Machine Learning-Based Prediction of Frequency Response Function for a Single-Degree-of-Freedom Vibration System

This study investigates the use of deep learning (DL) algorithms to predict the vibration amplitude response of Single-Degree-of-Freedom (SDOF) systems under harmonic excitation. The models were trained on synthetic datasets based on classical vibration theory, with the target being the logarithmic amplitude of the system's response. The DL models showed excellent accuracy, with $\mathbf{R}^{\mathbf{2}}$ values close to $\mathbf{1. 0}$ and low Mean Absolute Error (MAE) across different excitation force levels. The predicted amplitude curves closely matched theoretical values, including resonance peaks and amplitude decay. Minor deviations were observed near resonance, but remained within an acceptable range. This research validates DL for predicting vibration amplitudes in linear SDOF systems and suggests potential for applying DL models to more complex, nonlinear systems.