加权在线贯序极限学习机算法及其应用

Weighted online sequential extreme learning machine and its application

针对在线贯序极限学习机对所有数据等权处理这一缺陷,提出加权在线贯序极限学习机算法。依据运算过程中产生的网络均方根误差的差异,给新数据以及历史数据分配不同的权值,当网络均方根误差较大时减小其权值,较小时增大其权值。该算法实现了对新旧数据的不等权处理,利用航空发动机传感器数据验证该算法的可行性。验证结果表明,基于该算法所建的航空发动机传感器故障诊断模型要比基于传统在线贯序极限学习机算法所建模型的精度更高。

To solve the defect that the online sequential extreme learning machine uses equal rights to deal with all the data ,the weighted online sequential extreme learning machine （WOS-ELM ） algorithm was proposed .According to the different network root mean square errors emerging during the operation process ,different weights were assigned to the historical and new data . When the network root mean square error was big ,its weight was reduced ,and vise versa .This algorithm can implement the range of the historical and new data processing ,and the feasibility of the algorithm is validated by the simulation test on sensor data sampled from an aircraft engine .Results of the simulation test show that the precision of the sensor fault diagnosis based on the weighted online sequential extreme learning machine algorithm is higher than that based on the online sequential extreme learning machine algorithm for the aircraft engine .