COVER FEATURE | PUTTING BIG DATA TO WORK

F32

PUTTING
BIG
DATA
TO
WORK
BY AHMED NOOR

IMAGINE

WHAT FACTORIES WILL DO WHEN

THEY CAN TAKE FULL ADVANTAGE OF

ALL THE INFORMATION AT HAND

t a missile plant in Huntsville, Ala., for instance, Raytheon

Corp. monitors its assembly operations down to the turn of a
screw. If a screw is supposed to turn 13 times after it is
inserted in a missile component, but it turns only 12 times,
an error message flashes, and production is halted until the anomaly is
understood and rectified.
Just how much smarter factories can become will be determined
largely by the research and development of systems to manage information. Smart factories, after all, are part of the phenomenon popularly
known as Big Data.
The challenge of Big Data is that it requires management tools to make
sense of large sets of heterogeneous information. In the case of a factory,
sources of data include CAD models, sensors, instruments, internet
transactions, simulationspotentially, records of all the digital sources
of information in the enterprise. The data bank is large, complex, and
often fast-moving, and so it becomes difficult to process using traditional
database analysis and management tools.
Raytheons monitoring technology is often called manufacturing execution software, and several manufacturers are currently using MES to collect
and analyze factory-floor data. The systems enable the real-time control of
multiple elements of the production process.
Industry stands to reap many benefits from Big Data as more sophisticated
and automated data analytics technologies are developed. These technologies

TURN OF THE SCREW

If a screw in this missile

fails to complete its full
count of turns, Raytheon
will know about it immediately and be able to take
corrective steps.

Manufacturers already can monitor their operations in real time

and in close detail with manufacturing execution software.

BIG DATA AT WORK

will help extract value and hidden knowledge from large,
diverse data streams.
For example, the vehicle sensors in the Ford Focus electric
car produce vast amounts of data while the car is driven, and
even when it is parked. While the vehicle is in motion, the
Smart manufacturing aims at integrating these islands,
driver is constantly updated with information about acceleraenabling data sharing throughout the plant and among plants
tion, braking, state of battery charge, distance to charge points,
in a company.
and location. And while the vehicle is at rest, the sensors conA new wave of inexpensive electronic sensors, microprotinue to stream data about the tire pressure and battery system.
cessors, and other components enables more automation in
The driver gets useful, up-to-the-second data, while Ford
factories, and huge amounts of data to be collected along the way.
engineers in Detroit use the cars communication modules
Managers can get instant alerts about potential problems or study
and remote application management software to aggregate
the data to find ways to boost efficiency and improve perforthe information about driving behaviors in
mance. In addition, enhanced knowledge of
order to gain customer insights and plan prodthe workplace promises to improve worker
fyi
uct improvements. In addition, third-party
safety and to protect the environment because
vendors, like General Electric, AeroVironit can make zero-emission, zero-incident
RAYTHEON CO.
ment, and ECOtality, analyze millions of miles
manufacturing possible.
www.raytheon.com
worth of driving data to decide where to loIn automated manufacturing, Big Data
HEADQUARTERS: Waltham, Mass.
cate new charging stations, and how to protect
can help reduce defects and control costs
FOUNDED: 1922, as the American
the fragile utility grids from overloading.
of products. By tracking every detail about
Appliance Co.
Smart manufacturing has the potential of
every part that goes into a product, from
EMPLOYEES: 68,000.
REVENUES: $24 billion.
fundamentally changing how products are
its original manufacturer, to where it was
invented, manufactured, shipped, and sold.
stored, to when it was installed, the data
Raytheon is an international aerospace
A typical factory uses information
bank lets manufacturers retrace problems for
and defense company, whose products
technology, sensors, motors and actuators,
better resolution. Monitoring defect ratios
range from large-scale radars to cybersecurity technology. Its four principal
computerized controls, production manageand on-time delivery can help with supplier
business units are Integrated Defense
ment software, and the like to manage each
selection and performance assessment.
Systems; Intelligence, Information,
specific stage or operation of a manufacturInside the factory, having the ability to
and Services; Missile Systems;
ing process. However, each plant is an island
utilize
the mass of data on orders and maSpace and Airborne Systems.
of efficiency.
chine status allows production managers to

PROGRESS OF AUTOMATION

Automation makes factories more efficient and therefore comMECHANICAL

| OCTOBERintelligence
2013 | P.00 of
petitive, so the trend must
be for ENGINEERING
more. Increasing
machinery and production lines decreases the need for human
intervention in manufacturing processes. More and more, factory
jobs dont require use of the hands as much as use of the mind.

CARS THAT COMMUNICATE

Sensors in each Ford

Focus Electric produce
streams of data while
the car is driven and
when it is parked.
Not only is the driver
informed, but Ford and
others can analyze the
information to improve
the vehicle and to deal
with issues like placing
charging stations
or avoiding grid
overload.

VISUALIZING BIG DATA

A graph tracks how much

energy is produced by
more than 700 GE gas
turbines over a two-week
period. Each flash represents a turbine turning
on at some location in
the world. The wheel is
arranged according to
capacity of the turbines.
The height of the graph
represents the maximum
output of each turbine.

he exponential increase in the volume, variety, and

speed of data generated by simulations, sensors, and other sources is expected
to continue in the coming years. The data generated by large-scale simulation
is predicted to reach terabytes per second. There is a need for developing novel
intelligent software agents, or cybernetic assistants, to harness the power of BigData, and to
enhance the productivity of the engineering workforce. They can
help in crowd sourcing, and collaborative inventions, resulting in
pushing the limits of what is possible for engineering systems.
The cybernetic assistants combine machine learning, natural
language understanding, and novel modalities of human interaction to complement human intellect in new ways. Cybernetic assistants can leverage Big Data analytics tools and computational
knowledge technologies developed by the research teams at the
IBM Watson project, Wolfram Alpha, and other organizations.

Digital
Assistants
to Wade
Through
It All

They can exploit the new types of

predictive search facilities for dynamic
applications, where information is flowing
continuously. An example of predictive
search facilities is the MindMeld developed
by Expect Labs. It is a smart phone app that
listens to conversations and searches for
related material on the internet. As Expect
Labs describes it: As your conversation
changes, MindMeld continuously finds and
displays relevant pictures, videos, articles,
and documents so that any information you
may need is at your fingertips.

The cybernetic assistants may

eventually provide specific answers to
questions and summary reports on any
subject, or recommend a course of action
in addressing a problem. In addition, the
virtual assistants will have the capability
of tracking every action that the users
make on a computing device, and through
the use of state-of-the-art sensors (such
as gaze-tracking cameras, touchscreens,
and gesture detection) and predictive
analytics tools, deliver information
without the users needing to ask. n

P.36 | OCTOBER 2013 | MECHANICAL ENGINEERING

optimize operations, factory

scheduling, maintenance, and
workforce deployment.
A number of developments under way
now aim to realize the greater potential of
Big Data in smart manufacturing.
The Smart Manufacturing Leadership
MACHINE SHOPS
Coalition, officially incorporated as a nonApplications of data
profit organization in July 2012, aims to
management put
overcome the barriers to the development
people farther away
and deployment of smart manufacturing
from the factory floor.
People are outnumsystems. The coalitions members include
bered by machines in
stakeholders from industry, academia,
the shop below, and
government, and manufacturing. The
in a newer factory are
organization supports development of
almost gone.
approaches, standards, platforms, and
shared infrastructure that will encourage
adoption of smart manufacturing by providing easy access, lowering cost, and resolving serious technological barriers found in
todays manufacturing environment.
AUTOMATED MACHINES
Teams of members are currently developREQUIRE SOMEONE
ing a shared, open-architecture infrastructure
TO SERVICE THEM.
called the Smart Manufacturing Platform that
SOME MACHINE
allows manufacturers to assemble a combiOPERATORS
WILL
nation of controls, models, and productivity
BECOME
MACHINE
metrics to customize modeling and control
systems, making use of Big Data flows from
MINDERS, WHICH
fully instrumented plants in real-time. The
OFTEN CALLS FOR
development of a smart manufacturing
A BROADER RANGE
platform is intended to help management
OF SKILLS.
across the manufacturing ecosystem, and to
forecast activities that, for example, can be
used to slash cycle times and reduce the risks
of product development.
The infrastructure will enable third parties
or service centers when maintenance is needed. Data showing how customers use
to develop networked industrial applications
products can suggest improvements in design or manufacturing.
that provide tools for resource metering,
automated data generation,
mart manufacturing is likely to evolve into the new
intuitive user interfaces, and
paradigm of cognitive manufacturing, in which
fyi
Big Data analytics. Enhanced
machining and measurements are merged in order
computing power, an enabling
to form more flexible and controlled environments. When
FANUC CORP.
open-architecture infrastructure,
unforeseen changes or significant alterations happen,
www.fanuc.co.jp
and software advances can make
machining process planning systems receive on-line
HEADQUARTERS: Japan:
data management invisible to the
measurement results, make decisions, and adjust machining
Oshino-mura, Yamanashi Prefecture.
engineer, the bulk of it happening
operations accordingly in real time.
United States: FANUC America,
in the background.
It is conceivable that one day machines and processes in a
Rochester Hills, Mich.
factory will be equipped with capabilities that allow them to
Applications do not end at
FOUNDED: 1956
assess and increase their scope of operation autonomously.
the factory. Future products can
REVENUES: 489.4 billion
This changes the manufacturing system from a determinbe outfitted with sensors that
($5.1 billion).
istic one, where all planning is carried out off-line, to a dyconnect to the cloud, enabling
FANUC designs and manufactures
namic one that can determine and reason about processes,
after-sales service offerings.
automation technology for all stages
plans, and operations.
There could be an option in cars,
of factory production lines.
Because of smart and dexterous robots, and novel
for instance, that will alert drivers

Manufacturing will still need people.

Tools and Technologies

Traditional data warehouses are not able to handle
the processing demands posed by Big Data. As a
result, a new class of technology has emerged.

For more information

on Big Data, go to:

http://www.aee.odu.edu/bigdata

fyi

manufacturing technologies, many new producFORD MOTOR CO.

tion methods will require
corporate.ford.com
fewer people working
HEADQUARTERS: Dearborn, Mich.
on the factory floor. For
FOUNDED: 1903
example, FANUC RobotEMPLOYEES: 175,000
ics, a Japanese producer
REVENUES: $134.3 billion.
of industrial robots, has
automated its newest tool
Ford, one of the traditional Big
Three automakers in the United
plant to the point where it
States, has worldwide operations
can run mostly unattended
that include 65 factories. The
for 700 hours. Many other
company sells more than 5.5
factories use processes
million vehicles a year.
such as laser cutting and
injection molding that
operate without human intervention.
Manufacturing will still need people. This is because
automated machines require someone to service them.
Some machine operators will become machine minders,
which often calls for a broader range of skills.
Throughout history, new tools which make possible
new kinds of measurements and observations have led
to technological and scientific revolutions. Microbiology
was born in 1676 when the microscope discovered the
existence of microorganisms in a drop of water.
The emerging tools being developed to process and
manage the Big Data generated by myriads of sensors and
other devices can lead to the next scientific, technological,
and management revolutions. The revolutions will enable
an interconnected, efficient global industrial ecosystem
that will fundamentally change how products are invented,
manufactured, shipped, and serviced. ME

AHMED K. NOOR is Eminent Scholar and William E. Lobeck Professor

of Modeling, Simulation, and Visualization Engineering at Old Dominion
University in Norfolk, Va.

he Apache Hadoop, for example, is

an open-source platform that can
be used for processing a complex
mixture of data, including log files, pictures,
audio files, communications records, or e-mail,
regardless of native format. Hadoop keeps
all the data online for real-time interactive
querying, analysis, visualization, and business
intelligence. It is now widely used across
industries, including telecom, healthcare
finance, and entertainment, and in government.
GE uses Hadoop software to manage timeseries data, allowing it to scale data across multiple nodes, helping to manage the influx of tiny
pieces of data that come in almost constantly.
Microsoft has developed a suite of tools
for Big Data applications, including Microsoft
Upstream Reference Architecture created to
meet the needs of the oil and gas industry. It
provides a bridge between the needs of the business and the Microsoft technology that supports
IT. It offers statistical and analysis packages for
data mining, discovery, and reporting for diverse
information consumers.
It also integrates decision-making processes
and results to create and retain information
relationships, history, and context.
IBM developed predictive analytics products
for a variety of applications. The IBM platform
covers system management, application
development, and visualization and discovery.
Several automakers use the platform in projects
to develop and deploy connected car solutions
that enable vehicle-to-vehicle, vehicle-to-driver,
and vehicle-to-infrastructure communication
in real time.
Fast, Google-like search of data can be made
by using software systems like Clusterpoint.
BigQuery is a cloud-based interactive query service, developed by Google for massive datasets.
Cisco Industrial Smart Solution is a

PROCESSING HARDWARE

A chip designed by ETH Zurich and the

University of Zurich is comparable in size, speed,
and energy consumption to the human brain.

portfolio of validated IP-networking technologies

that connects factory automation and control
systems to enterprise business systems.
Grok, a cloud-based service company,
formerly known as Numenta, is building a
neuroscience-inspired engine to find patterns
in large data streams and to generate practical
predictions in real time.
A number of attempts are being made to
develop cognitive computers that can effectively
process Big Data, through emulating the human
brain's billions of interconnections, computing
efficiency, size and power usage without being
programmed. These attempts take advantage
of recent focus on understanding the human
brains ability to learn, recognize patterns, adapt,
and respond to changes in the environment.
They include the Defense Advanced Research
Project Agencys Systems of Neuromorphic Adaptive Plastic Scalable Electronics
(SyNAPSE) Project. The goal of the project is
to develop electronic neuromorphic machine
technology that scales to biological levels.
Researchers include teams from IBM, Hughes
Research Labs, and Academic institutions.

ALMOST LIKE THINKING In 2011, IBM demonstrated a building block for a chip architecture
based on a scalable and configurable network simulating brain activity.

3mm

The website, created

as a companion to this
Mechanical Engineering
magazine feature,
contains links to material
on Big Data information,
current activities,
technologies and tools,
educational and research
programs, including Big
Data University.

1024 Axons

learn more

1024 x 256
Synapses

256 Neurons
2mm

SPIKES OUT

SPIKES
TO USB

CPLD
SPIKES IN

AXONS
FROM USB