摘要
由于HMM本身的特点,在实际应用中如何找到一个合适的HMM模型始终是一个难点.采用传统的Baum-W elch算法对HMM进行训练,并提出了一种对训练结果进行评估的方法,构建了可自动评估的HMM训练体系.
According to the characteristics of HMM, it is difficult to find a suitable HMM model in the practical application. In this thesis, the Baum-Welch algorithm was used to train the HMMs and an evaluation algorithm of HMM was presented. Through it, an automatic evaluation HMM was constructed. The experiments on centrif- ugal pump show this novel evaluation algorithm can effectively improve the practicality of HMM diagnostic systems and the degree of automation on HMM training.
出处
《广州大学学报（自然科学版）》
CAS
2010年第4期13-17,共5页
Journal of Guangzhou University:Natural Science Edition
基金
国家863高技术研究发展计划项目(2008AA04Z407)资助
关键词
隐马尔可夫模型
评估算法
故障诊断
hidden Markov model
evaluation algorithm
condition recognition
