基于ANFIS的目标毁伤等级预测模型

分析了神经网络和模糊推理系统的优缺点,研究了自适应神经模糊推理系统（ANFIS）结构模型及后向传播和递归最小二乘算法相结合的混合算法．在分析了目标毁伤等级主要影响因素的基础上,构建了目标毁伤等级预测ANFIS模型,利用毁伤试验样本数据训练该模型,得到了与实际一致的目标毁伤等级,并将预测结果与基于BP神经网络的预测结果进行了仿真对比分析．仿真结果表明,该目标毁伤等级预测模型能够准确地预测出目标的毁伤等级,并且其预测精度较BP神经网络方法高,为目标毁伤等级预测提供了一种有效的方法．     

An experiment evaluating how the tiny mass eccentricities in spin-stabilized projectiles affect the position of impact points

This study investigates and quantifies some possible sources affecting the position of impact points of small caliber spin-stabilized projectiles (such as 12.7 mm bullets). A comparative experiment utilizing the control variable method was designed to figure out the influence of tiny eccentric centroids on the projectiles. The study critically analyzes data obtained from characteristic parameter measurements and precision trials. It also combines Sobol's algorithm with an artificial intelligence algorithm—Adaptive Neuro-Fuzzy Inference Systems (ANFIS)—in order to conduct global sensitivity analysis and determine which parameters were most influential. The results indicate that the impact points of projectiles with an entry angle of 0° deflected to the left to that of projectiles with an entry angle of 90°. The difference of the mean coordinates of impact points was about 12.61 cm at a target range of 200 m. Variance analysis indicated that the entry angle — i.e. the initial position of mass eccentricity — had a notable influence. After global sensitivity analysis, the significance of the effect of mass eccentricity was confirmed again and the most influential factors were determined to be the axial moment and transverse moment of inertia (Izz Iyy), the mass of a projectile (m), the distance between nose and center of mass along the symmetry axis for a projectile (Lm), and the eccentric distance of the centroid (Lr). The results imply that the control scheme by means of modifying mass center (moving mass or mass eccentricity) is promising for designing small-caliber spin-stabilized projectiles.     

ANN-FRIFS在多传感器信息融合故障诊断中的应用

对于多传感器故障诊断系统,必须根据环境的变化和传感器的状态等合理地运用各传感器的信息。文中设计了一种基于经验规则和模糊推理功能的ANFIS置信度判别器,此判别器能自适应地对各传感器数据综合判别,得到它们相应的故障置信度,进而对多传感器的信息进行融合,提高了诊断的准确度。     

ANFIS网络在舰船维修费用预测中的应用

利用自适应模糊推理系统(ANFIS)的模糊推理能力和学习功能相结合的优点,运用于舰船维修费用预测中.分析了影响维修费用的因素,采用变量投影重要性分析方法(VIP)对影响因素进行评估和筛选,使用ANFIS网络建立模型.通过实际算例进行分析,用历史数据训练ANFIS网络,预测舰船维修费用,在数据样本量小的情况下,较一般神经网络精度有明显提高.