Decomposition Method Of Based On Band-Pass Technique And Its Application To Engineering

With the rapid development of modern ocean industry, oceanographic information composed of all kinds of data series increase a lots. Due to increasing attention paying on the safety and reliability of ocean structures, the requirement to accurately forecast the marine environment and disasters becomes urgent for ocean and offshore construction projects, based on observed oceanographic data information. Fuzzy qualitative analysis is no longer satisfied, more accurate quantitative analysis results are wished to obtain. Low data processing speed characterized by inefficiently manual processing way changed into the one characterized by entirely automatic or intelligent data processing way. Therefore, more efficient data processing, eliminating measurement errors and systematic errors and enhancing data value become the focus of the current research.Several widely used data analysis techniques have been introduced in this dissertation, and their merits and defects of each analysis method have been objectively evaluated. Based on this analysis, two new data analysis methods have been developed, FFT band-pass filtering for empirical mode decomposition and FFT band-pass wave filtering smoothness technique. A new method for the empirical mode decomposition of digital signals or data sets named as FB-EMD, have been developed by virtue of a simple and fast band-pass filtering technique. Comparing with the shifting process proposed by Huang et al, the so-called FB-EMD technique adopts new method and decomposition rule to decompounded steady or non-steady data into some brief intrinsic mode functions exactly, consisting with the characteristics of those original data. The newly developed method has the following virtues: 1) a complicated digital signal can be more simply and fastly decomposed into a set of intrinsic mode functions(IMFs), by means of this method; 2) IMFs are rigorously orthogonal to each other, and therefore can prevent these IMFs from mixing. Consequently, the resultant Hilbert spectrum is of mono-value, rather than multiple-value functions according to the instantaneous frequency; 3) it is highly efficient and therefore capable for long digital signals.The efficiency of this method has been verified, by using it to process some typical signals. Most of the results show fairly good reliability and exactitude. By means of the proposed maximum intrinsic bandwidth, the method is prone to decompose a signal into a series of IMFs. They are meaningful and characteristic. After the comparison with other methods by decomposing complicated data sets, an important conclusion can be drawn that the FB-EMD is indeed superior to them. All problems that HHT signal analysis method solves can be substituted by the FB-EMD method.Another proposed method is mainly used to smooth the wave signals or data sets. The smoothness method using FFT and band-pass techniques can efficiently conquers the problems and defects which usually exist among several widely used smoothness methods. Comparing with those smooth methods, the smoothness method using FFT and band-pass techniques has more theoretical significance, and can be simply used in practical engineering application. On the other hand, its great merit is that it is extremely fit for automatically computer disposal.Generally speaking, both of the signal decomposion and analysis methods widen the application range even to nonlinear and non-steady data processing. We can believe that both of them may have wider future.