基于轮地作用参数和PLSDA方法的月壤力学性能评估

为评估在轨月球车周边环境月壤力学性能,以月球车辙信息、轮上载荷和滑转率作为基本参数,提出了16个二元及三元标识量,结合偏最小二乘判别方法(PLSDA)建立评估月壤力学状态模型。根据容重对模拟月壤力学状态进行分级,分别为松软状态、自然状态和紧实状态。应用轮壤相互作用试验台共采集247组试验数据,每种模拟月壤状态的试验数据按照2:1比例随机划分为校正集样本和预测集样本,最终校正集和预测集样本个数分别为166和81个。考虑到原始数据值相差大和所提出部分标识量包含冗余信息的特点,因此在建立PLSDA模型时,应用均值中心化预处理方法对原始数据进行数据预处理,并优选10个标识量建立识别月壤力学状态的PLSDA模型,对应的校正集的准确率和预测集的准确率分别为90.96%和90.12%。结果表明,应用PLSDA方法并结合月球车的车辙信息、轮上载荷和滑转率以及优选标识量所建立的评估月壤力学状态判别模型,其计算快速准确,可用于月球车在轨评估车轮前方月壤的力学特性和通过性能评估。

Mechanical performance estimation of lunar soil using wheel-soil interaction parameter and PLSDA

In order to estimate the mechanical performance of lunar soil nearby a lunar roving vehicle and terrain traversability, partial least square discriminant analysis (PLSDA) method is adopted to estimate mechanical parameters of lunar soil combined with 16 binary and ternary signatures which are proposed based on the available information of a lunar roving vehicle, such as wheel load, slip ratio and wheel sinkage. In this paper, mechanical state of lunar soil simulant is divided into three kinds of bulk density:soft, normal and hard. A total of 247 test data of wheel-terrain interaction test bed are collected. Then these data are assigned to calibration set and prediction set at the ratio of 2:1 randomly, and the number of calibration set and prediction set are 166 and 81, respectively. Considering that there are obvious differences among raw data and some signatures contained useless information, the optimized model of discriminant analysis is developed using the data pretreatment of mean center and 10 optimized combination of signatures, and the accuracy of calibration set and prediction set are 90.96% and 90.12%, respectively. The result demonstrates that wheel load, slip ratio, wheel sinkage and selected signatures combined with PLSDA method are suitable to estimate the mechanical parameters of lunar soil quickly. And the method can be used for trafficability prediction of lunar roving vehicle on line.