Evaluation Metrics: A Review of Decision Methods for Vibration Based Condition Monitoring

Fault detection is important for providing the reliability as well as efficiency of many different systems and processes. Choosing an appropriate assessment metric is important because it has a direct effect on how a model performance is evaluated and compared for detecting defects and other areas. Choosing an improper evaluation metric might lead to inaccurate assessments of model performance and suboptimal decisions regarding the performance of model. In this review article, we reviewed a variety of evaluation metrics used for vibration based fault detection. Our evaluation includes metrics created for machine learning algorithms, visualisation based techniques, and for error estimation methods. We provide a review of evaluation criterias for defect detection by combining knowledge from the literature, which helps with selection and decision-making of assessment methods in a variety of methods.