Prediction of remaining useful life by data augmentation technique based on dynamic time warping

•In the data-driven prognostics, lack of run-to-fail (RTF) data has been one of the bottlenecks that hinders practical applications in the field.•In order to tackle this issue, dynamic time warping (DTW) method is used to augment the data obtained from different operating conditions to the current system, which plays the role of virtual RTF data.•The virtual RTF data are used to train a neural network model to predict the remaining useful life of the current system. In the data-driven approaches for engineering prognostics, the lack of run-to-fail (RTF) data has been one of the bottlenecks that hinders practical applications in the field. In order to tackle this issue, the dynamic time warping (DTW) method is presented to augment the RTF data obtained from different operating conditions or systems to the current system, which plays the role of virtual RTF data under the current condition. Once the virtual RTF data are available, they are used to train a neural network model to predict the remaining useful life of the current system. When multiple RTF data are available with different behavior under different failure modes, an RMSE-based performance criterion is proposed that can adaptively choose the closest match to the current data and use it as the virtual RTF data during the prognosis process. Numerical examples are given to show that the proposed DTW-based data augmentation can predict the RUL with less uncertainty than the conventional neural network model without data mapping.