Study on Flame Combustion Stability Based on Particle Swarm Optimization Feature-Weighted SVM

Abstract

In order to achieve the automatic monitoring of the combustion stability of the boiler, and to quantitatively determine the stability of the combustion, the combustion stability evaluation model of the particle swarm optimization feature weighted support vector machine is proposed. The eigenvalues of the combustion state in the flame image is extracted, and the feature weight of each eigenvalue is obtained. Then, the kernel function of the support vector machine is modified by feature weighting vector. Particle swarm is used to optimize penalty factors and kernel parameters, and the same set of samples are used to test the classification ability of support vector machine and feature weighted support vector machine. The results show that the support vector machine model with feature weighting has higher recognition rate and can judge the combustion state accurately and effectively, which can meet the real-time requirement of stability judgment.