The Application Research Of Data Mining To Shield Fault Diagnosis

In recent years, in order to adapt to the needs of urban development and to meet growing travel demand of urban residents. Urban subway continues to accelerate the pace of construction, Shield is one of the special equipment for excavation of subway tunnel, due to the Shield is extra-large, complicated equipment which integrates optical, mechanical, electrical, hydraulic, sensor, information technology, in daily mining often meet some aspects fault of hydraulic and electric control, these fauls not only cause severe economic losses, even cause serious casualties. Therefore, how to timely and effective analysis the causes of the shield fault, and make correct judgment to the future operation conditions, this will have great economic benefits and academic research value.The shield data acquisition system stores large number of data, how to extract useful knowledge from these data is the key to shield fault diagnosis. This thesis introduces data mining technology into the shield fault diagnosis. Using the rough set theory of data mining technology to remove the redundant information which doesn't work in fault diagnosis decision. Aiming at the faults of traditional attribute reduction method has more repeated attribute. This thesis presents a kind of reduction method that based on attribute importance selection. With the reduction data as neural network's lead link, using threshold discriminant conditions of neural network algorithm to diagnosis the reduction data, diagnosed the fault reasons and parts.Secondly, because the complex and uncertain behavior when the shield run-time present, single forecasting model is difficult to reflect the real changes of operating state. Therefore this thesis adopts the least-square combine with neural network to the further prediction. The least-square can reflect the trend towards of linear sequence, neural networks could catch the variation of nonlinear time series, complement each other's advantages. Therefore realize to forecast future health of the shield.Finally, through VS2005 programming realize rough set attribute reduction algorithm,and combine with matlab simulation analysis the feasibility of fault diagnosis,prediction algorithm.