are added on the basis of ITICNN(Improved Convolution Neural Networks with Training Interference)and a high-precision convolutional neural network-Improved Convolution Neural Network Based on ITICNN(IBICNN)is proposed,and a rigid cage way fault diagnosis method is adopted that uses a directional vibration sensor to collect cage way vibration signals,uses the convolution layer of IBICNN to extract the digital features of vibration signals,and uses LSTM to extract the correlation information between multi-directional vibration signals.The fault diagnosis research is conducted on the cage way experimental model by building an experimental platform and a diagnosis rate of 99.4％is achieved.In order to solve the problem that it is difficult to diagnose faults due to a large amount of noise in the use of the cage way,the Adaptive Batch Normalization(AdaBN)algorithm is improved and the method of self-transfer learning is adopted to improve the anti-noise ability of IBINN.The diagnosis rate is 90.11％when the noise content is 100％.
国家科技期刊平台
NSTL
万方数据
