英文资料翻译-测试分析冲击振动信号进行水液压马达故障退化评估.doc

英文资料翻译英文原文（摘录）：Fault degradation assessment of water hydraulic motor by impulse vibration signal with Wavelet Packet Analysis and KolmogorovSmirnov Test H.X. Chen, Patrick S.K. Chua, G.H. Lim Abstract The machinery fault diagnosis is important for improving reliability and performance of systems. Many methods such as Time Synchronous Average (TSA), Fast Fourier Transform (FFT)-based spectrum analysis and short-time Fourier transform (STFT) have been applied in fault diagnosis and condition monitoring of mechanical system. The above methods analyze the signal in frequency domain with low resolution, which is not suitable for non-stationary vibration signal. The KolmogorovSmirnov (KS) test is a simple and precise technique in vibration signal analysis for machinery fault diagnosis. It has limited use and advantage to analyze the vibration signal with higher noise directly. In this paper, a new method for the fault degradation assessment of the water hydraulic motor is proposed based on Wavelet Packet Analysis (WPA) and KS test to analyze the impulsive energy of the vibration signal, which is used to detect the piston condition of water hydraulic motor. WPA is used to analyze the impulsive vibration signal from the casing of the water hydraulic motor to obtain the impulsive energy. The impulsive energy of the vibration signal can be obtained by the multi-decomposition based on Wavelet Packet Transform (WPT) and used as feature values to assess the fault degradation of the pistons. The kurtosis of the impulsive energy in the reconstructed signal from the Wavelet Packet coefficients is used to extract the feature values of the impulse energy by calculating the coefficients of the WPT multi-decomposition. The KS test is used to compare the kurtosis of the impulse energy of the vibration signal statistically under the different piston conditions. The results show the applicability and effectiveness of the proposed method to assess the fault degradation of the pistons in the water hydraulic motor.1. Introduction With the increasing environmental impact of operating oil-based hydraulic systems and the concern raised by environmentally conscious organizations, a most exciting area of development in the fluid power industry over the past few years has been in water hydraulics, which involves using tap water as a viable alternative to oil in fluid power transmission. Water hydraulics involves the use of a clean energy medium to transmit power and its use is in line with the global call to preserve our environment. Water hydraulic systems have been used in the farming, forestry, food, pharmaceutical and paper industries 13. The axial-piston motor is commonly used in water hydraulic systems to provide high torque and performance. Pistons are the principal operating elements of piston-type machines and their fundamental performance invariably depends on the smooth and efficient motion of pistons in the cylinder bore 4. Axial-piston swash-plate type hydraulic motor is comprised of a discrete number of pistons that reciprocate in a sinusoidal fashion for the purposes of torque output. It is impossible for an absolutely steady flow to be produced. The flow input of an axial-piston pump is not perfectly smooth. Some of the sinusoidal characteristics of the fluid displacement elements repeating at the sucking frequency are maintained. These sinusoidal characteristics are usually referred to as the flow ripple of the axial-piston motor, which generate thefluid-borne vibration (noise) and structure-borne vibration. Significant researches on axial-piston motors/pumps have appeared in the literature 5. The dynamic analysis of a swash-plate water hydraulic motor in a modern water hydraulic system was studied. A swash-plate mechanism was modeled as a system with three masses and 14 degrees of freedom (DOF). The simulated signals in X-, Y- and Z-axis showed that the vibration signals mainly consist of the hydraulic pump and hydraulic motor rotational frequencies. Much work has been done on fault diagnosis of oil hydraulics and other mechanical systems and this provides a useful reference for condition monitoring and fault diagnosis of a water hydraulic system. Crowther et al. 6 studied fault diagnosis of a hydraulic actuator circuit by using the output vector space classification approach. The diagnosis problem was illustrated by the developed theory in the second-order system. The presented method was effective on both the simulated data and experimental data from the laboratory-based hydraulic actuator circuit. Watton et al. 7 presented a method using rule-based concept to detect a number of the typical faults in hydraulic circuits. The method analyzed the on-line data and integrated both the qualitative and quantitative reasoning to detect the specific faults. The method was applied in an open-loop meter-in/meter-outflow control system with respect to leakages, check valves and flow control value behavior. Chen et al. 8 presented a novel method to analyze the vibration signals in fault diagnosis of water hydraulic motor, based on the second-generation wavelet. The multi-decomposition by lift scheme was developed. The new denoise method of the vibration signal was proposed by the lift scheme and the generalized cross validation (GCV).Stewart and Watton 9 described a program concerned with the development of a suite of modules, which can be applied to a variety of fluid power circuits. The approach can deduce the leakages using the real-time operational data and give some information on sensor integrity in a closed-loop cylinder position control system. Daley and Wang 10 reviewed the applications of fault detection and isolation (FDI) in fluid power systems. FDI used the nonlinear representation capabilities of neural networks to a simulated electro- hydraulic rotary drive system. Pietola and Varrio 11 studied the thermography in the condition monitoring, fault diagnosis and predictive maintenance of fluid power components and systems. The shortages and limitations of the thermal imaging in the fault diagnosis of fluid power were discussed. Gao et al. 12 investigated the real-time health diagnosis for hydraulic pumps by applying wavelet analysis. The simulation and experimental analysis was used to verify the capability of the wavelet method in diagnosing the health conditions of piston pumps. The types of the pump defects can be identified based on the pattern and amplitudes of the wavelet coefficients. The fault diagnosis of water hydraulic motor in the water hydraulic system was investigated based on adaptive wavelet analysis 13. The model-based method by AWT was applied to extract the features in the fault diagnosis of the water hydraulic motor. The magnitude plots of the continuous wavelet transform (CWT) showed the characteristic signals energy in time and frequency domain which can be used as feature values for fault diagnosis of water hydraulic motor. Wavelet Packet Transform (WPT) is the typical signal processing method for mechanical fault diagnosis. WPT can multi-decompose the signal into the different frequencies to obtain the localized impulse signals. The energy of the WPT coefficients is used for fault detection. Fan and Zuo 14 proposed a new fault detection method by WPT and Hilbert transform. WPT was an effective method to extract the modulating signal from the gearbox dynamics simulator to detect the earlier gear fault. The Wavelet Packet decomposition and wavelet residual analysis were used in hydraulic pump health diagnosis 15. A real-time pump health diagnosis system has been created by the pump discharge pressure signal analysis. Statistical analysis methods have been well developed and widely used in fault detection and diagnosis. The fault detection problem can be described as a hypothesis test problem and interpreted into an expression using specific models or distributions, or the parameters of a specific model or distribution. Test statistics are constructed to summarize the condition monitoring information using hypothesis testing. Chinmaya and Mohanty 16 applied the KolmogorovSmirnov (KS) test to analyze the motor current signature and vibration signal in the condition monitoring of a multistage automotive transmission gearbox at the different gear operations. The KS test of the motor current was verified to be effective method to monitor and detect the faults in gears. Zhan et al. 17 proposed a novel technique for the state detection of gearbox, whichfits a time-varying autoregressive model to the gear motion residual signals. The KS test was performed as a complementary statistical analysis to evaluate the state of the target gear, where the standard deviation of autogressive model residuals took its maximum in all tested gear motion residual signals for model order selection. The optimized statistical index was used to detect the piston condition of water hydraulic motor 18.The second generation wavelet was used to analyze the vibration signal and wavelet coefficients were calculated by the statistical index. The support vector machine was proposed to optimize the statistical index, which was used to detect the piston condition of water hydraulic motor effectively. The fault degradation assessment of water hydraulic motor is important for improving the water hydraulic system reliability and performance. The vibration signal from the casing of the water hydraulic motor is characteristic of the impulsive vibration signal, which is induced by the fluid in the water hydraulic system. This paper proposes a new method based on Wavelet Packet Analysis (WPA) and KS test to analyze the impulse vibration signal of the water hydraulic motor to assess the fault degradation of the pistons in water hydraulic motor. The proposed fault detection procedure is outlined as follows: (1) The vibration signal is analyzed by the multi-decomposition based on the Wavelet Packet. The level number of the multi-decompositions is 2. The soft-threshold is used in the wavelet and approximation coefficients to get the denoised coefficient. The reconstructed denoised vibration signal with higher signal-to-noise ratio (SNR) can be obtained by reconstructing the denoised coefficients in the multi-decomposition of the vibration signal. The reconstructed denoised vibration signal shows the impulsive energy of the vibration signal clearly; (2) kurtosis of the reconstructed signal is calculated from the wavelet coefficients and approximation coefficients and (3) the KS test is used to classify the kurtosis statistical probability distribution (SPD) each other under seven different piston conditions, which is used to detect the piston condition in water hydraulic motor.2. Impulsive vibration mechanism of water hydraulic motor The actuator studied here is a five-piston axial-piston motor used in a water hydraulic system. An accelerometer mounted on the casing of the MAH 12.5 water hydraulic motor was used to obtain the experimental vibration signal from the motor. The complete water hydraulic system was provided by Danfoss Inc. and Fig. 1 shows the general structure of the water hydraulic motor 19. An axial-piston motor consists mainly of a valve port plate with inlet and outlet ports, a swash-plate, an outer shell, a cylinder block, pistons with shoes, a bias spring, a port flange and a shaft. The piston fits within the bores of the cylinder barrel and is on the same axis as the output shaft. The swash-plate is positioned at an angle and acts as a surface on which the piston shoes travel. The shoes are held in contact with the swash-plate by the retaining rings and the bias spring. The port plate separates the incoming fluid from the discharging fluid. The output shaft is connected to the cylinder barrel. The pistons drive the cylinder block to rotate around the central axis of the motor shaft at a constant angular speed by the pressure in the supply port. Each piston periodically passes over the supply and return line ports on the valve plate. The slippers are held against the inclined plane of the swash plate. The pistons undergo an oscillatory displacement in and out of the cylinder block. When the piston passes over the supply port, the pressure causes the piston to withdraw from the cylinder block. The piston passes over the return line port and the fluid is pushed out of the piston chamber. These motions of the piston and cylinder block repeat and the basic task of the output torque is then completed. As the water enters the inlet and exits at the outlet of the hydraulic motor, the pressure in the cylinder chamber alternates from high pressure to low pressure. 1 2 3 4 5 6 7 8 15 14 13 12 11 10 91. retaining ring 6. port flange 12. piston2. bushing with spherical 7. hydrostatic bearings 13. shoe outer surface 8. shaft seal 14. swash plate3. cylinder block 9. motor shaft 15. bearing4. Spring 10. valve port plate5. thrust plate 11. outer shell Fig. 1. Swash-plate water hydraulic motor with five pistons. The total cylinder area inside a supply port is variable as a result of the cyclic variations of the piston passing through the supply port. It generates the variations of the axial output moment. This causes pressure ripple pulsation to occur. The pressure ripples in piston pumps and motor are mostly due to an intermittent impulsive back flow into the cylinder chamber in the neighborhood of the bottom and top dead center. The variations of the forces are applied from the piston to the swash plate and the valve cover. The force between the support for the swash plate and the valve cover is opposite in direction. The hydraulic motor body vibrates as a result of the pressure ripple. The characteristic frequencies of fluid-borne impulse vibration in hydraulic system have two flow and pressure pulsating sources generated by the pump and motor, respectively. The fundamental frequencies of water hydraulic motor and pump are determined as follows: f=zN (1)where f is the fundamental frequency of the hydraulic pump and motor, z is the piston number and N is the angular speed. In this work, fundamental frequency f was 52.5Hz, piston number z was 5 and angular speed N was 630rev/min.3. Wavelet Packet Analysis and KS test3.1. Multi-decomposition of WPA Wavelet Packet is defined as 20: （2） Wavelet Packet function is defined aswhere N is the set of positive integers and Z is the set of integers; n is the oscillation parameter; j and k are the scaling parameters (frequency localization) and the translation parameter (time localization). The first two Wavelet Packet functions are defined as: （3）where f(t) is the scaling function, and c(t) is the wavelet function. h1 and g1 are the low-pass and high-pass filters 20. The function is localized both in time and frequency. Each of them is a function of unit energy, with a scale of 2j, centered at 2jk, and with an oscillation parameter n. Wavelet Packet provides a library of orthonormal bases for the square-integrable space L2(R). For each n= 0, 1, 2, , the collection of functions : j, kZ forms an orthonomal basis for L2(R). Such a basis is called a Wavelet Packet basis and the decomposition of a signal into a Wavelet Packet basis is referred to as WPT. The orthonormal wavelet basis is a special case of Wavelet Packet basis. The space splitting operation for Wavelet Packets is different from that for multiresolution analysis. The multiresolution analysis by wavelets is obtained by splitting the subspace into and using the low-pass and high-pass filters, and doing the same for Vj recursively. The generation of Wavelet Packet by Eq. (2) corresponds the splitting operation not only on but also on , recursively as well. Owing to the low-pass and high-pass properties of h1 and g1, the splitting operation on divides the frequency interval represented by the subspace into a lower and an upper frequency part. Wavelet Packet thus allows finer decomposition in the frequency domain. In the mult-decomposition of the Wavelet Packet, a binary tree structure consists of the lower frequency and higher frequency as illustrated in Fig. 2. Eq. (2) defines the splitting relations in the Wavelet Packet orthogonal basis at the nodes. Here j is the depth, and n is the number of the nodes to the left of this particular node at the same depth. The scaling and wavelet coefficients in a depth and node can be reconstructed by setting Wavelet Packet coefficients of other depths and nodes to be zero. The reconstructed signal obtained from the Wavelet Packet coefficients corresponds to different frequency range. transforms the signal in the frequency range of , where is the sample frequency. In this paper, the sample frequency is 20kHz. At j+1 level, the two frequency ranges are 010kHz and 1020 kHz. At j+2 level, the four frequency ranges are 05kHz, 510kHz, 1015kHz and 1520kHz.The decomposition coefficients of a signal f(t) into Wavelet Packet are computed by applying the low-pass and high-pass filters iteratively. Let de