Data driven modeling and estimation of accumulated damage in mining vehicles using on-board sensors

Abstract

The life and condition of a MT65 mine truck frame is toa large extent related to how the machine is used. Damagefrom different stress cycles in the frame are accumulatedover time, and measurements throughout the life of the machineare needed to monitor the condition. This results inhigh demands on the durability of sensors used. To makea monitoring system cheap and robust enough for a miningapplication, a small number of robust sensors are preferredrather than a multitude of local sensors such as strain gauges.The main question to be answered is whether a low numberof robust on-board sensors can give the required informationto recreate stress signals at various locations of the frame.Also the choice of sensors among many different locationsand kinds are considered. A final question is whether the datacould also be used to estimate road condition. By using accelerometer,gyroscope and strain gauge data from field testsof an Atlas Copco MT65 mine truck, coherence and Lassoregressionwere evaluated as means to select which signalsto use. ARX-models for stress estimation were created usingthe same data. By simulating stress signals using the models,rain flow counting and damage accumulation calculationswere performed. The results showed that a low number ofon-board sensors like accelerometers and gyroscopes couldgive enough information to recreate some of the stress signalsmeasured. Together with a linear model, the estimatedstress was accurate enough to evaluate the accumulated fatiguedamage in a mining truck. The accumulated damagewas also used to estimate the condition of the road on whichthe truck was traveling. To make a useful road monitoring system some more work is required, in particular regardinghow vehicle speed influences damage accumulation.