故障预测模型的评价方法研究

Study on the Evaluation Methods of a Prognostics Model

在复杂系统健康管理的实现过程中,预测方法、预测模型起着非常重要的作用。评价一个预测模型的性能是对其优选的基本,也是推动故障预测良性发展的动力。为定量比较各种方法的预测效果,必须制定一套切实可行的评价标准体系,以便对预测效果进行科学、全面的衡量和评价。研究并建立了一套性能评价指标体系,并通过对一种基于神经网络的非线性组合预测模型的实例评价,验证了评价指标体系的有效性及适用性。

For health management of complex system, prognostics methods and models play a very important role. Therefore, evaluation of a prognostics model performance is fundamental for choosing a suitable model. It is also the driving force to promote the fault prognostics development. In order to compare the effect of various methods, a practical evaluation index system must be established to weigh and evaluate the combination of fore- casts scientifically and comprehensively. A set of performance evaluation index system is researched and built, and through specific examples of evaluation of the non-linear combination forecasting model based on neural network, the effectiveness and applicability of the evaluation index system are verified.