基于WSN与异常数据识别的拉线动态监测方法研究

Research on dynamic monitoring method of stay wire based on WSN and abnormal data identification

为了提升干字塔监测的数字化水平并保证施工现场的安全,文中对干字塔的拉线状态监测技术进行了研究。该研究在蓝牙、ZigBee等无线传输协议与拉力传感器的基础上,搭建了大规模无线传感网络(WSN),实现了监测数据的无线传输。针对WSN网络所采集的数据特点,对BP神经网络加以改进,且引入状态转移概率及一种基于置信区间思想的残差判定模型来进行WSN传输数据的校准,从而提升了模型的训练效率,避免了因WSN网络传输数据精度不足而影响网络泛化性能的问题。在实际工程数据集上进行的仿真结果表明,较传统的BP神经网络,改进后算法的平均训练时长降低了27.53%,迭代次数下降了26.61%,TPR提升了2.8%,FPR下降了1.98%,更适用于WSN传感网络的数据识别。

In order to improve the digital level of the monitoring of the dry type tower and ensure the safety of the construction site,this paper studies the pull wire condition monitoring technology of the dry type tower.Based on Bluetooth,ZigBee and other wireless transmission protocols and tension sensors,a large-scale Wireless Sensor Network(WSN)is built to realize the wireless transmission of monitoring data.According to the characteristics of data collected by WSN network,BP neural network is improved.The state transition probability and a residual judgment model based on the idea of confidence interval are introduced to calibrate the WSN transmission data,which improves the training efficiency of the model and avoids the insufficient accuracy of WSN transmission data affecting the generalization performance of the network.The simulation results on the actual engineering data sets show that,compared with the traditional BP neural network,the average training time of the improved algorithm is reduced by 27.53%,the number of iterations is reduced by 26.61%,TPR is increased by 2.8% and FPR is reduced by 1.98%,which is more suitable for data recognition of WSN sensor network.