Time series method for machine performance prediction using condition monitoring data

Accurate machine performance prediction is crucial to an effective maintenance strategy for improved reliability and to reduce total maintenance cost. In this study, a time series neural network based approach is introduced to achieve more accurate and reliable performance prediction of machine using condition monitoring data source. The proposed time series model utilizes the various measured condition monitoring data at the current and previous inspection marks as the inputs, and the machine output performance as the targets for the model. To validate the model, it considers a two-shaft industrial gas turbine as a case study. The collected condition monitoring data are used to train and validate the proposed model. Results showed that the proposed time series method could predict the performance of the gas turbine power output with more accuracy and better results.