基于混合遗传BP神经网络的城市系统作战能力评估

针对现有城市系统作战能力评估方法较少的问题,利用反向传播(back propagation,BP)神经网络在能力评估方面所具有的自适应、自学习、强容错性和泛化映射等优势,建立了评估指标体系并给出了指标的隶属函数。通过模拟退火遗传算法(simulated annealing and genetic algorithm，SAGA)优化 BP 神经网络的连接权重和阀值,弱化了指标评价中的人为因素,提高了评价结果的准确性、客观性和权威性,有效解决了传统遗传算法和 BP 神经网络易陷入局部极小值、收敛速度慢和抗干扰能力差等问题。仿真实例验证了该方法对城市系统作战能力评估的可行性和有效性。

Combat capability evaluation of city system based on#br# mix-genetic algorithm BP neural network

Aiming at the problem of less combat capacity evaluation methods of city system, the index system is brought forward and the subordinate function of each index is given. An assessment model based on BP neural network whose thresholds and connection weights are optimized by the simulated annealing and genetic algorithm (SAGA) is proposed to solve the problems. And the problems of the classical genetic algorithm and BP neural network trapping into the local minimum point, low convergence speed and with bad anti-jamming ability are solved. Using the characteristic dominances of self-adaptive, self-learning, efficient fault tolerant and wide mapping of BP neural networks, the model can weaken human factors of the index to improve the accuracy, objectivity and authority of assessment results. According to the simulation, the feasibility and validity of city system’s campaign capability assessment by this method are verified.