【作者】 杨帆；

【导师】 张为公；

【作者基本信息】 东南大学 ， 仪器科学与技术， 2016， 博士

【摘要】 本文围绕车辆地面通过性评价系统中试验方法、理论模型及评价方法三项关键技术进行研究,以实现军用车辆地面通过性的有效评价,切实提高军用车辆地面通过性能。针对其中所涉及的若干关键问题,首先研究了车辆地面通过性实车试验方法,提出并搭建了实车测试系统,实现了轮壤相互作用关键参数的有效测量；在实车试验的基础上,开展了车辆地面通过性理论模型研究,围绕典型理论模型—半经验模型进行了深入研究,致力于其中滑转沉陷问题及土壤特性参数实时获取问题的研究,以此来改进半经验模型并提高其模型精度及实用性,并用于预测地面通过性评价指标；通过引入统计学习方法,提出了统计预测模型,为准确预测车辆地面通过性评价指标提供了另一种可行思路；最后对车辆地面通过性评价方法进行了深入讨论,提出了多评价指标融合评价方法,并构建了车辆地面通过性评价体系,实现了通过性的有效评价,并用于指导实际评价工作。主要研究内容及成果包括：(1)针对车辆地面通过性试验方法主要局限于模型试验而实车试验缺乏的现状,提出并搭建了一套以车轮力传感器为核心,辅以轮速传感器、GPS等车载传感器的实车测试系统,结合实际应用需求,有针对性地进行了系统软硬件设计。在此基础上,设计并进行了硬地面与软地面场地试验,试验结果充分验证了实车测试系统的有效性及可靠性,试验中获取的大量实车试验数据为车辆地面通过性理论模型研究奠定了坚实的数据基础。(2)针对半经验模型中已有滑转沉陷模型适用性不明确的现状,基于实车试验数据,对典型滑转沉陷模型在军用车辆应用中的适用性进行了深入探讨。结合高斯牛顿迭代法及遗传算法,提出了滑转沉陷模型参数估计方法,实现了模型参数的准确估计；在此基础上,进行了模型适用性分析,选定了适用于本应用的最优模型,切实提高了半经验模型的模型预测精度,为实现车辆地面通过性评价指标的准确预测打下了基础。(3)为实现土壤特性参数的实时获取,围绕土壤特性参数在线估计方法进行了深入研究,提出了一种基于地面类型识别的土壤特性参数在线估计模型。试验验证结果表明,所提出模型的准确性和实时性都能够得到保证,配合实车测试系统,很好地实现了土壤特性参数的实时获取,从而提高了半经验模型的实用性,为通过性评价工作提供了准确的数据来源。(4)针对车辆地面通过性理论模型尚存在局限性的现状,以实车试验数据为基础,引入了统计学习方法,提出了一种基于改进相关向量机的挂钩牵引力统计预测模型。试验结果充分证明了统计预测模型的可行性与有效性,对比半经验模型,不仅降低了模型复杂度且提高了模型预测精度,为实现车辆地面通过性评价指标的准确预测提供了另一种可行思路。(5)针对车辆地面通过性评价策略及体系不完善的现状,分析并选定了军用车辆地面通过性实用评价指标,提出了一种基于模糊推理理论的多指标融合评价方法,给出了车辆地面通过性的综合性评价；在总结及串联全文研究内容的基础上,构建了较为完善的军用车辆地面通过性评价体系,实现了通过性的有效评价,对实际评价工作具有重要的指导意义。更多还原

【Abstract】 This paper focuses on the three key technologies of the evaluation system of vehicle trafficability on soft terrain, which are the test method, theoretical model and evaluation methodology. It aims to effectively evaluate and improve military vehicles’trafficability on soft terrain. To solve the involved key problems, firstly a research is done on the real vehicle test method, and a real vehicle test system is proposed and constructed to effectively measure the key parameters of wheel-soil interactions. On the basis of real vehicle tests, the theoretical models are deeply studied, especially the typical semi-empirical model. The problems of the slip-sinkage effect and the real-time acquisition of terrain parameters are focused on to improve the model precision and practicability.Then the improved semi-empirical model can be used to predict the evaluation indexes of vehicle trafficability. With the introduction of statistical learning methods, a statistical prediction model is developed, and it provides another practicable insight for the accurate prediction of evaluation indexes of vehicle trafficability. Finally, an in-depth discussion is done on the evaluation methodology of vehicle trafficability, and a multi-index evaluation method is proposed. Meanwhile, an evaluation system is set up to realize the effective evaluaton of vehicle trafficability, which is also helpful for the actual evaluation work. The main research work and achievements are summaried as follows:(1) For the test methods of vehicle trafficability on soft terrain are mainly limited to model tests, and for the lack of real vehicle tests, a real vehicle test system is proposed and constructed, which takes the wheel force transducer as the core and is supplemented by wheel speed sensor and GPS. Considering the actual application requirements, the hardware and software of the system are specifically designed. On this basis, field tests on rigid and soft pavements are designed and conducted. The effectiveness and reliability of the system is fully validated by the test results. A large number of real vehicle test data gained lays a solid foundation for the theoretical model study of vehicle traficability.(2) For the applicability of the existing silp-sinakge models in the semi-empirical model is not clear, typical silp-sinakge models’applicability in military vehicle applications is deeply analyzed based on real vehicle test data. An estimation model of the slip-sinkage model parameters is proposed with the combination of gauss-newton method and genetic algorithm. From the results, the model parameters are accurately estimated. On this basis, the applicability of slip-sinkage models is analyzed. Finally, an optimal model suitable for practical applications is confirmed, which helps to improve the prediciton accuracy of the semi-empirical model and lay the foundation for the accurate prediction of evaluation indexes of vehicle traficability.(3) To realize the real-time acquisition of terrain parameters, a research work is done on the online estimation methods. In this paper, an online estimation model is proposed based on terrain classification. It is indicated by the test results that the accuracy and instantaneity of the proposed model are both guaranteed. Cooperating with the real vehicle test system, the terrain parameters can be gained real-timely, which effectively improves the practicability of the semi-empirical model and provides accurate input data sources for the evaluation work of vehicle traficability.(4) For the limitations of the exsiting theoretical models of vehicle traficability, with the introduction of statistical learning methods, a prediction model of drawbar pull is proposed based on an improved relevance vector machine and real vehicle test data. Test results indicate that the statistical prediction model is feasible and effective. Compared with the simi-empirical model, the model complexity is reduced and the prediction accuracy is improved, which provides another practicable insight for the accurate prediction of evaluation indexes of vehicle trafficability.(5) For the imperfection of the evaluaton methodology and system of vehicle traficability, a series of practical evaluation indexes are confirmed after analysis; a multi-index evaluation method is proposed based on fuzzy inference theory to make a comprehensive evaluation of vehicle traficability. Moreover, on the basis of a summary and combination of the overall research contents, a relatively thorough evaluation system of vehicle traficability is proposed. It helps to realize the effective evaluation and it is of important guiding significance for the evaluation work.更多还原