Classification Tree-Based Wheel Unbalance Detection

In this paper, the problem of unbalance detection in vehicle’s wheels is analyzed. Starting from accelerations and angular velocities acquired from a wheel balancer machine instrumented with Micro Electro-Mechanical Systems (MEMS) sensors, meaningful features are extracted for different conditions of unbalance, namely, 30 and 40g. Then, classification trees, for both the static and pure couple unbalance, are trained for the simultaneous detection of severity and angular position of the unbalanced masses. The performance of the proposed approach is validated against experimental data and its effectiveness is compared to a cascade algorithm, previously published by the authors.