摘要
针对Bayes准则中先验概率定义困难的问题,提出了实时加权先验概率求解算法。算法不需要任何有关系统的先验知识,通过挖掘隐含在测试数据中的信息,结合系统的当前状态,实时地调节先验概率的取值。文中给出了算法的具体推导和试验验证,并结合实际使用,对算法中相关参数的取值给出了详细说明。结果表明:采用实时加权先验概率算法,对系统状态的判断更加合理,其判决结果的稳定性、可靠性,以及克服干扰的鲁棒性,比通常的算法有显著提高。
Aiming at that the prior probability of Bayesian criterion is difficult to assign, this paper puts forward a real-time weighted algorithm for prior probability assignment. The algorithm requires no prior knowledge and can adaptively adjust prior probability value according to conditions of the system by mining the hidden information in monitoring data. In the paper the detailed deducing process and validity experiments of the algorithm are described. The values of the related parameters of the algorithm are interpreted according to the practical applications. The results show that the new algorithm improves the rationality of conditions discrimination. The stability, reliability and robustness in overcoming the disturbance of the algorithm are all significantly better than the normal algorithm.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2008年第9期1951-1955,共5页
Chinese Journal of Scientific Instrument
关键词
Bayes准则
先验概率
状态监测
算法
bayesian criterion
prior probability
condition monitoring
algorithm
