The calculation of load path for vibration signals, described by accelerations, is often limited by a high implementation effort for simulation, caused by an estimation of existing non linearity e.g., or a need of a measurement campaign which is very time and cost consuming. Furthermore, component vibration tests, often using many different maneuvers to represent reliable and field equivalent load assumptions. With the usage of deep neural networks (DNN) we can simplify and accelerate the simulation and prediction of acceleration time series signals. Additionally, we estimate and compare the confidence of the measurement and the machine learning prediction to evaluate the results.
The developed python code for training and validation of the time series prediction using keras and TensorFlow is easily to adopt for other specific use cases.