Development of a new time domain-based algorithm for train detection and axle counting

This paper presents an innovative train detection algorithm, able to perform the train localisation and, at the same time, to estimate its speed, the crossing times on a fixed point of the track and the axle number. The proposed solution uses the same approach to evaluate all these quantities, starting from the knowledge of generic track inputs directly measured on the track (for example, the vertical forces on the sleepers, the rail deformation and the rail stress). More particularly, all the inputs are processed through cross-correlation operations to extract the required information in terms of speed, crossing time instants and axle counter. This approach has the advantage to be simple and less invasive than the standard ones (it requires less equipment) and represents a more reliable and robust solution against numerical noise because it exploits the whole shape of the input signal and not only the peak values. A suitable and accurate multibody model of railway vehicle and flexible track has also been developed by the authors to test the algorithm when experimental data are not available and in general, under any operating conditions (fundamental to verify the algorithm accuracy and robustness). The railway vehicle chosen as benchmark is the Manchester Wagon, modelled in the Adams VI-Rail environment. The physical model of the flexible track has been implemented in the Matlab and Comsol Multiphysics environments. A simulation campaign has been performed to verify the performance and the robustness of the proposed algorithm, and the results are quite promising. The research has been carried out in cooperation with Ansaldo STS and ECM Spa.     

Integrated fault detection and diagnosis system for longitudinal control of an autonomous all-terrain vehicle (ATV)

This paper presents a fault detection and diagnosis (FDD) method to enhance the reliability and safety for longitudinal control of an autonomous all-terrain vehicle (ATV). An integrated approach using decentralized and centralized FDD is proposed to optimize the tradeoff between sensitivity and robustness. While the decentralized approach is suitable for detecting faults in actuators and sensors directly connected to a single processor, it is sensitive to noises and disturbances and thus may result in false alarms. On the other hand, the centralized approach is based on information communicated between multiple processors, and it detects and diagnoses faults through analyzing concurrent computations of multiple hardware modules. However, its performance is still limited to isolating faults specifically in terms of components in the single hardware. To incorporate the advantages of both FDD approaches, a two-layered structure integrating both decentralized and centralized FDD is proposed and allows us to perform more robust fault detection as well as more detailed fault isolation. Finally, the proposed method is validated experimentally via field tests of an ATV.