Applied Mechanics and Materials Vols. 220-223

Abstract: Abstract. Nonstationarity feature representation and extraction method based on the wavelet decomposition and demodulation techniques are studied. Some component in special frequency band included faulty information is selected to reconstruct by wavelet analysis. The mono-components with fault feature in different frequency band would be captured and separated out. The demodulated and spectrally signals are analyzed by Hilbert transform, and it presents an approach to get the characteristic frequency of fault signals. So what kind of the fault mode is can be estimated. For the nonstationarity and modulation feature of rolling bearing fault signals, wavelet decomposition combined with Hilbert transform is effective in identifying the localized defects of rolling bearings.

Abstract: In order to analyze harmonic and inter-harmonic parameters more accurately, a new detecting method that combined the nonlinear least squares and fast Fourier transform(FFT) was proposed. The FFT was used to produce a rough estimated value of harmonic parameters and then the signal model was set. Finally, the parameters can be estimated more accurately by the use of optimization algorithm which based on damped nonlinear-least squares method. The combined method solves the problem of inaccurate estimate by FFT and solves the problem of how to set the signal model and to choose initial value. The simulation results show that the combined method is effective and far more accuracy than interpolated Hanning window method.

Abstract: Because frequency dynamic load identification method will confront with the illness problem of finding the verse of coefficient matrix, a new multi-source load identification algorithm based on least-squares in frequency is proposed. Based on the assumptions of linear time-invariant system and uncorrelated of each load, this new algorithm combines transfer functions, least-squares of generalized matrix inverse in frequency. According to response signals of multi-spots, it can identify multi-sources dynamic random loads in frequency domain at the same time. Formula derivation, application scope and steps of this method were summarized then. In order to valid its effectiveness and reasonableness, the author carried out vibration and acoustic simulation in cylindrical shell at last. Load identification results of simulations showed that this new method could basically meet the dynamic random load precision requirement of ±3db.

Abstract: In order to improve the recognition rate of the electronic nose system for small samples, an electronic nose pattern recognition algorithm based on support vector machine (SVM) is proposed in this paper. Identification experiments for three kinds of wine with similar odor were carried out. The sensor arrays are optimized by means of principal component analysis (PCA) method first. Then, make comparing experiment using different algorithms for different number of training samples of wine. The related results show that PCA-SVM based pattern recognition algorithms has high recognition accuracy, stronger classification capability, and has potential advantages in small sample classification and recognition experiments.

Abstract: This paper studies baseline correction algorithms for subtracting the background of real-word signal. A novel baseline correction algorithm is proposed that can be solved by random signal processing. With respect to generalized statistical features of the raw data, an appropriate threshold of standard deviation is set to extract the true baseline points unfailingly. Under the generalized meaning, the background at one signal point is substituted by the statistical features of its local window. By using this proposed algorithm, we established a time varying signal baseline independently and accurately. And performance evaluation shows that the proposed algorithm is more elaborate and tolerant of real-word data than the previous ones.

Abstract: The surface electromyography signal is often submerged by the noise background while being gathered and recorded. To some extent, the useful signal and noise signal can be separated by applying the wavelet de-noising method effectively to eliminate the noisy signal. However, there are discontinuous points and constant deviations in the traditional wavelet threshold methods. Taking these problems into consideration, the improved threshold method is proposed and based on soft threshold function to use a transition of the nonlinear function to deal with the threshold function and de-noise the SEMG. Experimental results show that the improved threshold method has better de-noising effect and higher signal-noise ratio than the traditional threshold method.

Abstract: The principle of quasi-synchronous discrete fourier transform (DFT) algorithm was introduced and the simulation on the algorithm was carried out. The quasi-synchronous DFT algorithm is independent of the first sample point and it is able to start sampling at any moment. The maximum relative amplitude error and phase error of the algorithm are 0.05 % and 5.4° when the frequency of the signal is between 49.5 Hz and 50.5 Hz. The accuracy of quasi-synchronous DFT algorithm can be improved by scores of times than that of DFT algorithm.

Abstract: In order to solve function approximation, a mathematic model of Rational Function Functional Networks (RFFN) based on approximation was proposed and the learning algorithm for function approximation was presented. This algorithm used the lease square method thought and constructed auxiliary function by Lagrange multiplier method, and the parameters of the rational function functional networks were determined by solving a system of linear equations. Results illustrate the effectiveness of the rational function functional networks in solving approximation problems of the function with a pole.

Abstract: Differential equations are important tools in studying of natural science, engineering technology, and the laws of social economic development. It is necessary to seek some new inequalities in order to study of boundedness, uniqueness, stability and boundary value problem of a differential equation. Motivated by Abdeldaim integral inequalities, in this paper, we establish a class of generalized retarded nonlinear Gronwall-Bellman-Type integral inequalities and give upper bound estimation of the unknown function by analysis skills. Finally we give an example to illustrate the effectiveness of our results in estimation of solutions of some differential equations with the initial conditions.

Abstract: By using information entropy to estimate the distribution uniformity of the pixels with a same gray level, an accurate salt and pepper noise detection method is presented based on the statistical property of salt and pepper noise. And then, a new modified mean filter is designed, which sets up noise-centre filtering windows, Moreover, the weighted means are calculated by merely using the non-noise points in each filtering window. The presented filter can efficiently preserve the details of images, avoid the affection of noise points on the restore points, and reduce the dimness of the noise points. Experimental results show that this algorithm has the better performance on noise detection, noise filtering, and the protection of detail.