论文部分内容阅读
针对航空飞行器故障的随机性、多层次性,造成飞行器的状态数据分布受到非线性干扰,采用传统的故障检测时,随机变量的取值范围受到这种干扰的影响,导致收敛性差、漏检率和误检率高的问题。提出深度学习的航空飞行器故障自助检测算法。依据受限波尔兹曼基原理构建深度学习网络的故障检测模型,在模型中引入能量函数,求解隐含层节点和可视节点的激活概率,采用极大似然的学习方法,遍历深度学习网络模型节点,获取最优解,实现对当前故障状态的有效判定,从而完成航空飞行器故障的自助检测。实验结果表明,采用改进算法进行航空飞行器故障自助检测,可以提高故障检测的准确率,减小误差,节约运行成本,具有更强的鲁棒性。
Due to the randomness and multilevel nature of aviation aircraft faults, the distribution of state data of aircraft is nonlinearly disturbed. When using traditional fault detection, the range of random variables is affected by such interference, resulting in poor convergence and missing detection rate And the problem of high false positive rate. A self-detection algorithm for aircraft faults based on deep learning is proposed. According to the limited Boltzmann principle, a fault detection model of deep learning network is constructed. The energy function is introduced in the model to solve the activation probability of hidden nodes and visual nodes. Maximum likelihood learning method is used to traverse deep learning Network model nodes to obtain the optimal solution to achieve the effective determination of the current fault status, so as to complete the self-service detection of aviation aircraft failures. The experimental results show that the improved algorithm can be used to detect the self-detection of aircraft faults, which can improve the accuracy of fault detection, reduce the error, save the operating cost and have more robustness.