Artificial intelligence is being used by engineers at the micro-levels of design and engineering, and for small- scale simulations. Now computing resources are
becoming available to deploy AI on large-scale development and
Math Works—the maker of popular Simulink and MATLAB—is
folding more AI tools into its portfolio for more efficient engineering design and simulation. AI will offload a good part of the
design process to the computer, with algorithms helping deduce
the next steps to take, much like an autonomous car deciding
the next step after it recognizes a person crossing a road. As a
result, engineers have more time to think up effective designs.
Another area of application for AI is predictive maintenance,
which came under the microscope recently after a fatal engine
explosion during a Southwest Airlines flight. AI tools could take
advantage of data collected by components like sensors to
create a more reliable maintenance model to prevent such accidents. The maintenance will improve as more data is collected
and fed into the learning model.
AI IN DESIGN
New engineering projects will benefit from artificial intelligence, but what about things that are already built? AI could still
provide a boost. For example, to keep aging wind turbines from
failing catastrophically, models and simulations can be used to
figure out when the turbines need to be serviced. The idea is to
collect and harvest data for predictive actions, said Andy Thé,
partner manager for AI and deep learning at Math Works.
Many are looking into AI implementation in engineering
projects, but deployments have been uncommon, more in single
digit percentage points, Thé said.
Math Works is adding to MATLAB—a programming language for technical computing—the AI framework tuned for
use on GPUs from Nvidia, which are the backbone of speech
recognition tools like Amazon’s Alexa, or image recognition via
Facebook or Google. Students and engineers can develop AI and
deep-learning models in MATLAB for faster development and
more effective engineering design, simulation, and validation.
“It’s a platform for scientists exploring an engineering
project. It enables ideas,” said Amnon Gai, manager for partner
marketing and development. Engineers can hook MATLAB to
Simulink, which is popular for graphical design and simulation.
Combined, the two can be used for full-scale prototyping.
AI in the form of machine and deep learning isn’t new to
Math Works. The company has possessed the hooks in the form
of neural network toolboxes, and the industry-specific algorithms have been around, Gai said. But GPUs have emerged to
enable supercomputer-like number crunching, and the availability of prebuilt functions and libraries makes the integration
of AI into engineering projects simpler. ME
Piping Systems Research & Engineering Company
PIPING AND EQUIPMENT
ANALYSIS & SIZING SUITE
Enabling new users to perform
piping and equipment analysis
in days rather than months
Smart Simulation & Sizing Tools
for Every Piping and
ASME PVP 2018 Booth 14
Photo: Getty Images