中文摘要：

持续载荷飞行模拟器（SLFS），是飞行模拟器坩‘无持续载荷模拟功能”到“有持续载荷模拟功能”的技术跃升，能够实现逼真的模拟实战飞行训练。因为有了持续载荷模拟功能，人体感知模拟环节提升到核心地位，逼真度评估方法必须与之相适应。利用数据挖掘中分类算法对模拟器的开环逼真度和闭环逼真度按不同权重进行层次评估。简化自适应谐振神经网络（SFAM）具有优良的分类性能和更快的收敛速度，但也存在当输入模式类属间差距较大时分类出错以及分类稳定性的缺陷。为了改善网络性能，采用自适应警戒参数的方法对网络进行优化；采用多SFAM分类器集成复合的方法克服分类缺陷。经试验，改进的多SFAM分类器在对影响持续载荷模拟飞行逼真度的各种因素分类过程中表现了较高的分类精度，是一种有效的适应持续载荷飞行模拟器的逼真度评估方法。

英文摘要：

Sustained Load Flight Simulator is an important technical progress from low gravity load to high gravity load. It can realize the vivid simulation training for air combat. For sustained load existence, Body Perception Simulation is becoming to the core position. The classification algorithm was applied to assess the open-loop and closed-loop fidelity of flight simulator by the different grading weight in the data mining. Simplified Fuzzy Adaptive Resonance Theory Map （SFAM） has the advantages of the classification performance and the faster convergence speed. But SFAM presents classification errors and stability error when category of input pattern has large gap. In order to improve network performance, the adaptive alert parameters was applied to optimize the network, and Multi-SFAM compounded classifier was used to overcome the classification defect. The experiments show that improved Multi-SFAM classifiers have better classification accuracy in the classified process of various factors of impacting flight simulator fidelity. The Multi-SFAM compounded classifier is an effective method for the assessment of flight simulation fidelity.