Real-Time Drilling Process State Monitoring And Diagnosis

Oil drilling are complicated, complex and changing work environment, poor working conditions, the drilling process real-time state monitoring and diagnosis is an important research direction. Drilling process real-time information acquisition, processing and transmission technology is key technologies for the drilling process real-time state monitoring and diagnosis. This paper focuses on information acquisition and information processing research work about automated drilling.Automated drilling is an important development direction of drilling technology in the future. In conclusion, the development of automated drilling technology on the basis of the status quo,Proposed the automated drilling information model, that the automated drilling system should include both ground and underground information processing systems, ground information processing system is automated drilling core, the downhole control system that include underground information processing systems, downhole measurement while drilling equipment and Combination of downhole tools will play an increasingly important role. Real-time state monitoring during drilling and abnormal state diagnosis is the core of two information processing systems, the complexity of drilling system and its working environment determines that we can not just rely on a single sensor information, we have more than one parameter to describe drilling process state. using surface and underground drilling process information from various sensors redundancy and complementarity, we used multi-sensor information fusion theory for the drilling process state monitoring and diagnosis, Based on multi-sensor information fusion method to improve the drilling process state monitoring and diagnosis robustness is feasible.Based on the complex characteristics of drilling system, Summarized multi-sensor information fusion, use multi-sensor information fusion method of neural network and evidence theory that does not rely on accurate mathematical model. Using multi-sensor information fusion to describe the system state and system behavior is better than a single sensor information, But when there is conflict or multi-sensor information fusion model imperfect, the information fusion results may be abnormal, in the drilling process may come to the wrong state estimation, this paper put forward such a case the basic framework for dealing with conflicting evidence, by amending the evidence combination rule, the allocation of the evidence of conflict can improve rationality of the evidence fusion results.Drilling ground systems information acquisition technology is not perfect, established a new type of wireless data acquisition monitoring system, drafted wireless data communication protocols, developed drilling ground data acquisition system. In accordance with the drilling expert knowledge, summed up the drilling abnormal state diagnosis criterion, established the feature extraction model for different parameters. Combined with drilling expert knowledge, established the drilling state-space, including drill piercie, well kick, lost circulation etc. nine abnormal state and normal state that is 10 kinds of drilling state. Using neural network to establish the drilling process model of multi-sensor information fusion to achieve drilling characteristics of state parameters to the drilling state space mapping, established the neural network training samples and sample of teachers, extracted the drilling process data under abnormal conditions, established abnormal samples, trained neural network to identify abnormal state, the diagnostic results show that neural network has good diagnostic capabilities.D-S evidence theory is an effective method for the fusion of uncertainty information, the multi-sensor information of the drilling system is uncertainty, try using D-S evidence theory to multi-sensor information fusion. In D-S evidence theory using basic belief assignment with the state space for the state identification, access to basic belief assignment is a prerequisite for evidence fusion, proposed based on neural networks combined with the D-S evidence theory method for drilling abnormal state diagnosis, that is a multi-sensor multi-level integration diagnostic model, the neural network recognition output is basic belief assignment for drilling state space, that is, the neural network output by the normalization is the input of D-S evidence theory fusion, along the timeline to produce decision-making output. The results show that the integration of multi-level state monitoring and diagnosis model has better diagnostic capabilities. Multi-level diagnosis model not only realize the current state of multi-sensor information fusion based on neural network, also achieved to fuse the historical status and the current state based on DS evidence, completed multi-dimensional and multi-level integrated fusion process.When the evidence is conflict, the traditional D-S combination model have defects, A conflict digestion model for evidence fusion is proposed to monitor the drilling process, implement not to affect the fusion result when evidence conflict, the fusion results is reasonable and correct, and improving robustness of integration multi-level fusion model.