摘要
开展月球探测对于提升我国综合实力具有重要意义.按照计划我国将在2017年左右完成月球采样并返回地球的目标.目前国内各科研院所对采样机具的研究多集中在钻取机具的设计及其仿真模拟上,对表层取样机具研究较少.基于表层取样研发了一套由直流电机驱动,并能通过检测其电流间接测算挖取运动扭矩的试验机构.利用该机构在6种不同的模拟月壤中进行不同试验参数的挖取试验后可知,在不同的试验条件下挖取机构承受的扭矩变化趋势大致相同,并能由4个特征点进行描述.4个特征点的取值随试验参数的不同而改变.完成试验后将试验数据进行归一化处理后导入BP神经网络进行学习和训练,建立了以运动参数(运动角度、机构悬挂高度)、模拟月壤类型(内聚力、内摩擦角)、模拟月壤密实程度(容积密度、孔隙度、相对密实度)为输入量,机具承受扭矩为输出量的神经网络模型.通过与实测数据对比可证明本文建立的BP神经网络挖取力学模型具有很高的拟合和预测精度.
It is very important to start the lunar exploration for the sake of the buildup of comprehensive strength of China. According to a national plan, the main task of Chang＇E-3 probe is to sample on the moon and return to the earth. At present, the researches on sampling tools are focused on the designing and simulating the drilling tool and there is little research on the surface sampling tools. A simulated lunar surface sampling tool, which is driven by a DC motor and could measure the torque during the experiments by using the current sensor, is designed for the experiments. Based on six types of lunar soil simulant and parameters, the experimental data shows that the change tendency of the tool torque under different conditions is nearly the same, and the torque curve could be described by four-feature points, whose values change with the different experiments parameters. All normalized data is imported to the BP neural network for learning and training in order to establish the model. The input parameters are divided into three types, including motion parameters （angle and hanging height）, type parameters （cohesion and internal friction angle）, and dense parameters （density, porosity and relative denseness） and the output parameter is the torque. By comparing the measured data and the output data of the model, it shows that accuracy in both fitting and predicting of the BP neural network is extremely high.
出处
《地球科学（中国地质大学学报）》
EI
CAS
CSCD
北大核心
2013年第6期1363-1370,共8页
Earth Science-Journal of China University of Geosciences
基金
北京空间飞行器总体设计部"采样机具与月壤相互作用平台的研究"(No.201205G007)
关键词
表层采样
模拟月壤
扭矩
神经网络
工程地质
surface sampling
lunar soil simulant~ torque
neural networks
engineering geology.
