Research On The Subsea Production System Equipment Fault Diagnosis Method Based On Bayesian Networks

With the continuous growth of global demand for oil and gas resources,the development of deep-sea oil exploration is receiving more and more attention.As the current mainstream of deep-sea oil and gas development mode,the safety and reliability of subsea production system is significant.So it is necessary to do the fault diagnosis research on the system.In this paper,the M-shape jumper,hydraulic power unit and subsea tree in subsea production system are used as the diagnostic objects.The research on fault diagnosis method based on Bayesian network is systematically carried out.The lack of equipment running data has always been an important factor that restricts the application of Bayesian network in fault diagnosis.The M-shape jumper is widely used in subsea production system as connecting component.The loading condition of jumper is bad so the jumper is easy to fail.However,due to environmental and economic factors,it is difficult to extract and save the operating data of jumper,making it difficult to carry out effective troubleshooting.In order to solve those problems,the simulation of jumper is done through the finite element analysis and obtain the data of jumper on both normal and fault conditions.Train the Bayesian network model by using parameter learning method.The proposed model solve the problem that build the effective Bayesian network for fault diagnosis in situation of lack of sample data.The number of the HPU components which has a strong relationship with each other is really huge.So it is difficult to do the accurate diagnosis for HPU by using traditional methods.Therefore the topology of HPU cause-effect Bayesian networks is established by fusing multisource information.The conditional probability of the diagnosis network is then calculated by using Noisy-OR/MAX.The results show that the network is perfect for the single fault mode.For the two-fault concurrent mode,the observation information nodes are added to the established single failure mode diagnosis network to assist the diagnosis of two simultaneous failures of the hydraulic power unit.The diagnosis results show that the proposed model can effectively diagnose a variety of faults occurring in the HPU.A fault diagnosis method of subsea tree system based on object-oriented Bayesian network is proposed.Introduce the concept of object-oriented in computer language to Bayesian network and establish multi-source information Bayesian sub-networks of the same or similar structure in the subsea production system according to the concept of class in object-oriented concept.Then repetitively call Bayesian sub-networks to form a complete Bayesian network of complex systems.The method proposed reduces the complexity and modeling difficulty of Bayesian networks.The sensitivity analysis and conflict analysis are performed to the proposed subnetwork and complete Bayesian network respectly and the results of the analysis as well as the diagnosis results of the subsea system validate the correctness of the proposed method.Design and manufacture physical model of M-shape jumper.Then the Strain test system is set up to collect the experimental data of M-type jumper under normal and fault conditions.Establish and train Bayesian fault diagnosis network with the experimental and simulation data.The fault diagnosis of the M jumper has achieved good results..