Please use this identifier to cite or link to this item:
|Title:||Noise modelling, vibro-acoustic analysis, artificial neural networks on offshore platform|
|Abstract:||Due to the limitations of the present noise prediction methods used in the offshore industry, this research is aimed to develop an efficient noise prediction technique that can analyze and predict the noise level for the offshore platform environment during the design stage as practically as possible to meet the criteria for crews’ comfort against high noise level. Several studies have been carried out to improve the understanding of acoustic environment onboard offshore platform, as well as the present prediction techniques. The noise prediction methods for the offshore platform were proposed from three aspects: by empirical acoustic modeling, analytical computation or neural network method. First, through evaluating the five-selected empirical acoustic models originated from other applications and statistical energy analaysis with direct field (SEA-DF), Heerema and Hodgson model was selected for calculating the sound level in the machinery room on the offshore platform. Second, the analytical model modeled three-dimensional fully coupled structural and acoustic systems by considering of the structural coupling force and the moment at edges, and structural-acoustic interaction on the interface. Artificial spring technique was implemented to illustrate the general coupling and boundary conditions. The use of Chebyshev expansions solutions ensured the accuracy and rapid convergence of the three-dimensional problem of single room and conjugate rooms. The proposed model was validated by checking natural frequencies and responses of against the results obtained from finite element software. Third, a modified multiple generalised regression neural network (GRNN) was first proposed to predict the noise level of various compartments onboard of the offshore platform with limited samples available. By preprocessing the samples with fuzzy c-means (FCM) and principal component analysis (PCA), dominant input features can be identified before commencing the GRNN’s training process. With optimal spread variables, the newly developed tool showed comparable performance to the SEA-DF and empirical formula that requires less time and resources to solve during the early stage of the offshore platform design.|
|Appears in Collections:||School of Engineering|
Files in This Item:
|Ji, X. 2018.pdf||Thesis||4.41 MB||Adobe PDF||View/Open|
|dspacelicence.pdf||Licence||43.82 kB||Adobe PDF||View/Open|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.