Acoustic Emission Health Monitoring For Cracked Tubular Joints Of Offshore Platform

As the most important infrastructure in the process of marine development, offshore platform needs huge investment costs and operational costs. Once a serious accident occurs, it will cause significant economic loss and adverse social influence. Thus, regular structure inspection and on-line monitoring have become the last and the most effective process to ensure normal production and safety of operators.Steel jacket platform is the most widely used type, and as its most pivotal constituent parts, tubular joints'structural failure is mainly caused by deformation and crack propagation. This paper is aimed to explore the crack development regulation, to achieve the purpose of health monitoring for cracked tubular joint and intelligent identification of crack states by acoustic emission monitoring for crack propagation.As a new nondestructive testing method, acoustic emission (AE) can realize the dynamic defects monitoring and damage condition prediction. By on-line monitoring, cracks can be monitored on time, which can prevent major accidents.On the basis of the theory study, the AE signals of various crack states are collected on the real time. By fatigue crack experiment, the fitted equations between accumulated energy and ring total are established in accordance with various crack states. It applies the feature extraction method of AE signal based on wavelet energy coefficient to health monitoring for cracked tubular joints of platform, and Back-Propagation (BP) neural network is used to the crack state intelligent identification.Firstly, crack growth process is monitored by AE technique. The trends of characteristics such as ring counts, root mean square (RMS), energy, etc. are extracted. It presents that AE signals generated by crack growth present certain periodic and intermittent characteristics, and the signal features vary with crack state, which has an important guiding significance for state recognition. Simulating the process of fatigue crack growth, removing AE signals generated by crack closure and unloading, the features of crack propagation are extracted, and the fitted equations between accumulated energy and ring total, which are suitable for the processes of crack initiation, crack propagation and fracture, providing theoretic foundation for offshore platform health monitoring and damage prediction.Secondly, characteristics of general wavelets and the determining method of maximum decomposition level of wavelet transformation are discussed. In terms of the rules, five-scale decomposition is suitable for AE signals collected, and the wavelet energy coefficients are extracted, indicating the features of signals. The result shows that the elastic wave generated by crack growth is composed of many high frequency signals, and changes along with different crack states. And the WEC is in accordance to crack states. Therefore, wavelet energy coefficients can be used to identify different crack states effectively.In the last place, the three-layer BP neural network is built up to recognize the crack states, and it takes the wavelet energy coefficients and five states of crack growth as its inputs and outputs. And then the optimum network architecture is determined by test. The sample signals are used to network training and simulating, realizing the intelligent identification to crack states effectively and the correct identifying ratio is 88%, achieving expected purpose.The research has great importance and applied value to promote the health monitoring on line for cracked tubular joints, to improve the real-time property, economy and safety of monitoring process, and to achieve the intelligent identification.