pmmi.

org

How to Utilize
Big Data to Enhance
Manufacturing Processes

How to Utilize Big Data to Enhance Manufacturing Processes 1

 CONTENTS
Introduction 3

How has big data analytics been utilized within the food 4
and beverage processing industry?

What are the main issues that manufacturers want to solve 6

through the use of big data analytics?

How should big data be collected, handled and processed? 8

What challenges are encountered in implementing big 11

data analytics?

How can outcomes / ROI be measured? 13

What Return on Investments have been achieved? 13

IHS Markit Recommendations for working with big 16

data analytics

© Copyright 2018 PMMI, The Association for Packaging and Processing Technologies, Inc. All rights reserved. The information
contained herein shall not be distributed or shared by the recipient. No part of this document may be reproduced without the
2 How to
express Utilize
written Big DataoftoPMMI.
permission Enhance Manufacturing Processes
Big data analytics along with terms such
as artificial intelligence, machine learning,
digitalization, the cloud etc. are necessitating
manufacturers consider the disruptive potential
on their markets and processes.
The potential impact of these technologies will vary by industry sectors,
which face different challenges, objectives and rates of adoption.
Industries such as automotive, electronics and aerospace have generally
been quicker in adopting technologies, while the food and beverage
processing sector has been somewhat slower.

As part of its coverage of the Industrial Internet of Things (IIoT), IHS

Markit has interviewed users of big data analytics across multiple
industries, including automotive, machinery production, electronics
manufacturing and power generation to understand how the solution
is being implemented to improve its business. These discussions
included a review of necessary technology investments, the challenges
faced in implementation, and returns on investment achieved. Findings
from these in-depth reviews have been consolidated to answer 6 key
questions companies at the start of a big data analytics program should
consider. IHS Markit also provides its recommendations for starting to
work with big data analytics.

This whitepaper starts by introducing some of the early applications of

big data analytics in food and beverage processing. It then consolidates
findings and best practices learned from other industry sectors, and is
used to answer critical questions any potential user of big data analytics
should consider prior to embarking on their own projects, including:

a Where is investment needed?

a How to deal with data?
a What are the challenges faced?
a What are the potential returns on investment (RoI)?

How to Utilize Big Data to Enhance Manufacturing Processes 3

 QUESTION:
How has big data analytics been utilized within the food and
beverage processing industry?

Opportunities for implementing big data analytics in processing are varied and
extend beyond generic applications applicable across a range of industries, such
as predictive maintenance or improved productivity.
The examples provide both applications and also links to more details of particular
companies initiatives. References and links to companies are not endorsements of
their projects (or the IIoT vendor), rather the analyst has included these as examples
to illustrate the potential applications of new technologies, and provide readers with
additional detail to the projects.

Condition monitoring and predictive maintenance –monitoring asset health

through the collection analysis of sensor data such as vibration and temperature can
significantly reduce machine downtime. Rather than a more reactive strategy, the
inclusion of “intelligence” allows companies to identify potential component faults
and rectify before the machine fails. This strategy can also be used to maximize the
efficiency of maintenance teams, allowing them to focus their time on degrading
assets. Tetra Pak a leading packaging machine builder has introduced this solution as
part of some of its machines for the food industry, also enabling it to benefit from new
business models from its customers.

Improve measurement – through improved sensing on a production line to monitor

variables such as pressure, temperature and other metrics of machine performance,
companies are able to better “right-size” products. Improved visibility of the production
process allows machines to adjust to minimize weight tolerances required and ensure
minimum product weights are adhered to. Hershey’s have employed this with the
production of Twizzler products.

Product sorting – use of cameras, vision systems and other inspection equipment
can be used to monitor shape and colour of produce to ensure it meets standards (e.g.
used in fruit and vegetable sorting). For example, TOMRA have used this technology
in selecting between potatoes for producing French fries vs. chips. This optimization of
products also helps reduce food wastage.

4 How to Utilize Big Data to Enhance Manufacturing Processes

Improve food safety – as well as product
sorting, combining machine vision with artificial
intelligence can improve the quality and safety of
food. As done by Kewpie, which utilized these
technologies in its production of baby food, it
can be used to identify and remove sub-par or
defective ingredients, through the identification
of visual anomalies.

Product customization – customer data

can also be collected to support tailoring
products to a customers requirements,
allowing for customization of foods. This has
been implemented by myMuesli, which
allows customers to create their muesli from
a selection of 80 different ingredients – this is
more a benefit of improved connectivity than big
data analytics.

Improve machine visibility – understanding

what is happening during the production
processes can be a challenge, especially on a IIoT vendors are also looking
multi-machine line. Tracking products through at other ways that big data
multiple stages of productions, improved analytics can be used to
sensing and data collection and processing i.e.
through the implementation of Manufacturing support manufacturing. One
Execution Systems (MES), can provide insights example of this is supporting
into bottlenecks and help optimize the process. recipe repeatability. Variations
in ambient temperature and
Creating new recipes – collecting information
on food (including flavor compounds), along with
humidity can impact the
analyzing online recipes has enabled artificial quality of a product. As such,
intelligence solutions to create new recipes and a company manufacturing
combinations that are both novel and (hopefully)
tasty. This has been introduced by Bear Naked
in multiple facilities based
who used IBM Chef Watson. in different meteorological
environments may struggle to
Machine cleaning – a possible application create an identical end product
that has been worked on by a team including even when a standard set of
University of Nottingham is that through
added sensing within the machine, operators ingredients and processes
can have visibility to monitor the amount of food are used. Use of analytics to
debris remaining within the machine and as more accurately monitor the
such optimize the cleaning process accordingly
– reducing cleaning time, energy and water processes can provide better
consumption. visibility to the conditions and
allow for adjustments in the
process accordingly.

How to Utilize Big Data to Enhance Manufacturing Processes 5

 QUESTION:
What are the main issues that manufacturers want to solve
through the use of big data analytics?

Driving factors for smart manufacturing initiatives

In the projects studied, each company was being faced with different challenges that
motivated them to introduce or expand big data analytics solutions. Themes across
these cases included the need for increased efficiency, reduced downtime, quality
management, flexibility and cost savings.
Smart technologies leveraged include:
• Networking and Connectivity
• The cloud and Edge data storage
• Remote monitoring
• Big data analytics algorithms
• Data visualization tools
• New user interfaces and solutions

Factors influencing need for data analytics solutions

Manual machine and equipment condition monitoring
• Increasing reliability demands
• Need for workforce optimization
• Switch from manual to automated data collection

Changing client demand

• Reduced lead times to market
• Increased flexibility of production to support greater customization of production
• Cost control
• New challenges in scaling team’s services to meet client demand

Data integration
• Unconnected sets of data, or platforms that aren’t interoperable increasing the
challenges of collecting data for analytics

• Minimal levels of data integration causing additional work, as well as labor

required to provide proof of outcomes in the planning phase

6 How to Utilize Big Data to Enhance Manufacturing Processes

Primary project objectives
identified REVIEW:
A common pain point identified
A successful IIoT project requires clearly defined objective(s) at was the need to support
the outset. Introduction of new technologies (i.e. a raft of new customers on increased demand
sensors) to support analytics, without a clear idea of what is
trying to be achieved is significantly less likely to succeed.
for product customization. This
is a growing challenge, in a
The examples that follow highlight the problems and pain points
companies identified at the outset of a project, that might be range of industries from CPG to
resolved through the use of IIoT related technologies. These span automotive to the electronics
across the product lifecycle, and often as a project commenced, industry.
the objectives were expanded to include resolution of additional
challenges in production. Managing and monitoring quality
• An on-going problem is reducing lead times to market for was also identified by multiple
customers, while at the same time increasing the flexibility of interviewees as a business
production to support greater customization of production. At
the same time manufacturers must continue to improve quality
challenge. Resolving this has
and reduce costs to remain competitive. multiple applications, not only
• Similarly, the need to reduce the effort required for production in supporting improved quality
planning of product variants as well as meeting increased and reliability of product but also
workloads with higher efficiency were identified. Other other benefits including:
objectives from projects included improving the quality of
individual project reports, which are important for management • Increasing speed of production
decisions, and to facilitate the understanding of production by identifying faults earlier in
process changes.
the process
• Within businesses with expensive and critical assets, such
as the power generation industry, challenges are faced in • Reducing costs through
monitoring the condition of machines manually. In one example, minimizing scrap/wastage
manual data collection is very labor intensive; using route-
(i.e. adding components or
based data collection, predictive maintenance specialists must
physically walk to each collection point to gather hundreds of material to faulty products)
data samples manually, returning to their computers to view and reducing wasted energy
and analyze their collected data. In one example, where almost through ongoing production
60,000 collections a month were taken, analysts were typically
spending 80% of their time collecting the data and only 20% of or processing of products that
their time analyzing it. This lead to both inconsistent diagnosis have become faulty.
and limited risk assessment, as well as a lot of walking!
• In contrast to traditional ways
• Transitioning from a manual to an automated process for data
capture was also identified as an important step in asset health
of monitoring quality, these
monitoring. This supports both the capture of more data, and solutions help automate the
removes human error in recording data, supporting improved entire process (reducing human
datasets from which analytics is conducted.
involvement and therefore also
• In another example, a company faced challenges meeting ever human error), as well as being
increasing customer demands such as flexibility as well as cost
control. Considering one major challenge - agility, it reviewed the able to monitor the quality of
advantage of being able to remotely manage operations. each product at each stage of
• One machine builder was faced with pressures of its rapidly a production process, rather
growing business, and the associated challenge of maintaining than just the first product in a
a quality after sales service to its customers. Here it partnered batch at a specific stage in the
with a leading IT firm and invested in a new predictive analytics
manufacturing solution aimed at better informing customers manufacturing process.
about the condition of their machinery.

How to Utilize Big Data to Enhance Manufacturing Processes 7

 QUESTION:
How should big data be collected, handled and processed?
Collecting, handling and processing data
u Infrastructure investment required
When considering the implementation of big data analytics
solutions; along with identifying preliminary objectives, users must
also consider existing infrastructure to understand the necessary
investments needed to enable a suitable solution.
Existing sensing capabilities, networks, data acquisition computers
and local servers should all be considered. The level of investment
required varies tremendously, some companies have already future-
proofed facilities to support a “digital path” strategy, others have
had to replace or retrofit large parts of the existing infrastructure.
In many cases, projects will require some upgrading of these
technologies, this could include upgrading bandwidth on networks,
installation of additional sensing technology, or investment in
local or remote data storage (local servers or access to 3rd party
infrastructure)
One company found that 75% of their project cost was not in
purchasing new software or sensors etc. as had been originally
expected, but rather in wiring the sensors to the data acquisition
computers.
While the use of data in manufacturing processes is not new, the
potential shift in scale of the type, volume and frequency of data
collection is. This opens the door for new applications but also
challenges in how larger volumes of data are used.

u Data collection
The volume of data collected is increasing as new opportunities are
identified that necessitate sensing placed on new assets as well as
measurement of new parameters (including, vibration, temperature,
pressure, corrosion, flow, level to name a few), requiring new sensors.
Inclusion of sensing on all assets can be extensive and quickly
reach a point of diminishing returns. For example, when considering
predictive maintenance applications, companies have gone through
a process of asset prioritization –identifying the impact of downtime
of certain pieces of equipment on the production processes vs. the
cost of adding sensing and networking and the ability to process
and work with the additional data.
Costs of data collection are not just in the roll-out of new sensors,
companies must also consider how this data is captured whether
manually, or via an automated process. The traditional route based
data collection involved employees walking from asset to asset
and manually recording data points – a costly, time consuming and
inaccurate method. Through improved networking an automated
data collection process can allow personnel to focus time on higher
value data analytics vs. data collection. This process also removes
the human error in data collection, and allows for more frequent
collection of greater volumes.

8 How to Utilize Big Data to Enhance Manufacturing Processes

 VARIETY

u Data aggregation and analytics VELOCITY

Changing how data is collected to support and how
frequently can be a challenge.
Some companies changed from collecting data once
every quarter to every 5 minutes. Creating a larger
volume of data for aggregation and processing, at a VOLUME
faster rate or velocity.
Additional sensors introduced new measurements
parameters increase the variety of data collected,
and to be analyzed.
Where companies didn’t have capability (or desire) to
collect data from all assets, asset prioritization was
used to select machines or products most crucial to VERACITY
the production process.

VISUALIZATION u VALUE

Collecting, handling and processing data

Data aggregation and analytics tool
examples:
New software tools are required to analyze data collected and
transform this into information.
Pattern recognition and prognostic software tools enable
solutions that extend across corrective, preventive,
conditional, predictive and proactive maintenance strategies.
A range of different software packages can be introduced to
identify anomalies in, for example mechanical, thermal, or
chemical data, to help identify potential failures and product
degradation.
While these tools provide a more extensive and easier to
use way of deriving value from data users of these tools
still require “big data analytics” skillsets. These skills of a
Data Scientist can be new to manufacturing organizations
and at the same are often in short supply – this challenge is
discussed later in the whitepaper.

160
Traditional Alarm
140

120
Actual
100
Predicted
80

First Pattern Recognition Alarm

60

40
Avantis PRISM makes it easy to identify anomalies and provides
notification of abnormal conditions before opreational alarms.

How to Utilize Big Data to Enhance Manufacturing Processes 9

 Software tools available
Visualization tools available
A lot of attention is paid to success of
algorithms and analytics tools however, many
end-users of IIoT technologies indicated that
development of the user interface (UI) was
equally challenging:
The challenges faced include:
• Many UI’s of analytics solutions are not
designed for workers without data science
experience
• The UI must be able to convert information
to insights to guide workers to the
actions they need to take (similar to repair
instructions that may be provided by
modern office printers)
• UI’s must be easy to use for IT illiterate staff
• Solutions must be adjustable to different
screen formats
• Designs that can be adjusted to support
regional preferences is also a benefit

The following diagram,

highlights how visualization
solutions support the
overall IoT “stack”of
translating data to insight.

10 How to Utilize Big Data to Enhance Manufacturing Processes

 7 Layers of the Internet of Things (IoT)

BIG DATA

5. DATA ANALYSIS 4. DATA INGESTION

Reporting, Mining, Big Data, Harvest & Storage
Machine Learning of “Thing” data

3. GLOBAL INFRASTRUCTURE
Cloud infrastructure (public,
private, hybrid, managed)
6. APPLICATIONS
Custom Apps built CLOUD
using “Thing” data

BUSINESS 2. CONNECTIVITY/ EDGE COMPUTING

VALUE Communications, Protocols, Networks,
7. PEOPLE & PROCESS M2M, Wifi, Telecom, HW Kits
Transformational decision
making based on “Thing”
Apps and Data

FOG

1. THINGS
Devices, Sensors, Controllers,
etc.

How to Utilize Big Data to Enhance Manufacturing Processes 11

 QUESTION:
What challenges are encountered in implementing big
data analytics?

A summary of problem encountered

Challenge Details Solutions Summary

There is still reluctance among many A team was established to understand and
manufacturers in sharing data with a third trial the technology before implementation.
Cloud usage
party. Many companies spoken with chose
local data storage.

Many have spent careers working in a particular Established “champions” work with teams
Employee way. Getting workers to use and trust in new to and help transition to the new technology.
buy in technologies and the results produced often
proved challenging.

With a large number of Advanced Pattern Strategize how to manage and prioritize alarms
Navigating Recognition (APR) models, the resultant numbers when there aren’t enough analysts to cope with
big data of alarms can be large. How to deal with this, all the models e.g. prioritizing assets based on
while retaining operational effectiveness, the criticality of the asset (i.e. steam turbine vs. fan).
needs consideration.

Creating convergence between IT and OT teams; Cross pollination of teams members, creation of
two traditionally siloed departments with different an intermediate “IoT” teams as well as Senior
IT vs. OT
(and sometime competing) objectives, is necessary Executive sponsorship to ensure sufficient and on-
when working on IoT solutions. going funding.

Digitalization creates challenges when it comes Prioritize and build a holistic solution to monitor
Cybersecurity
to protecting against possible industrial espionage. security levels of production facilities and to quickly
identify security incidents, at the start.

Many users realized they would not be able to These challenges were avoided by partnering with
Going it easily complete this project on their own without IT and other specialists upfront, as well as working
alone the specialized expertise and cooperation of with a supplier with a complete ecosystem of
other partners. partners.

12 How to Utilize Big Data to Enhance Manufacturing Processes

Problems Encountered –
Training People
Arguably the biggest challenge in introducing
an IoT solution is not the technology, but having
workers willing to use the solutions provided.
Many IoT projects have failed, not because of
limitations in the technologies introduced, but
from resistance of employees to fully commit to
the use of new technologies and solutions over
traditional ones.

One element of this is getting employees,

many of whom have spent their entire careers
practicing route-based data collection, both to Problems Encountered –
change to new methods of data collection, and Training People and OT and IT
also trust the data delivered. Even once the
new technology is in place, it is still common
Convergence
for specialists to receive a warning based on However, the benefit of training employees
the data and follow up by manually checking the “on the factory floor” is having operators and
equipment with handheld devices to verify the engineers involved in the process of determining
feedback. what works or doesn’t work in daily operations as
well as suggesting improvements. Not only does
To illustrate the importance of training people
this improve buy-in, but also empowers workers
to use technology, in one example a company
to unearth new opportunities and applications
trialled a predictive maintenance solution at
through the technology that previously weren’t
two of its facilities. While the facility manager
considered.
at one location was keen to trial the technology,
the manager of the other facility manager was It’s also important to have OT and IT departments
resistant. One year later, in reviewing the impact work together from the start of the project. These
on the business it was found that the first site departments, which often have their own, and
had begun to notice significant improvements in sometimes competing, priorities (and budgets).
reducing downtime and planning maintenance, Having senior leadership support of the project is
while at the latter facility the project had failed, important to ensure both continued funding – and
after limited trialling. to ensure conflicts of interest are resolved.
One company established leaders, who worked To support big data analytics in manufacturing,
with the maintenance specialists taking the some companies have partnered with
measurements, helping to improve familiarity Universities, to introduce a “Citizen Data Science
with the technology. program”.This identifies employees to receive
training in basic data science skills, familiar with
Some companies have looked into partnerships
the terminology and technology, and able to
with change management consultants in
support the introduction of solutions alongside
situations where there is a clear change in
their everyday role.
working processes and reticent to the shift to
new technology, in order to support a smoother
transition for their workers.

How to Utilize Big Data to Enhance Manufacturing Processes 13

 Problems Encountered – The Cloud and
Security
There were different attitudes when it came to the use of the
cloud. Some companies had worked through concerns around
security while others would only consider on-site servers. Long-
term companies are expected to both become more comfortable
with the use of a remote cloud, and also to use on-site storage
facilities where low latency response is necessary (i.e. control of
high speed lines).
• One company that adopted a cloud-first mentality, as part of their
digital plan did encounter difficulties when selecting where to
operate its main application (cloud or locally). The problem arose
when the main application, which was generally running on the
cloud, sometimes needed to be run locally. In order to run on a
mix of technologies, the user worked with a company offered
hybrid capabilities to support both local and on-cloud computing.
• For this company, the use of a cloud for storing and processing
data did concern some service customers. As a result, it offered
multiple options for storing and analyzing data as part of a hybrid
structure including a public cloud, a virtual private cloud, and on-
premises options as specified by the customer.
• To support secure transmission of data to the cloud, it can be
necessary that applications use a single sign-on process
(SSO), which allows it to have a strong security line back to the
company. Also, it can encrypt data at the data-packet level before
it is transmitted.
• Similarly solutions are available based on a holistic security
concept. With a monitoring system, industrial security specialists
can identify security-relevant events such as cyber-attacks on
the automation network and IT systems, and initiate effective
countermeasures. This approach makes it possible to identify
suspicious events and risks as well as typical attack patterns and
issue corresponding alarms.

Problems encountered – dealing with big data

Another challenge is handling the significantly larger volumes and
frequency of data collected. One company indicated that it had
To quote one project leader: switched from collecting data from assets once every three months,
to every five seconds, meaning huge increases in data volume.
With the large number of advanced pattern recognition (APR) models
running, in one example over 10,000, this can result in a high volume
of alarms received. To counter this, companies can set a strategy to
You need to make manage and prioritize alarms, especially in situations where there
solutions so simple that aren’t enough analysts to cope with all the models. One method is
asset prioritization based on the criticality of the asset (for example a
they want it and seek it steam turbine is rated as higher priority than a fan).
As well as analyzing the data, another challenge faced is developing
out because it’s so simple new user interfaces that are different from existing solutions in that
they needed to support data streamed back and forth from a cloud
to understand and read. (whether public, private or local). There is also the need to help users
better understand not just what is happening in the manufacturing
process, but also the next actions required.

14 How to Utilize Big Data to Enhance Manufacturing Processes

QUESTION:
How can outcomes / ROI be measured?
How to measure outcomes
Correctly measuring the outcomes of a big data analysis project is critical to identify whether the
project has been a success, whether it justifies further investment, as well as identifying possible
follow-on projects that might benefit from the solution.
Starting points for this include specifically defining the objective of the project and subsequently
measuring the performance or level of operation of the manufacturing process prior to introducing
big data analytics. Having these parameters then provides a benchmark against which a project’s
success (or failure) can be measured.

The point of measurement varies by application, examples include:

Quality

number of defects per million (the

number of faulty parts product
per million)

Productivity
the number of products produced in
a set time, or number of variants of a
product that can be produced Unplanned
downtime – the measure of the time
that an asset sits idle, outside of planned
maintenance projects

Efficiency

the minimization of use of resources i.e.

manufacturing materials, water, energy
etc.

Flexibility

the ability to expand the number of

SKU’s produced within a defined
timeframe

How to Utilize Big Data to Enhance Manufacturing Processes 15

 QUESTION:
What Return on Investments have been achieved?
Some general examples of RoIs achieved across the
product lifecycle - from other projects
DESIGN OPERATIONS MAINTENANCE SUPPLY CHAIN

Harley-Davidson Petroflow Energy Corp. CNH Industrial Sandvik Coromant

Reduced model New Site power consumption Average maintenance Monitors existence of any
Product Introduction (NPI) reduced by 43% time reduced by 50% bottlenecks in the overall
from more than a year to supply chain
1.5 weeks
Marathon Petroleum Fanuc
Company Saved US$2M from Shanghai CHILO
Maserati Alarm rate reduced by reduced downtime Press Company
Reduced number of approx. 90% Inventory error rate
prototypes, and time to Nova Chemicals reduced from 50% to 4%
Husky Injection Reactive emergency work within 6 months; inventory
market resulting in 30%
less development time
Moulding reduced by 47% and control converted US$95k
Productivity and cycle time time spent on proactive, loss to a US$158k profit
gained from 3% to 12% preventative maintenance
has increased by 61%

Detailed RoIs achieved by company reviewed

u Jabil
• Ability to monitor each product through each stage of the process; and based on the algorithms, it
can predict the likelihood of success.
• Eliminated the need for human involvement in conducting “first article inspection” at every step in
the process.
• Ability to predict and prevent failures of over half of all circuit boards at step two of the 32-step
process, with the remainder identified at step six – this was done with an 80% accuracy rate. This
early identification of faults allowed corrections to be made before additional, often expensive,
electronic components were added.
• Savings in excess of 15% in scrap and rework costs, as well as an additional 10% reduction in
energy costs.
Other expected benefits:
• Reduced down time waiting for the product to be tested before moving to the next step in the
manufacturing process.
• Improved overall equipment effectiveness in reducing the time that expensive assets sit idle
waiting for inspection to be completed. Optimized operation of individual machines in an
increasingly intelligent feedback loop.
• Lead time reduction, through both improved predictive analytics and efficient first article
inspection, enabling Jabil to meet the customers’ requirement for increased flexibility and faster
delivery of product.
• Support in forward scheduling of maintenance.
• Reduced warranty costs as a result of higher quality, reliable products, as well as greater
visibility of projected product life.
• Inventory level reductions as parts can be manufactured and delivered more quickly.
• Access to a new customer base with a more flexible manufacturing process. Ability to support
and service clients with lower volume requirements.

16 How to Utilize Big Data to Enhance Manufacturing Processes

 Specific outcomes
u Duke Energy
• Transition from collecting four readings from a data point per year, to
now collecting readings every five seconds. Operator rounds greatly
reduced while dramatically increasing the frequency of collection. This
allowed a shift from analysts spending 80% of time collecting data with
only 20% of time doing the higher value task of analyzing the data, to
analysts spending 80% of their time on data analysis tasks.
• Over 4 years, Duke Energy has avoided costs of 130% of their capital
budget spent with ability to avoid higher costs associated with failures.
• Improved reliability and lower operating costs resulting in increased
reliability and an optimized workforce that is
more analytical.
u Siemens
• The implementation of a fully networked production line allowed
Siemens to make large efficiency and quality improvements.
• The enabled a custom, built-to-order process involving more than 1.6
billion components for over 50,000 annual product variations, for which
Siemens sources about 10,000 materials from 250 suppliers to make
the plant’s 950 different products.
• Quicker delivery with 24 hour lead time
• Reduced time to market by up to 50%
• Cost savings of up to 25%
• Improved reliability and quality with 99.9% accuracy in product quality
• 100% traceability from component suppliers
• Equipment utilization rates average raised to 80% - 95%
Sandvik
u • Ability to guide its customers to optimize tool adjustments and
maintain a continuous workflow. As a result of alerting customers
when a tool should be changed and allowing optimum time for the
action to take place as well as helping to avoid unscheduled shutdowns
by setting alarms to take machines offline before failures occurred,
Sandvik Coromant was able to minimize idle time to 50%. This has
helped them to increase efficiency and generate millions of dollars in
savings for its customers.
• These technologies have also helped Sandvik Coromant save time and
improve quality for its clients in the design and planning phase with
Note: As with many machine tool and cutting data recommendations integrated directly into
its clients CAD/CAM environment.
companies exploring
Benteler
digital technologies, • Gained an end-to-end system that allows it to manage lines in a plant
u and move products between plants remotely. A seamless integration
the cases examined between the hardware components and the software applications
were also discovering was accomplished by leveraging a network of partners, each with a
particular expertise.
ancillary benefits • Achieved better production outcomes and lower costs in factories
where everything is connected to the network and each step in the
from their respective production process can be analyzed and controlled from the cloud.

projects.

How to Utilize Big Data to Enhance Manufacturing Processes 17

 RECOMMENDATIONS:
for starting with Big Data analytics

The 6 S’ of a successful IoT implementation

Based on these case studies and other discussions with companies that are
introducing big data analytics projects, below is a list of recommendations to
consider before embarking / accelerating big data analytics for its production
test process.
• Specify - What problem / pain point needs to be solved?
• Success - Can you define the success of a project?
• Start Small - Proof of concept projects on limited number of lines or assets,
then look to scale.
• Senior support - Need senior sponsorship, with a long-term vision of how the
technologies will be used in the business
• Shared responsibility – are both the IT and OT teams involved
• Support staff - Get your people involved in the project (they can support in
new applications)

u Review current capabilities and solutions

• The need for investment varied significantly between case studies. Before starting, a
company should understand the current state of its facilities.
• Is there a need for the introduction of additional sensors (and the type of sensors needed)?
• How is the facility connected and does this support quick collection of large amounts
of data?
• Is hard wired or wireless the best medium for collecting data?
• Is a cloud platform currently in place – and does it have the capability to support elements
of the project, such as data aggregation, analytics and visualization of data?
• What levels of cybersecurity are in place (for hardware, software and employee training)?
• What solutions are in place to enable information to be fed back to the user (i.e. having a
mobile devices policy, suitable and sufficient operator terminals etc.)?

18 How to Utilize Big Data to Enhance Manufacturing Processes

u People
• Providing training courses on big data for employees helps to accelerate awareness and
compliance of the technology down to the factory floor. This also improves the evolution
of the technology as workers can better contribute suggestions on where else solutions
can be implemented and how they can be improved.
• It’s important to have long-term senior management involved to promote adoption of
technology across all facilities. In some cases, projects have failed because factory
managers have not been convinced by the benefits and so have ensured compliance
and best practice at their factories. Having strong senior leadership involvement
supports adoption of best practices.
• If the organization has built big data analytics into part of its road map, this also avoids
the need for the project lead to apply annually for additional funding for continuation of
the project.
• Addressing the IT OT divide – there are very different attitudes to issues such as sharing
data and how quickly projects should be rolled out when contrasting the IT and OT
departments. Having the team responsible for the big data project including members
from both groups will help buy-in and also support identification of potential problems
and concerns early in the planning phase.

u Dealing with data

• Transitioning from a manual to an automated process for data capture was identified
as an important step in condition monitoring. This supports the capture of more data,
and removes human error in recording data, supporting improved datasets from which
analytics is conducted.
• A target for the frequency of data collection (for condition monitoring and predictive
analytics), is that for the average time between the production of each unit six sets of
data should be captured – to ensure a high confidence in the data used for the analytics.
So if a product is produced every 5 seconds, 30 data readings are taken).
• Tag data collected to the outcome of production – this produces labelled data which is
easier to analyse. This helps to more quickly identify data related to a faulty production run.
• Ensure readings are of a good quality (bad readings in = poor results out). This includes
correctly mounting sensors at the correct points on a machine, so that data collected is
consistent and accurate.
• Work with the cloud to ensure scalability of projects as you move from PoC (proof of
concept) to mainstream applications (and so as the size of data increases).

u Suggestions on partnerships with vendors

• Work with open and not proprietary software across different machine types. Specify
that suppliers must support open software / solutions that can be used on an open
platform. This is true both for hardware and software. Doing this reduces the need
for interfaces, and reduces the loss of quality of data. This is particularly important as
projects scale up and evolve to support new applications.
• Work with more than one IoT platform, trial platforms from different vendors on different
projects (in one example a platform for a quality management solution was from
Microsoft, while a predictive maintenance solution used IBM, at the same site).
• Select a supplier that is able to coordinate with an ecosystem of partners. No one
company can offer all the necessary solutions for an IoT project, so those companies
that have partners in place to address areas including cybersecurity, connectivity, cloud
platform, data aggregation and analytics tools, apps development, visualization tools,
systems integration and even change management.

How to Utilize Big Data to Enhance Manufacturing Processes 19

 pmmi.org
© Copyright 2018 PMMI, The Association for Packaging and Processing Technologies, Inc. All rights reserved. The information
contained herein shall not be distributed or shared by the recipient. No part of this document may be reproduced without the
20 How to
express Utilize
written Big DataoftoPMMI.
permission Enhance Manufacturing Processes