Abstract

The knowledge of the state of health of machinery gears helps developing cost effective maintenance plans, preventing costly down times caused by catastrophic failures. The widest spread strategy in industry to avoid faults and failures is based on preventive maintenance. Only its combination with a condition-based maintenance can detect early signs of potential machinery failures. Often, accurate information about the state of health of a piece of equipment is difficult to obtain. Strategies based on intelligent predictive maintenance could improve this situation. The most established method to gather information in mechanical systems using gearboxes relays in the use of accelerometers, which are expensive and whose installation is usually troublesome. The analysis of the electric signature of the electric motor that drives the gearbox provides a non-intrusive method, based on readily available information. Changes in the speed and load conditions of the gearbox produce correlated variations in the feeding current and voltage of the motor. A detailed analysis of these electrical signals can produce useful information about the state of health of the system. In this paper, a gear prognosis simulator (GPS) test bench equipped with a multistage gearbox is used to analyze different types of mechanical faults in the gears. Three fault families have been identified, high damage, moderate damage and low damage. Specific working conditions of the test bench have been selected to mimic the operation of different mechanical systems, such as machine tools or electro-mechanical actuators. The motor electrical current signature in the different conditions is analyzed to determine the health state of the gearbox. Signal descriptors (such as rms, kurtosis, peak-to-peak value, impulse factor, shape factor, etc.) are obtained from stationary speed. A selection of the most relevant descriptors has been carried, doing a one-way analysis. The results obtained reveal appreciable differences between the different faulty and nominal states of the gears, making possible the detection of the health state of the system using different advance data analysis techniques.