基于时间序列分析的装备可用度预测方法

An Equipment Availability Prediction Method Based on Time Series Analysis

摘要:

阐述了装备可用度预测的重要性,并以此为需求牵引,构建了具有非线性、非平稳的装备可用度时间序列。基于奇异值分解滤波算法,将原始序列分解为趋势成分和随机成分,分别应用粒子群训练的径向基神经网络模型和函数系数自回归模型进行组合预测,充分体现了 2类模型的各自的优势。实例分析,验证了模型和算法的有效性。实验与应用结果表明,该组合方法的预测性能和效果比单一使用 RBF和 FAR进行预测更好。

Abstract:

The significance of availability prediction of equipment system was proposed. The time series model of availability was put forward, which had characters of non-linearity and non-stationary. The trend element and random element of availability were decomposed based on the singular value decomposition filtering algorithm. Then, the radial basis function neural network trained by particle swarm optimization algorithm and functional co. efficient auto regressive model were proposed to solve the problem. This combining method showed the superiori. ty of each model, and its effectiveness was verified through example. The results of the prediction show that the function and effect of combining method was better than separate employment of RBF or FAR, and provided an effective means for the availability research.