Speech by Minister Khaw Boon Wan at the Fourth Joint Forum on Infrastructure Maintenance

Good morning Friends and Colleagues, 

1.     Last week (22 July), I read a thoughtful article in the New York Times. The title was “Let's get excited about maintenance!”  It is an American newspaper and article, jointly written by two American professors (Prof Lee Vinsel of Virginia Tech and Prof Andrew Russell of SUNY Polytechnic Institute). They were prompted to write this article following New York Governor Andrew Cuomo's declaration of a state of emergency for the Metropolitan Transportation Authority (MTA). MTA runs the metros in New York State and City, and largely relies on emergency repairs because it does not conduct sufficient preventive upkeep. It is not that they did not know, they just did not have the money. All these years and decades of neglected maintenance added up. As engineers, we all know that the most basic level of maintenance is to wait until it falls apart. This is the lowest level and most costly form of maintenance in the engineering profession. Of course, if you are a little better, you go for preventive maintenance - so you prevent or remove the problem before it happens based on data based on experience. And of course, in the field now is predictive maintenance, based on data, technology and sensors, trying to predict when we should replace and fix issues before anything happens. Otherwise, mindless predictive maintenance can be very costly. That means, you may replace things a little too early. While in this way, you can improve reliability substantially, it will be at a great cost. Using predictive maintenance allows you to optimise by achieving the same purpose for the lowest possible cost. But that was the sad state in New York. You may also have read about two recent train derailments there, as another example of the costs of neglecting maintenance.

2.     Sadly, the two Professors noted that the neglect of maintenance is not limited to New York, or to the metros. “All varieties of American infrastructure - roads, bridges, airports, sewers - are in decrepit condition…The American Society of Civil Engineers considers 17 percent of American dams to be 'high hazard potential'.”

3.     They asked a rhetorical question: “Why are we in this predicament?” Their conclusion: “Americans have an impoverished and immature conception of technology, one that fetishizes innovation as a kind of art and demeans upkeep as mere drudgery”. In other words, they treat innovation as God, but along the way, neglect maintenance. They stressed that “While innovation… is important, most technologies around us are old, and for the smooth functioning of daily life, maintenance is more important.” They quoted one study which estimated that “roughly 70 percent of engineers work on maintaining and overseeing existing things rather than designing new ones”.

4.     Against this reality, they lamented that the American society has not only failed to appreciate maintenance, it also fails to appreciate the maintainers themselves. “We do not grant them high social status or high salaries. Typically, maintenance is a blue-collar occupation: mechanic, plumber, janitor, electrician. There are white-collar maintainers (like the I.T. crowd) and white-jacket maintainers (like dentists). But they, too, are not celebrated like the inventor”. They urged Americans to think harder about maintenance, and to “grant it the prestige and the funding it deserves”.

5.     Actually, many successful corporations do. I am sure 3M does. General Electric and Boeing make heavy investments in tools and procedures for predictive maintenance, since their success depends on the reliability of their products and the existence of orderly routines to follow when things break down.

6.     I am sure the article has resonance with all of you.  While our infrastructure is in better shape than that in the US, we cannot say that Singaporeans view maintenance and upkeep as prestigious as say the Japanese, who have a strong culture of deep mastery and respect for craftsmanship. Although as I interact with my Japanese friends, I understand that the new generation of Japanese is not quite like the old generation, so in other words, there has been some erosion of this culture as well. Anyway, we are not here to debate which is more important: innovation or maintenance. The fact is both are critical.  But we can do more to invest in maintenance as good quality upkeep is absolutely essential for infrastructure.  And all the more so when Singapore is no longer young, and we have accumulated our fair share of legacy infrastructure.

Rail Reliability Performance

7.     In MOT, we experience the challenges of legacy infrastructure directly.  Much of our rail reliability problems are due to ageing infrastructure. The North-South and East-West Lines (NSEWL), for example, are turning 30 and due for a complete upgrade. I will find a separate opportunity to talk about the need of upgrading of NSEWL, what we need to do to complete this upgrading exercise and what is involved. We have completed sleeper replacement, third rail replacement is completing soon, the purchase of new trains, the power system - but it is more than that. This is why raising reliability for NSEWL takes time. That is why I also always stress and acknowledge that yes, we are facing problems, but please bear with us as we need time for the complete replacement of NSEWL. But I was also clear that Singaporeans cannot wait too long. When I joined MOT in 2015, I gave ourselves four years to up the reliability of our rail network. Originally I was clueless what the timeline should be, but my political timeline is five years. Of course I did some checks, four years would not have been sufficient to do a complete replacement but would have allowed us to make a major difference. We are now approaching the half-way mark. What is the progress?

8.     We have two indicators to measure our progress, which you are familiar with: (a) the MKBF (Mean Kilometres Between Failure) which tracks the frequency of any delay exceeding 5 minutes; and (b) the number of major delays exceeding 30 minutes. Of course, all delays are not welcome, but major delays are especially troublesome and disruptive. That is why we track them separately. Also, these are the two indicators that are used by major metros around the world, and therefore, it allows us the opportunity to benchmark ourselves against them.

9.     So what have we done on MKBF? In 2015, our overall network (averaged out) MKBF was 133,000 train-km, not even 150,000 train-km. That explains why Singaporeans were so unhappy. In the first half of 2017, it has happily gone up, to nearly 400,000 (393,000) train-km. This is based on preliminary data, because we are still scrubbing June data, and has also excluded the delays due to the testing of the new signalling system, that should be excluded because this is one of those one-in-thirty-year incidents. For an apple-to-apple comparison, it has to be excluded. It is a topic which I will spend some time talking about in the later part of my speech.

10.    We only have five MRT lines, so it is not a complicated network. It is useful to take a good look at the details of each line. They are different, but the common factor, fortunately, is that all five lines are making improvements, compared to one to two years ago. The best performer, the North East Line, has achieved nearly 1 million (978,000) train-km. The Downtown Line and the Circle Line have both exceeded 500,000 (518,000) train-km. Even the ageing North-South and East-West Lines also achieved 345,000 and almost 300,000 (282,000) train-km respectively. More importantly, look at last year and the improvements we have made.

11.    I consider this outcome satisfactory. This is not yet the steady state we are hoping for, but versus what we inherited two years ago, I think this is maybe a C+. Next year we should go for B. The following year, certainly before election comes, must be A. I consider this satisfactory against our targets of 300,000 MKBF this year, which we have now exceeded, 400,000 MKBF next year and 800,000 train-km by 2020. I set these targets last year, referencing the best performance in class, which is Taipei Metro. We met them, talked with them and understood how they did it. That time, their performance was 800,000 MKBF. So I thought, well, minimally we must close the gap with Taipei within this 4-year period. And we can't do it overnight, because I know we need time to replace the assets. Therefore, that is how we set these targets. Since then, Taipei Metro has made further improvements. They achieved 1,000,000 MKBF last year. I would not be surprised if next year, when we meet again, maybe they would have crossed 1 million too. So I am upping our 2020 target from 800,000 to 1,000,000 train-km. I think we have to do that. LTA and the Operators are working closely towards this new target: 1 million train-km by 2020.

12.    The second performance indicator is just as encouraging. The number of major delays exceeding 30 minutes has gone down from 10 in the 1^st half of 2016 to 3 in the 1^st half of this year. Not easy at all. While we are encouraged by the results, we all know we cannot be complacent. Even Hong Kong MTR, which is a global leader in rail reliability, recently has had to cope with two separate power faults. We know that maintenance excellence is a continuous process. The job is never really done.

Re-signalling on the North- South Line 

13.    Let me now spend some time discussing the on-going re-signalling project on the North-South Line (NSL). I thought I should do it because I don't like the media reporting, which has magnified the problem unfairly. I think they were being unfair to the team working their guts out on this re-signalling project. So let me go through slowly and carefully, what this is all about. Firstly, why are we changing the signalling system? The answer is simple.  The existing system is old, out-dated and in fact has been a cause of many train disruptions in the past.  The continuing problems on the East-West Line (EWL) that you hear reported in the media and sometimes played up, are partly due to this. Because we haven't started re-signalling EWL yet. Some of these problems you see on the EWL, in fact, are partly due to the old signalling system, which ought to be replaced. We will replace it but we have not started yet, which I will explain later. So that's the first reason, an old system that has to be replaced.

14.    Secondly, how will the new system look like? Will it be better than the old system?  Yes, of course. Whatever we do, we always find the opportunity to upgrade the system so it is not just more of the same. So we are using the opportunity to tap on the new technology because technology has changed during the last 30 years. For example, the new system uses frequency hopping. This is a much more robust system which can automatically change its radio communication frequencies whenever it detects any interference in its radio frequency band. Singapore is so digitally-wired and we are worried about cyber-attacks, vandalism or just interference from Singaporeans going about their daily lives. This frequency hopping technology will be a big help.

15.    Third, how will commuters benefit from the new system? The biggest benefit is that the new signalling system will allow us to pack the trains closer, or using the technical term, reduce the headway (the time between two trains) from the current 120 seconds to 100 seconds.  At 120 seconds, the current signalling system is supporting 30 trains per hour during peak period. So, even though we know the stretch is crowded and need to add more trains, we can't. That is the maximum we can do, 30 trains per hour. This is the situation today. But with the new system, at 100 seconds headway, we can add six more trains per hour during the peak period. 30 to 36, when the train is six cars, this is a huge increase in capacity. The experience should be visible and you can feel it. This is a 20% increase, a substantial improvement in capacity. It will make the trains less crowded and the commuter experience more pleasant.

16.    Fourth, when can commuters experience this improvement, when will this come about? For the NSL, as soon as the new signalling system and the performance of the train-borne signalling hardware stabilise, we will be able to realise this increase in train capacity progressively. In other words, we will be able to add more trains once this system stabilises. I shall elaborate a bit later. When will this be? Sometime this year, I hope sooner rather than later. It is now July, August, a few more months… Certainly we do not want it to stretch into next year, because we have the EWL to deal with.

17.    So why are we taking so long, from a commuter's point of view? Why are there so many problems? Because changing a signalling system is complex.  How complex is complex? First, all 141 trains have to be upgraded with new devices such as Vehicle On-Board Controllers (VOBC), antennas, speed sensors, proximity sensors, transponder readers. All 54 stations' existing relay interlocking has to be upgraded to computer-based interlocking.  All 95 km of tracks have to be upgraded with new cables, transponders, antennas, proximity plates, radios and radio signal equipment.  The three existing depots have to be upgraded too, and the existing Operation Control Centre at Victoria Street will also need to be upgraded. In fact, because it is too small, we actually have to relocate it to Kim Chuan Depot, and that in itself will add further complications. We are going to do all this while the trains are still running. In some of the projects that we have discussed with the supplier Thales, they said they had the luxury of actually closing down the line for three years while they changed the system. Whereas here, if we try three years, we will lose the government! So that cannot be done, we have to do it simultaneously. I hope even as a layman, one should be able to appreciate the difficulties involved. These are not simple excuses that we are giving ourselves.

18.    These are real challenges, and we were forewarned by London Underground and Taipei Metro, because they had done similar re-signalling works in recent years. They were the ones who first warned us there will be tons of problems, there will be flaming of you and everybody else on social media. We almost said let's forget about this project if we can. But we can't. So I said, how long is the agony? London said up to a year. They said it would take months, these are not problems that we can resolve in weeks. There will be many glitches and teething problems before the system fully stabilises. The system will stabilise, but you need to go through this painful process. Very quickly, I shared the experience of London and Taipei in Parliament at the beginning of this year, so that the MPs will be forewarned too, that yes, we have to do this, but expect trouble. We will do our very best to minimise the problems, but there will still be problems, so please bear with us. No pain, no gain. We have to go through this, things will get worse before they get better, but they will get better. Not only that, I advised our team, please don't rush, even though prior to that they had made a commitment that they will start NSL last year. I said, let's do more tests during engineering hours, identify the glitches, fix the software, install, test and retest until it runs properly, before we throw open to passenger service. That was the reason why we pushed back the timeline, to give us those precious few more months. But you know our difficulty? Engineering hours for running MRT lines are only three miserable hours every day. That's all. How many tests can you do during those three short hours? Why do other metros close down lines?  Because it gives them the entire 24 hours to run and do these tests, not just computer simulations. You have to actually run all 141 trains, and run it, load it, and see how it goes. Identify the glitches, upgrade the software, and reiterate. That is what we do. Even then, there will still be delays, but my strong belief is, if Singaporeans know that we have done our best to minimise the glitches, I think they will be understanding and I hope they will be forgiving. This, I think, is the case because as you know, I take trains quite often, and commuters come forward to express appreciation and gratitude. I know NSL because my constituency is in the north. Most of the users of NSL are my constituents, and I get it straight to me.

19.    This careful process of doing testing was indeed what the team did. LTA, SMRT, and the system supplier Thales, embarked on a careful and progressive testing regime. Our strategy was to first confirm its proof of concept, and we used the Changi Airport Line for that purpose. This was successfully done in 2014, before we decided to push ahead with this programme. Then we began the exercise to first re-signal NSL, which is now on-going. So EWL we have not started, that means we have some hardware installation but we have not started testing yet. So whenever you hear about little problems with EWL, it has nothing to do with re-signalling. The re-signalling has been blamed for everything now. Yes, it has been the cause of some problems, but it is not the cause of every problem. The delays that happened last week on the EWL have nothing to do with re-signalling, it is just a demonstration that ageing has affected EWL. I wish it was replaced five years ago, but this is what you have. Occasionally there are difficulties, but even then, as you see, the MKBF for EWL has improved from last year.

20.    So back to the testing of NSL. We are doing it in several stages, a cautious approach. Stage 1 was to progressively load the system, first for up to 40 NSL trains. Tests were then carried out on these trains during engineering hours, without any passengers. After more than 1,300 tests were done safely and satisfactorily, we moved to stage 2. Stage 2 was what we call shadow mode testing, as we loaded up all the remaining NSL trains. At this stage, we performed parallel diagnostics of the new signalling system. We were still running trains on the old signalling system but with the new signalling system that allowed us to collect data and to monitor the reliability of the new system. Only when we were fully satisfied with the stage 2 results, did we move into stage 3 of testing during passenger service hours. Even so, we did this very cautiously, starting with only the last hour of passenger service on weekdays from 28 March this year. So when it irritates, it won't irritate too many people. After we were happy with that, on April 16, we moved the testing up one notch by going into full-Sunday trials. On May 29, after the system, with new software patches, showed no further signs of problems, we moved into stage 4 of full weekday trials, which is the stage we are in now. This is to condition the new system to peak hour conditions. I read comments from journalists, “Why are we so stupid? Why are we doing trialling during full hours?” No choice, how else do you do it? We can do computer modelling and simulation, but it's never the same as when the trains are fully loaded. But we have done our best to make sure that the software is as good and robust as it can be before we started this full service. And as I said, this is the stage we are in now.

Disruption on 28 June 2017

21.    As the system is being stress-tested, as we have forewarned everybody, there were delays, there was congestion, as expected. They allowed the team to pin-point the glitches and fix them through software upgrade. It was a necessary iterative process of testing and discovery, with progressive software upgrades. It was unfortunately a necessary iterative process of testing and discovery. Overall, the experience has actually been less traumatic than what we had feared. After talking to the London Underground people and Taipei chaps last year, I and Tan Gee Paw and all the key people here were really very worried. We expected the sky to fall, as has happened elsewhere. But overall, it has turned out much better. Much smoother than what we had feared.

22.    The only one exception was a major disruption on June 28. Unfortunately, that was a major disruption. That major disruption coloured everybody's views about re-signalling. Let me explain what happened. On June 28 afternoon, they were configuring and loading the new radio software for the new signalling system on the EWL. By mistake. So it was a human error, a Thales engineering team connected an incompatible radio to the radio backbone network of NSL (and Tuas West Extension), which was still running, in fact it was afternoon peak hour. This caused a sudden congestion on the network which resulted in all NSL (and Tuas West Extension) trains losing radio communications. As a safety feature, when they lose communications, all trains stopped. The team realised the mistake almost immediately, but unfortunately, it took them a while to isolate the radio network and to restart the train service. We have over 140 trains, and to restart and re-launch every one of them took some time. That was the reason for what happened on June 28, it actually had nothing to do with the re-signalling project.

23.    It also explains that after re-signalling NSL, we will re-signal EWL and we have limited time to complete all these projects. So ideally, we just do all the NSL, stop, and then we start the process for EWL. That will take a long time. So teams are rushing and therefore, with only 3 hours a day, how are we going to complete EWL signalling? That's why they thought they will try to do things on EWL while NSL is running. But with this June 28 incident, we told the team, let's stabilise NSL, we continue to do some minor installation works for EWL but let's wait for the testing. It will take us longer for the EWL to complete, but under these circumstances, I think it is better this way.

Moving Forward

24.    Meanwhile, we continue to make steady progress with the NSL re-signalling project. In other words, if not for the June 28 incident, generally the project is doing well. A new software version (R6B4.A) was installed two weeks ago (July 14). The new software version has so far performed better than the earlier version (R6B3.B) and train operations are also smoother.  However, the reliability of the train-borne Vehicle On Board Controller (VOBC) remains an issue. There are some concerns there. There are still occasional momentary loss of VOBC redundancy and radio communications between train and track.  We have just deployed a new radio software version 3.4.6, to address the issue of radio communications. In addition, Thales is working on stabilising the system further by resolving the loss of VOBC redundancy. It will do so by looking into improving the hardware design for the various components within the system.

25.    Overall, we are confident that the new signalling system on the NSL should stabilise soon, certainly it will be before end of this year. Let me explain what we hope to achieve. As you know, we do testing in stages and when we start loading it onto the passengers, we try to make use of off peak periods. That's why for NSL, we did it during the June holidays before we reopened the schools. It's a stretch target, but we are working towards the December holidays. If things go on without new surprises, we certainly hope that the December holidays can allow us to begin the full testing of EWL with passengers. But in order to be able to do that, NSL must be stabilised before that. We don't want to have two projects going simultaneously. So if there are no new surprises, our hope is we can stabilise NSL over the next couple of months so we can proceed with this plan to begin the full testing of EWL using the December holidays. And as you observe from how we did for NSL, we cannot complete all the testing of EWL during December, so it will go on into next year. But this time round, we think the lessons learnt in NSL can be ported over because software patches which works on NSL should work on EWL too. So the learning curve will be much shorter for the EWL. And once we can do all that, we will be able to begin to launch more new trains on the NSL progressively during the peak period. Commuters can then enjoy the benefits of the new system, and we can then move on to upgrade the system on the EWL.

Conclusion 

26.    Friends and Colleagues, come next year, when we meet again for this forum, we will be able to report further progress. Not just on the re-signalling project, but also the overall efforts to improve our rail reliability. I know everyone on the ground is seized with this mission. Our rail networks support 3.3 million commuter trips every day.  We will not fail in this mission, but we need some time to complete the many tasks which are necessary. I thank you all for the perseverance and understanding of Singaporeans. Thank you.