基于RSOPNN的无线传感器网络节点故障诊断算法
发布时间：2018-11-08 10:07
【摘要】：针对无线传感器网络节点故障诊断中存在的冗余故障属性、噪声数据以及数据可靠性等问题,提出基于粗糙集-优化概率神经网络的无线传感器网络节点故障诊断算法(简称RSOPNN)。通过粗糙集从故障样本属性集合中求解故障诊断属性约简,从而去除冗余故障属性,降低冗余属性、噪声数据对故障诊断的影响,节省能耗。对于多个属性约简选择,以属性间的相关程度作为度量标准,代替常规的主观选择,从多个约简中确定最优故障诊断属性约简,解决主观选择的不合理性。以最优的故障诊断属性重构故障样本,作为优化概率神经网络的输入,建立故障分类模型,从而对故障进行诊断。实验结果表明,在不同的数据可靠性下,RSOPNN方法能够有效删减样本中的冗余属性和噪声数据,保持高效的故障诊断水平,符合无线传感器网络的需求。
[Abstract]:Aiming at the problems of redundant fault attributes, noise data and data reliability in node fault diagnosis of wireless sensor networks, A novel Node Fault diagnosis algorithm for Wireless Sensor Networks based on rough set and optimal probabilistic Neural Networks (RSOPNN).) The reduction of fault diagnosis attribute is solved by rough set from fault sample attribute set, so that redundant fault attribute is removed, redundant attribute is reduced, the influence of noise data on fault diagnosis is reduced, and energy consumption is saved. For multiple attribute reduction selection, the correlation degree between attributes is taken as a measure instead of conventional subjective selection, and the optimal fault diagnosis attribute reduction is determined from multiple reduction to solve the irrationality of subjective selection. The optimal fault diagnosis attribute is used to reconstruct the fault sample as the input of the optimized probabilistic neural network and the fault classification model is established to diagnose the fault. The experimental results show that under different data reliability, the RSOPNN method can effectively delete redundant attributes and noise data from the samples and maintain an efficient fault diagnosis level, which meets the needs of wireless sensor networks.
【作者单位】： 西北大学信息科学与技术学院;西北大学现代教育技术中心;
【基金】：国家科技支撑计划课题(No.2013BAK01B02) 国家自然科学基金(No.61373176) 陕西省重大科技创新专项资金项目(No.2012ZKC05-2)
【分类号】：TP18;TP212.9;TN929.5
[Abstract]:Aiming at the problems of redundant fault attributes, noise data and data reliability in node fault diagnosis of wireless sensor networks, A novel Node Fault diagnosis algorithm for Wireless Sensor Networks based on rough set and optimal probabilistic Neural Networks (RSOPNN).) The reduction of fault diagnosis attribute is solved by rough set from fault sample attribute set, so that redundant fault attribute is removed, redundant attribute is reduced, the influence of noise data on fault diagnosis is reduced, and energy consumption is saved. For multiple attribute reduction selection, the correlation degree between attributes is taken as a measure instead of conventional subjective selection, and the optimal fault diagnosis attribute reduction is determined from multiple reduction to solve the irrationality of subjective selection. The optimal fault diagnosis attribute is used to reconstruct the fault sample as the input of the optimized probabilistic neural network and the fault classification model is established to diagnose the fault. The experimental results show that under different data reliability, the RSOPNN method can effectively delete redundant attributes and noise data from the samples and maintain an efficient fault diagnosis level, which meets the needs of wireless sensor networks.
【作者单位】： 西北大学信息科学与技术学院;西北大学现代教育技术中心;
【基金】：国家科技支撑计划课题(No.2013BAK01B02) 国家自然科学基金(No.61373176) 陕西省重大科技创新专项资金项目(No.2012ZKC05-2)
【分类号】：TP18;TP212.9;TN929.5
【相似文献】
相关期刊论文 前4条
1 李冬妮,王亚沙,李U，
本文编号：2318147
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2318147.html
