摘要
自组织神经网络SOFM可以将多维数据快速地映射到二维平面上,但需要进行人工聚类划分。由于多尺度栅格技术可以对数据实现密度检测,相应的收缩聚类方法采用大小可变的栅格对相连单元进行标识,将点沿着数据的密度梯度进行移动,自动计算最好的簇作为聚类结果。以多维发动机飞行参数为应用对象,数据预处理提取特征后,按隶属度获得特征向量后输入SOFM网络,用收缩聚类算法(SCM)实现对输出的自动聚类检测,完成对发动机状态有效分析。实验结果和专家判读完全一致。
The SOFM can map muhi-dimensioanl data to 2-dimensional plane, but it needs manual cluster dividing. The shrinking-clustering method (SCM) can test the data density by marking the cells through an grid data adjustable and output the best-fit cluster automatically by moving the point along the density grads. This method has been successfully applied in aeroengine flight data processing. The flight data is imported to the SOFM after the analyziny of data features, and the system analyses the state affectivity of the aeroengine through SCM calculation.
出处
《传感器与微系统》
CSCD
北大核心
2006年第11期70-72,76,共4页
Transducer and Microsystem Technologies
关键词
飞行参数
收缩聚类算法
自组织特征映射
状态识别
flight data
shrinking-clustering method(SCM)
self-organizing feature map (SOFM)
state recognizing.