基于深度学习的体系作战效能智能评估及优化

采用仿真模拟评估方法进行体系作战效能评估包括配置想定、多次运行仿真系统和调用效能评估模型计算多个复杂步骤。针对此流程复杂、耗时长的问题,提出一种基于深度学习回归思想的体系作战效能智能评估方法,引入基于进化策略的作战效能优化流程。以全连接深度回归网络为预测模型,通过多个数据集的试验结果给出网络隐层数和样本量选择的指导原则。进一步结合遗传算法调整网络输入得到优化输出,迭代性能较好。以某仿真系统为例初步验证了其有效性和使用价值。

Intelligent Effectiveness Evaluation and Optimization on Weapon System of Systems Based on Deep Learning

In the process of conventional WSoS (Weapon System of Systems) operational effectiveness evaluation, multiple complex steps including reconfiguring the scenario, rerunning the WSoS simulation system, and conducting effectiveness evaluation model calculations are taken. For the problems of conventional effectiveness evaluation method, such as the complex process and time-consuming, the intelligent evaluation method of WSoS combat effectiveness based on machine learning regression is proposed, and the optimization process of combat effectiveness based on evolutionary strategy is introduced. Taking the full connection depth regression network as the prediction model, the guiding selection principles of network hidden layer number and sample size are given on the basis of the test results of multiple data sets. The optimal output is obtained by the genetic algorithm to adjust the network input, and the high iterative performance is got. Case study of a simulation system preliminarily verifies the effectiveness and value of the prediction model and the optimization method.