Slow and steady wins the race
It is easy to say so, because hindsight is 20/20.
This post was initially planned as a “6 months later” review of CBTC operations on the North South Line, but in light of recent events, I’ve decided to take a slightly different tack.
I read with great interest the final report by LTA on the Joo Koon train collision that happened last week. You can read it here: https://www.lta.gov.sg/apps/news/page.aspx?c=2&id=2cd1e503-041c-4a1e-be4e-fd7e8723f99e
However, the provided explanation gives more questions than answers to me. No doubt, the causes are technically sound. What has happened has already happened, so a lot of issues I mention here will mainly be “in retrospect”. Could things have been done better?
SMRT: Procedural error, not human error?
Yes, SMRT’s VP was right to say that there was no human error per se. But this is Singapore, we are very much sticklers to the book. So could it be an issue with the prescribed procedures? The given explanation makes me think, that why wasn’t the following train being run manually?
To side track somewhat (yes, pun fully intended), all trains are able to operate in three modes:
Fully Automatic Operation (FAO) — trains are driven automatically, including opening and closing of doors
ATP-restricted Manual Operation (ATPM) — trains are driven by a driver, who also has control over door operation, and drives according to speed limits provided by the CBTC system
Restricted Manual (RM) — driver drives at his own eyesight and according to signals on the track; he is hardware limited to 18km/h.
(ATP means Automatic Train Protection)
I would think that if a train with failed signaling equipment is on the line, the train behind it should be running in ATPM, so that the driver is able to react should there be an issue with the train in front. Perhaps such protocols should be put in place in future? After all, all NSL and EWL trains (as at this time of writing) still have a driver to supervise operations, so the driver can be called upon at times like this.
Also, one would think that did Thales fail to predict such a failure (we will get to this later), and as such not advise such protocol to be put in place? Or, if that’s not the case, did SMRT and LTA heed Thales’ warnings?
LTA: Too fast, too furious?
The second set of questions is for the transport regulator. As I detailed in previous posts, I am of the opinion that this entire mess was caused by LTA’s rashness in wanting to get the signaling system rolled out as soon as possible.
The real question here is, why commission the Tuas West Extension and the new C151B trains on only the new signaling system? The mainline NSL and EWL sure could use the capacity expansion (as is evident with expanded service on the NSL post-CBTC), so why was the decision taken to wait?
After all, the London Underground also took it slowly. The 2009 and S stock were all delivered with the capability to run on LU’s old signaling systems (an in-house ATO system on the Victoria Line, and tripcock signaling on the Sub-Surface Lines). Only after all the old trains were replaced, did work on the signaling system begin.
As such, I wonder if it was possible that some (or maybe even all) of the C151B trains could have been delivered with the capability to run on both the legacy ATP and new CBTC systems, so that it would not have been a sunk cost, and that not so many C151Bs would have to sit waiting in Tuas Depot for nearly 2–3 years?
Likewise, I am sure it would not have cost a lot to install the legacy signaling system on the Tuas West Extension — in fact, advance work could also have been done by Thales, so that when the time came to commission CBTC, they wouldn’t have to do so much?
The opening of the Tuas West Extension had been delayed because the resignalling project was behind schedule, and latest reports say that the plan was for the project to end in 2018. Had the TWE opened with legacy signaling, it could easily have operated for over 2 years before the time came to switch on the new signaling system — thus having the bet pay off.
I may be being naïve here, but perhaps if LTA had seen the state of the resignalling project and realized that it was going south so quickly, perhaps they could have let out a last minute tender for the old signaling system. There would still have been delays while the supplier did its work, but maybe not so much as now, and of course the suspension of service thanks to the collision may not have happened too.
This was what was done at the Jubilee Line Extension, which was supposed to come with a new modern signaling system, but then plans were changed and regular LU signals were installed “in the interim” — which it kept for around 15 years. Same goes for the 7 Subway Extension — old NYCT signals at opening in September 2015, to receive CBTC by the end of this year.
Thales: Expertise at stake
I reserve the most questions for Thales. Let me start off by saying that I do not doubt Thales’ experience in resignalling programs. However, following such an incident, even if we were to attribute things to “bad luck”, questions must be asked.
Firstly, shadow mode, as well as phased transitions, should be a well-tested, well-run operation. After all, it was done first in London, and then New York, and who knows, in countless other places. MTR Corporation should also currently be perfoming shadow-mode testing on the Tsuen Wan Line, or if not, they should be starting soon.
So why did it fail here? Could the first failure (the removal of the “protective bubble” by the trackside device near Clementi) be partly due to Thales not fully understanding the legacy ATP system run on the NSEWL, and how to interface it with the new SelTrac CBTC?
For London, resignalling was already a complex, difficult task even with WW2-era tripcock-based mechanical trainstop technology. On the Singapore MRT, built in 1987, we have much more advanced systems than that.
Also, this is not the first time that SelTrac has caused some rather odd issues on the NSL. The great disruption of the evening of 28 June, caused by a mistake on Thales engineers’ part, also resulted in evening peak mayhem for hundreds of thousands. One would think that Thales had already learnt from that, so why are things still happening?
Besides that, it is very perplexing that when one of the two Vehicle On Board Computers failed, it resulted in a change of the train state, from six-car to three-car. One wonders whether unforeseen side effects (like a derailment, god forbid) could have happened had the train proceeded elsewhere on the line, seeing as the signaling system sees it as a three-car train whereas it’s actually a six-car train. Is that a design limitation?
Fortunately, it was a fault noticed by the train captain, and the train was already preparing to go out of service, but one wonders why it can still happen. A NSEWL train consists of two 3-car EMUs put together semi-permanently — one would assume that the opportunity be taken for a single VOBC should control the entire 6-car train.
For the interested, Land Transport Guru has also reported about other incidents of deadlocks and wrong track setting on their page here (http://landtransportguru.net/cbtc-implementation-review/), among other issues.
The importance of teamwork
That said, a complex and difficult project like the resignalling project would require lots of teamwork. Much has been said about a rather difficult working culture within SMRT, including a rather interesting thread (now deleted, so don’t try looking) by a user claiming to be a Station Manager.
Putting that aside, though, one wonders what is the dynamic between the staff of LTA, SMRT, and Thales. Are they a highly integrated, cohesive team, able to quickly put heads together to solve any problems that may arise, or are they a loosely-coupled operation, with a firm client-supplier relationship, and hence causing a mentality of “the customer is always right” or “the supplier knows best”?
I believe it all falls to a matter of all parties having the right expectations of each other. And that can only be achieved by good teamwork, knowing what each party is able to deliver, and having realistic goals.
The signaling project was already delayed — as pointed out by both LTG and me previously, it is very difficult to pull it off “seamlessly” and “smoothly”. I believe that what the team should have done was to close ranks, learn from their mistakes, and never repeat it again.
Ultimately, almost 6 months later, were any lessons learnt? How have they been learnt? Will they be applied in the rest of the project down the road? I guess, only time and travels will tell. “Accountability” is an easy word to throw about — all the more since SMRT CEO Desmond Kuek alluded to “deep-seated cultural issues” in the company that he has not been able to fix in his five years there.
But what use is accountability when there’s so much work left to be done? What use is asking for heads to roll when what the hands need most now is a steady leader? I believe we should just close ranks, work together, and finish the work ahead of us first. Only then should we settle the bill.
Of course, there is no crying over spilt milk, what has happened has happened. It’s important that everyone learns from the incident, and make sure that it can never happen again.