摘要
针对常规故障预测方法难以分析复合故障的情况下各个故障对系统的交互作用、难以分析装备数据复杂特征、难以实时准确预测故障等现状,对现代大数据和人工智能方法应用在故障预测领域进行研究,提出基于深度学习的故障预测技术,将系统故障预测可分为动态预测和静态预测;利用深度学习算法处理装备状态监测和试验验证获得的海量故障数据,通过故障模型训练、故障特征识别、故障演化规律获取来对系统进行在线动态预测;针对软件故障突变特性,利用软件质量特征属性进行静态故障预测;同时,提出使用开源深度学习框架TensorFlow进行系统研制方法;通过基于深度学习的故障预测技术,能够提高装备故障预测能力。
To solve the problem analyzing the effect of complex faults and the complex characteristics of the equipment data, to predict the fault in real time, this paper presents a scheme of fault prediction based on deep learning. On--line dynamic fault prediction is carried out on the basis of mass data obtained from equipment condition monitoring and test verification by methods of fault model training, fault feature recognition, fault evolution. Based on the attribute information of software quality, software static fault can be predicted. Fault prediction system can be developed based on TensorFlow to improve the fault prediction ability of equipment.
出处
《计算机测量与控制》
2018年第2期9-12,共4页
Computer Measurement &Control
关键词
深度学习
故障预测
故障演化
软件静态故障预测
deep learning
fault prediction
fault evolution
software static fault prediction
