A $500 Open-Source Sensor Could Give Urban Railways a Low-Power Way to Count Wheels

Urban rail operators may soon have a simpler way to record passing wheels, thanks to an open-source detector that uses a spring-loaded mechanical pedal, a magnet, and a reed switch rather than the more complex electromagnetic systems commonly used in railway monitoring. The prototype, described by researchers Eren Erdi, Emrah Sarioglu, Baris Oguz Gurses and Aysun Baltaci, is designed for non-safety-critical applications such as temporary wheel-passage logging, maintenance surveys and experimental Internet-of-Things monitoring. In field tests conducted at speeds up to 90 kilometres per hour, the device registered all 500 observed wheel passages, with no missed detections or false positives. Its estimated construction cost is about €500, and the complete mechanical, electronic and software design is released under a CC BY 4.0 open-source licence.

The device addresses a practical gap in railway infrastructure. Certified train-detection systems, axle counters and track circuits are essential for signalling and safety, but they can require permanent installation, specialized hardware and substantial infrastructure work. The new detector is not intended to replace those systems. Instead, it acts as a compact event sensor that can be attached to an existing rail and used where engineers need information about wheel movements but do not need a safety-certified train-detection function. Possible uses include counting traffic during a maintenance campaign, triggering cameras or vibration sensors when a wheel passes, estimating train frequency, and collecting localized data at sites without a fixed electrical supply.

At the heart of the design is a small vertical pedal housed in an aluminium carrier mounted to the rail. When a wheel flange rolls over the pedal, the contact pushes it downward. A permanent magnet embedded near the top of the pedal then moves closer to a reed switch positioned inside the housing. A reed switch is a magnetically actuated electrical contact: when the magnetic field reaches the required strength, two internal ferromagnetic contacts close and produce a discrete electrical signal. Unlike an electromagnetic coil, which generates an analogue waveform as a wheel passes through its magnetic field, the reed switch produces a simple on-or-off event. That binary output reduces the amount of signal processing required and is particularly compatible with a battery-powered device that spends most of its time asleep.

The mechanical geometry was tuned to urban railway wheels. The pedal surface sits 14 millimetres below the top of the rail head, allowing it to contact a commonly standardized 28-millimetre wheel flange while retaining tolerance for manufacturing and wear differences. The pedal can travel as much as 14 millimetres during a wheel passage, although its total motion is mechanically limited to prevent contact with the sensing assembly. Two compression springs provide the restoring force and are preloaded by 24 millimetres during assembly. That preload is not merely intended to return the pedal to its starting position. It also reduces rebound after impact, limiting the possibility that a single wheel could make the pedal oscillate through the switching position several times and create spurious counts.

The researchers used MATLAB and Simscape to model the interaction between a wheel flange and the pedal at different speeds. They treated the pedal and springs as a lumped mechanical system and used Hertzian contact theory to estimate the forces and pressures at the wheel–pedal interface. Hertzian theory describes how curved elastic bodies deform when pressed together: the contact area grows with load, while force increases nonlinearly with indentation. Under the model’s fully elastic assumption, peak contact pressure rose from about 330 megapascals at 10 kilometres per hour to 711 megapascals at 90 kilometres per hour. Those pressures exceed the approximate yield threshold of the aluminium pedal, meaning the highest-speed impacts enter an elastic–plastic regime rather than remaining purely elastic.

That apparent weakness is a deliberate feature of the prototype. The pedal is made from Al 6063-T6, an aluminium alloy that is weaker and lighter than railway wheel steel. The researchers designed it to function as a replaceable mechanical fuse: if an unexpected overload or misalignment occurs, the detector component should deform before the wheel does. Once the aluminium begins to yield, the contact surface can flatten and spread the load over a larger area. The analysis estimated a hardness-like pressure limit of roughly 510 megapascals for the alloy, substantially below the approximate 800-megapascal pressure associated with yielding in the lowest-strength railway wheel rim steel considered. This material hierarchy is intended to protect the vehicle’s wheel flange while confining irreversible damage to an inexpensive, replaceable part.

The simulations suggested that the 24-millimetre spring preload was the minimum value that eliminated secondary penetration of the pedal after the initial contact. At the most demanding modelled speed, 90 kilometres per hour, the pedal became fully damped within 17.076 milliseconds. The researchers selected a 30-millisecond firmware validation window, or debounce interval, to reject rapid repeated signals caused by switch chatter and mechanical rebound. Even when the assumed structural damping ratio was varied from 0.3 to 0.7 per cent, the predicted settling time remained below 30 milliseconds. The wheelbase travel time at 90 kilometres per hour was approximately 60 milliseconds for the modelled 1.5-metre wheel spacing, leaving a predicted interval in which the pedal could reset before the next wheel arrived.

The electronics are built around an STM32L432KC microcontroller, a low-power device that remains in deep sleep until an electrical transition from the reed switch triggers an interrupt. The prototype’s measured sleep current, including its voltage regulator and input circuitry, was approximately 10 microamperes. The system accepts an input supply between 3.7 and 13 volts direct current. When the pedal moves, the reed-switch signal passes through an input-conditioning stage and wakes the microcontroller. The firmware then checks the transition during the 30-millisecond validation period before recording one event. A relay output is included, while future versions could add wireless communication without changing the basic sensing mechanism. The design files include STEP models for the mechanical parts, KiCad electronics files, fabrication outputs, enclosure geometry, firmware and MATLAB, Python and CSV validation materials.

The researchers also examined whether repeated wheel impacts might permanently push the pedal out of alignment. A reduced-order, two-dimensional elastoplastic finite-element model represented the aluminium using a von Mises yield criterion with isotropic hardening, a standard approximation in which the material’s yield surface expands as plastic strain accumulates. Under the modelled repeated loading, most of the residual indentation developed during the first few cycles before approaching a shakedown-like state. After nine load–unload cycles, the predicted total indentation was approximately 0.70 millimetres, while the remaining increment was only 0.000008 millimetres per cycle. That value remained well below the 5-millimetre triggering allowance in the magnet–reed geometry. However, the result is not a guarantee of long-term service life. The model did not include wear, fatigue, three-dimensional contact evolution or the Bauschinger effect, which can influence metals subjected to repeated loading and unloading.

The strongest evidence came from a real railway installation. The prototype was mounted on one rail and tested with a single vehicle of known axle configuration over a speed range from zero to 90 kilometres per hour. Direct visual observations were compared with the detector’s output for 500 wheel-flange passages. Every observed passage produced a valid recorded event, and the system reported no missed detections or false positives. Inspectors also found no visible deformation of the vehicle’s wheel flanges after testing. The results show that the mechanical trigger and low-power electronics can work together under the tested conditions, but they do not establish performance across every railway environment. Temperature changes, water, dirt, vibration, lateral wheel motion, wheel-flange variation and long-term reed-switch ageing could all affect reliability. Extended field trials will be needed before the detector can be considered for sustained deployment, but its open design gives researchers and operators a low-cost platform for adapting and testing wheel-monitoring concepts without rebuilding an entire railway sensing system.