用于倾角监测中的MEMS加速度计补偿方法  

Compensation Method of MEMS Accelerometer for Inclination Monitoring

作　　者：杨小平[1,2] 谭凯 蒋力 刘光辉 李哲宏 Yang Xiaoping;Tan Kai;Jiang Li;Liu Guanghui;Li hehong(School of Information Science and Engineering,Guilin University of Technology,Guilin 541004,China;Guangri Key Laboratory of Embedded Technologyand Intelligent System,Guilin University of Technology,Guilin 541004,China;Guangri Zhuang Autonomous Region Geological Environment Monitoring Station,Nanning 530029,China;Guilin Saipu Electronic Technology Limited Company,Guilin 541004,China)

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541004 [2]桂林理工大学广西嵌入式技术与智能系统重点实验室,广西桂林541004 [3]广西壮族自治区地质环境监测站,南宁530029 [4]桂林赛普电子科技有限公司,广西桂林541004

出　　处：《微纳电子技术》2022年第9期911-919,965,共10页Micronanoelectronic Technology

基　　金：国家高新技术研发计划(863计划)(2013AA12210504);广西壮族自治区科技攻关项目(AC1638012,AD18281068);广西壮族自治区南宁市青秀区科技局科技计划(RZ19100041)。

摘　　要：针对在山体滑坡倾角监测中微电子机械系统(MEMS)加速度计存在误差的问题,传统方法的补偿效果欠佳,且无法很好地对时间序列数据进行分析。为了提高山体姿态监测的精度,采用了一种基于一维卷积神经网络(1D-CNN)与长短期记忆(LSTM)网络相结合的MEMS加速度计误差补偿方法。将采集到的加速度数据转换成角度数据,然后通过1D-CNN与LSTM网络模型进行训练,设计了误差补偿的硬件系统,从而实现实时误差补偿。实验结果表明,与卡尔曼滤波和反向传播(BP)神经网络相比,X轴的均值和标准差分别为0.000 057°和0.000 033°,误差下降了一个数量级,说明1D-CNN与LSTM相结合的网络对MEMS加速度计具有更好的补偿效果,为将来应用在山体滑坡倾角监测中奠定了基础。For the error problem of micro-electromechanical system(MEMS) accelerometer in mountain landslide inclination monitoring, the compensation effect of traditional methods is not good, and the time series data cannot be well analyzed. To improve the accuracy of mountain attitude monitoring, a MEMS accelerometer error compensation method based on one-dimensional convolutional neural network(1 D-CNN) and long short term memory(LSTM) network was used. The collected acceleration data were converted into angle data, and then trained by 1 D-CNN and LSTM network models. The hardware system of error compensation was designed to realize real-time error compensation. The experimental results show that compared with Kalman filter and back propagation(BP) neural network, the mean and standard deviation of the X-axis are 0.000 057° and 0.000 033°, respectively, and the error is reduced by an order of magnitude, indicating that the network of 1 D-CNN combined with LSTM has better compensation effect for the MEMS accelerometer, and lays a foundation for future application in landslide inclination monitoring.

关 键 词：微电子机械系统(MEMS) 加速度计 误差补偿 一维卷积神经网络(1D-CNN) 长短期记忆(LSTM)网络 倾角监测 

分 类 号：TH703[机械工程—仪器科学与技术] TH824.4[机械工程—精密仪器及机械]