基于改进MDLP算法和分子布朗运动优化的传感器节点故障诊断  

Fault Diagnosis of Sensor Node Based on Improved MDLP Algorithm and Molecule Brownian motion Optimization

作　　者：韩小祥[1] 王雪梅[1] 

机构地区：[1]紫琅职业技术学院,江苏南通226002

出　　处：《计算机测量与控制》2013年第12期3212-3214,3218,共4页Computer Measurement &Control

基　　金：江苏省自然科学基金项目(BK2010192);南通市科技计划项目(BK2012035);院特色专业建设项目(201104)

摘　　要：针对监测区域中的传感器节点由于环境因素易于出现各类故障而影响了监测精度,提出了一种基于改进MDLP(Minimum Description Length Principle,MDLP)算法和布朗优化运动优化算法的节点故障诊断方法;首先,在经典MDLP的基础上设计了一种改进的全局数据离散化方法能有效地根据属性重要度对所有连续属性值进行离散化,在此基础上采用差别矩阵进行属性约简以减少特征维数,采用适合小样本数据故障诊断的LSSVM(Least Square Support Vector Machine,LSSVM)作为故障诊断模型,为了进一步提高其诊断精度,采用改进的分子布朗运动优化算法对其参数进行优化,从而得到最终的故障诊断模型;仿真实验表明:文中方法能有效实现传感器节点的故障诊断,具有较小的诊断误差的优点,与其他方法相比,在具有较高诊断效率的同时具有较高的故障诊断精度,具有一定的优越性。Aiming at the sensor node owing to the environment factors and always leading to varieties of faults then affecting the preei sion of diagnosis, a fault diagnosis method based on improved MDLP algorlsm and Molecule Brownian motion was proposed. Firstly, all the attribute values were discrete by improved MDLP （Minimum Description Length Principle） based on the significance of attribute, and then the differene matrix was used to reduce the data dimension. Finally, the （Least Square Support Vector Machine, LSSVM） was used o diag- nose the fault for it has suit small sample data. In order to improve the diagnosis precision, the improved Molecule Brownian motion was used to optimize the parameters of LSSVM. The simulation experiment shows our method can diagnose node fault effectively, and it has high diag- nosis accuracy, compared with the other methods, it not only has high diagnosis efficiency and precision. Therefore, it has big priority.

关 键 词：传感器节点 故障诊断 分子布朗运动 支持向量机 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]