On Wednesday morning, November 5, 2014, the TTC suffered two major delays on the subway system. One was a complete shutdown of service between St. Clair West and Union Stations, and the other was a period of very slow operation approaching Broadview Station westbound.
Updated November 10, 2014 at 5:00pm: The TTC has now provided an explanation for the delay on the University subway. See the body of this article for details.
The morning commute was painfully difficult for everyone on the subway, and these incidents inevitably raise calls for “someone to do something” so that they won’t happen again. That’s an easy political call, but one requiring a deeper understanding of the underlying problems. This is not just about the physical state of the signal system, or the TTC’s ability to respond to major events, or the long-standing question of subway capacity, but a mixture of all of these. Quick fixes would be nice, but if they were available, Toronto would not be in the transportation mess it faces after years of inaction, denial and pandering for votes to the detriment of transit everywhere.
The problem on the University line has not been fully explained yet by the TTC, but the general impression is that there was a failure of the signal system caused by a blown fuse or faulty relay [Andy Byford interview on Metro Morning, not available online].
Updated November 10, 2014: The following explanation of the problem was provided by the TTC’s Brad Ross:
We found a fault with a non-vital relay that did not pick up due to lack of power. A while ago an old code board was removed from St George, together with its power supply. The power supply, however, was still required for other equipment still remaining related to signals. A permanent feed has now been installed and processes are being reviewed to make sure something like this does not occur again.
Because St. George is itself a turnback point, service had to be cut at the next available location to the north and south.
Although there is a crossover at Spadina Station, this is not a practical location to terminate trains coming south from Downsview because of constraints on station capacity for subway-to-bus transfer operations. Therefore, the TTC chose to use St. Clair West as the northern turnback point. To the south, the next available location is Union Station, and the Yonge line was operated from there north to Finch. Buses provided a shuttle service in between.
On the Danforth subway, a track circuit failure on the Prince Edward Viaduct near Broadview Station westbound caused the signal system to think that a train was present when, in fact, none was there. This is a standard design feature of railway signals in that they “fail safe” and show red signals for a “ghost” train rather than giving a clear signal even with a train present. When this happens, manual operation past the red signals is possible, but the process is somewhat more complex at a location like Broadview because of the three-track section between there and Chester Station.
Regular “block signals” (the ones with only one signal aspect) are designed to allow trains to bypass a red display by pulling right up to the signal, stopping, and then proceeding. This is possible because the track circuit for the “next” block actually extends back far enough ahead of the signal that a train occupies it before passing the signal. When this happens, the automatic train stop (the “trip arm”) will descend and let the train pass, but this only works if the train approaches very slowly.
At any location where there are switches, and hence the possibility of a collision between conflicting movements, the signals operate more restrictively than on the “ordinary” section of the line between stations. For a train to bypass the signal, Transit Control must allow a “manual key by” (this causes a rarely seen red-over-red-over-solid-amber display at the signal). The approaching train’s driver must open the cab window and press a lever mounted on the tunnel wall to complete the bypass process and lower the trip arm. This ensures that there is no doubt about what is happening, and that everyone is aware that a train is being allowed past a point where it would normally be forced to stop. There are two such signals westbound between Chester and Broadview.
The effect of the need for manual dispatching and the manual signal bypass processes was to very considerably slow train progress through this area, and that quickly led to a backlog of service on the route.
An explanation of the problem from the TTC’s Brad Ross:
The PEV (Broadview) incident was due to the failure of an insulated joint on the bridge rail.
Insulated joints in the rail electrically separate two adjacent signal circuits which allow for the tracking of train movements throughout the Subway system. The bridge rail is a secondary restraint rail that protects the train from leaving the guideway in the unlikely event of a derailment. While the train does not mechanically run on the bridge rail, the rail itself is electrically connected to the signal circuitry through the base plates that secure the running rail and bridge rail to the bridge deck. This particular insulated joint was serviced the night before by a Subway Track Corrective Maintenance crews. Servicing of the joint involves replacing broken components and the tightening existing fasteners. The joint was in service for approximately one hour before it electrically failed. Crews dispatched to the scene disassembled the joint and effected repairs. While still under investigation, it appears the older fish plates used on the joint did not have the isolating properties necessary to maintain the integrity of the electrical circuit and the joint failed electrically. At no time was the running track integrity compromised mechanically and the train fleet was never in danger of derailment due to the failed joint in the bridge rail.
Subway Track Maintenance staff are looking into new processes that will allow for the quick identification similar conditions in the future necessary to prevent a reoccurrence. [Email from Brad Ross, November 7, 2014]
Toronto’s Aging Signal Systems
The signal systems now operating on both the YUS and BD subways date from their construction. The original Yonge line (Eglinton to Union) opened in 1954, and the University extension to St. George in 1963. The original Bloor-Danforth line (Woodbine to Keele) opened in 1966. This puts the systems definitely in the “elderly” framework. Parts are difficult to obtain, and although the systems are repaired, some system-wide components such as cabling don’t lend themselves to simply swapping out one part for another. This is further complicated by aging track structures and tunnels where rail-to-rail insulation may fail, or where water on tunnel floors or poorly-drained roadbeds can lead to short-circuits.
[I will pause here while those who claim that subway systems last “100 years” quietly leave the room.]
Toronto faces the a problem common to any major industrial investment — it wears out eventually, and the ongoing cost of upkeep of a 50-year old plant is much greater than when it is new. Just as with steel mills and other industries of the past century whose owners lived off of the value of the initial capital construction, Toronto had not yet reached the point where major subsystems within the subway network needed replacement. (The city has a similar problem with roads, water and other utilities.) Now, facing a growing demand for capital repairs, the city (and its funding partners) plead poor and underfund repair budgets. It is much preferable to tell people about a new subway somewhere than to trumpet how you will repair the one we already have.
The TTC has a project to completely resignal the YUS from end-t0-end and to implement automatic train control by 2019. The work has progressed to the point it is fairly complete on the “southern U” from Bloor to St. George Station, and work is underway from Bloor north to Eglinton. None of the new systems has been turned on yet, and Wednesday’s failure was in the old equipment.
The BD project is not even scheduled to begin until 2023 with a completion date maybe 5 years later. By that time, the original line’s signals will be over 60 years old. The timing is as much about keeping an as-yet unfunded project from adding to the 10-year capital deficit as it is about good planning.
On July 2, 2014, Leslie Young at Global TV published an article looking at the reason for TTC subway delays in 2013. This piece includes several charts including an interactive map where one can filter for type of delay, and another that allows selection by station and time of day.
Unsurprisingly, the greatest number of delays occur in the AM peak when the system is under greatest stress. Ill/injured customers and door problems are the top two culprits, something that should surprise nobody who rides the system. Note the much more even distribution of delay causes in the PM peak. Signal problems (the pale blue band in the chart) are a relatively small portion of the total.
[Chart by Global News. Click on the chart for the full article.]
My one quibble with the charts is that they report on the number of delays, but not their length and, hence, their relative effect on service. This will be important in allocation of resources to reduce delays so that effort is concentrated where it will have the most benefit.
Riders are much more likely to be affected by crowded cars, and doors are much more likely to be held open when doorways are packed with just one or two more hopefuls trying to squeeze onboard. These are both problems related to overcrowding, and unless the TTC can find some way to deal with that, the problems will not go away.
The universe had a quiet joke at my expense on Thursday morning as I was enroute to the CBC for a Metro Morning interview. As I changed trains at St. George to go south to St. Andrew, an announcement about trains holding at Queen’s Park for an ill patron came over the speakers. No time to take chances that it would be a short delay. I left the station and took a cab to the studio.
It must have been a very short delay because there was no notice via TTC’s eAlert that it ever happened.
The Problem of Communications
A common complaint by riders is that they never know when there are problems on the system. “Never” might overstate the case, but when this happens often enough, the typical perception is that good info is the exception, not the general rule.
The TTC’s eAlerts come thick and fast on some days, and on others they are rare or stop altogether. This isn’t just a variation in the odds of a delay because it’s easy to experience one for which no eAlert ever appears. Meanwhile, collectors sitting at station entrances are often totally at sea about what might be happening one or two levels “downstairs”. There has been talk of monitors in the booths where this info could be displayed, and there are now video screens at station entrances where notices are supposed to appear. (These monitors may or may not be positioned so that collectors can read whatever might be posted on them.)
For the record, I use the subway a lot, and I have never seen an outage notice on any of these screens.
A further problem is that many riders enter the subway through bus and streetcar connections, not at the main street entrance, and they don’t even pass by these monitors. At platform level, the news, if it is posted at all, runs along the bottom edge of screens that cannot be read over the full length of the platforms. Advertising takes priority over notices except in extreme emergencies when the TTC invokes an override feature.
For the past few years, in an era of spending cuts and service reductions, the TTC focused on looking better – cleaner stations, renovated washrooms, uncluttered collector booths – but communication remains a big problem.
An Industry Peer Review of the TTC’s Subway
Back in 2010, the TTC received two reports comparing its subway operations to other major systems. Neither of these reports is available online in the meeting agendas, only one-page placeholder reports.
- January 10, 2010: A benchmarking report on subway cost and performance
- June 2, 2010: A Peer Review report from the Nova group of metro systems, an organization of which TTC CEO Andy Byford is now the head.
I will deal with the Nova report first as it looks at high level stats while the benchmarks look at specific aspects of operations.
The Nova presentation is a general review of the TTC subway relative to other members of the Nova and CoMET metro groups [map]. Both the size of Toronto’s network and its annual ridership are lower than others, particularly in the CoMET group which includes such heavyweights as New York, London, Paris and Beijing. However, the report claims that the TTC is one of the only two systems outside Asia and South America that covers its operating costs from the farebox.
That’s an intriguing statement because, of course, past attempts to divide revenue among TTC routes have proved to be a complete waste of time. In a flat fare, free transfer system, any scheme to divvy up fares will have inherent biases toward or against certain modes and types of journeys.
Without going into the details of the calculation, it appears that the study took the 2008 revenue for the subway at about 45% of the total, while the operating cost was only about 28%. Overall, administrative costs at TTC are low as a proportion of the total allocated cost, service is inexpensive to operate, and maintenance costs are “moderate” by comparison with other systems.
The TTC ranks third out of all systems in car kilometres per staff hours, and fourth in capacity operated versus route kilometres. It also ranks 4th last in passenger km vs capacity km due to the high level of off-peak service and the low load factors over much of the system’s mileage at that time. These rankings are a direct result of large trains running frequent service over the entire network almost 20 hours/day.
The report concludes:
When compared with other metros in the world, Toronto’s subway offers excellent value for money. [Presentation, page 23]
The TTC may pat itself on the back for the financial ratings, but these are affected both by assumptions of how revenue should be allocated, and by the basic question of how operating costs and spending stack up against other systems. Being “profitable” might not be the badge of honour it seems. We know that the system loses money overall, and that the surface routes are an integral part of the whole – without them, subway ridership would collapse. Moreover, the context in which this report appeared – the first year of the Ford administration where the appearance of businesslike efficiency was essential to the the TTC’s image – is difficult to ignore.
The benchmark report reviews four areas of subway costs and operations: track maintenance, civil structures, fleet maintenance, and signalling. It notes the importance of local practices such as the relative proportion of preventative and reactive maintenance, and the balance between capital and operating funding sources. However, the report is silent on how these might affect the TTC’s position relative to other systems.
This is not a trivial problem because in a system where “capital” dollars appear out of thin air by way of subsidies, there is an incentive to push as many costs as possible into that budget rather than booking them as ongoing operating costs. Further distortions can occur when subsidies to each budget come from different sources and at different rates. For Toronto, the migration of capital funding from senior governments to the city’s budget is a painful shift that the city has not yet absorbed. This is directly responsible for the large deficit in known funding for future capital maintenance projects.
Performance ratings for each area are based on failures that affect service, and these are reported both with a 3 and a 5 minute delay threshold. The larger value screens out minor delays, although these are important on a very busy system where even a short delay can bring chaos because of close headways and a service operating with no spare capacity. This shows up in a real-life example on the TTC described later in the article.
For track maintenance, TTC values (different for the YUS and BD subways) are less than half of the “average” for peer systems in the review, although the values lie in the bottom end of a “good practice” range. These are inherently contradictory statements especially considering that “good practice” itself is a band well below the peer average. There is no explanation of why costs are so high on several (unnamed) systems, whether this represents a different accounting approach to costs, or if the systems in question had a particularly bad set of conditions to deal with.
What does show up is that Toronto’s rate of delays at the 3-minute threshold is more than six times that of its peer systems with only a few (again unnamed) coming close. Astoundingly, this situation is reversed when a 5-minute threshold is used and Toronto is well below the average value. Without further data to explain this shift or to put the severity of the delays into context, these are meaningless statistics.
Delays due to failures of civil structures (e.g. the compromise of structural integrity in a bridge) are extremely rare throughout the industry, and there is no data to allow comparison between systems. Moreover, there are two classes of potential incidents: those due to deterioration through age, and those due to external factors such as vehicle collisions with structures. What is odd about this “benchmark” is that the most common “civil structures” of a subway, almost by definition, are the tunnels through which it runs, but the report is silent on this topic.
Fleet maintenance is another area where Toronto ranks below the average of peer systems, but again the “good practice” band is defined well below this average and it overlaps the range of Toronto values. Again, one must ask which target we should be aiming at – the “average” or a “good practice” level. The cost/car kilometre is broken down by equipment type and route with the T1 fleet coming in lowest (the TRs did not exist when the report was produced), with the H6, H4 and H5 fleets having higher costs in that order. This also begs the question of whether Toronto’s cost vs peer averages represents “good” practices or simply an aging fleet that is expensive to maintain.
Delays related to fleet issues are at a level comparable to most peers especially for the T1 cars on the three-minute threshold, and are substantially better on the five-minute level. This implies that there may be many small delays caused by the Toronto fleet, but relatively fewer major ones.
On signalling costs, Toronto has among the lowest cost per track mile figures of its peers with several spending at five-to-ten times the rate of Toronto. As with the other benchmarks, the “good practice” range is defined well below the average value and the range of costs reported by other systems is so broad as to not be credible as a valid metric. Something else, such as relative age and reliability or simply accounting practices (capital versus operating spending), is at work here.
Failure rates in Toronto show the same relative standing to other systems – Toronto’s signal-related delays are much more common (ironically with the Sheppard line being much higher than YUS/BD) on the three-minute scale, but much lower on the five-minute scale. These numbers simply do not make sense, and beg the question of how each system has recorded and categorized its delays.
To me, the benchmark numbers are almost worthless because there are too many unexplained differences and, most importantly, because the average cost level for the industry is well above what the authors consider “good practice”. I have included this review to show the challenges Toronto faces in deciding where it stands relative to other systems and how much we should strive to improve. Even being “average” may not provide the quality of service and reliability riders need so that the city can count on the TTC as its primary transportation mode.
Improving Subway Capacity
Over the past decades, the TTC’s attitude to subway capacity has evolved considerably. Through the 1980s, the system was bursting with riders, and provision of more capacity both with subway service and with expansion of the Bloor-Yonge interchange were high-priority issues. In the 1990s, Toronto lost 20% of its system ridership (annual trips fell from about 450m to 360m) and the subway’s capacity faded as a political issue. Riding for 2014 is expected to be about 535m, and fixing subway capacity is at the top of everyone’s agenda. This will not be as easy as it looks because, in part, for years the TTC persisted in saying that there was no problem, and capacity would be provided from various sources including:
- Larger subway cars (the TR fleet now operating on YUS can carry 8-10% more riders than the T1 fleet serving the BD line).
- Opening the Spadina extension will divert riders who now use the Yonge route, notably commuters funneled into Finch terminal, to the extended Spadina line.
- New signals will allow trains to operate closer together through automatic train control and moving block technology (in use on the SRT for decades).
The conversion to TRs has already been accomplished and so that does not represent future capacity, at least with the current fleet. There have been proposals to add a seventh car to the TR trainsets, but TTC’s subway planning appears to see this as part of the next round of car purchases. (New trains will be needed to replace the T1s now on the BD line in the mid-2020s. What could happen is that the existing TR fleet would shift to BD, and the new generation of 7-car trainsets would go to YUS.)
Through all the talk of greater capacity on YUS, the TTC was adamant that there was no need for another subway, a “relief line”, into the core area. For a time, the TTC claimed that future growth including that from a Richmond Hill extension could be handled with the existing infrastructure. Politically what was (and is) happening is the fear that a DRL will soak up so much funding that nothing else will be built, especially in the vote-rich and subway-starved outer 416 and inner 905.
That’s a topic for a separate article, but this puts the whole question of subway capacity and reliability in a wider context. How long can we expect to handle growth in demand to travel to Toronto’s core without more capacity in the network and better reliability of what we already have? The TTC’s position now is that a Downtown Relief Line is its top priority. It is ironic that after this near miraculous conversion, the TTC’s priority may be eclipsed by other regional priorities such as the Metrolinx RER network and John Tory’s SmartTrack. Had the TTC advocated much sooner for a DRL, we might actually be building one by now.
Recently, the TTC revised the schedules on the YUS and BD lines. This was publicized as “adding trains”, but what was going on was a more subtle change.
- Two trains were added to each line in the peak period, but the scheduled headways did not change. In other words, on paper, the amount of service operated would be the same with the old and the new schedules, but trains would have longer to make their trips.
- Crew practices at terminals were changed to use “step backs” so that an incoming train’s crew did not leave with the same train, but “stepped back” two trains. This built in a relief break for the crews and eliminated situations where they would have to depart almost immediately upon arrival.
- Carhouse trips at the end of the peak periods have been changed so that trains leave service outbound (northbound at Wilson, eastbound at Pape) so that they do not compete with the post-peak daytime or evening service for slots at terminals.
I asked the TTC how effective this change had been, and they provided an interim report that was recently presented to the management executive committee.
There are three goals for revised subway operation:
- improve terminal operations and reduce instances where incoming trains are held in queues;
- improve train throughput;
- improve journey times by reducing holds for across-the-platform crew changes.
In a before-and-after comparison, dwell times and departure delays are down at all terminals. This is especially notable at Downsview in the AM peak where only half of the scheduled service operates due to the short turn at St. Clair West.
The all-important metric of line capacity has definitely improved with the average trains/hour count at Bloor rising from 23 to 25.22 (10%) for an extra capacity of 2640 riders/hour (based on 1200/train). The count at St. George is up from 21 to 23.78.
The scheduled headway is 2’21” or 25.53 trains/hour on both the YUS and BD lines during the AM peak period.
The Yonge line is up almost to its scheduled level while University has some distance to go. This could be related to reliability problems with the short turn at St. Clair West, but the report does not comment on that issue.
On the Bloor-Danforth line, there have been similar improvements, and the trains/hour count is now at the scheduled level westbound at Yonge. Eastbound to St. George, the level is still a few trains short of the scheduled headway.
PM peak and midday improvements are also recorded, although at lower levels. During the PM peak, the scheduled trains/hour is 23.84, but the actual values observed were 20.7 and 23.1 for YUS and BD respectively with the new schedules. Clearly, there is still some distance to go on YUS.
Off-peak values are odd and I suspect that they cover more than the midday period because the trains/hour values are well below the weekday midday scheduled levels. What is missing from the comparison charts are the target, scheduled values.
Also included is a chart showing week-by-week values for the before and after periods. What shows up here is a substantial swing in day-to-day values. Although the “highs” are impressive and may even include the insertion of “gap” trains over and above the scheduled service, the “lows” are troubling.
For the purpose of estimating reasonably available capacity, the TTC cannot yet cite the theoretical scheduled level of service because it does not reliably achieve this. Moreover, as trains become more crowded, the associated problems of blocked/jammed doors and passenger illness will become more frequent. Small delays can quickly cascade into major problems because of follow-on effects from losing even one train’s capacity in the peak hour.
To their credit, the TTC regards this as only an interim success with more work to be done. Whether riders will notice the difference, especially when major disruptions are so memorable, is hard to say.
The TTC, Toronto, Metrolinx and the Toronto Region face huge challenges with moving people and a long era of less-than-adequate investment both in expansion and in renewal of the existing system. Riders want “more” at a time when budgets don’t even cover what we already have. Politicians want to talk about bold, new, “smart” moves, but not so much about reversing past errors.
The focus of many proposals is the core area where transit demand is concentrated and the alternative – driving – is impractical for most. But problems of congestion and travel demand extend well beyond Toronto’s core – all the new subway lines and regional express trains and SmartTrack will not eliminate congestion for travellers going region-to-region around the city’s perimeter. Indeed, the very definition of “regional” travel begs the question of where the boundary with “local” travel actually lies.
Transit is transit although the requirements for various journeys will differ. Toronto cannot afford to ignore any of these trips.
Improving subway service is an important part of the overall plan, and getting the most out of what we already have will be vital as we wait for something, anything to relieve the capacity pressure. That pressure will not vanish thanks to growing population and the returning attractiveness of downtown as a place to work.
Toronto’s budget planning must not be highjacked by one or two signature projects to the detriment of the system as a whole, and “regional” spending will not replace the need for better “local” service.