基于免疫和神经网络的航空设备故障预测研究

摘要

To reduce the entire life cycle cost of weapon system and enhance their affordability, technology of fault prediction is studied for aeronautic equipment.The immune algorithm is used to ameliorate the activation function of hide layer. Then Immune Neural Network (INN) is got to track and predict the characteristics parameters of equipments.Results show that the improved neural networks can archive fault prediction 3 hours before the time point of faults respectively, and the networks' performances are improved significantly compared with the BP neural network.%为解决武器装备全寿命周期费用高、经济可承受性差的难题,开展航空装备的故障预测技术研究.采用免疫算法改进隐含层激励函数得到免疫神经网络,用以进行装备特征参数的跟踪预测,结果表明免疫算法改进的神经网络可在故障前3小时实现预测,较BP网络性能有较大改善.