Technology Study On Adaptive Active Noise Cancellation For Ship Self-Noise

With the development of society, the subjects of noise control have attracted more and more attention in both civil and military fields. Great progresses have been made up till now, however there are still many problems remaining to be solved. The technique of active noise control (ANC) in the presence of useful signals is just an important subject which arises recently.In this dissertation, some aspects of the subject have been studied: theories of new ANC technique with SNR (Signal-to-Noise Ratio) processing gain in the environments of noise and signals; algorithms and systems for noise cancellation while retaining signals undisturbed; methods of online path modeling encountered in real sound field; and cross-coupled multi-channel ANC systems for practical applications and acoustic simulation experiments.The noise is always hoped to be cancelled completely in the conventional ANC system because it does not have any useful information in most cases. But in some special application environments, such as the naval ships where it is extremely sensitive to newly introduced sound field, and the noise to be cancelled (mostly, the ship's self-noise) contains useful signals (target signals). The ANC system is expected to have the highest level of noise reduction while to retain the signal energy in the greatest extent. This kind of system is defined as ANC system with SNR processing gain. In the dissertation, the system physical model was built; concepts of "noise processing gain", "SNR processing gain", "signal distortion" and "signal distortion measurement" for the system were introduced and mathematically defined; a relatively complete theoretical system was established; and the computer simulation tests had been conducted as well. The results show that the newly developed system based on classical ANC principles works but the signal will be distorted in some degree, and the most important factor that influences the system performance is the SNR of reference input. In other words, if the reference input is designed to be picked up the properly, the system can used to cancel the interference noise, obtaining great SNR processing gain while retaining a low level signal distortion.Based on the above theories, several novel systems are proposed to solve the problems concerning noise cancellation and signal distortion. These systems include two kinds of ANC systems with pretreated reference input using independent component analysis (ICA) in time and frequency domains, respectively and an ANC system with pretreated reference input using crosstalk resistant adaptive noise canceller (CRANC). Theoretical analysis, computer simulation tests and processing of raw data collected from sea run tests are accomplished. The results show that these preprocessing techniques can reduce the cross interferences between signal and noise. Examined by the performance indexes of the ANC system with SNR processing gain, these new systems have better capabilities than existing systems.Aiming at the lacking capability of optional time and frequency (high frequency details and low frequency contour) resolution for existing time domain, frequency domain and wavelet decomposing domain ANC systems, a new system structure is proposed to solve this problem. This dissertation proposed a new filtered-X (FX) ANC system based on wavelet packet decomposing (WPD). LMS and RLS algorithms are derived under the filtered-X (FX) structure. Theoretical analysis, computer simulation tests and processing of real data collected from sea run tests are accomplished. Results show that the proposed scheme has inherent technical advantages and favorable performances which the conventional ANC system does not have.The existing of real sound path is one of the essential differences between ANC and adaptive noise cancellation. The online path modeling technique of ANC system has been studied intensively in this dissertation. Disadvantages of current techniques are that there exists disturbance between modeling filter and control filter, and that the exciting source for online secondary path modeling will increase the energy of sound field after system convergence Three new online path modeling ANC systems are proposed in this dissertation: modified FX structure online secondary path modeling ANC system, online secondary path modeling ANC system based on peak prefiltering technique, and ANC system of simultaneous equations which considers the existing of reference input path. The theoretic analysis, computer simulation tests and processing of real data collected from sea run tests are carried out. The results show that these systems have advantages on their own, and can be used to solve the two problems mentioned above in a certain extent.Multi-channel ANC system is the inevitable choice for regional noise cancellation. The problem of cross-couple of multi-channel has been studied in this dissertation. As an example, the modified FX structure is generalized into multi-channel ANC system. Algorithm of multi-channel FXRLS is developed. Theoretical analysis and computer simulation tests show that the multi-channel ANC system has good capabilities.