Signal cleaning techniques and anomaly detection algorithms for motorbike applications.

This paper outlines the results of the curricular internship project at the Research and Development section of Ducati Motor Holding S.p.A. in collaboration with the Motorvehicle University of Emilia-Romagna (MUNER). The focus is the development of a diagnostic plugin specifically tailored for motorcycle applications with the aim of automatically detecting anomalous behaviors of the signals recorded from the sensors mounted on-board. Acquisitions are performed whenever motorbikes are tested and they contain a variable number of channels related to the different parameters engineers decide to store for the after run analysis. Dealing with this complexity might be hard on its own, but the correct interpretation of data becomes even more demanding whenever signals present corruption or are affected by a relevant degree of noise. For this reason, the whole internship projects is centered on a research around signal cleaning techniques and anomaly detection algorithms which aims at developing an automatic diagnostic tool. The final goal is to implement a preliminary processing on the acquisition that allows an understanding of the quality of the signals recorded and, if possible, applies strategies that reduce the impact of the anomalies on the overall dataset.