Excavation performance is a term used in underground construction to describe the influence of several parameters on the cutting rate of mining machinery. The main factors affecting the excavation performances are:
- The geology (rock and rock mass mechanical properties and conditions).
- The process (in terms of operation, maintenance, logistics).
- The methodology (the type of machine, tools and ground support chosen).
The Doan Valley Project represents a unique scenario where, within the same geological features and operational boundaries, different excavation methodologies have been successfully implemented and can now be evaluated for the performance achieved:
- The 3.6-m (12-ft) main-beam tunnel-boring machine (TBM) for the total 9,170 linear feet (LF) (2,795 m) of Woodhill and Martin Luther King Jr. Conveyance Tunnels (respectively, WCT and MLKCT).
- The roadheader for the excavation of the total 520 LF (158 m) of the Doan Valley Storage Tunnel (DVT) Starter Tunnel (ST), MLKCT starter tunnel and tail tunnel.
- The 6.4-m (21-ft) single shield TBM for the 9,670 LF (2,956 m) of DVT.
Given the same geology and process, this article will analyze and quantify the efficiency of each methodology based on the results achieved versus some of the assumptions made at the estimating stage.
The NEORSD program and the DVT project Located in the cultural center of the University Circle neighborhood on the east side of Cleveland, OH, Doan Valley is a major component of the Northeast Ohio Regional District’s (NEORSD) Project Clean Lake program, a $3-billion, 25-year program with the ultimate goal to ensure 98 percent of wet weather flows entering the combined sewer system receives treatment, thereby drastically reducing raw sewage discharge into Lake Erie and associated waterways. As part of the Easterly Service CSO discharge area, the annual combined sewer overflow (CSO) capture for the Doan Valley Project is 365 MG/year through three tunnels ranging from 2.6 to 5.5 m (8.5 to 18 ft) finished diameter and six tunnel shaft sites with associated near-surface sewer structures. On April 6, 2017, McNally/Kiewit DVT Joint Venture (MK DVT JV) was awarded the DVT contract by the NEORSD for $142,320,000. Construction commenced on July 10, 2017.
The geologic setting for all three tunnels is primarily the Chagrin Shale bedrock, except for a short reach at the end of the WCT drive in Cleveland Shale. In general, the Chagrin and Cleveland shales rock masses are characterized by weakly bedded plans including clay-filled bedding joints and weathered interbeds, nearvertical joint sets and a tendency for slaking. A summary of the three main tunnel geological and design features is provided in Table 1.
Due to the typical logistics constraints of an urban jobsite, each TBM tunneling operation was planned and had to be executed in sequence, one at a time. The WCT tunnel started first in January 2018, the DVT tunnel in January 2019 and the MLKCT tunnel in October 2019. With few exceptions (mainly among the laborers), the same workforce has operated both machines and built one tunnel after the others adapting to the different methodologies.
The DVT starter and tail tunnel mining operation between April and August 2018 was arranged on three, eight-hour shifts, working five days a week. The roadheader excavation was performed mainly during the second shift, in an effort to keep the shotcrete supply and application on the day shift.
The WCT/MLKCT and DVT tunnel operations were arranged in two main stages: the TBM initial mining and the TBM full production mining. For the WCT, the first stage was on one eight-hour shift per day. The second was on two eight-hour shifts per day. For the DVT, the first stage was on two eight-hour shifts per day. The second was on three, eight-hour shifts, always working five days a week. For the MLKCT, both initial mining and full production mining were executed on two, eighthour shifts.
It must be noted that while TBM operation repeats itself across each shift, and crews have all the same composition, the roadheader operation crew changes during the production cycle based on the sequential task to be performed (excavation and muck removal, scaling, rock-bolting and mesh installation, shotcrete).
Common to all the tunnel headings and tunnel operations performed at Doan Valley is the muck haulage system, which is made up of three stages. First, the muck is placed in muck boxes by bucket-style equipment (roadheader operation) or a conveyor (TBM operation). Once the box is positioned at the bottom of the shaft, it is then hoisted up using a crawler crane and dumped into a temporary muck pile on the surface. There, a front-end loader transfers the muck into dump trucks for transportation off site.
With a low range of rock strength, very low abrasivity (as far as TBM practice) and a tunneling length always less than 3,050 m (10,000 ft), two muck trains were determined to be adequate for each TBM heading to achieve, at reasonable cost, the average advance rate of 1.1 m/h (3.5 ftph) (84 LF/day at DVT; 56 LF/day at WCT and MLKCT).
For all the shaft operations on this project, crawler cranes were utilized primarily for dumping boxes, delivering rib and lagging sets/segments/pipes, delivering supplies/materials, and transferring equipment. The crane also served as a secondary means of emergency egress with a man cage always staged within proximity to the shaft.
The roadheader is a highly mobile excavation machine, however it attacks only a portion of the face at any one time and is therefore limited to a fraction of the performance of the TBM. For this reason, its use at the Doan Valley Project had to be minimized to the tunnel length and horseshoe profile exactly needed to allow a smooth assembly and launch of the 6.4-m (21-ft) TBM.
Instead, the shielded TBM mines the full face, providing:
- The maximum possible grade of safety for the workers.
- A much higher grade of automation of the production process (excavation and lining).
- High performances out of mechanization and full-face excavation.
- A finished tunnel, in case of segmental lining (a single-pass tunnel lining).
Roadheader specifications and starter tunnel (ST) production cycle. The AQM 150HR roadheader was shipped to site completely assembled, except for the operator roof support and swing extension conveyor assembly on the back side of the machine. The overall mobilization for this machine took four shifts, which is a major advantage of using this equipment on short tunnel drives. When the power supply and the site logistics are setup for it, the roadheader becomes almost a plug-andplay operation. The initial mining stage for this operation is irrelevant as it does not compare to the other methods.
The excavation cycle began with the entire 102 m (335 ft) length of starter tunnel top heading. This was limited by design criteria to maximum of 3 m (10 ft) advance and the roadheader backed away from the face and off to one side of the tunnel, to allow installation of the ground support during the following two shifts (rock bolts, wire mesh and shotcrete).
Once the 102 m (335 ft) of top heading was completed, the roadheader was moved back to the shaft to excavate the additional 3 m (10 ft), down to the DVT invert elevation. From the lowered shaft bottom elevation, the roadheader excavated the 102 m (335 ft) of DVT starter tunnel bench, the 47 m (155 ft) of tail tunnel and the 9 m (30 ft) of MLKJCT starter tunnel, with excavation heading alternation (so as to maximize the machine utilization).
The roadheader daily production chart in Fig. 4 shows the overall production of the entire operation, which includes all three headings. The first portion of the chart, which only represents the DVT starter tunnel top heading, is the only section where the roadheader operated from start to finish without alternating between multiple headings. Therefore, for this paper, this data best compares to the other excavation methodologies.
After completion of the DVT, the cast-in-place final lining will be installed in the DVT starter tunnel. This structure will match the inside diameter of the DVT segmental lining (5.4 m or 18 ft).
3.6 m (12 ft) TBM specifications and WCT/MLKCT production cycle. A 3.6 m (12 ft) Robbins main-beam open-gripper TBM was used for the excavation of the WCT and the MLKCT. To date, this TBM has completed more than 49,000 m (160,000 ft) of rock tunnel in both Canada and the United States, including several previous projects for the NEORSD in Cleveland, OH. In 2009, Herrenknecht refurbished this machine and added a floating tail shield for a wider range of applications (e.g. unstable ground) than a traditional open shield gripper TBM.
The muck from the cutterhead is conveyed by the machine and back-up conveyor toward five muck boxes, which are pulled under the conveyor discharge point by a 13.6-t (15-st) diesel locomotive.
Both the WCT and the MLKCT have a two-pass lining system consisting of:
- Tunnel initial support — expanded steel rib and timber lagging to prevent block falls, slabbing and raveling into the tunnel, installed at completion of each 1.5-m (5-ft) mining stroke.
- Final lining — precast reinforced concrete pipe (2.6 m or 8.5 ft ID with 21.6-cm or 8.5-in. wall thickness) designed in three different lengths (1.2, 1.8 and 2.4 m or 4, 6 and 8 ft) to negotiate different tunnel radii. Each pipe section is lowered in the shaft and then transported into the tunnel using a pipe carrier. Next, the annular gap between the extrados of the pipe and intrados of the rib and lagging initial support is backfilled with cellular grout to provide good lining to rock contact and reduce voids.
6.3 m (21 ft) TBM specifications and DVT production cycle. A 6.3 m (20.75 ft) diameter single-shield TBM was specifically designed and built for the Doan Valley Tunnel.
Based on lessons learned in the past, in similar geologies, the machine design implemented some job specific features:
- Larger loading buckets to minimize cutterhead clogging.
- Stepped design shields to prevent entrapment in softer and potentially unstable shale.
- Active articulation between front and center shields for better control of the cutterhead lookup in softer shale (where the typical machine tendency is to dive).
- Tail shield circumferential joint fully welded to the rear shield (instead of articulated) to strengthen the rear section of the machine and minimize any deformation of the tail can.
- Continuous TBM/back-up conveyor belt to minimize the number of dumping points and reduce the potential for dust.
- Extended erector longitudinal travel to easily reach and expose the tail seal brushes for inspection.
- Extended verification drill and platform longitudinal travel with the ability to operate probing and contact grouting along a five-ring span.
- Lift pedestals welded on each side of front and center shield to jack and swiftly push the whole TBM to the face, in the 102-m (335-ft)-long starter tunnel.
The initial support and final lining for the DVT is a one-pass system consisting of fiberreinforced precast concrete segments assembled into rings, concurrently with the TBM advance. Each universal design ring is made up of five 25- cm (10-in.)-thick — 1.5-m (5-ft)-long segments plus a key. The annular gap outside of the segmental lining is filled while the segments are leaving the shield with bi-component grout through four injection ports in the tail shield.
The machine was designed for a maximum advance rate of 100 mm/min, and during the entire DVT drive it was consistently operated in the 80-100 mm/min range and in the 5,000-7,000 kN range of thrust force. The total duration of the excavation cycle was typically between 40 and 50 minutes (15-20 minutes for the 1.5-m (5-ft) mining stroke and 25-30 minutes of ring build). The next cycle however could not start until several minutes later, being the muck train re-set cycle at the shaft 55 to 65 minutes long (5 to 25 minutes more than the excavation and ring build time).
Only one (gage) cutter was changed during the whole drive. In such a soft rock the challenges were typically represented by the diving tendency of the machine, the over breaks in the crown and the clogging tendency of the loading buckets especially in the presence of ground water at the face (in addition to the water from the lined tunnel, being DVT a 0.17 percent downhill drive).
Initial mining and learning curve analysis
For the scope of this paper, the initial mining is defined as the period of time and linear footage of tunnel that goes from the day the excavation commenced to the day the whole production system reached its final configuration (in terms of machine fully operational, two production trains able to cycle themselves and crew fully deployed).
If irrelevant for the roadheader drive, this transition between assembly time and full production mining represents a crucial phase for any TBM project. At the initial mining stage, the machine can be tested under load and fine tuned, the crew can finally familiarize themselves with the new equipment and operation, and the whole supply chain gets calibrated to the TBM needs.
Machine mobilization. A substantial difference between the three methodologies is in the mobilization time. The roadheader traditionally requires minimal field assembly effort and can be made ready to mine in very few shifts.
The 3.6-m (12-ft) main beam TBM is essentially made of a front section of the main beam (with cutterhead, main drive, tail shield and gripper carrier) and a rear section of the same beam carrying the operator station, the main electrical cabinets, and the hydraulic power unit. When powered up, this core front section of the machine can be made ready to bore in less than a week. The backup decks with miscellaneous equipment are then added one at a time and directly towed by the front section of the machine during the initial mining stage.
After the shop tests, the 6.4-m (21-ft) TBM was broken down into 49 major components and six containers of miscellaneous parts, for a total of 55 truckloads. It took two weeks to receive it all and then three months to make it ready to bore (including five weeks of round-the-clock field weld and two eight-hour shifts per day for pre-assembly on the surface, assembly at the bottom of the shaft, skidding to the launch cradle, hydraulic/electric connections and testing).
The machine was launched with six of the eight backup decks completely assembled in the starter tunnel. With the exception of ventilation and dewatering, every other system was in its final configuration and all temporary set-ups (typical of the launching phase), were kept to a minimum.
In summary, the mobilization time (from receiving the first component to “ready to bore”) was:
- Roadheader for ST: 1.5 work days.
- 3.6 m (12 ft) TBM for WCT: 5 work days.
- 6.4 m (21 ft) TBM for DVT: 76 work days.
- 3.6 m (12 ft) TBM for MLKCT: 3.5 work days.
Advance rates and downtimes. The initial mining stage for all three TBM operations started with a singlerail track and a single train arranged in temporary configuration depending on the launch layout. The initial mining duration was different among the three operations based on the launching logistics and jobsite restrictions at each shaft.
Initial mining ended when:
- The switches and double rail tracks were installed (shaft re-set).
- The second production train was put into service.
- The crew was fully trained and all the scheduled production shifts safely deployed.
Below are the initial mining durations and advance rates for each operation:
- WCT — 176 m (580 lft) advance (4.1 m/d (13.5 lfpd) or 0.5 m/h (1.7 lftph) average) in 43 work days (1 x 8 hr shift).
- DVT — 189 m (620 lft) advance (4.7 m/d (15.5 lftpd) or 0.27 m/h (0.9 lftph) average) in 40 work days (2 x 8 hr shifts).
- MLKCT — 88 m (290 lft) advance 5.8 m/d (19.3 lftpd) or 0.37 m/h (1.2 lftph) average) in 15 work days (2 x 8 hr shifts).
As shown in Fig. 7, the MLKCT initial mining required less than half the duration and the length of WCT, which was bored using the same TBM. Both the jobsite parameters and the experience of the crew directly affected this outcome. The WCT was launched from a 23 m (75 ft) long x 4.5 m (15 ft) wide x 9 m (30 ft) deep portal shaft which was supported by a soldier pile and waler frame construction. For this drive, the required initial mining advance was 88 m (290 lft) further than the MLKCT due to the size restriction of this portal shaft. During the MLKCT, the entire 47 m (155 lft) of tail tunnel and 9.1 m (30 lft) of starter tunnel were utilized for the initial mining operation. This was an advantage as the machine did not have to mine as far before the double rail tracks and second production train could be put into service. From a crew standpoint, the key players that were part of the WCT and DVT drives were kept for MLKCT and strategically placed from the beginning on the two shifts. There was an adjustment period for the workforce, from the comforts of the highly automated DVT TBM back to the tight spaces offered by a 3.6-m (12-ft) diameter TBM, with construction material largely handled manually and the increased potential for injuries.
Similarly, during the DVT initial mining (on a singlerail track) only one production train was utilized. After six weeks and 189 m (620 lft) of advance, the switches and double rail tracks were installed, the second production train put into service, and the third eighthour shift implemented.
A faster advance rate was initially achieved at WCT, even working one shift less than at DVT. In fact, in the first 20 work days at WCT the advance was suspended only for back-up deck installation and ventilation setup. On a larger and more complex machine as DVT the first four weeks have been heavily affected by:
- Blocking and backfill of the first eight rings built above the launch cradle, in the horseshoe section.
- Testing/commissioning of the annular grout system and systematic verification drilling.
This is evidenced in Fig. 8, where delays in the process not specifically related to the respective machines (external delays such as shaft re-set and segmental lining backfill) are captured under “Other.”
The same chart also shows a substantial difference between excavation time at WCT and DVT. At the early stages of the initial mining, the main-beam TBM experienced gripping issues in the soft shale that limited the cutterhead thrust to a minimum. As a consequence, the penetration was reduced, and each push extended in time. In addition to this, for the entire time on a singlerail track only two muck boxes could be loaded, exposed and lifted out of the short launching shaft. This resulted in a fragmentation of the 1.5 m (5 ft) mining stroke into two or three sub-strokes per push (about 1 lft advance per each 8 cu yd box) and an overall extension of the excavation time.
None of this was an issue at DVT, where the single shield TBM was thrust forward against a reaction frame, and the length of the starter tunnel (combined to the co-axial tail tunnel) allowed the full muck train (seven boxes) to be utilized from the beginning.
Based on the chart in Fig, 7, 23 work days into the start-up process the DVT daily advance rate overtakes the WCT rate and the gap gets wider after the shaft reset.
Full production mining analysis
As anticipated, the roadheader operated within the full production mining range from the beginning and for the whole excavation of the starter tunnel top heading. In the case of the TBMs, the production records fall into this range when two essential conditions are met: the operation is run with two production trains and the crew is fully deployed (on 2 x 8 hour shifts in case of WCT; on 3 x 8 hour shifts in case of DVT).
No MLKCT data will be shown in this section, due to the fact the full production mining phase just started at the time this paper was submitted.
Advance rates. In terms of advance rates, all three curves in Fig. 9 demonstrate a very consistent operation throughout the duration. The average advance rate recorded at WCT was 17.7 m/d (58.1 lftpd) or 1.1 m/h (3.6 lftph) (on 2 x 8 hour shifts), with 27.7 m (91.1 lft) mined on the best day and a total of 472 m (1,549 lft) advance in the best month. This was achieved in October 2018 (second month of full production mining after the flood) in the range of 914-1,371 m (3,000-4,500 lft) distance from the shaft. The only deflection in the WCT curve is due to the single shift advance in the week of Aug. 27, 2018, which was the re-start of the operations after the tunnel flood experienced on April 15 of the same year.
The average advance rate recorded at DVT was 24.4 m/d (80.1 lftpd) or 1 m/h (3.3 lftph) (on 3 x 8 hour shifts), with 33.5 m (110.1 lft) mined on the best day and a total of 587 m (1,925.8 lft) advance in the best month. The top production records were all achieved in May 2019 (second month of full production mining) in the range of 914-1,828 m (3,000-6,000 lft) distance from the shaft, similar to what was observed at WCT. It was predicted for both tunnels that the two muck train logistics (with muck boxes dumping time always on the critical path of the production cycle) would have performed at its best in the second half of the WCT drive and in the second third of the DVT drive. That is when the excavation + ring build + train transit time was expected to equalize the time required to dump the muck boxes and re-set each train (40-45 minutes at WCT; 55-65 minutes at DVT).
The average advance rate recorded during the excavation of the starter tunnel top heading was 2.3 m/d (7.48 lftpd) or 0.09 m/h (0.3 lftph) (on 3 x 8 hour shifts), with 3.4 m (11.42 lft) mined on the best day and a total of 43.2 m (141.95 lft) advance in the best month. This was achieved in June 2018 (second month), with best day in the range of 70-75 m (230-245 lft) distance from the shaft.
As largely expected, irrespective of the excavation methodology, the muck removal and the supply chain soon became the main factors driving the pace of the tunneling cycle (Fig. 10). With careful planning and surface/underground crew coordination, some nonboring activities have been worked on simultaneously, so that the actual downtime was less than the sum of the parts. Specific to DVT, ordinary maintenance, survey tasks, utility extension, grout ports cleanup, annular grout verification drills and contact grout would normally be performed as needed during ring build and while waiting for the next muck train. Cutterhead inspection and grout system cleanup are performed during scheduled maintenance time (typically once a week).
The large amount of time under “other” for the roadheader process is related to the frequent changes in operation within the production cycle (from mining to bolting, from bolting to shotcreting, from shotcreting back to mining).
Machine availability and utilization. Maximizing the overall advance rate of a tunneling system is dependent not only on the excavation speed that can be achieved, but also on minimizing the time that the system is not operating (downtime). The efficiency of the system is usually measured by the “utilization,” defined as the percentage of the total shift working time or total production cycle in which the productive capacity of the machine is used. Nonworking shifts, weekend days and holidays are not included. A related term is “availability,” to express the percentage of time that the machine is available for use, divided by the maximum amount of time it would be available if there were no downtimes for repair or unplanned maintenance. The availability has a direct bearing on the utilization.
At Doan Valley, 62 percent utilization was achieved on the 3.6-m (12-ft) TBM (excavation and ring build) during the WCT full production mining. This is about 7 percent lower on the 6.4-m (21-ft) TBM, mainly because at higher penetration rates the utilization tends to decrease (as excavation time is reduced). The low 33 percent for the roadheader, reflects the fact that the machine was utilized for one-third of the whole production cycle (excavation only), while the remaining time was spent operating different equipment for installation of ground supports. The high availability (always more than 75 percent, up to 87 percent on the 3.6-m (21-ft) TBM) reflects the effectiveness of the maintenance work mostly performed during the routine downtimes, built into each production cycle (rock support installation in the case of the roadheader or waiting for the next production train in the case of the TBMs).
It is interesting to note that within the same geology, operating process and crew, there is almost no difference in terms of availability (also read as reliability) between the brand new TBM (DVT) and the repeatedly rebuilt TBM (WCT).
Material and labor cost analysis
This review is confined to the material placed and left in the tunnel (ground support and/or final lining) and the direct labor hours involved with each construction process. From the estimating stage, the labor cost and the permanent material combined represented the first and biggest contribution to the construction of the three tunnels and was therefore a major target for reduction with improved performances.
In general, the cost of the materials left in the tunnel tends to be higher with the size of the tunnel. Considering that only the DVT was a single-pass tunnel, Table 5 provides an actual measure of how the WCT and the ST compare to the permanent material cost in the main tunnel (in absolute terms, assuming DVT=100 per linear foot).
The main benefits of a single-pass approach are in terms of earlier availability of the tunnel and overall reduction in labor cost per linear foot of tunnel excavated. Table 6 provides an actual measure of how the WCT, DVT and MLKCT compare to the labor cost in the Starter Tunnel (in absolute terms, assuming ST=100 per linear foot).
The man hours per foot spent at MLKCT initial mining are half of what was invested at the same stage of the WCT drive. This is in line with what was already pointed out analyzing average advance rates, distances, and durations of the respective initial mining stages. It is remarkable but not unexpected the amount of hours that go into moving a larger machine forward up to the point the second production train can be implemented, especially when those (DVT) hours are more than double if compared to WCT, more than four times if compared to MLKCT.
At the full production mining stage, the low efficiency and high labor cost of the roadheader operation are the reasons why the length of the service tunnels must always be minimized, with careful reflection. A starter tunnel only five feet shorter would have meant no room for back-up deck no.6, and a whole set of labor-intensive temporary arrangements that would have heavily impacted the TBM launch.
The disproportion between the TBM tunnels and Roadheader tunnel becomes even more striking if considering the man-hours required to bring the 23’ wide by 23’ high horse shoe section of the starter tunnel down to the 18’ dia. of the DVT final lining. At the same time, after the direct labor cost for the second pass (pipe installation and cellular grout backfill) is factored in, the delta between WCT and DVT becomes minimal.
At Doan Valley, due to the relatively short length and duration of each tunnel drive, the start-up phase of each tunneling operation (initial mining) had to be reduced to a minimum, so as to achieve the scheduled advance rates as quickly as possible. At this stage, in terms of hourly footage, the 3.6-m (12-ft)-diameter main-beam TBM performed up to two times better than the 6.4-m (21-ft)-diameter single-shield TBM, and in terms of labor cost up to four times better than the 6.4- m (21-ft) TBM.
Following the initial mining, steady system performances had to be maintained throughout the entire “full production mining” phase, turning the downtimes built into each cycle (such as the muck boxes handling at the shaft) into opportunities for concurrent non-mining tasks. In this scenario, the machine utilization achieved by both TBMs was twice the one recorded for the roadheader with hourly advance rates 10 times greater than for the roadheader.
In terms of material cost per linear foot of finished tunnel, the one-pass tunnel lining solution result was comparable to the cast-in-place option but two times more expensive than the two-pass. In terms of labor cost per linear foot for the finished product, the starter tunnel turned out to be almost 10 times more expensive than both TBM tunnels. This is the reason why every effort was made to optimize the cross section and length of the service tunnels.
Across the whole process, a key factor was the detailed planning and engagement of the key field players from the early stages. Then, the same crews were deliberately kept together for more than two years on three separate, but in many ways comparable, tunneling operations, to minimize the learning curves.
As a result, what was accomplished at each drive was not record-breaking production, but a performance consistency in the scheduled distance/time and within the budgeted costs.
Thuro, K., and Plinninger, R.J. 1999. Roadheader excavation performance – geological and geotechnical influences. In Proceedings of the 1999 9th ISRM Congress Paris, Inc. 1241-1244. Eskikaya, S., Bilgin, N., Balci, C., and Tuncdemir, H., 2005. From research to practice “Development of Rapid Excavation Technologies”. Metall. Underground Space Use: Analysis of the Past and Lessons for the Future. Taylor and Francis Group, London, ISBN 04 1537 452 9. Schneider, E., Spiegl M., Turtscher M., and Leitner G., 2011. Hard-rock TBM performance prediction. Tunnels & Tunneling International, January 2011.
Scialpi, M., Comis, E., and Clark, J., 2012. System efficiency in one of the world’s longest TBM tunnels. In Proceedings of the 2012 World Tunneling Congress, San Francisco.