Smartphone based sensing enables automated vehicle prognosis

This paper presents a smartphone based application whereby a vehicle owner can obtain a reasonable prediction of the vehicle's potential failure time. Through a hybrid model-based and data-driven approach, one can obtain a predictive maintenance suggestion; given the current state of degradation. The smartphone is used both for sensing and computation. The proposed minimal-sensing approach is only meant to indicate Level-1 failure - the first step in identifying the existence of fault. Here, one assumes that the vehicle's vibration continues to increase over time thus indicating progressive degradation in its ability to absorb shock. The vertical vibration is measured using accelerometer and an appropriate measure is deduced for each completed trip. Furthermore, a trend line is obtained that continues to compute time-to-failure with respect to a pre-determined breakdown point. In addition, the effects of potentially rough driving style is also qualitatively accounted for. The results of its deployment in an office bus is presented. Detailed analysis on the data captured for passenger cars is in progress.