Proper operation of a hydraulic system used in a fatigue test machine (FTM) is crucial. This is because a fatigue test may take well over hours and is not necessarily supervised. Any system failure may result in specimen destruction or experiment failure. In this study experimental data is collected and analyzed to prognoses the hydraulic system. Prognosis may be used to set an alarm level when the predicted values of failure fall within the warning region. This paper presents an approach to predict the operating conditions of a hydraulic system a few increments ahead in time, otherwise known as multi-step ahead (MS). The approach is further validated using experimental data. To do this, applied force on standard aluminum specimen is recorded in time series. Wavelet soft thresholding is used to filter and reduce the effect of noise and sharp edges in the measured applied force data (time series). Embedding dimension and time delay are determined using Cao\\\\\\\'s method and auto mutual information (AMI) technique, respectively. These values are subsequently utilized as inputs for constructing prediction models to forecast the future values of the machines’ operating conditions. The results show that the neural network (NN) prediction model can track the change in machine conditions and has the potential to be used as a machine fault prognosis tool.

Keywords