Big Data, Writ Large: Technology to Advance Tunnel Boring

Modern TBMs are data-driven systems, from ground investigation tools ahead of the machine to touch-screen technology in operator’s cabins, to integration with programs on the surface. Today’s TBMs, paired with cutting-edge data collection and monitoring, can efficiently bore in even the most demanding circumstances.

Nearly all the parameters of a TBM can be monitored today, and this data can be transmitted via fiber optic cables to offices on the surface or even mobile phones. This data has the power to turn a project around. Simple observations, for example cutterhead RPM and penetration rate in a given geology, can result in altered operational parameters and reduced thrust that can speed up advance and increase cutter life. All that is required is proactive analysis by management and engineers, and good communication with the TBM operator.

But how do we accurately interpret the data, and ask the right questions of it so that we can improve our projects? And how do we then work with TBM operators and other crew members to make the best choices on a given project? The answer, explains Aaron McClellan, Tunnel Superintendent for Kiewit Underground, comes back to the human element. “Whether it’s Artificial Intelligence or data algorithms, everything we’ve built has a human hand in it. It’s going to take people to ask the right questions, and the people are going to take this where it needs to go to benefit our industry.”

Technology Gains Traction

Big data has been around for a long time, but the questions humans are asking of it are reshaping the possibilities. With ever-increasing network, server and wifi capabilities, the sheer amount of data can be overwhelming. What we need to do, says McClellan, is “work smarter not harder, and do more with less. We need to push this into the new era and get on board with increases in data. Otherwise, in the contracting world you’ll be left in the dust. About 35% of the world’s GDP was spent on the construction industry in the last 20 years, but in that time production in the construction industry only increased by 1%. The whole end game is to get more out of the data so we can better apply our knowledge to improve.”

In a climate where data is being used to answer questions in near real time, delays and inaction are the enemy. The best way to ensure the data is being used correctly, says McClellan, is to have someone “to analyze the data on site. You have to have the foresight early on, especially with data logging, of what types of questions you will try to answer throughout the project. Each question will require different tools to solve the problems, and the tools will also be different depending on the equipment, whether we are looking at hard rock TBMs, slurry, or EPB. The goal is for people to be able to adjust quickly in order to react quickly.”

Data logging systems, McClellan adds, are widespread and several companies offer interpretation software that combines data from the equipment with monitors and instrumentation as well as geotechnical data to determine trends and any problems. But the data must be in near real time in order to get the reaction time needed underground to make a real difference.

“Overall, there are three different goals with the data we are collecting: One is to reduce risk to be more competitive, and to go along with that, not add additional risk to said work through automation. Second, we are also looking to increase our production: faster ring builds, faster advances. Lastly, there is optimization, and when I think about optimization it’s not about production, it’s about doing more with less. I think about eliminating our waste and doing the best with what we have.”

Data-Driven Technology Applications

Various data-driven applications can be seen within the lens of the three goals, says McClellan. “To reduce risk there are some new developments regarding foam control and muck consistency. EPBs can be prone to blockages in the muck chamber, plenum, cutterhead, and screw conveyor, which causes EPB sensors to give erratic readings. New developments recognize that these pressures are erratic and work at controlling the foam to be able to control the pressure gradient from the face through the plenum and into the screw conveyor. This achieves better muck consistency, which should make the sensors read much better as well. This is something being developed right now.”

“Another way to look at it is from the production side, with locomotives and rolling stock. It’s 100% possible to be able to use telematics to understand our cycling of our locomotives. Ideally, we should be able to understand easily and quickly that, for example, in the next 1,000 ft I’m going to need to install my switch, and if I don’t then the TBM will have to wait on the train. All of these things, even simple things that you can work into your hoisting or mucking equipment, can help us optimize logistics cycles.”

“In terms of optimization, I am thinking along the lines of less waste. For example, there’s been a couple of projects that have used ventilation on demand. Basically, this involves ramping up the ventilation relative to the people and equipment underground to better reduce kWh charges on a job.”

Artificial Intelligence

Artificial Intelligence (AI) is another aspect of technology that is a hot topic in the industry today, and it could be a game-changer. But AI has a long history, explains Robbins Vice President of Operations Steve Chorley: “In 1992, I worked on a project in Africa, the Lesotho Highlands project, where we monitored four machine parameters to create an automatic boring mode on a TBM. It did work, but there were many problems with it. I think with the development of soft ground machines, particularly in Japan, they have been at the forefront of developing this big data and using it for their advantage. How far have we come? Well, if you look at recent developments in Malaysia, they have developed the Autonomous or A-TBM. I think that’s a big step forward, to control steering and advance the TBM in an automatic or semiautomatic mode.”

Even with rapid recent advancements, both Chorley and McClellan urge for slow and steady progress. “It’s coming to us all and we need to get used it. If you can keep the TBM steering in line, and increase performance as a fully automated machine, then that’s the future. But right now, we’re not 100% ready for it. The crew in Malaysia, they are the pathfinders and they’ll show us the way,” said Chorley.

“The way that I look at it goes back to the risk,” said McClellan. “We need to use AI to our advantage; from the contractor standpoint, it’s going to be standard operating procedure. But we need to consider the impact of reducing crews by implementing more AI. Are we creating more risk by doing so? The way I look at it is like in an airplane or car, there are algorithms for automated operation but there’s always a human element behind it and rightfully so. I’m sure both of those industries can name countless incidents where human intervention was needed in order to be able to correct a problem. We should take a lot of this with a grain of salt, and make sure we can back up what we say. I think we’re years away from true AI in tunneling, and we’re rightful to take our steps forward cautiously. We need to make sure we’re doing the right things for the industry.”

Operator Training

The human-machine interaction is a key point of ensuring that big data is used correctly and efficiently. But in an industry where there are no international standards for those driving the TBMs in the operator’s cab, implementing any changes is a challenge. “In boring machines, I’ve seen lots of sinkholes and ground heave worldwide. Mainly it comes down to human error; the cost of creating a sinkhole can’t be underestimated. I’ve observed a sinkhole in London where a householder reported a TBM in his backyard. How did that happen? Well, the operator simply saw muck coming out of the screw conveyor and presumed he hadn’t finished the cut. Better training could have prevented that incident,” said Chorley.

But is it better training on the machine that’s needed, or better training off the machine in a safer situation? Since 2000 the International Tunnelling Association (ITA) has identified the need for training courses, but offerings for training are few to date. “In a lot of countries today, when you visit the jobsite you’re required to take several safety courses, even before you’re allowed underground. But we’ve got a machine that costs millions of dollars, and for the actual crew we let them figure out operation from just a few interactions, on-the-job, with someone saying ‘these are the controls, this is the pressure,’ etc. Yes, typically the operators have experience, but sometimes an operator will tell me ‘well it wasn’t like this on the other machine. It’s not like that on this machine; I thought this would have happened.’ If you can take that operator off of the machine and give him or her controls for the machine in an office environment, I think we can prevent a lot of these major incidents.”

To that end, simulator programs are being developed, including one run by the Colorado School of Mines that focuses specifically on EPB TBMs. The simulator currently exists as a standalone system operating on a PC, but it could be linked to a system in an actual operator’s cab on a TBM.

There are an endless number of scenarios that can be created for EPB operator training, but at the moment five scenarios have been developed. These scenarios focus on EPB operations during excavation and standstill (ring building) within each scenario. Geological parameters and tunnel depth can be modified, and include the following scenarios: Boring in sand; boring in clay; boring through transitional geology; boring in mixed face conditions; and boring in homogeneous conditions.

While the simulator is only for EPBs, plans are underway to expand training for hard rock tunneling in the future.

Command Chair

Related to the EPB Simulator is the idea of a Command Chair—a chair that is self-contained and can be used on the surface or in an operator’s cab, and can be loaded with simulations so the operator can get experience in different scenarios ahead of actual machine operation. “If I can incorporate the data from the machine into the chair, I can get real time training,” said Chorley. “I think that’s the next phase of the program, to integrate the software into an operator training cabin, similar to an airline pilot or even excavator operation.”

The Command Chair contains all systems necessary to control the TBM within its design. Robbins designers, with input from onshore and offshore command chair manufacturers, and field service personnel (those who actually operate the machine), came up with an innovative design based around a helmsman’s chair. The two panels on either side of the operator have been designed with modular control panels that are interchangeable based on the type of machine the chair will be installed on. Each modular panel is specific to that type of machine and is designed to be instinctive for the operator to start and control the machine. Functions on the control panels have been reduced to the absolute minimum with all other functions to control, monitor or stop and start systems being transferred to two touchscreen panels also incorporated in the command chair.

“The advantage of this chair is that it is self-contained. We can send this chair to site months before the TBM is ready, and we can have the operator undertake training. He or she can be ready to be on that machine, trained, and know how to react to the conditions because we will have them in the simulator, whether it’s a sinkhole scenario or ground heave or something else. They can also be trained on proper machine operation, know which controls to use, how to close gates on screws, what to do in case of incidents, how to shut down the machine etc. I think the command chair, being self-contained and able to be transferred from an office to a machine, is a good advantage for anybody,” said Chorley.

Robbins is currently testing the Command Chair at a project in Mexico. The functionality of the chair is good, but Chorley adds that for it to work properly in other tunnels, there must be high-speed internet: “It’s got to be a fiber optic connection. If we really want to transmit big data, we really have to support those high-speed connections.”

In terms of remote machine operation, Chorley adds that it’s still in the future: “If you have a high-speed connection and an operator’s chair on the surface, there is an argument to say ‘Why don’t we operate on the surface all together? Why do we need a person on the machine?’ I don’t see it being ready yet for remote operation. In terms of the technology we may be ready to do it, but in terms of the safety of the personnel on the machine I don’t think we’re quite ready for remote operation.”

Real Wear Headsets

Another use of data in the tunnel can be seen in the use of headsets to get real time feedback from underground to remote locations on the surface. Robbins Field Service technicians currently use headsets, manufactured by RealWear, to accomplish this task. The headsets are essentially hands-free, voice-controlled, head mounted tablets, designed to be dust-tight, waterproof, and drop-proof. The devices are also compatible with safety equipment so they can be worn with a hard hat and safety goggles. The headsets are also undergoing testing by Robbins at a project in Mexico.

“We’ve had reasonable results. The biggest problem we’ve had is that we need to improve our internet throughout the machine, and make sure we have wifi or hotspots to get constant communication,” said Chorley. “The advantage of these headsets is that the technician is able to talk to our offices here in the United States in real time. The headset has a built-in camera, built-in audio, and built-in video. The technician can be looking at a problem on the machine and relaying real time information back to me. We can actually send that technician data and drawings to help him or her to troubleshoot a problem.”

While the technology is still being perfected, Chorley sees a lot of potential: “In the future I think wifi along the entire length of the tunnel will be needed, because you’re going to require this kind of headset technology as a standard. It makes it easy to transfer data from or to an individual person on the machine.”

Data in Action: Drilling Data on a Recent Project

On a recent project, a Single Shield TBM was equipped with a unique setup to drill in limestone where water was expected: two drills in the shields were used for drilling through the head in 16 different positions and a third drill on the erector was able to drill through the shields in an additional 14 positions. To add to that, water-powered, high pressure down-the-hole (DTH) hammers allowed for drilling 120 m ahead of the machine at pressures up to 20 bar if necessary.

“We actually had a lot of data; we had data coming through our data logging system, data coming from engineers out in the field. This data was amassed but it was tough in the moment to be able to visualize these data. Later on, we were actually able to go back and make visualizations with it, incorporating data relative not only to the depth of the drill but also relative to the water we would pick up along the drilling path,” said McClellan.

Using the visualizations, the contractor was able to pick up on patterns, identifying discreet features along the tunnel alignment. “We looked at drill depth, water ingress, and grout injected. Drill depth was recorded using a data logger and instrumentation installed on our drills. This would then be verified with very detailed field engineer recordings. Not just depth, but drill penetration rates, chuck times, and utilization were also recorded. Water ingress was only recorded through intermittent stops where flow tests would be done with bucket or barrel. These stops would be usually done on 50 ft intervals, but were more frequent if we knew a feature had been hit; usually the driller would have an idea if that was the case. Grout injections were recorded at the underground batching unit that had its own paired data logger. This would also be field verified with our field engineers. The spheres in the images were scaled relative to a 2:1 grout mix with its test bleed and the center of the sphere was placed in a location that represented the average of where the water was picked up along the drill length.

McClellan added that while in this example the analysis was done post operation, the data can be used for future tunnels and the process could be used during tunneling. “All the elements of the data are there, it’s more about what you want out of this data, and what you are trying to achieve. You can answer questions like ‘If we don’t grout this ground, what can we expect from the tunnel face when we push through this zone?’ It’s much better than brainstorming on the fly during a tunneling operation.”

Conclusions

Big data is here to stay, and it will only get bigger. But there are some key takeaways that can help our industry leverage the advantages it offers. “All of these scenarios with data applications and interpretation require people; good people with experience, who are able to get these analytics and algorithms as priorities out front where they can make an impact. A lot of times when you’re trying to figure out these processes on the fly in the tunnel, you’ve already missed the boat. You need to be proactive in your approach,” said McClellan. “Verification is key with all this data management. It is easy to just sit back and say data loggers are collecting the data, but there are plenty of times in my career that data collected was incorrect and humans realized it, not computers. This is one of my main arguments for keeping people in this data management cycle.”

For Chorley, part of gaining bigger acceptance in the industry is for people to realize that data can be used on any equipment, whether new or refurbished, and can be just as effective as long as the staff are there to interpret it. “We have a machine that was used recently in France on the Galerie des Janots project. It was first used in 1983, but for this project we installed a small PLC. It was not too cost prohibitive, and we were able to get that PLC integrated into a data logging system, so we could get the same data we’re getting from modern TBMs (built in the last 5 years) off of an older machine built in the 80s. It’s doable, and it’s a real advantage.”

“The other key to being successful is having someone on staff who knows how to analyze and interpret the data. The crew also needs training: We need to have this interaction between engineers and the crew to get the best out of the machine, in order to better understand the problems, monitor face pressure, etc. When you get that interaction working smoothly the advantages far exceed the investment.”

RELATED: Liebherr, Robbins Team Up to Ease Congestion in Mumbai