Modern Wheel Slide Prevention Systems

[GC class B1]

I have taken an interest in modern WSP systems and the principles of their design. I have a basic understanding of how the original Girling system worked on Mk3/HST vehicles, but not the more sophisticated systems fitted to modern DMUs and EMUs. I have read that unlike the original Girling system which was independent, modern systems allow one wheelset to rotate at the full rotational speed for reference purposes and the WSP system aims to ensure that the other wheelsets are braked to rotate within 20% of the speed of the reference wheelset. I would appreciate anyone with technical knowledge of these systems helping me understand in more detail how they work, and specifically answer a couple of questions. 1. Is there a reference wheelset on each vehicle of a unit or only one per unit. 2. What is meant by the 20% of the reference wheelset - does this means that its average rotational speed is at least 80% of the rotational speed of the reference wheelset and how is this achieved.

 

[D365]

AFAIK, each vehicle of a multiple unit typically has its own WSP system.

 

[GC class B1]

AFAIK, each vehicle of a multiple unit typically has its own WSP system.

Yes. That is my understanding however with an effectively unbraked axle on each vehicle would that mean that only a maximum of 75% of the brake force will be available during WSP operation?

 

[millemille]

Yes. That is my understanding however with an effectively unbraked axle on each vehicle would that mean that only a maximum of 75% of the brake force will be available during WSP operation?

Modern WSP systems on fixed formation multiple units communicate between vehicles, typically splitting the train into two halves, so each vehicle doesn't need a reference speed axle.

But you are correct in saying that if the WSP systems are vehicle specific then the best stopping performance that can be achieved is 75% of clean dry rail.

However, it was proved nearly 30 years ago that better low adhesion stopping performance is achieved by allowing an axle to become unbraked and recover to true speed over ground for the WSP rack to use to control the other three axles than allowing all four axles to slide or lock with no true reference speed.

Effectively three controlled axles are better than four uncontrolled...

 

[GC class B1]

Modern WSP systems on fixed formation multiple units communicate between vehicles, typically splitting the train into two halves, so each vehicle doesn't need a reference speed axle.

But you are correct in saying that if the WSP systems are vehicle specific then the best stopping performance that can be achieved is 75% of clean dry rail.

However, it was proved nearly 30 years ago that better low adhesion stopping performance is achieved by allowing an axle to become unbraked and recover to true speed over ground for the WSP rack to use to control the other three axles than allowing all four axles to slide or lock with no true reference speed.

Effectively three controlled axles are better than four uncontrolled...

Thank you for your information, I was pretty sure you would be able to help me. One reason that I asked the question was I am interested in understanding how to estimate the Wheel/rail coefficient of friction from the deceleration rate achieved under WSP activity.

 

[100andthirty]

Thank you for your information, I was pretty sure you would be able to help me. One reason that I asked the question was I am interested in understanding how to estimate the Wheel/rail coefficient of friction from the deceleration rate achieved under WSP activity.

Estimating the coefficient of friction under the train is tough. The very action of WSP alters the coefficient of friction, so the value at the front of the train will generally be lower than the value at the back of the train. Indeed some suppliers are exploring whether the brake demand could be varied along the train. So, if the driver demands a "medium" brake, the WSP might conclude that a "low" brake rate is appropriate at the front of the train, "medium" in the middle, and WSP might have cleaned up the rail enough to allow a "high" rate at the back of the train giving an average "medium". This has been simulated but I don't know if this has been tried for real.

Another challenge is managing the electrodynamic brake (regenerative or rheostatic brake). Some organisations determine that it's simpler to dump the electrodynamic brake is WSP activity is detected and use the friction brake with per axle WSP.

 

[GC class B1]

Estimating the coefficient of friction under the train is tough. The very action of WSP alters the coefficient of friction, so the value at the front of the train will generally be lower than the value at the back of the train. Indeed some suppliers are exploring whether the brake demand could be varied along the train. So, if the driver demands a "medium" brake, the WSP might conclude that a "low" brake rate is appropriate at the front of the train, "medium" in the middle, and WSP might have cleaned up the rail enough to allow a "high" rate at the back of the train giving an average "medium". This has been simulated but I don't know if this has been tried for real.

Another challenge is managing the electrodynamic brake (regenerative or rheostatic brake). Some organisations determine that it's simpler to dump the electrodynamic brake is WSP activity is detected and use the friction brake with per axle WSP.

Thank you. I understand that it is impossible to get an accurate value of adhesion from deceleration and I am only looking to estimate an average value. As well as the available adhesion, the maximum retardation that can be achieved for each wheelset will also be influenced by variations in axle weight and more so by weight transfer under acceleration and braking.

 

[100andthirty]

A great deal is done in train design a) to match brake forces to axle weight and b) to minimise the effect of weight transfer. Weight transfer is a particular issue for freight locos starting heavy loads. As an example, the class 69 locos (class 56 structure/bogies/traction motors, class 66 engine and control system) can't deliver as much tractive effort as a class 66 because of weight transfer with the ex-class 56 bogies which aren't as good in this regard as the class 66 bogies.

 

[niggill617@gma]

I believe

Modern WSP systems on fixed formation multiple units communicate between vehicles, typically splitting the train into two halves, so each vehicle doesn't need a reference speed axle.

But you are correct in saying that if the WSP systems are vehicle specific then the best stopping performance that can be achieved is 75% of clean dry rail.

However, it was proved nearly 30 years ago that better low adhesion stopping performance is achieved by allowing an axle to become unbraked and recover to true speed over ground for the WSP rack to use to control the other three axles than allowing all four axles to slide or lock with no true reference speed.

Effectively three controlled axles are better than four uncontrolled...

When initially braking with active dynamic brakes the WSP light will often illuminate for a brief period, even in perceived good adhesion conditions. I presume this is due to a higher level of mu requirement, regarding less dynamic braking axles. Once the dynamic brake is automatically deactivated the friction brake takes over and WSP ceases as more braking axles require less friction.

 

[millemille]

Thank you for your information, I was pretty sure you would be able to help me. One reason that I asked the question was I am interested in understanding how to estimate the Wheel/rail coefficient of friction from the deceleration rate achieved under WSP activity.

In post incident On Train Data Recorder analysis I've carried out in the past you can provide an estimate, and it is only an estimate, of rail head coefficient of friction as follows.

Firstly, you need the gradient profile for the track that corresponds to the OTDR data under review.

Note the time and train speed and brake demand at the time of initial brake application and note the time and speed at which WSP activity first occurs.

Subtract the anti-jerk timing ( the delay built into the brake system which smoothly builds brake force to prevent jerking) from the period between brake application and WSP activity starting.

Subtract the train speed at the start of WSP activity from the train speed at the initial brake application.

Divide the train speed delta by the time delta and you've got a declaration rate.

Correct this for the gradient of the track, a falling gradient will reduce the declaration rate and rising vice versa so if the train is climbing a rising gradient at the time of the brake application then the gradient has helped the train to slow and it's help needs to be removed from the deceleration rate value.

So if you've calculated that the train was able to achieve 1%g retardation rate before WSP activity started, and the brake demand should have produced a higher rate of deceleration than that then it is reasonable to suggest that the rail head coefficient of friction was in the order of 0.01.

But this does not recognise the trains rolling resistance (parasitic friction in axle bearings etc. and aerodynamic drag) or which axle/axles were experiencing WSP activity.

And it assumes that static and dynamic friction are inversely and directly proportional to each other.

You can get cleverer than this and use the leading axle of the train, as it sees true rail head coefficient of friction, but this data isn't generally easily available on any train system.

 

[GC class B1]

In post incident On Train Data Recorder analysis I've carried out in the past you can provide an estimate, and it is only an estimate, of rail head coefficient of friction as follows.

Firstly, you need the gradient profile for the track that corresponds to the OTDR data under review.

Note the time and train speed and brake demand at the time of initial brake application and note the time and speed at which WSP activity first occurs.

Subtract the anti-jerk timing ( the delay built into the brake system which smoothly builds brake force to prevent jerking) from the period between brake application and WSP activity starting.

Subtract the train speed at the start of WSP activity from the train speed at the initial brake application.

Divide the train speed delta by the time delta and you've got a declaration rate.

Correct this for the gradient of the track, a falling gradient will reduce the declaration rate and rising vice versa so if the train is climbing a rising gradient at the time of the brake application then the gradient has helped the train to slow and it's help needs to be removed from the deceleration rate value.

So if you've calculated that the train was able to achieve 1%g retardation rate before WSP activity started, and the brake demand should have produced a higher rate of deceleration than that then it is reasonable to suggest that the rail head coefficient of friction was in the order of 0.01.

But this does not recognise the trains rolling resistance (parasitic friction in axle bearings etc. and aerodynamic drag) or which axle/axles were experiencing WSP activity.

And it assumes that static and dynamic friction are inversely and directly proportional to each other.

You can get cleverer than this and use the leading axle of the train, as it sees true rail head coefficient of friction, but this data isn't generally easily available on any train system.

Thank you for your explanation. Assuming I have understood your calculation correctly, I have a couple of questions. Your calculation seems to ignore any deceleration during the brake build up time although my understanding is that there will be some, albeit small deceleration during this build up time. When calculating stopping distances I use the formula in the standard which assumes half the build up time is free running and half is at the achieved deceleration rate, although as this time is measured to 95% of the maximum for the brake step, I add half a second to take into account the tapering off as the maximum Brake Cylinder pressure is reached.
You state that your method of calculation assumes that static and dynamic friction are inversely and directly proportional to each other. Are you referring to the static and dynamic wheel/rail coefficient of friction or am I misunderstanding the terms. I understood that the static wheel/rail coefficient of friction would be applicable until the wheelset rotational speed was lower than the true speed and slide began to occur therefore only static friction would apply until WSP activity commenced. Also I believe that dynamic friction will be less than static friction, but not necessarily in inverse proportion.

 

[millemille]

I believe

When initially braking with active dynamic brakes the WSP light will often illuminate for a brief period, even in perceived good adhesion conditions. I presume this is due to a higher level of mu requirement, regarding less dynamic braking axles. Once the dynamic brake is automatically deactivated the friction brake takes over and WSP ceases as more braking axles require less friction.

It could be.

It could also be differences in the allowable wheel creep ratio (vehicle speed over ground vs. wheel rotational speed) between the traction control electronics and the friction brake control.

Peak tractive/braking effort is achieved when a small creep ratio is allowed so WSP systems don't intervene as soon as any creep ratio is detected, they intervene once the acceptable threshold is breached or the rate of change of creep ratio indicates the axle is out of control.

On at least one fleet I've worked on the traction control electronics allowed a much lower creep ratio than the WSP system before attempting to control the axle.

So in your scenario the traction electronics are attempting to control a perceived unacceptable wheel slide and then they throw in the towel, so to speak, and hand brake control to the friction brake and the WSP system is fine with the level of slide that the traction electronics got their knickers in a twist over....

 

[niggill617@gma]

It could be.

It could also be differences in the allowable wheel creep ratio (vehicle speed over ground vs. wheel rotational speed) between the traction control electronics and the friction brake control.

Peak tractive/braking effort is achieved when a small creep ratio is allowed so WSP systems don't intervene as soon as any creep ratio is detected, they intervene once the acceptable threshold is breached or the rate of change of creep ratio indicates the axle is out of control.

On at least one fleet I've worked on the traction control electronics allowed a much lower creep ratio than the WSP system before attempting to control the axle.

So in your scenario the traction electronics are attempting to control a perceived unacceptable wheel slide and then they throw in the towel, so to speak, and hand brake control to the friction brake and the WSP system is fine with the level of slide that the traction electronics got their knickers in a twist over....

Just for argument sake, a rail wheel has a diameter of around 0.8 m. The wheel would rotate at 16 revs per second, at around 90 mph. If the wheel slides by 20 percent of the circumference within one revolution, the WSP lamp will illuminate and possibly latch for a short time. The speedo theoretically swings by 2 mph within 0.012 of a second. If the wheel slides by 20 percent on each revolution for 16 revs, the speedo will swing by around 18 mph within 1 second. The train is still in control, as long as the driver does not overreact.

Two doctrines of defensive driving during low adhesion: Brake early and light, or get the brake in and let sand and WSP do the job. Aside; ( some units may now have sanding in step 1).

In my opinion this means brake at the 3 percent g braking point nominal step 1. If after a period, the train is not slowing down select higher, possibly full service / emergency. Defensive driving is not initially selecting step 2 at the 6 percent g braking point, from line speed, if low adhesion is probable. The classic example being just after windy conditions, during the Autumn on a damp rail.

A train with dynamic brakes may indicate WSP for a short time. The friction brake will become active, with the associated build delay, this may give the wrong perception to the driver.

 

[D365]

Modern WSP systems on fixed formation multiple units communicate between vehicles, typically splitting the train into two halves, so each vehicle doesn't need a reference speed axle.

But you are correct in saying that if the WSP systems are vehicle specific then the best stopping performance that can be achieved is 75% of clean dry rail.

I suppose the 90s-era multiple units that I work on have a more primitive WSP system.

In post incident On Train Data Recorder analysis I've carried out in the past you can provide an estimate, and it is only an estimate, of rail head coefficient of friction as follows.

Firstly, you need the gradient profile for the track that corresponds to the OTDR data under review.

Note the time and train speed and brake demand at the time of initial brake application and note the time and speed at which WSP activity first occurs.

Subtract the anti-jerk timing ( the delay built into the brake system which smoothly builds brake force to prevent jerking) from the period between brake application and WSP activity starting.

Subtract the train speed at the start of WSP activity from the train speed at the initial brake application.

Divide the train speed delta by the time delta and you've got a declaration rate.

Correct this for the gradient of the track, a falling gradient will reduce the declaration rate and rising vice versa so if the train is climbing a rising gradient at the time of the brake application then the gradient has helped the train to slow and it's help needs to be removed from the deceleration rate value.

So if you've calculated that the train was able to achieve 1%g retardation rate before WSP activity started, and the brake demand should have produced a higher rate of deceleration than that then it is reasonable to suggest that the rail head coefficient of friction was in the order of 0.01.

But this does not recognise the trains rolling resistance (parasitic friction in axle bearings etc. and aerodynamic drag) or which axle/axles were experiencing WSP activity.

And it assumes that static and dynamic friction are inversely and directly proportional to each other.

You can get cleverer than this and use the leading axle of the train, as it sees true rail head coefficient of friction, but this data isn't generally easily available on any train system.

Ah ok, this explains why "WSP activity" has crept into the group standards for on-train data recorders. I suppose that a WSP signal could be added to the OTMR that is fitted on each driving vehicle.

 

[ComUtoR]

I suppose that a WSP signal could be added to the OTMR that is fitted on each driving vehicle.

I think all of the OTDRs at my TOC have a WSP signal.