碾压施工质量监控的径向神经网络拟合高程研究

Height Fitting by Radial Neural Network for the Construction Quality Control of Face Rockfill Dam

针对面板堆石坝摊铺层表面及测量机器人采集摊铺层离散数据密且精度高等特征,利用径向神经网络方法,采用复二次函数作为径向基,根据数理统计理论和逐步趋近法自适应优选平滑因子拟合摊铺层高程,以达到与实际高程的最佳吻合。以某面板堆石坝的测量机器人采集的数据为例,对该方法进行了验证。结果表明,其拟合精度明显优于常规拟合方法,更加适合面板堆石坝摊铺层表面的高程拟合。

According to the characteristics of face rockfill dam surface and the points collected by the surveying robot are much and their height difference is small,to achieve optimal fitting effect compared with real height,a method was proposed for fitting height by radial basis function（RBF） neural network based on multiquadric（MQ） function and optimal smooth factor by means of mathematical statistics and progressive approach method.To verify its feasibility,we carried out the surface height fitting of face rockfill dam with the data from surveying robot.The results show that the method can get better fitting effect and has higher precision than many common fitting methods.So the method maybe more suitable for surface height fitting of face rockfill dam and can provide the reliable data base to monitor the rolled thickness and construction quality for face rockfill dam.