Impact detection using a machine learning approach and experimental road roughness classification

Abstract

Abstract First, this publication presents the experimental validation of a road roughness classification method. Second, an impact detection strategy for two-wheeled vehicles is proposed including a classification of service loads, mild special events, and severe special events. The methods presented utilise the vehicle’s onboard signals to gather field data. The modular road roughness classification system operates with the vehicle’s transfer functions, and continuously classifies the road profile, according to ISO 8608. The method was successfully validated on test tracks with known road profiles. The impact detection strategy was developed using a supervised machine learning technique. Six road obstacles were ridden over using different velocities to invoke mild and severe special events. The most popular classifiers were trained for comparison and prediction of future observations. The developed impact detection strategy shows a high accuracy and was successfully validated using a k-fold cross-validation. The combination of the road roughness classification system and the impact detection strategy, enables a holistic field data acquisition of customer usage profiles, in the context of durability engineering. The collection of customer usage profiles improves vehicle design targets and enables a virtual load acquisition.