Condition monitoring of a gearbox through the Internet

Abstract

Condition monitoring of rotating machines used at factories and mines could significantly reduce production losses to allow continuous production. Unexpected machine failures could be eliminated, and specific problems then identified and solved as a predictive maintenance approach. In this study a cost-effective continuous condition-monitoring system for a gearbox was developed. The monitoring system measured vibration and temperature at short time intervals, and these parameters were effectively used as early warnings when problems started and developed over time. In addition, electrical current was also measured for changing power and torque which were then also determined for the gearbox regarding different applied loading conditions. A mathematical model was developed to describe all the different defect frequencies for each of the ball-bearings in the gearbox. This model also described the gear-mesh frequencies for the gears in the gearbox. This mathematical model was implemented in a computer program, and the information obtained used in the analysis of measured data regarding possible damage at the bearings and gears. The condition-monitoring system comprised electronic hardware coupled to appropriate sensors and also to a computer. A second computer program was written so that the measured parameters could be captured and indicated. In addition, a second mathematical model was developed to convert measured time domain acceleration data into the frequency domain by using a series of Fourier coefficient amplitudes at certain frequencies and corresponding phase angles. Acceleration signals were also converted to corresponding velocity and displacement signals. The time domain signals were used to determine single overall RMS vibration magnitudes. This model was then also implemented in this computer program. The computer programs were both implemented in a Matlab environment. By using certain input parameters for the bearings and for the gears respectively as characterised, the first program computed bearing-defect frequencies and gear-mesh frequencies. This was regarded as necessary because amplitudes at these frequencies normally represent wear.