Study Of Fault Diagnosis Of Offshore Platforms Based On Soft Computing

The health monitoring and fault diagnosis of offshore platform structure plays animportant role in economic development and social progress. Once the structure happens fault,it will have unintended consequences if we can’t find and do some remedial measures. Theload capacity of structure will decrease when the fault is not serious. But when the defect isvery serious it will affect work and production or even collapse. Not only cause hugeeconomic damage, but also bring adverse environmental and social impact or even losepeople’s life. Therefore, the fault diagnosis of offshore platform has been a hot research topicin the field at home and abroad. Meanwhile, the offshore platform has complex structuralfeatures and poor working conditions, and the structural main force is in the underwater. Afterthe structural fault occurs, we can not find easily, which made the traditional fault diagnosis ofstructure appeared to be inadequate. Therefore this feature makes the fault diagnosis ofoffshore platform become a difficult problem. With the development of computer technology,the soft computing theory has also been unprecedented progress. Soft computing hasincomparable advantage compared with the traditional method of calculating, so softcomputing has been studied and used extensively in the field of structural fault diagnosis.Therefore, the soft computing is used to study the fault diagnosis of underwater structures ofoffshore platform in this paper.We combined with vibration modal analysis theory, neural network and genetic algorithmto study the fault diagnosis of offshore platform. For the offshore platform structure, it will bevery difficult to determine the location and extent of fault one time because of the lager-scalenetwork systems. And this way will take a long time and seriously affect the diagnosis results.So in this paper the author according to the characteristics of offshore platform used a newway to do fault diagnosis of offshore platform step by step. This new method can determinethe damage location quickly and it is a very important innovation in this paper. Firstly thecomplex offshore platform underwater structure is divided into several simple substructures.The changing rates of mode curvatures caused by structural damage are put into PNN todetermine the damaged substructure. After determining the damaged sub-structure, we shoulddetermine accurate damaged bar of structure. If we only use BP neural network to diagnose there will be some defects such as a lager-scale input samples, slow convergence and BPneural network is easy to fall into local minimum. So in this paper the genetic algorithmoptimized BP neural network weights to make up the shortfall. In the damaged substructurethe comprehensive changing rates of mode shape are put into GA-BP neural network todetermine the accurate damaged bar. Lastly the changing rates of natural frequency of squareare put into GA-BP neural network to determine the extent of structural damage.The simulation model of offshore platform is built by the finite element. The parametersof networks are calculated by ANSYS finite element analysis software. The structuraldamages are simulated by reducing the elastic modulus.In the paper an offshore platform for the numerical example was simulated using theabove new method. There were enough training samples and test samples in the threediagnostic network systems. And under different damage conditions we do some simulationanalysis. The numerical examples showed that: the study of fault diagnosis of offshoreplatform based on soft computing is feasible and effective. At last we determine the locationand extent of fault of offshore platform using different network systems and differentstructural parameters. The individual damaged component of structures was diagnosed usingthis method. And the result showed that this method is good and it can determine the locationof fault, avoid measuring a large number of modal data, save diagnosis time. It also had alower rate of misdiagnosis and a strong fault-tolerance and anti-jamming.