Research On Blind Separation And Noise Source Identification For The Vibro-acoustic Signals Of Vehicle And Engine

With the fast development of automobile and high-speed train industries recently in China, the traffic noise problem is getting increasingly serious. The internal combustion engine (ICE) and traction system are dominant noise sources of automobile and high-speed train, separately. The combustion noise induced by combustion excitation and mechanical noise induced by valve-slap and piston-slap are generated around the Top Dead Centre (TDC), both of which are typically transient time-varying signals. Therefore it’s difficult to separate the mechanical noise from combustion noise. Meanwhile, it’s difficult to identify the noise source of traction system concerning the combination of mechanical noise, electromagnetic noise, fluid noise, aerodynamic noise, etc. Therefore, it would be unable to identify and separate such complex noise souces by traditional spectral analysis methods. The following work has been done focusing on the vibro-acoustic characteristic analysis and noise source identification of the ICE and high-speed train’s traction system.An adaptive generalized S transform (AGST) based on concentration measure is introduced. The numerical and experimental ICE vibration signal analysis results both indicate that the time-frequency representation (TFR) obtained by AGST has a better time-frequency resolution. Then, the AGST method is adopted to analyze the vibration and noise signals from the critical parts of a four-stroke and four-cylinder ICE. The results indicate that, the vibration and noise components induced by mechanical excitation (such as valve-slap, piston-slap and reciprocating inertia force) and combustion excitation can be exactly identified in the TFR with the help of AGST.A modified ensemble empirical mode decomposition (MEEMD) method is proposed base on the recent studies of the improvements of ensemble empirical mode decomposition (EEMD). The numerical and experimental ICE vibration signal analysis show that MEEMD is a better adaptive signal decomposed method, which not only restrains the disadvantage of mode mixing problem of the empirical mode decomposition (EMD) method, but also overcomes the non-IMFs, mode splitting and noise residue problems of EEMD method.The MEEMD-AGST method is adopted to analyze the vibration signals of a four-stroke and four-cylinder ICE, which is validated by the vibration experiments under motored condition. The analysis results indicate that the mechanical excitation components and combustion excitation components in the vibration signal of ICE can be effectively separated by the MEEMD-AGST method.The MEEMD-AGST method is adopted to study the noise contributions of a four-stroke and four-cylinder ICE’s vibration components and identify the dominant noise sources. The results show that the noise contributions of different vibration components are obtained, and the vibration sources which generate the dominant noise are identified. The research achievements are of great significance to the noise control and structural optimization for ICE.The AGST method is adopted to analyze the vibro-acoustic characteristic and identify the noise source of high-speed train’s traction system. The research achievements indicate that the wheel-rail noise, electromagnetic noise from the traction converter, fan noise from the cooling module and aerodynamic noise are the major noise contributors. Then, the speed-varying characteristic of traction system noise is analyzed, and optimization suggestion for low-noise traction system is proposed.