Abstract:
Derailments, which frequently occur in several developing countries, are often caused by uprooted or faulty rail blocks. If happens, derailments result in huge loss of property along with massive lethal consequences. Devising a real-time automated wireless sensing system for detecting uprooted or faulty rail blocks to prevent derailments is strenuous. This particularly holds over the developing countries considering their low-resource and insufficient network infrastructure in remote areas where trains frequently run. Existing studies in this regard are yet to present a pragmatic solution worth of mass deployment, as the solution demands planning of deploying suitable sensors. Accordingly, in this thesis, we first explore suitability of sensors being capable to cope with the fact that stopping a train after sensing an uprooted or faulty rail block ahead needs a considerable response time and distance. This demands exploring separate methodologies in road to enabling long-distance sensing and networking to confirm having the considerable response time and distance. Accordingly, we explore vibration sensing exploiting multi-sensor fusion and multi-hop networking exploiting LoRa with elevated positioning to adopt suitable sensing and networking solutions. We perform several iterations of real on-field data collection and rigorous analyses to evaluate the performances of the solutions. Our evaluation demonstrates more than 95% accuracy in sensing an incoming train from a 1km distance and a networking delivery ratio of 80% from a 2km distance while having a speed of 50km/h.
After adopting suitable sensing and networking solutions, it becomes extremely important to investigate how the sensor nodes (subsuming the sensing and networking solutions) could be cost-effectively deployed in real life while minimizing the possibility of derailments. The deployment strategy should realize the notion of selective deployment in accident-prone areas, which needs to be investigated based on system-level parameters (such as sensing and transmission ranges, as already explored) of the adopted solutions as well as statistics on derailments over the deployment premise that is the railway track in Bangladesh. However, deploying the sensor node in an optimized fashion is very challenging. Existing studies in this regard are yet to present a pragmatic solution that realizes an optimal way to deploy the sensor nodes along the rail track while minimizing both the expense of deployment and the vulnerabilities of derailments. As a remedy, in this research, we first formulate a new multi-objective optimization problem for the intended mass-scale sensor deployment. Then, we employ a meta-heuristics based approach to solve the formulated optimization problem that can eventually optimize deployment cost and vulnerability to derailments. Finally, we analyze the effectiveness of the generated solutions on optimized sensor deployment topologies.
In our analyses, we found that the sensors can be deployed throughout the railway by saving 50% of the deployment cost comparing to the brute-force deployment, and can save more than 65% of the potential railway accidents.
