will this work

[laxman289]

working-in the normal state(no cracks) no emf is induced in secondary coil.when crack is there the magnetic flux changes hence emf is induced in the secondary which turns on the thyristor and hence buzzer.will this circuit work as intended

[Joatmon]

First thoughts - what about where the rails join together and deliberate breaks such as points?

[studiot]

Why would significant flux pass through the rail. Surely your coils need reorienting?

 

The detector design need improvement as well.

 

What is the distance between the wheel centres and the rail?

Edited June 12, 2012 by studiot

[John Cuthber]

Have you ever tried listening to am radio anywhere near any electric train tracks?

It might be OK for a non-electrified one, but I suspect that noise would still be a problem

[ewmon]

Why not simply measure the resistance between the wheels because cracks would cause higher resistances? A Wheatstone bridge might work well.

 

Also, cracks are more of a deep V shape (versus a gouge), so the two sides of the crack would act as a capacitor (even though they're connected at the bottom of the V, pretty much the same way two adjacent turns of a coil act as a capacitor). This means that the crack can be measured using AC rather than DC.

 

How do the "cross-sectional areas" of cracks that you want to detect compare with the natural variance of a rail's cross-sectional area due to their manufacturing process?

[laxman289]

First thoughts - what about where the rails join together and deliberate breaks such as points?

yeah thats right...breaks cause a problem..but here in india two rails are joined by large thick metallic plate.so i assume that the change in flux will be much lower in case of breaks than cracks...if we r adding a suitable resistance then gate current can b made lower than min gate current required to trigger

 

Why would significant flux pass through the rail. Surely your coils need reorienting?

 

The detector design need improvement as well.

 

What is the distance between the wheel centres and the rail?

but there isnt any other closed path through which flux can pass

[studiot]

but there isnt any other closed path through which flux can pass

 

It's a question of sensitivity. Why do you think search coils in metal and other detectors are oriented with their axes at right angles to your drawing?

[Enthalpy]

A few considerations about the design as it is first sketched here...

 

Rail joints will give a much bigger signal than any crack, whatever they look like, for sure. You need software to discard detection at the joints. The joints can be detected automatically by the software if their spacing is regular enough for some travel distance.

 

The described electronics is too rudimentary, but the real challenge is to get initially a clean signal; then, processing it is banal engineering.

 

You would better clean the rail before passing the detector on it. A static or rotating brush before the detector could remove loose object; depending on the induction you desire, you have to remove ferromagnetic dirt as well, which might be done by rotating ferromagnets as in waste sorting - but ferromagnetic dirt stuck at your magnets might always be a worry.

 

I agree the flux path goes through the rail in your design. You can add static wedges around the wheel where the contact with the rail is narrow, or offer a path elsewhere: an airgap can be acceptable if broad and thin. The coils that create the induction would better be a pair near the ends. And to pick a cleaner signal, magnetic designs tend to have symmetric picking coils (or sometimes emitting coils), with the useful signal being a difference between induced voltages.

 

What sort of defect do you want to detect: is it a notch, several millimetres long and rather shallow? Is it a rail-wide and centimetre-deep crack where the butt faces are separated by 100µm?

 

What operation mode do you need: a special waggon added to a commercial train running at its normal speed of 20m/s? Or can you have a special train running much slower?

 

Skin effect is a serious limitation. Imagine the magnetic flux increases over 1m (using very long poles at the air gap), stays for 1m, decreases over 1m, with the waggon running at 20m/s: it correspond roughly to 5Hz. In copper, skin depth would be 30mm, but in steel (take µ_r=5000) it's only 1mm unless the flux suffices to saturate superficial steel. So apparatus for non-destructive testing use to observe how well metal parts repel the induction, not how well they conduct it. A flux-conducting design would limit the frequency of an AC field even more seriously.

Edited June 22, 2012 by Enthalpy

[Enthalpy]

My alternative proposal uses GHz radio waves.

 

Power at 40GHz (7.5mm wavelength) is radiated with linear polarization oriented at +45° versus the rail. A reflector or lens of D~90mm concentrates it on a D=40mm spot at 0.2m down distance on the rail. At the same focus, an other primary source with linear polarization receives the reflected wave; it's oriented at -45°, adjusted to near-zero coupling with the transmitter.

 

• Reflection on a smooth sound rail surface keeps the +45° polarization which the detector doesn't sense.
• Cracks more or less perpendicular to the rail, even if thin, give a signal in the detector.
• The rail's inner and outer edges would give a signal, but they are smooth, and the illuminated spot avoids them.
• Post-detection software senses quick variations in the reflected signal, discriminating further the rail's sides. Like:
  compare the measured value with a reference interpolated from measurements 100mm and 200mm before and after.

Possible improvements:

• Components exist for 94GHz, to reduce the illumination of the rail's sides.
• A precise smooth illumination function (like Gaussian) reduces reflection by the round rail's sides.
• Have several detectors and cross-check to avoid unjustified detection.

Small and cheap enough to equip one or more railway engine on each line.

Marc Schaefer, aka Enthalpy

[Enthalpy]

With polarizations like here under, sound uniform metal reflects a signal that the detector attenuates, while a more or less transverse crack in the rail, which reduces the E component parallel to the run, gives a signal favoured by the detector.

 

 

One interesting location is at the bogies (cut for view), where the vehicle's weight distends the cracks:

 

 

The antenna example here is a horn with a lens, sketched at a bigger scale than the railway engine.

More sensors would avoid false detections; their polarization can be offset if someone expects cracks at 45°.

 

Marc Schaefer, aka Enthalpy

Edited June 23, 2012 by Enthalpy

[Enthalpy]

This other proposal illuminates the rail at an angle that lets the reflected light escape if the rail is sound. A crack scatters light to abnormal directions, which a detector (close to the source in this sketch) monitors:

 

 

A laser diode for DVD burner is a candidate for the source, as it allows fast modulation and narrow wavelength filtering, both useful against stray light.

 

The target position on the rail is preferably where the vehicle's weight induces bending and shear in the rail, widening possible cracks.

 

Here as well, I'd put several such sensors to avoid unjustified detections. And again, software can use successive measurements of the received background signal and compute an interpolation as a reference to compare the received signal with.

 

Rust could scatter light as well, so I expect this method to work only with rails used regularly. Light shall target the part of the rail cleaned by the wheels.

 

Marc Schaefer, aka Enthalpy

[imatfaal]

A bit off-topic - but when testing for hidden cracks and corrosion an easy method is using sound. Ultrasound readings can cover a vast area (I am thinking ship hulls) and provide a very good and reliable way to spot cracks and corrosion that are either within the body of the metal or covered by heavy paint.

 

The method discussed of shining light would only detect top surface flaws - which could easily be found by touch sensitive devices; the trick is finding the problem that is hidden from view.

[Enthalpy]

I don't expect cracks to appear deep in a rail during its service. The big stress is at the contact with the wheel, at the surface.

If a rail has a void from its manufacture, this has to be detected at the production site, where ultrasound is an excellent technique indeed.

Here I concentrated on wear mechanism for rails in service, and I ilke the sensors to work on a train at commercial speed.

 

I considered sound for this purpose, but it looks difficult. The train itself creates huge noise in the rail, especially at rail joints, and the rails produce strong echoes. This would rather need to send a pedestrian team when the traffic is stopped, meaning a survey once in many years, as compared with permanent monitoring made possible if regular trains are equipped.

[imatfaal]

Hey Guys

 

Saw this when I was following a link Moontanman posted for something completely different.

 

http://www.gizmag.com/automatic-road-repair/23000/

[Enthalpy]

Failures of the fishplates that join some rail ends caused several catastrophes
https://en.wikipedia.org/wiki/Fishplate
Recent tracks have continuous welded rails but keep fishplates at railroad switches, and older tracks can have one every 18m, so monitoring automatically the fishplates would improve the safety, by detecting loose bolts or loose plates early.

This could be done with cameras mounted on railway vehicles and automatic image analysis. While I have no ideas for the software, here are some lights on other aspects.

Cameras sit at railway engines better than on special vehicles, as this brings routine monitoring and adds no traffic.

Slower trains give software more time. Some tracks carry 40m/s passenger trains during the day and 10m/s freight trains at night; cameras better sit at the freight engine then.

Inexpensive flat tilted mirrors can provide the adequate viewpoint to cameras mounted at a protected location. Four cameras see both sides of both rails.

2mm resolution suffice. It gives only 200µs frame time at 10m/s and less exposure time. Spinning mirrors improve the exposure time but are imperfect, complicated and less reliable. Instead of a 2D camera retina, I suggest a line retina. It takes at each frame an image of a line perpendicular to the track, and the vehicle's movement creates a 2D image from successive frames. 5k frames/s is easy for such sensors, and 100 pixels need 500k pixels/s only.

Strong light is required for 200µs exposure, but only at 200mm*2mm or a bit more. 8We on 200mm*20mm are as good as 2kWe on 1m².

Accelerometers or microphones near the fore and aft axles can give their own information about a railjoint's health and tell when the retina at a known position on the bogie observes a joint. The retina collects permanently data which is kept and analyzed within the pertinent time spans.

The retina's frame rate can vary with the vehicle speed inferred from the accelerometers or microphones or some other means.

Two joints per 18m observed from both sides at 10m/s let analyze 2.2 images/s containing 4-6 boltheads each. Reasonable challenge.

Once this works, one can consider monitoring also the parts that fasten the rails on the sleepers.

Marc Schaefer, aka Enthalpy