| Railway is the important infrastructure of transportation, and the train is the specific carrying vehicle of rail transport. The operation condition of train is a key factor to the rail transport safety. Therefore, the safety operation is a top priority to protect human life and property security, even to the economic development which has a very important practical significance.As the wheel-set components are the supporting parts and traveling parts of the train, due to its poor operation condition such as long period of high-speed and heavy loads situation, the contact surface of the relevant rolling components is always attacked by the long-term alternating stress. The rolling components are easily caused to fatigue, crack or other faults. If such faults are continued developing will bring additional shock and vibration to the train, and lead to bearing heating, axis cutting, and even lead to a serious crash accident. It's necessary to apply condition monitoring and fault diagnosis methods to the key components of wheel set, detect the early fault and avoid the accident which is dangerous to the railway transportation.This paper focuses on the train wheel-set's early defects of during operation, by theoretical analysis the definition of the frequency-varying fault is put forward. The fault mechanism, condition monitoring methods, fault signal processing algorithms of train wheel-set's frequency-varying fault are researched based on the detailed analysis of the key components structure and kinetic characteristics of traction gear fault, axle box bearing fault and the motor shaft bearing fault. By theoretical analysis, simulation studies, laboratory analysis and actual applications, a combination of monitoring and diagnosis method is systematic studied. The main research content and results are as follows:(1) The structure and mechanical characteristics of the train wheel set component is analyses in this part. The vibration characteristics, fault mechanism and fault spectrum characteristics is researched. The common fault frequency characteristics of the train wheel set component are summarized. Through the detail analysis of the external influences which is different from the common rotating machinery, the concept of frequency-varying factors is summarized and presented. By entirely think of the operation factors, the general fault mechanism model is established.(2) By the definition of the fault characteristic coefficient K which is unrelated to the frequency-varying factors, the conventional spectral analysis method can be normalized to the characteristic domain signal, and the characteristics spectral analysis method is achieved. The principle and some key technologies (such as:synchronous sampling design, sampling parameters selected et al.) of characteristic spectral analysis method is analysed in detail. Through simulation analysis and experimental study on the train wheel set fault test system, the accuracy and usefulness of characteristics spectral analysis method is verified.(3) In this part, the train likely-cycle vibration signal is analysised. By research the cyclostationary properties of the frequency-varying fault signal, the spectral correlation density function analysis method of train wheel set vibration signal's cycle statistics parmeter is selected. From the circulating the auto-correlation function of train wheel set vibration signals, the noise reduction analysis is proposed. The noise characteristic of the train wheel set vibration signal is researched through spectral correlation density function based on the auto-correlation function analysis of additive noise. And the noise reduction effect is verified by the simulated signals and experimental data analysis.(4) The cyclic statistics method of is applied to extraction the fault characteristic of the frequency-varying fault. For calculating the cyclostationary properties as cycle frequency α, a cycle frequency a calculate algorithm based on the full frequency band sweep extraction method is proposed. By calculate the spectral correlation density function value where the cycle frequency a is get from the frequency band sweep method. The fault characteristic is extracted from the wheel components to achieve a precise diagnosis for such fault. In the last of this part, a higher order bispectrum analysis is applied to the train wheel set fault. Be combined with the characteristic domain signal processing method proposed in this paper. The characteristic domain's higher order bispectrum is put forward. And the characteristic domain bispectrum diagonal slice is used to extract fault characteristic. Through the practical application research, it shows that such analysis method has a certain practicality.(5) The train wheel set frequency-varying fault signal's local mean decomposition method (denoted by LMD) is studied in this part. Full account of the multi-component AM-FM signal collect from train wheel set fault, the local mean decomposition method is applied to decompose this signal. The new local mean decomposition method based on the average processing with window sliding extraction technology is proposed and will be applied to extract a multi-component signal into several single-component signals. In order to achieve the no speed-tracking characteristic domain signal analysis techniques, the train wheel set instantaneous frequency of rotating frequency is extract from the several signal-component signals. By analyzing the principles and causes of the end effect in the local mean decomposition, a extension characteristic wave method based on the statistical characteristic of the original signal ending localized waveform is proposed. In order to eliminate the end effect of the decomposition algorithm, an intensity formula of the end effect's quantitative analysis is defined. All the method put forwarded above is verified in the real case application, and achieved satisfactory results.(6) For the non-stationary and non-linear characteristics of the fault signal, a time-frequency analysis method is researched in order to extract the fault signal's time-frequency characteristic. A variables improved method of the kernel function is researched. And an amendments polynomial Wigner-Ville distribution for the train wheel real-time status monitoring is proposed. For the polynomial kernel function of the signal frequency distribution map is more complicated, a frequency map optimal path search algorithm based on the Viterbi algorithm is proposed. Through the experiment on the train wheel set fault test system and the actual train running data analysis, the methods researched above is verified, and obtain a satisfactory result.The fault diagnosis analysis methods which are proposed in this paper are all undergoing by a rigorous theoretical analysis, simulation and real experiment has a strong practical application of engineering. This paper presents analytical methods is account of the frequency-varying factors, the precise diagnostic algorithm is put forward and verified. Such monitoring and diagnostic methods and techniques are not only can used to the train wheel set components, but also can promote the use of any possible speed fluctuations rotating machinery monitoring and diagnosis. |
PDF Full Text Request