Civil Engineering May 2021 | Vol 29 No 4

32 May 2021 Civil Engineering using non-contact displacement measure- ment systems, Wi-Fi enabled data logging, and light-dependant resistors to trigger wetting and drying cycles. This novel apparatus has allowed for the characterisation of the SWRC of un- saturated railway formations. This enables the implementation of unsaturated soil mechanics to address the issue of climate change and its impact on heavy haul railway formations and transportation infrastructure. THE INTERNET OF RAILWAY THINGS (IoRT) A dedicated Data House has been constructed adjacent to the N4 highway, serving as a centralised data acquisition and transmission platform for transdis- ciplinary research projects in and around UP’s Innovation Africa campus. The potential of an edge intelligence solution was successfully demonstrated in 2020, whereby a low-cost, low-power computer hardware platform detects and classifies passing vehicles in real-time along the freeway with an accuracy of 96%. These statistics are aggregated alongside envi- ronmental data (ambient air temperature, pavement surfacing temperature, soil moisture, air quality, CO 2 concentration, and weather station data) measured by a wide range of LoRaWAN (long-range wide area network) sensors. The LoRaWAN provides the capacity to serve hundreds of these devices with coverage extending across the entirety of the Hillcrest campus. To date, more than 40 of these low-cost, battery-powered sensors have been installed across the Innovation Africa campus, with custom- ised sensor platforms located as far away as Port Alfred due to join the network in 2021. Data from this distributed sensor network is aggregated within the central- ised Innovation Africa IoT Data Platform. This scalable sensor network serves as the next evolution in real-time condition monitoring of infrastructure, in particular railway environments; characterised by remote, distributed assets of significant economic importance where continuous measurement and identification of tem- perature, rail breaks, impact loads, and permeant settlement effects are crucial for safe railway operations. To address the disparity between state- of-the-art advancements in deep learning and efficient condition monitoring of the digital railway, the Chair in Railway Engineering established a collaborative research project with 4Tel, based in Australia, in 2019. Neural network-based multi-view stereopsis (MVS) reconstruc- tion pipelines are incorporated with millimetre-accurate geolocation services to measure the condition of railway assets (track geometry). A low-cost, mobile, real- time kinematic (RTK) geolocation service was developed at UP which provides 14 mm accurate geolocation capabilities up to 13 km away from the stationary reference antenna installed at the Data House. Figure 2 The newly developed SWRC apparatus incorporating wireless technologies for the accurate measurement of soil sample mass Figure 3 Camera frame (left) with GPS antenna installed and on the righthand side the synthetically rendered rail, a depth map and finally a point cloud of the railway profile