Prognostics Of An Industrial Gas Turbine: A Time Series Forecasting Data Driven Approach

Abstract

Prognostics assesses and predicts future machine health, which includes detecting incipient failures and predicting remaining useful life. Several studies approached prognostics as a time series forecasting problem. The main goal of this study is to evaluate the performance of a set of methods in the prediction of future values from a dataset of time series collected from sensors installed on an industrial gas turbine. Methods tested include the use of ensembles of feedforward neural networks, ensembles of long short-term memory networks, exponential smoothing, and Auto Regressive Integrated Moving Average (ARIMA) models. Results show that the use of ARIMA models to forecast on the dataset is the best default method to apply, and is the only method that consistently beats a simple naïve no-change model.

Keywords

• prognostics
• time series
• forecasting
• neural networks
• arima