Manufacturing industry incurs a large portion of energy consumption and carbon emission in the economy. Traditionally, smart energy management depends on aggregated measures from billing information, as well as physics-based models, empirical results derived from extensive experiments, which tend to be limited in the ability for real-time monitoring of energy efficiency. There is an urgent need to develop energy monitoring solutions for more transparency about energy use. This paper presents a new sensor-based approach for recurrence analysis of continuous power signals and multi-state modeling of energy efficiency in the machining process. First, we leverage the recurrence plot to characterize the nonlinear variations in power signals and further help delineate different states in the machining process, thereby providing statistics of energy consumption in each state. Second, we compute the composite index of energy efficiency for each workpiece and then develop multivariate statistical control charts for process monitoring of continuous production of workpieces. Third, after an anomaly is detected, we propose the orthogonal decomposition approach to diagnose the root cause of abnormal states in the energy use. The proposed methodology is evaluated and validated on real-world manufacturing of shaft-like parts in a machine shop. Experimental results show that the prediction error of energy efficiency from sensor-based models is within 5% from the ground truth, which show great potentials to implement sensor-based monitoring and analysis of real-time energy efficiency in the manufacturing process.
All Science Journal Classification (ASJC) codes
- Computer Science(all)
- Materials Science(all)