Jianliang Li

An Accelerating Method of Training Neural Networks Based on Vector Epsilon Algorithm

This paper studied the accelerating convergence of the vector sequences generated by BP algorithm with vector epsilon algorithm, and presented the conclusion that the algorithms is not only convergent but also accelerated. Finally, we tested them for three classical artificial neural network problems. By numerical experiments, results shown that can reduce CPU time for computation and improve the learning efficiency.

Simulation Analysis of Transmission Loss in the Vibrating End Plate of an Expansion Chamber Silencer

The boundary element method (BEM) is frequently employed to predict the transmission loss (TL) of an expansion chamber silencer with a rigid walled boundary, while the effect of the vibration of the end plates of the expansion chamber silencer is usually ignored. However, when the airflow or acoustic pressure in a duct is large and the end plate of the expansion chamber silencer is not thick enough, the effect of the vibration of the end plate should not be ignored. In this paper, based on a coupling of the BEM and the finite element method (FEM), the numerical results illustrate the existence of the effect on the TL of the vibrating end plates. Furthermore, numerical experiments also show that for an expansion chamber of constant size, the thicker the end plate, the less influence the vibration of the end plate has on the TL. And this may be helpful for the design of silencer.

Components Locating in PCB Fault Diagnosis Based on Infrared Thermal Imaging

The paper studied a statistical correlation based algorithm in components location in Printed Circuit Board (PCB) fault diagnosis. As integration and micromation complexity increases, interaction between thermal sources and thermal image drift of circuit components bring new difficulties into the applications. It’s pointed that feature points algorithm is a feasible method in inconsistency registration. The results indicated that thermal image drift can be effectively reduced and lots of computation load are saved.

Prediction of the performance of the expansion chamber in air ducts

An expansion chamber is frequently used as an acoustics silencer for the control of duct-borne noise in ventilation systems. The flow of air and the effect of sound pressure may cause the vibration of the end plates of the expansion chamber and in turn will influence the performance (usually in terms of the transmission loss (TL)) of the expansion chamber. By combining the plate bending element and the plane stress element, BEM (for acoustics systems) and FEM (for the structure analysis) are applied to predict the performance of the expansion muffler. A method for calculating the TL for acoustics silencer using BEM-FEM coupled model is therefore presented, and an improved method based on the four-pole parameters used in the BEM is obtained.

Simulation Analysis for Peak Pressure of Shock Wave Based on Lax-Friedrichs Method

FEM software has been well-developed in shock wave simulation. As the software usually is huge and complicated, subdivision mesh needs lots of computation, and the precision is not assured, finite-difference methods are highly regarded. Difference schemes are related to the stability and precision. In this paper, a Lax-Friedrichs based difference method is presented. Different grids were discussed and the numerical results were compared with Henrych empirical formula. Furthermore, comparing the data from numerical simulation of TNT explosive and the data from DYNA software, we concluded finite-difference methods have advantage of FEM methods.

A Neural Network Training Algorithm Based on Collinear Scaling Quasi-Newton Method

Based on collinear scaling and local quadratic approximation, quasi-Newton methods have improved for function value is not fully used in the Hessian matrix. As collinear scaling factor in paper [1] may appear singular, this paper, a new collinear scaling factor is studied. Using local quadratic approximation, an improved collinear scaling algorithm to strengthen the stability is presented, and the global convergence of the algorithm is proved. In addition, numerical results of training neural network with the improved collinear scaling algorithm shown the efficiency of this algorithm is much better than traditional ones.

Simulation Analysis of Acoustic Attenuation Performance for Different Shape of an Expansion Chamber Silencer

This paper studied the internal laws of silencers’ acoustic attenuation performance, when the cross-sectional shape, length and diameter of the expansion chamber changed. Based on the BEM-FEM coupled model and to combine with the improved Four-pole Parameters Method, the transmission loss (TL) was calculated. It is pointed that the flatter the expansion chamber is, the worse the performance is. When the length decreases, the domes of the TL become broader, and the number of domes decreases. When the diameter increases, the TL increases, but the growth rate and the inactive frequency decrease. The results indicate that by using computer modeling analysis, it is sensible to optimize the shape to meet the needs of the eliminating frequency band and the applied circumstance.

Research and New Development of Application of the Estimation of Distribution Algorithms

The Estimation of Distribution Algorithms (EDAs) is a novel class of evolutionary algorithms which is motivated by the idea of building probabilistic graphical model of promising solutions to represent linkage information between variables in chromosome. Through learning of and sampling from probabilistic graphical model, new population is generated and optimization procedure is repeated until the stopping criteria are met. In this paper, the mechanism of the Estimation of Distribution Algorithms is analyzed. Currently existing EDAs are surveyed and categorized according to the probabilistic model they used.