Forecasting the Dynamic Response of Rotating Machinery under Sudden Load Changes

This paper analyzes vibration data that shows sudden amplitude changes due to non-stationary load conditions. The data were recorded in a wind turbine that operated under gusty winds and showed high peaks during short periods. Data were analyzed with the auto-regressive integrated moving average (ARIMA) algorithm, and the results were compared to the exponential forecasting method. Other methods have been applied for forecasting vibration data, but the simplicity of this method makes it suitable for rotating machinery with high variable loading conditions. The analysis of the method’s parameters is included in this paper, and the results showed that the optimum configuration depends on the data variations and the existence of significant trends. Forecasting vibration data is challenging; it depends on the source data quality, the preprocessing algorithms, and the deterioration of the mechanical elements. Predictions become less accurate when the machine operates under sudden changes, and evaluating damaging effects caused by the sudden event is difficult to estimate.