基于LMD-SNN的随钻异常振动识别

Identification of Abnormal Vibration while Drilling Based on LMD-SNN

针对旋转钻井过程中对钻头异常振动(横向振动、纵向振动、扭转振动)信号识别准确率低,效率差,导致钻具损坏甚至井眼报废的问题,提出一种基于局部均值分解(local mean decomposition, LMD)和脉冲神经网络的异常振动识别方法。首先,采用局部均值分解将微机电系统(Micro Electro Mechanical System, MEMS)加速度计信号分解为具有振动特征的PF分量;然后,从分解的乘积函数(Product function, PF)分量中提取随钻异常振动特征,并将提取到的加速度计振动特征编码为脉冲;其次,使用改进的学习规则训练脉冲神经网络对随钻异常振动进行识别。设计模拟实钻实验和仿真,训练后的脉冲神经网络可以对随钻振动的识别度达到99.39%,显示了该方法在实际应用中的巨大潜力。

In order to solve the problem of low accuracy and low efficiency of abnormal vibration( transverse vibration,longitudinal vibration and torsional vibration) signal recognition in rotary drilling,which leads to drilling tool damage and even hole scrap,a method of abnormal vibration recognition based on local mean decomposition( LMD)and spiking neural networks is proposed. Firstly,the MEMS accelerometer signal is decomposed into pf components with vibration characteristics by local mean decomposition;Then,the abnormal vibration feature while drilling is extracted from the decomposed product function( PF) component,and the extracted accelerometer vibration feature is coded as pulse;Finally,the improved learning rules are used to train the spiking neural networks to recognize the abnormal vibration while drilling. The results show that the recognition degree of the trained PNN can reach99.39%,which shows the great potential of this method in practical application.