Condition Monitoring of Manufacturing Production Lines Using Fractal Analysis of Energy Consumption Datasets

Many plants use a combination of maintenance strategies across their production facilities. Under condition monitoring strategy, a continuous analyze of a combination of machine health parameters is performed, aiming the observation of changing trends. Online condition monitoring systems generally include measure of vibration, thermography, ultrasound, oil analysis or energy parameters. In the case of energy analysis, an abnormal behaviour of the manufacturing units generates deviations from the reference models by altering the energy consumption footprint. In this paper an approach based on fractal analysis of energy consumption datasets is evaluated on a real testbed assembly line. By computing the Hurst exponent for each data subset which corresponds to different assembly operations, for several manufacturing cycles, it was found that it is possible to perform an energy consumption pattern matching for each operation.