Merlin found at fault for Smiler accident

Riders =/= a safe ride.

Another issue is that 2 trains are now gone, and the lengths of track where the collision occurred, trains valleyed, and where the trains finally stalled will need to also be examined thoroughly. It definitely won't be operating for the rest of the season because of this.

In my opinion, I just don't see Alton Towers keeping this thing around. :/

[gouldy]

Nah, the absolute worst thing that will happen is that it'll only run one train operation for the rest of time.

I very much doubt it'll be open again this season, but during that time Gerst will be on site and will be sorting things and putting other failsafes in place so that it can't be manually overridden which is clearly what's happened here. The most extensive the work could get would be if they had Gerst rework that element and refit it. The number of stalls the ride has had there is beyond a joke, and it has now played a part in a major accident.

I didn't know it was prone to valleying there. Too bad. I can't even imagine one train operation on that ride. Since there are four trains and two are gone, hopefully they'll still run two. If it was running only one train it might as well be closed lol.

[Mikey]

We were talking about this on twitter but I have an educated guess about how this situation happened.

First a little backround:

Ride Control systems work off a series of inputs and outputs, inputs are made of push buttons, proxy switches, photo eyes, mechanical limit switches, encoded motor feedback which is generally a serial counter input (Pulse Width Modulation) or any number of circuit of data input types. The outputs on the PLC handle the control of the subsystems such as opening air valves to open brakes or lower mag jets, starting feed motors by engaging the contractor (Lower voltage circuit that engages a high voltage/amperage device like a lift motor). A PLC program is developed that manipulates these devices based on the inputs it is getting from the sensors. The computer is essentially blind unless one of these sensors is triggered to indicate a units position which is why start up with units in a home position is crucial to the rides operation.

The computer does not care that there are two units in service, it only cares that when proxy switch A and B are activated that brakes C and D remain closed. It knows that once proxy switch C and D are activated and deactivated 8 times and then off for more then two seconds straight that it should turn the lift output off. It knows that the wind sensor is sending it via serial the integer 09022 which means the winds are out of the east at 22 MPH and it needs to launch the train with a little more oomph or its going to roll back. It knows these things because it was programed by a human to react in a certain and predictable way to the data it has.

I think this situation started with the ride Trouble Lighted because the safety or main PLC caught something it didn't like. (Computer errored out and triggered a stop event.) Maintenance comes to investigate and resets the trouble light. A trouble light in most modern systems is the equivalent of an E-Stop but is generally more graceful in its shutdown then simply removing power from the ride like an E-Stop and letting mechanical fail safeties kick in. They are looking at the ride and the trouble light error does not indicate the real underlying problem but rather a code code that is a symptom of the problem. Maybe block related. Maintenance resets the trouble light but can't see, forgets to check, or needs to over ride the safeties to get units back into a home position to restart in automatic mode.

The problem is when the PLC is reset during a trouble light or E-STOP, the PLC starts at the top of the program again. This means that any unit position information has to be relearned based on currently triggered sensors. The computer may see the train in the station, in the safety brakes, at the top of the lift, and in the barn, but it can't see that train that vallied because there is no sensor there. Sensors generally are only installed near block zone equipment and not out in the track unless one is trying to determine an input speed or exit speed for a trim break. But this part of the logic would not generally connect into the logic that handles blocking.

Once the original zone blocking data was gone, the ride PLC did what it was programmed to do and because there was a physical ghost train it couldn't see and with out human intervention they had an accident. This could have been avoided, and is generally not a problem when you are only running two or three trains. On a ride like the smiler that runs 5-6 trains positioning data becomes very important as does the process for restarting the ride during a stop event.

Different ride manufacturers have different ways of addressing this problem including only permitting block resets to be completed from a keyed service panel that has a clear line of sight to the section of track, to requiring that all the E-STOPs be pushed and pulled back out in order of zone priority in order to insure that someone has inspected all zones for rogue trains. Personally I have seen other people get into trouble very quickly with the ride in maintenance mode with maintenance over ride enabled. The later mode is not designed to be used when units are loaded with live cargo because it disables all of the safeties. You use this mode to do heavy maintenance, to stack all your boats on lift to drain the water, to short launch trains to warm the system up, to offload trains into a sheltered single place for the winter.

Most places would rather unload the guest before going into maintenance over ride because of the high risk of something working unexpectedly. None the less, this is all speculation and the root cause analysis could determine that this is a result of solar flares.

[TTD03]

^ Wow Mikey. I think this could have been completely avoided if all rides were like Dragster, and have a number of trains switch from, for example, 1-6. That way the ride would know that if you were running 5 trains, 5 different blocks should have been triggered, and if only 4 were triggered, then the ride should have detected an error and not have started.

[JAMMY]

Great write up Mikey. I too think that is the most probable cause, which will require a definitive reevaluation of operational procedures.

I don't know if it's because I followed construction so closely but this whole incident has made me feel ill. :/ Still won't stop me riding, we all know the risks, but enjoy the coasters anyway as we should.

I was waiting for Mikey to weigh in on this, haha. So you're saying it was human error. Dang. That's quite unfortunate. How could one not think to check the track OR at least wait for the empty one to return before sending a fully loaded train? =/

In related news, Merlin's stock plummeted today. This accident is and will really make the company look bad. The ride may be SBNO for a long time, but I can't say if they will ultimately remove the ride or not.

I am only guessing but with a 2 year old coaster that cost $18 million, I don't think its going anywhere. Then again, I didn't think Excalibur at Valleyfair was going anywhere either.

Updated to SBNO on RCDB

http://rcdb.com/11062.htm#

[mkingy]

Mikey wrote:

-explanation snipped-

Thanks for the detailed insight Mikey, if I were a betting man that's the theory I'd put my money on.

If I understand you correctly - would a sensor installed at the location of the stalled train by sufficient to prevent this occurring following a reset if what you described did actually happen?

And if so - with the history of the train struggling in that area would/should this have crossed the designers mind or is it a case of captain hindsight?

[gouldy]

"Human error likely in Alton Towers rollercoaster crash, former staff claim"

http://www.theguardian.com/uk-news/2015 ... taff-claim

[Coasterkidmwm]

Mikey wrote:

[lots of relevant text here involving a non-Estop shutdown]

Does Six Flags allow bottom-tier ride ops to clear errors? When I worked at CP we were allowed to do that (after being told to by the mechanics) and I'm wondering if an employee didn't understand something correctly just went ahead and cleared an error at the worst time ever.

I was told by other employees that this was an unusual practice and places like King's Island and Carowinds don't do that, but given that it's not hard to do I am not sure if I believe that.

"...the Smiler was manufactured by German firm Gerstlauer, which also built the trains for a rollercoaster in Six Flags Park in Texas from which a woman fell to her death in 2013." Media!!! Sucks for Gerstlauer--as this is almost certainly not their fault if it was in maintenance mode or whatever. That's a necessary override and nothing seemed to have failed mechanically, they designed it properly. More details could surface though. I can't imagine the headache of an environment at Gerst's firm from all of this.

[Paradox]

I hate how people still blame Gerstlauer for NTG when it was mainly human error that caused the accident.

Regardless of the issues it's had, I don't think you can just give up and remove a record breaking attraction after a few accidents.

[gouldy]

^ The British public seems to think differently. Social media is a very frustrating place to be at the moment for a UK enthusiast it has to be said...

^ They are going a little bit hysteric. Temporarily remove the train to investigate or permanently remove the track itself... they sadly seem to choose an unrealistic solution. Optimistic or pessimistic.

[slosprint]

Since what Mikey says is probably accurate, why don't rides simply save last known train positions on non-power reliant memory, like a hard drive, during a reset as an industry standard, or have sensor data not be reliant on the power cycle of the rest of the system? We are talking about less than 1 Kb of data here, and assuming the train had valleyd at the time of the reset, ride operators would have been told exactly what section the train was on, or what section had not picked up data, and been forced to count trains. Maybe I'm just being captain hindsight here, but it seems like an easy and inexpensive fix.

[Coasterkidmwm]

rcking04 wrote:

This is among the most serious of issues this ride has had.

The trains stalling.
The track pieces breaking apart to reveal gaps between them.
This major accident due to a failure in a computer system that will result in a multitude of lawsuits and settlements, and during which the ride will most likely not be allowed to operate at all while under investigation and pending court cases.

It just seems like this is an example of a project too ambitious to be viable...

This ride just might end up along the lines of Windjammer Surf Racers.

Are you trolling or what is your deal?
1. "Established" rides like Raptor are prone to valleying as well. I saw it Estopped halfway out of the block brake at least 3 times last year in the morning, and I wasn't even around when it's cold and more likely to happen.
2. Construction company
3. They don't know the cause of the accident yet.

[mkingy]

Further updates naming two of those injured.

Source: http://www.bbc.co.uk/news/uk-england-32997703

Two teenagers have been named as victims of the Alton Towers rollercoaster accident.
Joe Pugh, 18, and his girlfriend Leah Washington, aged 17, both of Barnsley, were riding on the Smiler on Tuesday.
Two carriages crashed on the ride at the Staffordshire theme park, leaving some passengers trapped for four-and-a-half hours about 25ft (7.6m) up in the air.
Mr Pugh and Miss Washington and another man and a woman suffered leg injuries.
A man who operated the Smiler ride when it opened in 2013, who gave his name only as Mark, has expressed his frustration the accident happened.

Meanwhile, Oaks Working Men's Club, in Barnsley, where Mr Pugh works said: "We at The Oaks WMC would like to extend our prayers and best wishes in the speedy recovery of our barman Joe Pugh and his girlfriend Leah Washington."
Joe Pugh
Joe Pugh, 18, works at The Oaks Working Men's Club in Barnsley
Mr Pugh's friend Alice Holmes spoke of her shock to the BBC's Mark Ansell.
She said: "I text saying 'Are you ok? Smiler has crashed' and I never got a reply.
"I text him again and again and he still didn't reply. Then, this morning I saw a clip on the news and I knew it was him from his outfit because he dresses quite nice."
The BBC's Declan Wilson spoke to friends of Miss Washington, who said she was the more badly injured. One friend said she had injuries to her legs, hips and hands.

The carriage was travelling at 20mph when it crashed
Alton Towers said on social media that the park would remain closed on Thursday, as an investigation into the cause of the accident continued.
Nick Varney, chief executive of the park's owners Merlin Entertainments, said: "A full investigation is now under way and we will continue to work closely with the emergency services and the Health and Safety Executive to better understand the cause of this terrible accident."
The accident happened at around 14:00 BST on Tuesday, leaving 16 people stuck in the air at an angle of about 45 degrees.

People at the theme park reported hearing "a loud crash" when the collision happened
The park said the carriage carrying the people was travelling at 20mph when it crashed.
Four people were airlifted to major trauma centres across the West Midlands.
People at the theme park reported hearing "a loud crash" when the collision happened.
'Get well soon pal'
Mr Pugh, a textile design student at the University of Huddersfield, has been sent supportive messages on social media from friends.
"Wishing Joe Pugh and his girlfriend a speedy recovery," Kay Eggleston wrote.
Ross William Gibson wrote: "Can't believe when I wrote this I didn't know one of the four was my friend Joe Pugh and his girlfriend. Get well soon pal."

The Smiler was the first of three Infinity Coasters built by German manufacturer Gerstlauer
The other two coasters are at PowerPark in Finland and Erlebnispark Tripsdrill in Germany
The coaster in Finland, called Junker, has only been open to the public since Saturday and is running as normal today
Mikko Kiviluoma from PowerPark said: "Our rollercoaster Junker is a brand new model with the latest technology, so as we know it is a different model. Junker was built up during this spring 2015 and tested over 2,000 times before it opened. Safety is number one thing in PowerPark and we do normal checks every day for all our rollercoasters and other equipment."

Mark, who operated the Smiler ride in 2013, said: "I am frustrated that such a simple error has been made, knowing myself how the system should function and that it is one of the safest rides I've ever worked on.
"One carriage should only be in a certain block section at a time.
"When the empty carriage didn't make it to the next block section after a certain amount of time, the ride should have stopped any other carriages being sent and entering that block."

Coasterkidmwm wrote:

Are you trolling or what is your deal?
1. "Established" rides like Raptor are prone to valleying as well. I saw it Estopped halfway out of the block brake at least 3 times last year in the morning, and I wasn't even around when it's cold and more likely to happen.
2. Construction company
3. They don't know the cause of the accident yet.

We'll see what happens! Personally, I don't want to this ride go.

This ride though from here on out may not operate the same though if it did remain. The computer system behind operation may change. The trains could change. The ride may be reprofiled.

By the looks of the media there, there's a ton of bad publicity circulating.

I'm just being doom and gloom and betting the ride will be removed.

[mkingy]

Link for your optimism rcking http://www.bbc.co.uk/news/uk-england-st ... e-32993933

[Mikey]

Coasterkidmwm wrote:

Does Six Flags allow bottom-tier ride ops to clear errors?

No, they can't even unlock the restraints when a trouble light happens. There are some ride specific exceptions to this policy but all of those involve unloaded units and a safety check procedure.

slosprint wrote:

Since what Mikey says is probably accurate, why don't rides simply save last known train positions on non-power reliant memory, like a hard drive, during a reset as an industry standard, or have sensor data not be reliant on the power cycle of the rest of the system? We are talking about less than 1 Kb of data here, and assuming the train had valleyd at the time of the reset, ride operators would have been told exactly what section the train was on, or what section had not picked up data, and been forced to count trains. Maybe I'm just being captain hindsight here, but it seems like an easy and inexpensive fix.

While this sounds like a solution that would be easy to implement, ride computers are not like your home PC. Think of a ride computer as a giant calculator. It's very good at giving you correct data reliably so long as the inputs are correct, but if you take the battery out, use the clear button then all the data you just entered is lost. PLCs are very simpleton, they have no persistent storage outside of a flash card that a working program is written. The PLC cannot write to this card, but it can output data to a separate storage for logging but it can't ready back data from the system it outputs to.

PLCs use lader logic as a programing language. The program is stored in flash and loaded when the PLC_START function is called. When a safe stop needs to occur, a PLC_STOP function is called. This unloads all data from memory. By default if power is removed, the PLC reverts to PLC_STOP until PLC_START is called in a start up sequence. There is no way to preserve the set values once a reset has commenced. You can only check against the existing values you have at start up.

I know this sounds like a limitation, but these controllers are designed to handled millions of cycles and do so with out fault or error. They have been trusted in service for over 35 years.

Alton towers have tweeted that the park is closed for a third day due to the ongoing investigation.