Abstract: The fault diagnosis model of coal mill was established based on least squares support vector machine. When using the model for fault diagnosis, support vector machine was greatly affected by kernel function parameters and penalty factors. To solve the problem, the model parameters were optimized by beetle swarm optimization (BSO) algorithm, and a coal mill fault diagnosis method based on support vector machine (SVM) was proposed. Firstly, the longicorn whisker search strategy was introduced to improve the position update rules of particle swarm optimization algorithm. Then, the fault characteristics were screened by partial mutual information method. Combining with the measured data of a power plant, the improved algorithm was used to optimize the kernel function parameters and penalty factors of support vector machine. Finally, BSO-SVM(beetle swarm optimization-support vector machine), SVM(support vector machine), PSO-SVM (partical swarm optimization-support vector machine)and GA-SVM(genetic algorithm-support vector machine)models were used to diagnose the fault of coal mill, which were compared with the actual fault types, and different levels of noise interference were added to the four models to test the stability of the model. The results show that the classification accuracy of BSO-SVM model is the highest, reaching 96.88%. In the five levels of noise interference, although the evaluation index F1aveand accuracy Accaveof the model decreased slightly, they can still maintain the highest level. Comparing with SVM, PSO-SVM and GA-SVM models, BSO-SVM can identify faults more stably and accurately, and provide practical reference for mill fault diagnosis.

Key words:  fault diagnosis; beetle search optimization (BSO)algorithm; support vector machine (SVM); coal mill; parameter optimization