【作者】 李卉；

【导师】 郭立红；

【作者基本信息】 中国科学院研究生院（长春光学精密机械与物理研究所） ， 机械电子工程， 2012， 博士

【摘要】 现代高技术条件下的战争，对光电防御系统的研制到作战使用都提出了更高的要求，作战效能的高低是衡量武器系统优劣的重要指标。本文分别就光电防御系统在设计阶段和战场使用阶段的作战效能评估方法展开深入的研究。目前，国内外开展作战效能评估多是针对作战飞机、无人侦察机、反舰导弹等方面，本文在对光电防御系统作战效能评估参考较少情况下开展研究，具有实际意义。本文对国内外作战效能评估方法的发展和研究现状进行了归纳和总结，介绍了作战效能评估的基本理论、基本流程、作战效能评估指标体系建立的基本原则。建立了以评估最大作战效能设计方案为目的的作战效能评估指标体系和以评估光电防御系统作战效果为目的的作战效能评估指标体系。在系统研制时，希望得到作战效能最大的光电防御系统，此时作战效能评估问题是一个多属性决策问题，本文提出将灰色局势决策原理与灰模式关联决策原理由实数拓展到区间数；引入专家咨询法与信息熵法相结合的主客观组合赋权法，完善了这两个模型，使之更适用于光电防御系统的作战效能评估。算例证明这两个模型的有效性。在作战使用时，影响作战效能的因素关系复杂，且呈非线性，本文将适用于分类问题的BP神经网络和支持向量机（SVM）两种方法应用于作战效能评估中，提出了对光电防御系统作战效能值进行“分类”的思想，即把待评估数据通过神经网络映射到“很高”、“高”、“一般”、“低”、“很低”五种类别，进一步评估光电防御系统作战效能。经算例分别验证了上述方法的有效性，克服专家决策系统不易修改完善、自适应能力差的缺点。提出采用蝙蝠算法优化BP神经网络初始权值和阈值，得到最优权值和阈值，构造BP神经网络，解决了BP神经网络结构难以确定的问题。针对复杂战场环境下获得的信息具有不完全性和不确定性的问题，采用粗糙集与支持向量机相结合的方法，进行光电防御系统作战效能评估。采用粗糙集理论对属性进行约简，将约简后的特征输入支持向量机，所得分类结果优于未进行属性约简的分类结果，通过算例验证了该方法的有效性和实用性。更多还原

【Abstract】 The war of modern high technology conditions puts forward higher request tophotoelectric defense system from design to combat use. The discretion of the combateffectiveness evaluation is an important index of weapon system quality. Thereforethis dissertation makes further research on combat effectiveness evaluation methodsof photoelectric defense system in design and actual battlefield use stage, respectively.At present, the combat effectiveness evaluation is more for combat aircraft, the UAVand anti-ship missile, etc. This dissertation makes research on combat effectivenessevaluation of the photoelectric defense system under the condition of less references,which has practical significance.This dissertation summarizes the present situation of the development of combateffectiveness evaluation methods at home and abroad, introduces the basic theory ofcombat effectiveness evaluation, basic process, and the basic principles of setting upcombat effectiveness evaluation index system. Then, this dissertation establishes thecombat effectiveness evaluation index system for the purpose of evaluating thedesign scheme of the largest combat efficiency and the combat effectivenessevaluation index system for the purpose of evaluating the fighting effect ofphotoelectric defense system.In system design, we hope to get the photoelectric defense system of the biggest combat effectiveness. At this time, the combat effectiveness evaluation problem is amultiple attribute decision making problem. This dissertation puts forward the greysituation decision-making principle and grey model associated assessment principleexpanded by the real number to the interval number; introduces the combination lawof subjective and objective weighting of experts consultation method andinformation entropy, perfects the two model, makes them more appropriate for thecombat effectiveness evaluation of photoelectric defense system. Examples provethe validity of the two models.When used in combat, the influence factors of the combat effectiveness arecomplex and show a nonlinear relationship. This dissertation will apply to BP neuralnetwork and support vector machine (SVM) method to combat effectivenessevaluation, propose the thought of make the photoelectric defense system combateffectiveness value to different "classification", that is to map the data to the "veryhigh","high","normal""low" and "very low" five categories through the neuralnetwork, further evaluate the photoelectric defense system combat effectiveness. Theexample verifies the validity of the methods above, respectively, which overcome theweakness of the expert decision-making system not easy to modify and the poorquality of the adaptive ability. This dissertation puts forward to optimize the BPneural network weights and threshold by use of the bat algorithm, gets the bestweights and threshold, constructs BP neural network, and solve the problem of BPneural network structure which is difficult to determine.For the problem of the information acquired with incomplete and uncertaintiesunder the complex battlefield environment, this dissertation adopts the combinationmethod of the rough sets and support vector machine (SVM) to evaluate the combateffectiveness of the photoelectric defense system. Using rough set theory to attributereduction, this dissertation inputs the characteristics after attribute reduction to thesupport vector machine. The classification results are better than the classificationresults of no attribute reduction. The calculation example shows that the method iseffective and practical.更多还原