The improved particle swarm breaker fault status parameter optimization of SVM classification
Yihang Sun · 2016
signal，PSO，energy method，fault diagnosis Abstract：In order to improve the mechanical structure of the type of fault resolution precision high voltage circuit breaker spring mechanism, the paper analyzes the characteristics of the circuit breaker and the combination of mechanical vibration signal PSO algorithm (PSO) SVM parameter optimization method proposed collaborative dynamic acceleration constant inertia weight particle swarm optimization (WCPSO) optimization support vector machine (SVM) analysis breaker fault classification parameters and kernel function parameters.The vibration signal circuit breaker empirical mode decomposition, the total intrinsic mode components through energy analysis to obtain the required fault feature vectors and support vector machine as input, the use of dynamic acceleration constant synergy inertia weight PSO support vector machines penalty factor C and radial basis kernel function parameters σ optimize the fault feature vector signal input test samples after SVM training sample trained optimized for fault classification, fault status classification.The experimental analysis of this method can effectively improve the resolution of the breaker failure signal type Accuracy. SVM parameters and their kernel function parametersSupport vector machines can be divided into two-dimensional space as shown in Fig( 1), rectangular and oval dot points represent two types of data samples, H is the sorting line, H0 and H1 represent samples classified by line and parallel to the nearest sample classification line spacing interval between them became classification.When the total sample linear separability exists a