Wireless sensor network for derailment detection in freight trains

The development and installation on freight trains of systems able to early detect and communicate to the train driver a derailment event is a key issue for railway safety, because the train length and the environment noise often prevent the driver from immediately realizing about the occurrence of these dangerous events. Mechanical devices such as the EDT101 by Knorr Bremse, which stop the train by directly acting on the brake line when too large vibrations are detected, are expensive, difficult to install, and do not leave to the driver the freedom to choose the best course of action in case of an accident. In order to overcome these drawbacks, which up to now have limited a large scale diffusion of such devices, we have implemented an electronic system based on a set of nodes that can be easily installed on freight cars and are able to detect an ongoing derailment by measuring the vertical acceleration and to rely the information to the driver in the locomotive [1]. Since freight trains do not have data and power lines, we have used wireless communication among the nodes and we have proposed to power them scavenging energy from vibrations, using supercapacitors for the energy storage. To this end, we have used low-power devices and we have designed the system in such a way that each node will stay in the ultra-low-power sleep mode for most of the time. The main components in each node (Fig. 1) are a LIS3DH accelerometer (from ST Microelectronics) and a Texas Instruments CC1110 system-on-chip (including a microcontroller and a transceiver). When the train starts moving and the supercapacitors are fully charged, an initialization phase takes place. Each uninitialized node transmits a packet containing its identification code; in sequence, each node (starting from that on the locomotive and proceeding towards the train tail) receives these packets and, on the basis of the received power, identifies its nearest neighbor. At the end of this phase, the network configuration is complete and a schedule is established for the periodic propagation of the information from the tail to the head of the train. The nodes are synchronized in such a way that, during each cycle, each node remains in the ultra-low-power sleep mode, apart from a short time interval in which it reads the data from the local accelerometer (and, optionally, from other sensors, e.g. temperature or humidity sensors), receives the information packet transmitted in that moment from the previous node, and then transmits its own packet to the following node. The packet contains a derailment alert if this event has been locally detected, or otherwise the information transmitted by the previous node; the packet is used also to rely, with lower priority, the data from local sensors. We have tested the operation of the nodes and the protocol both in the laboratory and in an actual railway environment (Fig. 2). An average power consumption of about 0.5 mW per node, compatible with the performance of electromagnetic energy scavengers, has been measured (Fig. 3), keeping each node active for about 20 ms in each 2 s cycle. Tests on an actual train and for different node positions have shown no packet loss over distances above 65 m. References [1] M. Macucci, S. Di Pascoli, P. Marconcini, B. Tellini, “Derailment Detection and Data Collection in Freight Trains, Based on a Wireless Sensor Network”, IEEE Trans. Instrum. Meas., DOI: 10.1109/TIM.2016.2556925.