A comparative study of power-based health assessment models

In the industrial energy monitoring domain, several platforms are offered to acquire real-time power data and monitor energy consumption at tool, component, machine and system levels. The motivations of employing such platforms are majorly cost-effective purposes or environmental issues. This paper proposes and applies a power-based approach with non-intrusive sensing technique to evaluating machine health status. Data of effective power and reactive power are acquired synchronously and segmented based on machine cycles. Exhaustive feature extraction and reduction is conducted to reduce the dimension of the datasets. Two health assessment models are then used to model baseline behavior from the training set and measure the distance between training and testing set to provide health status information of the testing set with respective distance metrics. The models are applied in two case studies on two machines in the same manufacturing line.