Improving Machine Learning Diagnostic Systems with Model-Based Data Augmentation - Part A: Data Generation

Various diagnostic systems based on artificial intelligence or machine learning algorithms are already being used today to monitor electrical equipment in power supply systems. The challenge of these data-based diagnostic approaches lies in dealing with the limited fault-condition data available. One possible solution to this problem are data augmentation techniques that generate synthetic data from existing data. In this paper, we develop a model-based data augmentation approach that uses computer-implementable, electromechanical models to generate synthetic data. This approach uses statistical information extracted from the available data to sample model parameters and generate synthetic normal- and fault-condition data. It is shown for vibration measurements of a power and distribution transformer that the proposed model-based data augmentation can generate realistic synthetic normal- and fault-condition data.