Non-Intrusive Electrical Monitoring for Real-Time Estimation of the Production Parameters in a Sheet Metal Stamping Line

Metal forming processes play a key role in modern manufacturing, but they are characterized by high energy consumption and low overall efficiency. In this context, precise methods for monitoring the operational state and cycle-dependent metrics of manufactured parts are essential to implement energy optimization strategies. This article presents a data-driven and non-intrusive methodology to identify, in real time, the part under production and to estimate both cycle time and energy consumption per part. The method relies exclusively on electrical measurements taken at the main switchboard and at the first process-stage switchboard. These signals are used to calculate electrical quantities such as root mean square (RMS) current and active power, and a machine-learning (ML) approach is proposed to automatically identify the part in production. To this end, time-domain features are extracted directly from the signals, while time-frequency features are extracted using Continuous Wavelet Transform (CWT). These features are employed to train Support Vector Machine (SVM) classifiers optimized via grid search. Experimental results show that the model achieves a test accuracy of 99.9%. Once the production state is identified, the system estimates cycle time and energy per cycle in real time. Approximately 58,000 production cycles, corresponding to several part types were characterized.