Model Order Determination And Noise Removal For Modal Parameter Estimation Of Offshore Platform Structures

As the oil drilling is gradually developing to deep-sea, a number of offshore platforms have been used. However, the damage in the platform structures can not be avoided during their service. To protect the safety of offshore platform structures and to prevent the occur-rence of major accidents, the periodic/aperiodic detection and safety assessment is needed for offshore platform structures during the service. Especially for aging platforms, the im-plementation of effective health monitoring has a more important significance. The tech-nology of structural health monitoring based on the response information of the vibration test is a global detection method, and has its own advantages compared with conventional non-destructive testing method. Modal parameter identification of which is a fundamental and critical component. Therefore, the accuracy of the modal parameter identification for large-scale civil engineering structures such as offshore platform is particularly important.Model order determination and noise removal for measured data are very important for modal parameter identification. Modal parameter identification methods require prior information on model order. Model order determination has an impact on the computa-tional efficiency and the accuracy of real modal parameter identification. Meanwhile, due to the harsh environmental conditions of offshore platforms in service, measured response data inevitably contain noise. The modal parameters extracted from the measured data are not accurate, which is bound to affect the security situation in offshore platform structures evaluation results. Therefore, noise elimination from measured response signals becomes extremely important.This focuses on modal parameter accurate identification technology based on output response of vibration testing of the offshore platform structure, which includes the associ-ated model order determination method, the noise removal method, and improvement of the existing methods of modal parameter identification. The main contents and contributions are as follows:1.An appropriate estimate of the order of the model has become the single most impor-tant problem in modal parameter estimation. Several traditional methods for model order determination are summarized and pointed out the flaws in the application.A new model order determination method——rank estimation of Hankel matrix is pro-posed. The impulse response function, rank of Hankel matrix, singular value decom-position, as well as the relationship between rank and model order are theoretically explained. The new method makes up the defects of similar methods which are lack of theoretical support. A numerical model and an experimental model are used to demonstrate the performance of the new method;2.The theory of signal de-noising is explained. For the first time the structured low rank approximation (SLRA) technique is introduced into the field of modal parameter identification. Based on the results of model order determination, Cadzow's method is used to eliminate the noise, which is suitable for engineering applications.And the choice of the matrix dimensions, calculation efficiency and evaluation of de-noising effect, etc. are also discussed. A numerical model and an experimental model are applied to verify the effectiveness of the method;3.To improve the accuracy of modal parameter identification of offshore platforms, a modal parameter identification approach based on model order determination and noise removal is proposed. Firstly the model order is accurately determined prior to the parameter estimation; then the noise is eliminated by SLRA method; and finally complex exponential method (Prony) is applied for modal parameter identification. This approach overcomes the weakness of complex exponential method for noise-sensitive. An experimental study of offshore jacket platform model is performed to validate the feasibility of the method for the practical offshore platform structures;4. Traditional MIMO identification methods make use of stability diagram to distinguish the true and computation modes. The stability diagram method is implemented by repeating the same analysis many times with a varying model order until a stable result emerging. However, it's still difficult to distinguish the real modes among a large number of computation modes, especially when the SNR is low. A new MIMO modal parameter identification method based on model order determination and noise removal is proposed. The method is implemented by using Prony method to identify modal parameter based on a unified model order and noise removal.We can iden-tify the true modal parameters by marking each of the de-noising signal estimation results in the signal—frequency stability diagram. The traditional MIMO method and the new method are used for the one dimension cantilever beam model and two dimensions plate model experimental data, respectively. The results show that the new MIMO modal parameter identification method reduces the influence of human factors and improves the identification efficiency and accuracy, in particular, improves the es-timation accuracy of damping ratios. This article extends the traditional Prony method from SISO to MIMO and provides a new idea of accurately identifying the modal pa-rameter, which has good prospects for further practical engineering application.