ID |
原文 |
译文 |
6274 |
首先,危险分析技术(system theoretic process analysis,STPA)可以从系统的角度分析影响战机飞行安全的风险源,基于人为可靠性动态分析(information,decision and action in crew,IDAC)模型是目前最全面的人为可靠性分析模型,结合两者的优势提出了IDAC-STPA分析方法。 |
First of all, the risk analysis technique (system theoretic process analysis, STPA) can affect the aircraft flight safety from the Angle of system analysis of risk source, based on the human reliability of dynamic analysis, information, decision and action in crew, IDAC) model is the most comprehensive human reliability analysis model, combined with the advantages of both IDAC - STPA analysis method is proposed. |
6275 |
然后,利用该方法分析战机飞行风险源,并在此基础上建立了飞行员操纵-战机机体模型,叠加不同飞行条件下飞行参量的风险度可以得到相应机动动作的安全谱,在此基础上得到该飞行动作的风险度。 |
Then, by using the method of analyzing aircraft flight risk source, on the basis of pilot - aircraft body model is established, the superposition of different flight conditions risk degree can get corresponding maneuver flight parameters of security, this flight is obtained on the basis of the action of risk degree. |
6276 |
最后,通过对某型战机眼镜蛇机动的安全性分析,发现该型战机飞行中的不安全控制行为和潜在风险。 |
Finally, through the security analysis of a certain type of aircraft cobra maneuver, found that the type of aircraft in flight safety control behavior and the potential risk. |
6277 |
通过安全谱提供一种直观的分析事故演化的方法,在此基础上计算的风险度提供了一种定量分析事故演化的方法,该方法可以为飞行员的飞行训练提供一定的借鉴。 |
Through security spectrum provides an intuitive method of analysis of accident evolution, on the basis of calculating the risk degree of provides a method of quantitative analysis of accident evolution, the method can provide certain reference for the pilots of flight training. |
6278 |
通过对大量数据进行训练并采用分布式深度学习算法可以学习到比较好的数据结构,而传统的分布式深度学习算法在处理大数据集时存在训练时间比较慢或者训练精度比较低的问题。 |
Through the study of the training of a large amount of data and USES the distributed deep learning algorithm can learn better data structure, and the depth of traditional distributed learning algorithm when dealing with large data sets exist training time is slow or low precision problem. |
6279 |
提出Bootstrap向下聚合随机梯度下降(Bootstrap aggregating-down stochastic gradient descent,Bagging-Down SGD)算法重点来提高分布式深度网络的学习速率。 |
Put forward the Bootstrap aggregated stochastic gradient descent down (the Bootstrap aggregating approaches stochastic gradient descent, Bagging - down SGD) algorithm is the key to improve the depth of the distributed network learning rate. |
6280 |
Bagging-Down SGD算法通过在众多单机模型上加入速度控制器,对单机计算的参数值做统计处理,减少了参数更新的频率,并且可以使单机模型训练和参数更新在一定程度上分开,在保证训练精度的同时,提高了整个分布式模型的训练速度。 |
Bagging - Down SGD algorithm through add speed controller, in numerous single model for computing the values of the parameters do statistical processing, reducing the parameters update frequency, and can make the single model training and parameter updating separately, to some extent in to ensure that the training accuracy at the same time, improve the training speed of the whole distributed model. |
6281 |
该算法具有普适性,可以对多种类别的数据进行学习。 |
The algorithm is universal, can be to study a variety of categories of data. |
6282 |
针对存在系统不确定性和外界干扰的接近绕飞阶段跟踪航天器相对目标航天器的姿态与轨道一体化控制任务,设计了一种具有预设性能的鲁棒反演控制器。 |
On existing system uncertainties and external disturbance near orbit phase tracking spacecraft relative target spacecraft attitude and orbit control task, we design a robust inverse controller with the default performance. |
6283 |
该控制器能预先设计系统的稳态与暂态性能,保证相对姿轨跟踪误差满足预先设计的性能指标要求; |
The controller can advance design steady state and transient performance of the system, ensure the relative pose rail track error can satisfy the demands of upfront design performance indicators; |