Wang Huaying

Information processing in digital holographic imaging and measurement

According to Fresnel diffraction theory, the principles of imaging and shape measurement are analyzed concisely. The model of phase aberration in numerical reconstruction is built. The zero-order diffraction and unwanted first-order image can be removed by using frequency filtering. Based on lensless Fourier transform digital holographic setup, the experiment is conducted by using a US Air Force test target as microscopic object. The accurate phase image is obtained by combining the automatic procedure for phase aberration compensation with manual adjustment of reconstructing parameters. The lateral resolution of the intensity image of 3.10mum without any pre-magnification is demonstrated experimentally, and higher resolution of phase image has been achieved. The results show that in the case of strong noise, the inexact reconstruction parameters can be gained only by the automatic procedure. In order to obtain the accurate phase information, the finer manual adjustment is necessary subsequently. It is the committed step that recording high quality digital hologram in digital holography.

Comparison of the reconstruction algorithms in digital micro-holography

To improve the quality of the reconstructed image, the common used three reconstruction algorithms in digital holographic microscopy, Fresnel transform algorithm, angular spectrum algorithm, and convolution algorithm, all based on fast-Fourier-transform (FFT) are investigated and compared. By using off-axis lensless Fourier setup the digital hologram of a USAF test target is recorded and reconstructed numerically with the three algorithms at different reconstruction distances. The results show that by Fresnel transform algorithm the lensless Fourier transform digital hologram can be reconstructed at any distances. For convolution and angular spectrum algorithms, there is an optimal reconstruction distance. For convolution algorithm, when the reconstruction distance is different from the optimal distance, the image resolution is decreased, particularly for small distance. When the reconstructing distance is slightly smaller and very larger than the optimal one the high quality image can also be obtained by using angular spectrum algorithm. Angular spectrum algorithm is better than convolution algorithm. The Fresnel transform algorithm is the optimal numerical reconstruction algorithm in digital holographic microscopy.

A novel intelligent system for analysis and recognition of power quality disturbance signal

The power quality disturbances and the resulting problems have emerged as an important research. The power system industries with sensitive electrical loads have become more dependent on the quality of power supply system. The power quality disturbances analysis is becoming an essential issue because of the widespread use of electronic nonlinear loads that have affected the operation of distributed power system network in residential and industrial areas. A novel approach to detect and locate short duration disturbance in distributed power system combing neural network is presented. The paper tries to explain to investigate feature extraction of transient signal and to analyze the disturbance signal. The feature information obtained from wavelet decomposition coefficients acts as input vector of wavelet network for power quality disturbance pattern recognition. The power quality disturbance recognition performance is completed and the improved back-propagation algorithm is used to fulfill the network parameter initialization. By means of simulation data training, the disturbance pattern can be obtained from the trained wavelet network output. The simulation results and analysis indicate that the wavelet transform combining with neural network is sensitive to transient signal singularity detection.

Application of Wavelet Transform and Fuzzy Theory for Turbo-Generator Fault Mode Classification

A novel combined method based on wavelet transform and fuzzy neural network for concurrent vibrant faults of turbo-generator sets is presented. The fault feature is distinct and the high frequency components in the process can be employed to reveal fault characteristics. Fuzzy neural networks show good ability of self-adaption and self-learning, wavelet transformation or analysis shows the time frequency location characteristic and multi-scale ability. The fault diagnosis model of turbo-generator set is established and the improved least squares algorithm is used to fulfill the network parameters initialization. By means of choosing enough samples to train the fault diagnosis network and the information representing the faults is input into the trained diagnosis network, and according to the output result the type of fault can be determined. The computation of wavelet fuzzy network is dynamic and global optimized, therefore the convergence speed and the error precision are improved. The test results of demonstrates that the proposed method has its advantage in dealing with concurrent fault situations and is featured by a high probability of accuracy, proving the method to be effective.