基于神经网络和证据理论集成的钻井过程状态监测与故障诊断

Drilling state monitoring and fault diagnosis based on integrating neural network and evidence theory

钻井过程状态监测与故障诊断是钻井系统安全运行过程中的重要保障。基于信息融合原理,先建立钻井过程参数子空间和子神经网络进行初级融合,形成对钻井故障辨识框架中各故障模式的证据支持,再利用D-S证据理论将子网络输出所形成的证据进行融合,得到各故障模式的置信区间,很好地实现了钻井状态识别。试验结果表明,基于神经网络和证据理论集成的融合算法降低了神经网络的复杂性,提高了神经网络诊断过程的效率,集成融合算法可以很好地提高钻井参数融合的准确性。

State monitoring and fault diagnosis of drilling process is the significant support for safe working of drilling system. Based on information fusion theory, parameter subspace about drilling process parameters and neural subnet for primary fusion was firstly established： So the evidence supporting for different fault mode in drilling fault frames of discernment can be obtained. Then by using D-S evidence theory, the confidence interval of fault diagnosing results was improved by fusing the evidence body that was outputted by neural subnet, the status of drilling process was identified very well. The experimental results show that the complexity of neural network can be decreased and the efficiency of neural network can be improved by the fusion arithmetic of integrating neural network and evidence theory, the veracity about drilling parameters can be improved by fusing integrating fusion arithmetic.