水轮发电机组轴系故障趋势预测与评估.pdfVIP

  华  中  科  技 大 学 硕 士 学 位 论  文  Abstract Songjianghe cascade is northeast area seven five during the hydropower construc- tion projects, located in southeastern Jilin province a mountainous area, in recent years, Songjianghe cascade turbine increasingly large-scale, complex, automation, because the fault led to the unexpected shutdown will result in serious losses. Fault prediction tech- nology can effectively avoid accident occurrence and development, to provide advanced predictive maintenance to provide scientific methods. Therefore, this paper to Songjianghe cascade hydropower station as the background, the turbine shaft fault trend prediction method and unit operation state evaluation methods are studied, results of fault diagnosis and expert decision support system research and development is an important part of the scientific and reasonable evaluation unit, health status, realize the unit condition mainte- nance and predictive maintenance lay a solid foundation. This paper expounds the turbine generator shaft structure, and the fault characteristic undertook thorough analysis. Research Based on time series autoregressive model analysis of shafting fault prediction method, the frame vibration and bearing temperature is pre- dicted and the results were analyzed according to the shafting status data structures, based on BP neural network unit shafting fault prediction method. In order to overcome the BP nerve network training slow convergence speed, easy to fall into local minimum point, LM algorithm is applied to replace the traditional gradient method, so as to improve BP neural network prediction model of efficiency and accuracy. BP neural network prediction model and autoregressive prediction models were compared and analyzed, the results show the improved BP neural network model to predict the effect better than the auto- regressive model. In shafting