h a l f b a k e r yTip your server.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
The condition of a railway track is critical to the safety
of
the trains (and passengers) who travel over it.
Defects in rails tend to develop over time, and if left
unnoticed can become catastrophic - cracks and rail
breakage do occur, and kinks (vertical or horizontal
displacements) can
give rise to high accelerations,
excessive loads, rollingstock damage or even derailment.
Railway owners spend a great deal on regular human
inspection (boots on the ballast), and on expensive rail
measurement trains, to identify defects for fixing before
they become dangerous.
I travel by train quite a lot, and I notice particular places
on my route where the train lurches alarmingly, or clunks
heavily... and I wonder whether the track maintenance
team have been out to have a look recently.
I also use the train operator’s mobile phone app to buy
my
tickets, check train times, and check for disruptions.
Given a train full of passengers with mobile phones, each
of which has an accelerometer (and GPS) built in, it
shouldn’t be difficult to collect acceleration/location
data
from a large number of devices and extract useful
condition data from it.
Each individual phone might have a significant error, but
interpolating across the population (ie Kalman filter)
should produce reasonable accuracy.
When a passenger signs up to the ticket/disruption app,
they’re asked if they’d also contribute (their data) to
the
reliability and safety of the railway.
Network Rail New Measurement Train
https://www.youtube...watch?v=WhVdTXh5XoA
Also known as the Flying Banana [Frankx, Sep 14 2019]
Smartphone Sensing Capabilities for On-Board Railway Track Monitoring
https://doi.org/10.1155/2019/1729153
André Paixão, Eduardo Fortunato, and Rui Calçada, 2019 [Frankx, Sep 15 2019]
[link]
|
| |
I like the idea of using cheap consumer accelerometers to
monitor track conditions. However, if it worked, wouldn't it
be better to take one such accelerometer and mount it rigidly
on either the train or the suspension? That would give you
more accurate data than the average of a few dozen phones,
each one only loosely coupled to the train and all wiggling
independently. |
|
| |
//wiggling independently// |
|
| |
Yes, true. But for each train you’d have to buy a
big rugged “railway-approved” box and bolt it onto
an appropriate bit of train structure, and provide a
power supply and rugged aerials - all of which gets
quite pricey. |
|
| |
Whereas your passengers collect the data without
even knowing it, with ready-charged and freely
communicating phones. |
|
| |
I think you can collect the data from them
essentially for free, and do some clever signal
extraction, to get good data out of the “wiggly”
data |
|
| |
But... if the railway is prepared to use passenger data to
guide its track maintenance, it could get (and use) the same
data much more cleanly using the same equipment. |
|
| |
What I mean is - it's probably harder to get a railway to rely
on [badly-coupled mobile phones + multiple passengers who
move independently] than to get them to rely on [a single
well-coupled mobile phone]. If the former worked, the latter
would work somewhat better and would be easier to justify. |
|
| |
Also, re. the wiggliness and data averaging. |
|
| |
Suppose a typical minor track fault causes a jolt of
magnitude 1 to the seated passenger. Suppose also that the
movements of a passenger cause jolts (noise) of magnitude
10. If you average 2 passengers, the data will have a noise
of about 10 over root2, or around 7. If you average over 10
passengers, you'll have a noise of around 3. If you average
over 100 passengers, you'll have a noise of about 1 - and so
you still won't be able to detect minor track faults. |
|
| |
You might do better if you harvested sound, since many
track faults produce a distinctive sound; but passengers will
not be happy with having their phones transmit their audio
to the rail network. |
|
| |
Hmm. Thanks [MB], i’ll have a look at some
figures. |
|
| |
Not needed in places like Japan, where the tracks
are engineered to total perfection, allowing their
amazing trains to have timekeeping to the second.
Would be useful in places like the UK or North
America where trains and track technology hasn't
advanced from the Victorian era. |
|
| |
Doesn't accelerometer hardware have three axises? to tell what sort of action is being vectored. Shouldn't a train joint stick out in the norm of passengers sitting, standing semi-motionless data? Time referenced with 100 passengers, the waveform should travel through the phones down the train. |
|
| |
//won't be able to detect minor track faults// |
|
| |
Well, looks like I'll have to learn some of the maths around
Kalman filters. I think they can do this - as you point out,
error decreases with the number of input sources. |
|
| |
Also, the same section of track can be measured by
multiple trains, so data can be summed across many passes. |
|
| |
For the railway owner, this would be no-cost real-time track
data with no requirement for additional
boxes/infrastructure. |
|
| |
I think that's quite an attractive proposition |
|
| |
///Not needed in places like Japan, where the
tracks are engineered... |
|
| |
Ahem, recent shinkansen blunder after a
cleaner left the door open, kind of windy inside.
And, the whole non-shinkansen train timetable
goes to pot late Friday night, due to drunks
falling over and getting stuck in the doors etc |
|
| |
[not_morrison_rm] Personal experience? |
|
| |
Safety on the Japanese rail system is comparable
with the UK. We have plenty of embarrassing (and
alarming) incidents too. |
|
| |
// You might do better if you harvested sound, since many track faults produce a
distinctive sound; but passengers will not be happy with having their phones transmit
their audio to the rail network. // |
|
| |
In Calgary, the train cars have signs on the doors notifying passengers that video and
audio recording is going on within. They could use that audio. |
|
| |