Song Zhiyong

Hybrid optimization method for parameter selection of support vector machine

Parameter selection of support vector machine (SVM) is a key problem in the application of SVM, which directly has influence on generalization performance of SVM. By using the support vector bound which is modified by F measure as criterion function and using genetic simulated annealing algorithm to select kernel parameters and penalty factor, a method of parameter selection of SVM based on hybrid optimization strategy is proposed in this paper. The proposed method adequately combines the optimization advantages of genetic algorithm and simulated annealing algorithms. The experiment results demonstrate that, compared with cross validation method, this proposed method improves accuracy of SVM parameter selection and generalization performance of SVM.

A two dimension distributed MVI-CFAR detector

MVI-CFAR detector has a strong adaptive characteristic in both homogeneous and inhomogeneous distributed background, which tackles the problem of degraded VI-CFAR detection performance when interference targets appear in both leading and lagging sliding windows. This paper focuses on target detection using ultra-band radar and presents a two dimensional distributed MVI-CFAR detector by exploiting target environment information contained in Range-Doppler matrix and combining detection advantages of MVI-CFAR as well. By conducting MVI-CFAR detection of the same target detection unit in both Range dimension and Doppler dimension and then fusion decision of two dimensional detection result under the guidance of the principle of distribution, target detection under the condition of complex clutter background/multi-target is realized. By Monte Carlo simulation, this paper analyzes the detection performance of this algorithm in both homogeneous and inhomogeneous background and verifies the effectiveness of it.

Study on contour error related to dynamic parameters of CNC machine tool

As the contour error on cutting part is closely related to the dynamic parameters of CNC machine tool, which are usually refered to the dynamic accuracy, this paper presents a way to identify the shift of dynamic parameters with special shape feature. By modeling control system of CNC machine tool, the dynamic accuracy parameters of CNC machine tool are selected and the contour error curves are simulated. The contour errors are the results of unsuitable machine parameters and there is certain regularity. Using multiple shape features, the uncoordinated parameters of CNC machine tool causing the contour error are identified. The simulation shows that it is an effective way to identify the machine dynamic accuracy with the characteristic of contour error.