A machine learning based data-driven model is developed in this research as a digital twin of a pipeline resonance bending fatigue testing equipment for monitoring and prediction of anomaly. In this work, the first-order vibration mode of the pipeline, the stress measures of the test pipeline at different frequencies and angles, and the stiffness of the exciter bearing at the pipeline fulcrum are achieved first based on the partial differential equations considering resonance bending of the pipeline due to vibration. Then the three-dimensional model of the whole testing equipment is established, and the boundary conditions are defined for finite element analysis. The data set is subsequently obtained based on simulation through finite element analysis, and the data are used to train the model based on the support vector machine (SVM) method for prediction of anomaly. Effectiveness of the model is verified through tests for monitoring and prediction of anomaly.