Underground Construction - October 2022 - 14

Artificial Intelligence
Going digital
Rise of A.I.
The pace and scale of Houston's challenge is so daunting that
Rabbi and others in its water department were early to recognize
the essential role that technology must play in achieving
their objectives.
Houston's wastewater infrastructure is staggering, with
more than 6,000 miles of pipes, 133,000 manholes, 381 lift
stations and 39 treatment plants serving a population of more
than 2.2 million, along with some smaller area municipalities
that contract with the city for treatment services
Rabbi, who first joined the city's Public Works department
in 2005, said his group has been working to improve Houston's
wastewater systems since he rejoined the group in 2012 after a
four-year consulting hiatus.
" It's not that we just started as part of the consent decree.
We were negotiating the consent decree over many years, and
we knew that we had to be more efficient, " Rabbi said, describing
the lead-up to adding A.I. in its workflow.
" So, it's not new for Houston to be proactive in evaluating
and adopting cutting-edge technologies whenever something
new is available, " he added. " This is a continuation of the culture
our city has. "
One of the early, ongoing initiatives has been the installation
of digital monitoring sensors. More than 300 of them
had been installed underground before the consent decree
was signed, and Rabbi said the city negotiated to increase that
number ten-fold under the deal. More than 600 have been installed
to date, he said.
Houston has long embraced the use of robotic CCTV inspections
of its sewer lines operated by above-ground specialists,
along with other technologies to better manage its systems and
help replace a retiring workforce with digital-native millennials.
" Everything has to be digital, and we are always exploring
new imaging technologies that may help us improve and manage
our wastewater systems better, " Rabbi said.
One area in which his department has been on the leading
edge of inspection technology is the early adoption of GoPro
360-degree cameras in a quest to inspect some 10,000 manholes
per year.
Using this method, small crews can capture data using a
sub-$500 camera that would otherwise require sophisticated
manhole scanning systems at a capital investment of around
$150,000, said Eric Sullivan, Sewer AI's director of business
development.
" They can attach the GoPro to a pole and, with the right
lights, capture data that we think is actually superior to what
those other rigs can offer, " said Sullivan, a 10-year veteran
NASSCO trainer. " It's all about the software that one would
use to process this data. "
Employing two fisheye lenses that slightly overlap each other's
field of view, the GoPro cameras capture millions of pixels
of video data that Sewer AI's photogrammetry software processes
to produce a full, geometric 3D model of the structure.
14 OCTOBER 2022 | UCONonline.com
Houston's growth, terrain and weather create unique challenges
for the city's wastewater systems, but its aging water
infrastructure is a problem that plagues cities throughout the
United States, where the average U.S. water-network pipe is 45
years old and many cast-iron pipes are more than a century old,
according to a 2021 report by the American Society of Civil
Engineers (ASCE).
Artificial intelligence technologies moved into the wastewater
management space to help system owners address the challenges
of both aging infrastructure and the loss of experience
with an aging workforce.
Traditional CCTV inspections require that the individual
who operates the robot also be an expert in how sewers deteriorate
and how to do condition assessment. It can be hard to
find people who are knowledgeable, consistent and know how
to operate the equipment.
Sewer AI, Sullivan said, was formed to provide solutions to
reduce the cost and increase the speed and data quality of inspections
and provide cloud-based tools to collaborate in ways
that weren't possible before.
" Since the dawn of the CCTV equipment, the most common
workflow is that the operators are driving the camera in
the pipe, and when they see something in the live video, they're
going to stop the camera and type something in to code it, "
Sullivan said, citing a German study of CCTV operators that
found about 20 percent of structural defects don't get coded
using a manual method.
" Our point at Sewer AI is that the operator should never be
taking their hands off the camera controls. They should never
be wasting time figuring out if it's a crack circumferential or
a fracture multiple, " he said. " They should just make sure it
got on the video, and then they should get that video to the
experts who use AI. "
When it does go to the experts, their artificial intelligence
systems are looking at every pixel in every frame of every video.
And each of the different AI tools is designed to look for
distinct types of conditions in a way that Sullivan says is " much
more rigorous than you would ever get from people manually
watching or trying to annotate at that level. "
It's also dramatically faster, a key benefit for Houston's aim
of inspecting more than 5 million linear feet of pipe per year.
And, as Rabbi pointed out, one of the biggest advantages to
using artificial intelligence technology is that it continues to
learn and improve the more data it processes.
" The A.I. tools that we use are processing about 25,000 to
50,000 feet per hour, and we can run multiple inferences in
parallel, " Sullivan said. " So technically speaking, there's very
little limitation to it. " UC
FOR MORE INFORMATION:
Sewer AI: (866) 631-1323, sewerai.com
HR Green: (713) 965-9996, hrgreen.com
http://www.sewerai.com http://www.hrgreen.com http://www.UCONonline.com
Underground Construction - October 2022

Table of Contents for the Digital Edition of Underground Construction - October 2022
