With autonomous driving, the driver’s perceptual ability for irregularities in the chassis system will be decreased. Therefore, monitoring the chassis system for possible defects will be necessary. This paper analyzes the suitability of the four unsupervised learning algorithms for novelty detection Local Outlier Factor, Angle-based Outlier Detection, k-nearest neighbors and One-Class Support Vector Machine. The investigation is conducted using actual driving data with damper defects emulated using semi-active dampers. Aside from using manually generated features or using FFT datapoints as features, two automatically generated feature datasets using Autoencoder and Sparse Filter are investigated. Furthermore, the influence of different scaling methods and algorithm specific parameters is analyzed. Results show that a precision of up to 80 % is possible.