A twist fault is a measurement, not an opinion: the difference in cross level between two points three metres apart, in millimetres, off a train that measures the whole route every fortnight. This one is fourteen millimetres, at the same place, every run, and gone by the time anybody stands on it.
I am a track section manager and the recording train's output for my section lands on my desk.
What twist is and why it matters more than the other faults.
Twist is the change in cross level over a short length. If one rail dips relative to the other over three metres, a bogie can unload a wheel and a derailment starts that way. So there are limits in millimetres, and a twist over the limit is a fault with a response time on it.
The recording train measures it with an inertial system and lasers, continuously, and it is accurate to about a millimetre. When it flags something, we go and measure by hand with a gauge and a level, and the hand measurement is what decides the repair.
At one chainage on the down line, in a cutting, the train has recorded a twist of between twelve and fifteen millimetres over three metres on every run for four years.
That is over the limit. It generates a work order every fortnight.
Every fortnight somebody walks out and measures it by hand and finds nothing.
Not a small twist. Nothing. Cross level within a millimetre or two along the whole ten metres either side, which is a well maintained piece of track, which it is, because it has been tamped three times in four years for a fault nobody can find.
What has been done.
The tamper. Three times. A tamper lifts and packs the track to a design line and it removes any twist that is there, and the recording train has found the fault on its next pass after each one.
The recording train itself. Two different vehicles have measured that section and both record it. Their calibration is checked against a reference site every week and their output on the rest of my forty miles agrees with the hand measurements to about a millimetre, which is why I believe them.
The formation. A soft spot under the track would move under load and not under a man with a gauge. So a trolley mounted device was pushed over it with a loaded axle, and then in 2022 the engineer's train ran over it slowly with a colleague watching from the cess with a level.
Nothing moved. The rails sat still under a hundred tonnes.
What the recording train shows that a man with a gauge cannot.
I asked for the raw trace rather than the exception report, which took some persuading.
A twist fault from a soft spot has a shape: it develops over a few metres, peaks, and recovers. This one does not. The cross level is flat, and then in the space of one sample, a quarter of a metre, it steps fourteen millimetres and holds for three metres and steps back.
Track does not do that. A rail is a continuous steel beam and it bends in curves, not in steps. A step in cross level over a quarter of a metre would need the rail to be broken, and it is not, and the ultrasonic test car says so.
The engineer who does the train's data looked at that trace with me and said it looked like the track was fine and something a metre and a half wide was sitting on one rail.
What is done about it.
The work order comes in and I close it with a hand measurement showing no fault, every fortnight, and my closures have been audited and upheld.
The line speed there is unchanged. I have thought hard about whether it should be and I have decided, with my manager and with the data, that a measured twist that is not present when measured by hand is not a fault in the track, and that slowing every train on a hand measured piece of good track would be responding to an instrument rather than to a railway.
What I do.
I walk it myself on the fortnight now, rather than sending somebody, because I would rather the closure was in my own name.
The one thing I changed is the time.
The recording train passes at about half past two in the morning, and the fault is in its data at that time, and I used to send the gang out first thing after the report landed, at about seven.
In September 2023 I went out at two, to be standing on it when the train came through, with a level on the rail and a lamp.
The cross level was flat. The train came through at forty and I stood back in the cess and watched it pass, and its report the next morning showed fourteen millimetres.
What I have not put in a closure is that between the train passing and my walking back to the level, about a minute, the level had moved. Not the reading: the level. It was a metre further along the rail than I had left it, lying on its side in the four foot, and I was the only person in that cutting.