Search Results

Although excellent progress has been made recently in powertrain fault diagnosis based on vibration signals, most of them are based on the assumption that the fault features of the training and test data are drawn from the same probability distribution. Due to the limitation of the domain shift phenomenon, the performance of the current intelligent fault diagnosis methods is significantly reduced. Even many existing transfer learning methods have the problem of low generalization ability. Inspired by sparse representation theory, a novel cross-domain fault diagnosis method based on K-means singular value decomposition (K-SVD) and long short-term memory network (LSTM) is proposed in this study. First, K-SVD can convert source domain data into a sparse dictionary and sparse coefficient. The domain-invariant features are explored in the sparse dictionary, which contains redundant features. The sparse coefficients are input into the LSTM to obtain a primary classifier.

As a critical power source, the diesel engine is widely used in various situations. Diesel engine failure may lead to serious property losses and even accidents. Fault detection can improve the safety of diesel engines and reduce economic loss. Surface vibration signal is often used in non-disassembly fault diagnosis because of its convenient measurement and stability. This paper proposed a novel method for engine fault detection based on vibration signals using variational mode decomposition (VMD), K-means, and genetic algorithm. The mode number of VMD dramatically affects the accuracy of extracting signal components. Therefore, a method based on spectral energy distribution is proposed to determine the parameter, and the quadratic penalty term is optimized according to SNR. The results show that the optimized VMD can adaptively extract the vibration signal components of the diesel engine. In the actual fault diagnosis case, it is difficult to obtain the data with labels.

Noise source identification and separation of internal combustion engines is an effective tool for engine NVH (noise, vibration and harshness) development. Among various experimental approaches, noise source identification using signal processing has attracted extensive attention because of that the signal can be easily acquired and the requirements for equipment is relatively low. In this paper, variational mode decomposition (VMD) combined with independent component analysis (ICA) is used for noise source identification of a turbo-charged gasoline engine. Existing algorithms have been proved to be effective to extract signal features but also have disadvantages. One of the key problems in presently used method is that the main components of the signal, i.e. the main source of the noise, are unknown in advance. Thus the parameters selection of signal processing algorithms, which has a significance influence on the identification result, has no uniform criterion.

As a major power source, diesel engines are being widely used in a variety of fields. However, because of complex structure, some faults which cannot be detected by direct signals would occur on engines and even lead to accidents. Among all kinds of indirect signals, vibration signal is the most common choice for faults detection without disassemble because of its convenience and stability. This paper proposed a novel approach for detecting multiple engine faults based on block vibration signals using variational mode decomposition (VMD) and echo state network (ESN). Since the quadratic penalty has a great influence on adaptable VMD that may make expected component signals cannot be extracted exactly, this paper proposed a dynamic quadratic penalty value, which will change with decomposing level. This paper selected a best dynamic quadratic penalty value by analyzing a large amount of data and results showed that this approach can decompose signals more exactly.

Knock threshold detection is the key of closed loop control of ignition in gasoline engine, and it is also the difficult point in knock measurement. In this paper, an investigation of knock detection in turbocharged gasoline engine using bispectrum slice and ensemble empirical mode decomposition (EEMD) based on the engine cylinder head vibration signals is presented. By adding some finite amplitude Gaussian white noises to the signal, EEMD keeps the signal continuous in different time span, and therefore the mode mixing inhering in the classical empirical mode decomposition (EMD) method is alleviated. Power spectrum density (PSD) estimation is used to determine the band range of the resonance frequency generated by knock component. EEMD is used to decompose the original signals, the time-frequency characteristics of the Intrinsic Mode Functions (IMF) are analyzed using Continues Wavelet Transform (CWT) due to its excellent time-frequency resolution.

This paper reports an investigation of knock detection in spark ignition (SI) engines using EEMD-Hilbert transform based on the engine cylinder block vibration signals. Ensemble Empirical Mode Decomposition (EEMD) was used to de-compose the signal and detect knock characteristic. Hilbert transform was used to analysis the frequency information of knock characteristics. The result shows that for cylinder block vibration signals, the EEMD algorithm could extract the knock characteristic (include light knock), and the Hilbert transform result shows that the instantaneous of knock characteristics concentrate in 5000-10000Hz. At last, the knock window is then determined, based on which a new knock intensity evaluation factor K is proposed, and the results show that, the parameter K is reasonable and effective.