基于BP神经网络的雨滴谱仪设计

王珲 葛益娴 刘清惓 韩上邦 孙启云
关键词: 雨滴直径; 输出电压; 导电圆环; BP神经网络; 梯度下降法; Sigmoid函数
中图分类号: TN02?34; P412.13 ? ? ? ? ? ? ? ? ? 文獻标识码: A ? ? ? ? ? ? ? ? ? ? 文章编号: 1004?373X(2019)03?0180?04
Abstract: In order to reduce the size, weight, and cost of the traditional raindrop spectrometer, and improve the measurement accuracy of the raindrop spectrum, a raindrop spectrometer composed of new raindrop spectrum probe, Cortex?M3 ARM processor, and high?precision and low?noise measurement circuit is proposed, which is based on isometric conductive ring structure. The output voltage difference caused by different raindrop diameter is used in the raindrop spectrometer for measurement. The Sigmoid function is taken as the transfer function, and the gradient descent method is used to establish the BP neural network model of output voltage and raindrop diameter of the raindrop spectrometer. The algorithm based on this model is embedded into the ARM processor to obtain the raindrop diameter. The raindrop spectrometer can realize the observation of raindrop whose diameter within 2.67~5.56 mm, and the measurement error is less than ± 0.1 mm.
Keywords: raindrop diameter; output voltage; conductive ring; BP neural network; gradient descent method; Sigmoid function0 ?引 ?言
雨滴谱反映了雨滴数量随直径的分布情况,是降雨观测的重要组成部分。文献[1]在研究层状云和对流云降水雨滴谱特征时发现,层状云降水的谱宽小于对流云降水,且对于这两种降水而言,降雨强度越大,降水中的大粒子越多。文献[2]在评估人工降雨系统效能时发现,通过提高人工降雨的雨滴直径和自然状态的相似性,可以提高人工模拟降雨的质量。此外,雨滴中值粒径和坡面产沙量具有较高的相关性,雨滴中值粒径越大,土壤侵蚀越严重[3]。因此,雨滴谱的观测具有重要意义。
目前,应用较为广泛的雨滴谱仪有冲击型雨滴谱仪Joss?Waldvogel(JWD)和Parsivel激光雨滴谱仪。JWD冲击型雨滴谱仪由传感器、处理器和电缆组成,根据雨水滴落在传感器表面产生的电脉冲的大小测量雨滴直径。其观测结果受雨滴的大小、速度、形状以及背景噪声的影响,会低估暴雨中小雨滴的数量[4]。Parsivel激光雨滴谱仪主要由发射机、接收机、控制、运算和存储电路组成,根据雨滴通过采样区域时的光信号的变化测量雨滴直径。激光雨滴谱仪使用简单,维护方便,但是体积大、成本高,存在重叠误差[5]。为了减小雨滴谱观测仪器的体积、重量和成本,提高测量精度,本文设计一种基于BP神经网络算法的雨滴谱仪,利用等距导电圆环结构的雨滴谱探头、Cortex?M3 ARM处理器及高精度低噪声测量电路实现对雨滴的观测。


5 ?结 ?语
针对目前市面上雨滴谱仪存在体积大、成本高等问题,结合BP神经网络算法,本文设计了一种基于等距导电圆环结构的新型雨滴谱探头、Cortex?M3 ARM处理器及高精度低噪声测量电路的雨滴谱仪。实验结果表明,该雨滴谱仪实现了对直径为2.67~5.56 mm雨滴的观测,测量误差小于±0.1 mm。但该雨滴谱仪仍然存在一些问题,如探头表面积水也会影响雨滴观测。因此,比较和分析热蒸发、超声波雾化和机械振动初始等方法,选择合适的方式对探头除水是本文未来的研究重点。
参考文献
[1] SUBRATA K D, MAHEN K, KAUSTAV C, et al. Raindrop size distribution of different cloud types over the Western Ghats using simultaneous measurements from micro?rain radar and disdrometer [J]. Atmospheric research, 2017, 86: 72?82.
[2] 郭东静,陈锡云,马晶,等.基于雨滴谱的人工降雨系统效能评估[J].水土保持学报,2015,29(4):85?90.
GUO Dongjing, CHEN Xiyun, MA Jing, et al. Reliability ana?lysis for artificial rainfall system based on raindrop spectrum detector [J]. Journal of soil and water conservation, 2015, 29(4): 85?90.
[3] 盛世博,王瑄,盛思远,等.沈阳地区天然降雨雨滴特征对坡面产沙量的影响[J].水土保持研究,2017,24(2):12?16.
SHENG Shibo, WANG Xuan, SHENG Siyuan, et al. Influence of natural rainfall on slope sediment yield in Shenyang region [J]. Research of soil and water conservation, 2017, 24(2): 12?16.
[4] 朱亚乔,刘元波.地面雨滴谱观测技术及特征研究进展[J].地球科学进展,2013,28(6):685?693.
ZHU Yaqiao, LIU Yuanbo. Advances in measurement techniques and statistics features of surface raindrop size distribution [J]. Advances in earth science, 2013, 28(6): 685?693.
[5] 胡子浩,濮江平,张欢,等.Parsivel激光雨滴谱仪观测较强降水的可行性分析和建议[J].气象科学,2014,34(1):25?31.
HU Zihao, PU Jiangping, ZHANG Huan, et al. Feasibility analyses and recommendations of parsiveloptical distrometer in measurements of strong precipitation [J]. Journal of the meteorological sciences, 2014, 34(1): 25?31.
[6] ALSHAWABKEH A N, SHEAHAN T C, WU X. Coupling of electrochemical and mechanical processes in soils under DC fields [J]. Mechanics of materials, 2004, 36(56): 453?465.
[7] 焦李成,杨淑媛,刘芳,等.神经网络七十年:回顾与展望[J].计算机学报,2016,39(8):1697?1716.
JIAO Licheng, YANG Shuyuan, LIU Fang, et al. Seventy years beyond neural networks: retrospect and prospect [J]. Chinese journal of computers, 2016, 39(8): 1697?1716.
[8] 罗云芳.一种BP神经网络校正算法的实验室智能温控系统研究[J].现代电子技术,2015,38(20):84?91.
LUO Yunfang. Research on laboratory′s intelligent temperature control system based on BP network correction algorithm [J]. Modern electronics technique, 2015, 38(20): 84?91.
[9] 王蒙,常胜,王豪.一种自适应训练的BP神经网络FPGA设计[J].现代电子技术,2016,39(15):115?118.
WANG Meng, CHANG Sheng, WANG Hao. FPGA based design of BP neural network with adaptive training [J]. Modern electronics technique, 2016, 39(15): 115?118.
[10] 舒小健,高太长,刘西川,等.基于降水为物理特征测量仪的雨滴形状观测与分析[J].气象,2017,43(1):91?100.
SHU Xiaojian, GAO Taichang, LIU Xichuan, et al. Observation and analysis of raindrop shape based on the precipitation micro?physical characteristics sensor [J]. Meteorological, 2017, 43(1): 91?100.