Joo Koon train collision: 'No human error' involved, SMRT says

Joo Koon train collision: 'No human error' involved, SMRT says

There was “no human error” involved in the Joo Koon train collision, train operator SMRT said on Tuesday (Nov 21). Instead, the accident had been caused by the unexpected disabling of a protective feature” on the train that was hit, SMRT and the Land Transport Authority (LTA) confirmed in a joint media briefing.

SINGAPORE: There was “no human error” involved in the Joo Koon train collision, train operator SMRT said on Tuesday (Nov 21). Instead, the accident had been caused by the unexpected disabling of a protective feature” on the train that was hit, SMRT and the Land Transport Authority (LTA) confirmed in a joint media briefing.

Last Wednesday, 38 people were injured after a software glitch in the East-West Line signalling system caused two trains to collide at Joo Koon MRT station. LTA earlier said that the glitch arose when a faulty train was transiting between the old and new signalling systems.

Authorities explained that two signalling systems actively operate on the EWL - the new Communications-Based Train Control (CBTC) system between Tuas Link and Pioneer MRT stations, and the legacy signalling system, which controls train movements between Pioneer and Pasir Ris stations.

Following the accident, a total of 19 MRT stations will be fully closed on two Sundays in December to "significantly accelerate" resignalling work on the East-West Line.


Giving more details on the cause of the collision, SMRT and LTA said the train in question, which was launched at 5.38am from Ulu Pandan Depot, passed a trackside device at Clementi which had yet to be modified for compatibility with the CBTC system. As the train travelled east from Ulu Pandan depot to Pasir Ris MRT station, and then back west towards Pioneer MRT, it was controlled by the old signalling system. The CBTC system took over active control once the train entered the CBTC-controlled area west of Pioneer MRT station.

Findings from the investigation indicated that an "abnormal condition on a train-borne CBTC signalling equipment" disabled a feature designed to apply a protective "bubble" around the affected train to ensure safe distances between trains, said authorities.

When this happens, the CBTC system should immediately apply a second protective "bubble" to replace the first one, authorities added.

However, this second bubble was "unexpectedly disabled" when the train passed by a "trackside device" at Clementi, which had "yet to be modified for compatibility" with the new CBTC system.

When the train arrived at Pioneer MRT station, where it transitioned to the new CBTC system on the Tuas West Extension, the train captain "correctly detected" the abnormal condition and reported it to the Operations Control Centre, authorities added.

Alvin Kek, SMRT’s senior vice president for rail operations on the North-South and East-West lines, explained that by procedure, when the abnormality was reported, the train would be routed to Joo Koon for passengers to alight.

When it stopped at the station, a "closed track" protection was automatically activated to prevent other trains from entering or leaving the platform. As the train had an abnormality, the doors could not open automatically and the train could only move in what was termed a Restricted Manual mode, which has a maximum speed of 18 kilometres per hour.

The driver of the train had to step out of the driver’s cabin to operate the platform screen doors at Joo Koon manually. Once passengers had detrained, he would then have to manually close the platform screen doors, before returning to the cabin to operate the train manually.

“There is time needed for the driver to do this,” he said.

The moment the platform screen doors closed, the “closed track protection at the station was lifted, he added, and the rear train did not sense any protective bubbles around the front train, and started to move forward automatically, resulting in the collision.

The second train, which was in automatic mode, was not under the control of the driver.

Thales, the French company which supplied the new signalling system, noted that the incident happened because the day before, the passive CBTC system - which runs in the background to collect data for performance monitoring - was connected to the active CBTC system for the first time. "Before that, there was an isolation, so if the incident had happened before, by the time the train made it to the active system, it would have created a new bubble," said senior engineer Marcin Dedys. 

"That was a unique situation and Thales had rectified that, so that is the only time where the train passed by the particular point, where the NCO (Non-Communicating Obstruction, which is the second protective bubble in the CBTC) was actually disabled," added LTA chief executive Ngien Hoon Ping. "That particular case has not happened, and will not happen again."

Thales added that while there are similar systems in other countries, there have been no problems in any of them. 

Speaking after the news conference, Transport Minister Khaw Boon Wan commented that Thales "could have done better", even as he acknowledged that the firm was faced with a "challenging situation" having to deal with two different signalling systems on the line.


Asked if the driver of the second train would have been able to prevent the collision in the 10 seconds it took for the train to hit the one in front of it, SMRT said that, by its assessment, the driver “did not have enough time to react”.

“We have now confirmed that the second train stopped at a distance of 36 metres away, but the duration for it to move from a stationary position to hitting the front train was 10 seconds,” said Mr Kek.

“We assessed that within this 10 seconds, it would be very difficult and challenging for the train captain to understand what is happening,” he added. Train drivers, he said, are trained to react when they feel that the trains are coming too close.

“However, given that this happened during a period of 10 seconds, our assessment is that our train captain has to first orientate himself towards what happened, because he is controlling a train that is in automatic mode,” he said. “He has also been taught in courses that in all train systems, there will be at least one protective bubble around it.”

“So he now has to orientate, and tell himself that it is coming too close, while at the same time, there should be a bubble, but then again the bubble may not have been in place,” he added.

“This is why we think he did not have sufficient time to react in this incident.”


Despite the explanation, transport analyst Park Byung Joon said there are three big questions arising from the incident.

The first, he said, is with regards to the driver’s reaction time, and the way drivers are trained.

“Drivers were trained in a way where they believe there will be one (protection) bubble there,” said Dr Park, who is a senior lecturer at the Singapore University of Social Sciences.

“Ten seconds is quite a long time. If the driver did not do anything because he was waiting for the system to automatically extend and stop the train, then this driver training manual is wrong. The revision for this driver training manual needs to be done immediately.”

Dr Park added that the very fact that the driver was waiting for the protection system to kick in automatically is “not really satisfactory”.

“It could have been a man on the track,” he said. “Whether the train is on the system or not, the driver must be authorised to jump in and stop the train when he feels is necessary.”

Another question, he said, is about incident management and how SMRT prepares for contingencies.

In this case, he said, the contingency was that the safety protection measure under CBTC was lifted. In this scenario, how does the train recognise it, and how can it be detected by trains on other levels of the system, Dr Park asked.

“How much preparation and tests were carried out for this particular contingency?” he added.

Another issue arising from the train collision is a “fundamental issue”, Dr Park said, noting that currently, it appears that safety measures are entirely handled by software.

“Are we prepared for a situation where all software fails, and do we have physical protection as an ultimate backup?”

Additional reporting by Calvin Hui

Source: CNA/dt