НАПОЛНЕНИЕ БАЗЫ ДАННЫХ СПЕКТРОГРАММ И ВЕЙВЛЕТ-ПРЕОБРАЗОВАНИЙ ВИБРОСИГНАЛОВ ТОПЛИВНЫХ ФОРСУНОК CRIN2

A universal mathematical scheme for hidden information detection from various original n-dimensional signals is presented. The key points include the establishment of functional expression and the employment of cascade neural networks. The former indicates the detection of hidden information is dependent upon the shape of whole signals, not the specific values and the sampling number of the signals, the latter refers to the architecture of neural networks consisting of two neural networks, the first extracts main features of signals, and the second detects hidden information from the extracted features. Since wavelet functions play important role in signal analysis and feature extraction, thus wavelet neural networks constitute the first neural networks. The general scheme of feature extraction from original signals in L/sup 2/(R/sup M/) by wavelet neural networks is presented. The application in signal processing of chemical chromatographic spectra of solution shows satisfactory results.

With regard to fault diagnosis of rolling bearing, the envelope demodulation method is usually used to analyze the original vibration signal of faulty bearing, then the fault location of the bearing is determined by examining the distributions of fault characteristic frequencies, harmonics and sideband on the envelope demodulation spectral. The fault characteristic frequency with its harmonics could not be extracted by this traditional method when the original vibration signal is contaminated heavily by background noise. Besides, this method needs high professional knowledge and will expose drawbacks such as complex work, low diagnostic efficiency and so on while dealing with a large number of faulty bearings in engineering application. To solve the above problems, this paper proposes an automatic harmonic feature extraction method. Firstly, a series of bandpass filters are obtained based on fast Kurtogram, and then the original signal is filtered by the series of bandpass filters. The series of filtered signals are subjected to envelope analysis and noise reduction processing. Finally, the denoised series of envelope results are processed by the proposed algorithm for extracting the number of harmonics, harmonic intensity, and harmonic intensity index, and the target feature frequency and its harmonic components hidden in the original signal is extracted automatically. A simulation case and an engineering case verify that the proposed method can not only automatically calculate the number of harmonics of the characteristic frequencies, but also calculate the corresponding harmonic intensity, providing more effective and efficient feature support for fault diagnosis of bearing.

Experimental investigation of the oscillation characteristics of natural circulation of R134A in a two-phase loop

The power system load equipment is more sensitive to power quality disturbances than equipment applied in the past. Therefore, the electric supply quality has become a major concern of electric utilities and end-users. A novel approach to detect and locate power quality disturbance in distributed power system combining wavelet transform with neural network is proposed. By performing decomposition of transient waveform, the original signal is divided into two parts: the low-frequency and the high-frequency, corresponding to approximation part and details part respectively. The paper aims at complex wavelet analysis, and then explores feature extraction of disturbance signal to obtain dynamic parameters, superior to real wavelet analysis result. The characteristic vector obtained from wavelet decomposition coefficients are input data of neural network for power quality disturbance pattern recognition. The improved training algorithm is used to complete the network parameter identification. By means of simulation and experimental data, the disturbance pattern can be obtained from the neural network output. The simulation results show that the proposed method is effective for transient signal analysis, taking advantage of complex wavelet transform and neural network.

Vibrational signals resulting from tool wear have non-linear and non-stationary features. It is also difficult to acquire large numbers of typically worn samples in practice. In this work, a method of predicting the wear of milling tools is proposed based on ensemble empirical mode decomposition (EEMD) and the use of a support vector machine (SVM). The EEMD method is used to decompose the original non-stationary vibration acceleration signals into several stationary intrinsic mode functions (IMFs). The energies of the signals in these different frequency bands change when the tool is worn. Thus, the tool wear state can be identified by calculating the EEMD energies and energy entropies of the different vibrational signals. The correlation coefficients between the IMF components and original signal were calculated and wear-sensitive IMFs chosen. A SVM is then established by considering the energy features extracted from a number of wear-sensitive IMFs that contain primary information on tool wear. These are considered as the inputs to judge the wear state of the tool. The results show that the method is capable of predicting the wear state of the milling tool to good effect. Furthermore, the predictions made using an LS-SVM based on EEMD method are more accurate than those made using FFT, Wavelet analysis and EMD methods.

Vibration signals of tool wear are proved to be non-stationary ones. They usually carry the dynamic information of tool wear and are very useful for tool wear condition recognition. The wavelet analysis is especially suitable for non-stationary signal processing and the artificial neural network is a very good tool for signal identification. In this paper, a new efficient tool wear monitoring method based on the wavelet packet transform and the artificial neural network was presented. The cutting vibration signals in different milling wear conditions are decomposed and reconstructed by using the wavelet packet transform and feature extraction with conditions information is obtained by using proper and scientific methods to extract from signals. Wavelet pretreatment distills feature information from original signal as input vectors of decision net in order to reduce input data dimension, optimize the net construction, compute complexity and decrease the decision the errors. The theoretical background of wavelet packet transform and grey relational degree analysis optimization radial based function network (RBFN) was given. The effect of cutting parameters on energy parameters was taken into account while extracting energy parameters of tool wear characteristic vector, this made the extracted parameters of characteristic vector be more sensitive to tool wear and the sensitivity to cutting parameters be the minimum. The recognition method for tool wear condition was studied through artificial neural network and wavelet packet analysis, and relevant RBFN was established. The results showed that RBFN can be successfully used to recognize and analyze tool wear conditions.

Non-parametric modelling of a rectangular flexible plate structure

Observer based active vibration control of flexible space structures with prescribed performance

Bearings are necessary rotating machinery and plays an important role in the modern industrial systems for its safety and reliability. Timely fault diagnosis of bearings can reduce the probability of failure, thereby reducing economic losses and casualties. Many signal processing methods for fault feature extraction and fault recognition have been applied by scholars and engineers. Although numerous current methods identify and diagnose bearing faults correctly, they rely on a lot of existing information and experts experience, so it is not possible to establish a one-to-one correspondence between the original signal and the failure mode. Furthermore, the structure and parameters of the artificial intelligent neural network need to be optimized through experts and current knowledge. Alexnet neural network improves the learning ability and provides inspiration and direction for the above problem. The ensemble empirical mode decomposition (EEMD) solve the problem of mode mixing. The wavelet transform could impose the time and frequency features. Combing the prior of EEMD with continue wavelet transform, an adaptive fault feature model has been constructed that can directly provide the information to corresponding with the fault classified neural network. In this approach, fault signals are enhanced by extracting envelope decomposition and frequency signals. Numerous bearing data which containing different fault signals are used to verify the effectiveness and accuracy of the proposed method. The diagnosis results show that the novel alexnet neural network classifies bearings fault with high accuracy and robustness under complexity environment.

Purpose. Development and improvement of a method for analyzing vibration signals from rolling bearings based on wavelet analysis for the detection and identification of equipment defects. Research methods. Wavelet analysis was employed for processing vibration signals from rolling bearings. Threshold wavelet filtering was applied to highlight weak impulse components in the signals, and Morlet wavelet was used to ensure effective filtration. Results. The research results indicate that the proposed method, utilizing wavelet filtering, enhances the speed and reliability of vibration diagnostics for bearings. This allows for the efficient extraction of characteristic frequencies associated with various types of rolling bearing defects. In comparison with other signal analysis methods, the use of the developed method based on continuous wavelet analysis has proven to be particularly effective in extracting characteristic diagnostic frequencies. This method not only allows for the identification of specific types of defects in rolling bearings but also ensures universality, enabling its successful application for analyzing other types of non-stationary signals. Experimental studies have confirmed the high efficiency of the developed method, especially in the early stages of defect development. The application of this method is evident not only in its ability to effectively highlight the characteristic frequencies of bearings but also in its capacity to conduct signal analysis for the identification of equipment defects as a whole. This makes the proposed method a promising and versatile tool for the diagnosis and monitoring of the condition of technical systems. Scientific novelty. Application of the proposed method for processing vibration signals from rolling bearings based on wavelet analysis to improve the effectiveness of defect detection and identification in equipment. Practical value. The developed method can be applied in the industrial sector for the analysis and diagnostics of rolling bearings in equipment. It enables the timely detection of defects, reduces the risk of equipment failure, and lowers operational maintenance costs. Thus, this method has practical value in enhancing the reliability and productivity of industrial equipment.

In view of the wavelet neural network input number to determine the lack of theoretical guidance, the number of hidden layer nodes to determine the defects of difficult, put forward a kind of based on multivariate time series and grey relational -sensitivity of the wavelet neural network structure optimization method. This method firstly before the wavelet neural network learning, number of multivariate time series is used to determine the network's input, and then using grey relation in the process of learning -sensitivity pruning method to determine the neural network hidden layer node number, to achieve the goal of structural optimization. Through the model simulation results in the short-term wind power prediction, the results show that the method the optimized wavelet neural network to improve the wind power prediction accuracy and verified the effectiveness and feasibility of this structure optimization method, for the determination of wavelet neural network structure provides reference.

A nonlinear approach for recovering an undistorted signal which is distorted by any system is presented. This method is based on the techniques of neural networks. When trained with the original signal as target and the distorted signal as input vector, a neural network can be used to estimate the original signal instantly and can be used in real time for online signal recovery. The proposed approach has been tested in the case when the original signal is a random binary time sequence and the distortion is obtained by passing the original signal through a low-pass filter plus a white noise. It is shown that this approach provides excellent recovery of the original signal.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

With the growing number of applications of structural health monitoring systems on civil engineering structures, different types of sensors have been used to collect both the vibration data and environmental conditions. Signal loss is an unavoidable issue during transmission that significantly affects the quality of signals and consequently affects the practical applications of structural health monitoring techniques. This paper proposes a novel approach using deep learning technique for recovering the lost measurement data. The used convolutional neural network is a fully feed-forward convolutional neural network with residual connections. It constructs the nonlinear relationships between the incomplete signal with data loss and the original signal by training with the prepared training datasets consisting of pairs of incomplete and original vibration signals. The trained network then extracts the robust higher representation features of the input incomplete signals layer by layer using the compression layers, and expands those features gradually throughout the reconstruction layers to recover the original signals. The effectiveness of the approach is evaluated by using the data from Dowling Hall footbridge. The network exhibits the good capability of lost data recovery, even when the signals have severe data loss proportion up to 90%.KeywordsData recoveryStructural health monitoringTransmission lostDeep learningConvolutional neural networks

The condition monitoring and fault detection of rolling bearing are of great significance to ensure the safe and reliable operation of rotating machinery system.In the past few years, deep neural network (DNN) has been recognized as an effective tool to detect rolling bearing faults. However, It is too complex to directly feed the original vibration signal to the DNN neural network, and the accuracy of fault identification is not high. By using the signal preprocessing technology, the original signal can be effectively removed and preprocessed without losing the key diagnosis information. In this paper, a new EEMD-LSTM bearing fault diagnosis method is proposed, which combines the signal preprocessing technology with the EEMD method that can get clear fault feature signals, and LSTM technology to extract fault features automatically that improves the efficiency of fault feature extraction. In the case of small sample size, this method can significantly improve the accuracy of fault diagnosis.

Effectively leveraging the spatial features of time series signals to improve the accuracy of bearing fault classification in neural networks presents a significant challenge. To address this issue of different operating conditions, a novel model termed spatial pyramid pooling residual network-deep belief network (SPRout-DBN) is proposed. First and foremost, the Gramian angular difference fields (GADF) are utilized to encode original vibration signals of bearings. Secondly, two-dimensional images transformed by GADF from original signals are input to a novel designed residual network with spatial pyramid pooling to extract fixed-size temporal fusion feature vectors. Finally, a deep belief network is employed for classification and cross-domain learning, enabling the identification of fault samples under varying operating conditions. The proposed method is validated by two sets of datasets from Case Western Reserve University and Jiangnan University, achieving accuracies of 99.81% and 99.0% under identical operating conditions, and 99.41% and 98.43% under different operating conditions with 40 samples. Comparative analysis indicates that the proposed SPRout-DBN remains more robust and effective compared with other methods such as K-nearest neighbors, support vector machines, LeNet-5, ResNet-18, domain adaptation networks, and domain-adversarial neural networks in diverse operating environments.