MAINTENANCE

FALL ED I TI ON
Contents

3 — How predictive maintenance impacts preventive

maintenance

11 — Digital Difference -- Video Rebrand 2021

12 — Getting started with AI-based predictive maintenance

26 — Preventive maintenance software: Doing preventive

maintenance the right way

34 — Natural Gas Compressor Application Optimizer

38 — How to avoid COVID-19-related schedule delays

using Lean

44 — Six sigma principles could improve additive manufacturing

quality control

2
How predictive maintenance
impacts preventive maintenance  Back to TOC

You can’t shut down every time there’s an alert

T raditional preventive maintenance (PM) prescribes routine-based asset or system

maintenance, regardless of whether performance has degraded or not. By estab-
lishing predictive maintenance (PdM), manufacturers can better define the optimum
window of time when maintenance should be performed, based on the predicted fu-
ture health condition generated from equipment data.

IIoT-based predictive maintenance feeds data into plant MES and ERP, to improve
business performance.

MES manages manufacturing processes. Data derived from sensors and scanners may
require manual data entry by operators and maintenance personnel. ERP primarily
focuses on financial transactions tied to production, enterprise and supply-chain per-
formance. Many ERP systems interface smoothly with MES, as opposed to competing
or overlapping.

MES is essential to a plant predictive maintenance and scheduling. MES tend to be

industry specific; an aerospace industry MES has different concerns than one for the
beverage industries.

Combining predictive maintenance with MES ensures:

• An integrated system that allows correlating OEE and machine health

3
How predictive maintenance impacts preventive maintenance

• Improved maintenance scheduling leading to better production scheduling and  Back to TOC
reduced unplanned downtime

• Integrated maintenance management, including predictive work orders.

A unified machine dashboard for health, maintenance and other key metrics, derives
and distributes an optimal maintenance schedule.

Examples and case studies of predictive maintenance

From our experience developing predictive maintenance solutions, a 5% reduction in
unplanned downtime can deliver $500,000 in annual savings for an automotive industry
manufacturer customer and a 5-10% improvement in overall equipment effectiveness
(OEE) saved an aerospace manufacturer $300,000 per year.

For a manufacturer looking to implement predictive maintenance, CNC machine tools

are a great place to start.

The first step is to monitor front and rear spindle bearings. Vibration data can be mea-
sured by adding an accelerometer, typically installed on the spindle housing. Other
data, such as spindle rotation speed and load or program number, are captured from
the machine controller through protocols such as OPC-UA or MTConnect.

Controller data provides useful contextual information for understanding vibration

data and leads to development of a corresponding predictive algorithm. Machine con-
troller data alone is sufficient enough to build the model.

4
How predictive maintenance impacts preventive maintenance

 Back to TOC

Important model parameters include position, speed Maintenance and scheduling

and load of the actual axes, as well as the correspond- dashboard. It makes sense to
combine production constraints
ing commanded values for those parameters. In ad-
and maintenance constraints in a
dition, we focus on detecting pre-load loss and poor single execution schedule. Source:
lubrication of linear axes or ball screws. Predictronics

To enhance predictive maintenance with predictive scheduling, additional data from

MES and ERP is needed. This data can be categorized as either dynamic or static.

Dynamic data, such as the maintenance and production schedule is gathered from
the MES, including start time and end time for each activity and the production and 5
How predictive maintenance impacts preventive maintenance

operation IDs. Meanwhile, static information is collected from ERP, including spare  Back to TOC
part and shipping costs, supplier-provided lead times when ordering parts, as well
as repair and downtime-associated costs. These are needed to determine an optimal
maintenance schedule that takes heed of predictive maintenance alerts of an im-
pending failure.

Enhancing predictive maintenance with predictive scheduling results in a unified dash-

board showcasing machine health plotted against scheduled production orders. It also
displays actionable information, such as when to order spare parts and when mainte-
nance is required.

Through continuous machine monitoring, orders change from green to yellow as the
urgency to replace or repair critical parts builds. With an integrated predictive mainte-
nance and scheduling, maintenance not only uses machine health information to pre-
dict failure, but the team can now consider the risks of completing scheduled produc-
tion orders versus accomplishing maintenance tasks.

Transactional maintenance
Predictive maintenance and monitoring can make an existing computer maintenance
management system (CMMS) more effective, on a per asset basis or across the en-
terprise. Machine parameters and sensor data pulled for each machine’s dashboard
and the master health, predictions, and diagnosis reports can be pulled into a master
CMMS dashboard.

If machine warning thresholds are close to being exceeded, the predictive mainte-
nance system can automatically generate work request forms documenting asset,
6
How predictive maintenance impacts preventive maintenance

logged-in operator making the request, diagnostic needing attention, asset status  Back to TOC
(working and needing inspection or not working and needing repair), as well as detec-
tion and work-request dates and times.

The potential of data-based predictive maintenance saving hundreds of thousands of

dollars per year may inspire some to think big, but many times it is more efficient to
start small, prove value and scale fast. An effective strategy covers:

• Assessment: Identify a project, assemble a team and assess the digital maturity of
existing process and participants.

• Documentation: Confirm the existing data-collection strategy, document the busi-

ness case and calculate return on investment.

• Analysis: Ascertain the ideal end state for the project, its people and technology.

• Statement of work: Choose the components (hardware, software, sensors and

software) needed for implementation and estimate the time required.

• Implementation: Build the templates, conduct the project and compile achieve-
ments.

• Scale: Identify lessons learned and build a center of excellence around the
project to identify opportunities to repeat the success in other departments or
plants.

7
How predictive maintenance impacts preventive maintenance

 Back to TOC

Proving value by thinking big, starting small and The ISA-95 automation pyramid is a
classification of software systems particular
scaling fast moves a plant’s productivity, mainte-
to manufacturing. Source: Predictronics
nance and scheduling from a hope-for-the-best
approach to a can’t-fail approach because data is
flowing, and obstacles are identified before they become problems. You are now work-
ing in a smart factory.

A look at the predictive landscape

Predictive maintenance can serve as the eyes, ears, tactile sense and collective experi-
ence of the operator, shift supervisor and maintenance department.

The maintenance staff traditionally intuited when something didn’t look, sound or
smell right. When something didn’t “seem” right, sure enough, many times it wasn’t. In
other cases, fail-and-fix was the model. The results were unexpected downtime, upset
customers, damage control and run-ups in costs.
8
 How predictive maintenance impacts preventive maintenance

Think of the Industrial Internet of  Back to TOC

Compression
Things (IIoT) as the convergence

Optimization
of information technology (IT) and
operational technology (OT). At

is Hard,
the bottom of the pyramid are the
shop-floor devices (CNC machine
tools, welding robots, conveyors

Our Experts and more) able to convey data to

Can Help!
the cloud. The next layer up is a
supervisory control and data ac-
quisition (SCADA) system for ob-
taining data from sensors or ma-
Quad O has increased scrutiny on compressor
methane leaks. Identify flow measurement chine controllers. To engage the
solutions to combat fugitive emissions as part
of your LDAR program and optimize your gas business, the data may be fed into
compressor. a manufacturing execution system
Discover Solutions to:
(MES) and the results shared with
• Emissions issues due to packing
vent leakage the enterprise resource planning
• High fuel consumption, allocation issues,
inaccurate fuel consumption tracking
(ERP) system to improve data cap-
• Inaccurate emissions data during a ture and applied analytics.
blowdown event
• Environmental fines due to methane
emissions (due to misreporting) Download Your Guide Now!
Innovations in condition
monitoring
Help ensure your compressor’s mechanical seals are healthy True predictive maintenance
by adding a flow measurement to your packing vent line.
based on machine data and
Make Dowtime a analytics can lead to near-zero
Thing of the Past foxthermal.com
downtime, reduced spare parts
9
How predictive maintenance impacts preventive maintenance

inventory and more effective maintenance through more accurate mean time be-  Back to TOC
tween failures (MTBF) and mean time to repair (MTTR).

Continuously monitoring manufacturing assets for improved predictive maintenance

leads to greater understanding of machine health, performance and failsafe produc-
tion scheduling. A successfully implemented multiple-template approach enables
an organization to build predictive models with ease, as well as incorporate machine
learning and industrial AI to rapidly deploy them. Such templates involve:

• Develop models by collecting data from machine tools, robots, pumps, conveyor
motors and other systems to develop data models

• Identify features for use

• Select and integrate machine learning and AI methods

• Establish repeatable analysis flow charts

• Deploy solutions rapidly and monitoring systems continuously for improved pre-
dictive maintenance and production scheduling.

Mo Abuali and Edzel Lapira

Mo Abuali, Ph.D. is managing partner, IoTco, Cincinnati, OH. Edzel Lapira, Ph.D. is
president & CEO, Predictronics Cincinnati, OH.

10
Digital Difference -- Video Rebrand 2021

 Back to TOC


Digital Difference -- Video Rebrand 2021
Now that you know that thermal mass flow measurement is right
for your application, explore the ways that Fox Thermal has the
advantage over competitors --- the Digital Difference!

11
Getting started with AI-based
predictive maintenance  Back to TOC

Artificial intelligence (AI), and particularly machine learning (ML), provide

effective tools for implementing predictive maintenance.

P redictive maintenance is not a new concept. The technology for implementing

predictive maintenance has been available for years. The benefits are well-
known, particularly given the high cost of downtime in most industries. Despite these
points, adoption has lagged, particularly in the end-user community. It hasn’t been
simply a question of capital outlay – data capture and logging/transfer capabilities
are already built into many components. For example, drives with current and tem-
perature monitoring capabilities or human-machine interfaces (HMIs) that double as
data loggers.

Security concerns aside, the sticking point often has been lack of skills for converting
data into actionable information and the cost and time associated with either bring-
ing that in-house or dealing with third-party services. Artificial intelligence (AI), and
particularly machine learning (ML), provide effective tools for implementing predictive
maintenance and saving big. Indeed, according to McKinsey & Company, AI-based
predictive maintenance can boost availability by up to 20% while reducing inspection
costs by 25% and annual maintenance fees by up to 10%.1

One of the primary challenges of predictive maintenance is combing through massive

volumes of data to extract only meaningful, actionable information. Particularly given
the rapid growth in the industrial Internet of things (IIoT), organizations can find them-
selves data rich, yet information poor. ML is an organized methodology for extracting
12
Getting started with AI-based predictive maintenance

insights that can be used to detect developing defects before they become major  Back to TOC
problems, determine the remaining usable life (RUL) of even troubled assets, allow re-
pairs to be scheduled during minimally disruptive windows, and conduct a root-cause
analysis to prevent similar failures in future.

The basics of AI
Certain types of predictive maintenance modeling can be easily addressed with fair-
ly simple, even manual, computations. The true value of ML is its ability to take into
account large volumes and diverse types of data in the context of complex machine
dynamics and real-world operations to arrive at a greater understanding of asset op-
eration and health.

ML is part of a class of applications known as narrow AI. This refers to functions that
are written and trained to perform specific tasks. That chatbot you just interfaced with
during your online banking session, for example, was probably a narrow AI applica-
tion set up to present a specific response to a specified set of inputs (and to elevate
to a human being in the case of other requests). In the same way, ML can run sensor
data through a statistical model to detect conditions defined as corresponding to a
developing defect. “Machine learning is not intelligence in the true sense,” said Scott
Genzer, data scientist at RapidMiner (Boston, Massachusetts). “It’s really nothing more
than the good old fashioned mathematical modeling we’ve been doing for decades.
The difference is that we have the computing power to mung massive amounts of
data to find patterns, to find signal in a lot of noise that we used to do by hand.”

ML solutions are already in broad deployment for use cases like fraud alerts and pre-
dictive maintenance. In contrast, general AI, which encompasses the types of sen-
13
Getting started with AI-based predictive maintenance

tient machines that  Back to TOC

are the mainstays
of pop culture, is
enormously com-
plex and will most
likely remain a lab-
oratory curiosity for
some time to come.

In ML, one or more

algorithms operate
on a set of training data intended to describe fac- In supervised machine learning, algorithms
operate on a manually labeled training data
tors like asset condition and performance, failures,
set to create a model. This model can be
maintenance processes, environment, records of used with production data to return results
failure or maintenance, etc. Using this data, the or predictions. Courtesy: A3

algorithm creates a mathematical model that de-

scribes the complex system and its interactions. The data is organized such that there
is a dependent “target” variable to be predicted, in this case to describe asset health,
remaining lifetime, etc. The goal is that when fresh data is put into the model, the
model will return a status, prediction, etc. (Figure 1)

Machine learning can be divided into supervised learning and unsupervised learn-
ing. In supervised ML, a function (model) is trained to act on new input in a defined
manner using large amounts of manually categorized data. Supervised learning for
predictive maintenance is commonly addressed using classification or regression. In
classification, discrete input maps to discrete output; with enough of the right kind of
14
Getting started with AI-based predictive maintenance

data, the model can classify an asset as healthy or not healthy, for example, or a prod-  Back to TOC
uct as acceptable or not acceptable (Figure 2). Output is not always binary; the model
may return a range of possible outcomes. Regressions take quasi-continuous input,
like time series temperature or vibration data, and returns a continuous output value
in the form of a trend that can be used to predict future function. While classification
might be used to determine whether an asset has a defect that can lead to unsched-
uled downtime, regression would draw on historic behavior plus current data to pre-
dict the remaining useful lifetime of the asset and estimated time to failure.

In unsupervised ML, the algorithms work with unlabeled data to seek out patterns
through clustering (what information belongs together?) and correlations (what events
happen together?) Consider a bottling line. A supervised machine-learning classifier
might detect an elevated temperature on a motor and based on the model, send an
alert to maintenance to investigate it as a potential developing defect. An unsuper-
vised ML model might discover that the temperature always rises on this motor when
the machine is packaging a more viscous liquid during August, so maybe there is no
developing defect after all. Unsupervised ML can uncover unexpected patterns that
lead to valuable insights. For example, maybe the temperature doesn’t rise as much
on days when Joe is running the machine. Now, the company not only avoids unnec-
essary replacement of a healthy asset but can review Joe’s techniques to potentially
discover a way to improve machine operations across shifts, production lines, and
even facilities.

Both approaches can be useful for predictive maintenance. Sometimes one can be
used to inform the other, such as when correlations uncovered by an unsupervised
learning model are used to update the supervised learning model.
15
Getting started with AI-based predictive maintenance

Begin with the  Back to TOC

business case
One of the biggest
mistakes with ML
in predictive main-
tenance is diving
directly into data
gathering and model
building. For suc-
cess, projects need
to start with the
Supervised machine learning is generally
business understanding. What is the pain point used for two tasks: classification (left) and
in terms of costs and business impact? How is regression (right). In classification problems,
it addressed today and how would ML improve the algorithms take input data to produce
discrete output data (e.g., healthy asset or
on that? What are the objectives, e.g., decrease not healthy asset?) In regression problems,
unscheduled downtime, optimize product quality, the model takes input data and produces a
continuous output value that can be used,
boost throughput, etc.? Clearly articulating the
for example, for predictions (how soon will
goal is the first step to achieving it. “AI is a tool, this asset fail?) Courtesy: A3
not an outcome,” said Saar Yoskovitz, CEO of Au-
gury (Haifa, Israel). “Focus on the use cases that you want to address and decide what
is the best technology stack you need to use to do that.”

Be sure to quantify objectives–it’s important to not only define success but to also
understand the cost basis to evaluate return on investment (ROI). Finally, don’t forget
the human factors. Starting off with simple projects likely to provide rapid benefit will
streamline approvals. “The computer side is not difficult,” Genzer said. “The hard part
16
Getting started with AI-based predictive maintenance

is getting the engineers to buy in, getting a manager to say yes, and getting it on the  Back to TOC
plant floor.”

Understand and investigate the data

Once the business case is established, the next step is to gather and investigate the
available data. Can the current data be used to answer the questions identified during
the business analysis? Is there sufficient detail and context? Here, OT needs to work
with IT and the data scientists to clarify how data maps to physical phenomena and
make sure that input is coming from across the breadth of the organization. For ex-
ample, motion control is essential to product quality in additive manufacturing. As a
result, ripple and discontinuities showing up in a run can indicate systematic motion
problems that can be tracked back to problems with components, even if sensors
don’t directly reveal the concern.

Because of the data crunching capabilities of ML algorithms and models, the data
can–and should–come from a wide variety of sources, so long as they are relevant.
”Machine learning works best when you have a lot of data,” said Genzer. “That is ab-
solutely the golden rule. Storage is cheap. Start with the big data lake and cull it from
there.”

“Data isn’t free, but it is much, much cheaper than down time,” said Paul Ardis, tech-
nology manager for machine learning at GE Research (Niskayuna, New York). “So,
if we consider the trade-offs and the decision-making process, it is often worth the
time, trouble, and cost to set up a design that polls as much data as possible; even
if we don’t necessarily know what it is going to be capturing from a failure-modes
perspective.”
17
Getting started with AI-based predictive maintenance

One of the challenges of predictive maintenance in discrete automation is that indus-  Back to TOC
trial equipment is built to last. That means even older machines have a limited failure
history, while new designs, of course, will have little or none. That’s good for opera-
tions but can be problematic when the goal is to capture large amounts of data on
machine behavior around degradation and failure. Running equipment to failure just
to gather data isn’t practical. Fortunately, there are alternatives.

• Historic records: Digitizing and formatting historic records on machine condi-

tion and maintenance can be time and labor intensive but is absolutely essential.
Launching an ML project from scratch squanders all of the expertise and insights
that the organization has accumulated over time. Moving legacy records into the
digital domain should be a major priority from the beginning of the project.

• Development data: For OEMs wanting to field ML-based predictive maintenance

solutions, either as a feature for customers or a way to maintain their equipment
“fleets,” data may be available from initial designs. Prototype testing can serve as
effective run-to-failure, or at least run-to-degradation, exercises.

• Surrogate modeling: Development data can be useful but limited, particularly in

its ability to generalize conclusions to other assets. The solution is surrogate mod-
eling, a data-driven process for streamlining computer simulation. Conventional
computer simulations are expensive and time-consuming because they require
extensive training to closely approximate real-world performance. In surrogate
modeling, the process works back from simulation results from a baseline model
to generate the “surrogate” model, which is then trained by running a full compu-
tation for only a limited set of conditions rather than comprehensively. “The intent
18
Getting started with AI-based predictive maintenance

is to come up with an effective model that can be developed at reasonable speed  Back to TOC
for whatever processing is needed, Ardis said.

• Transfer learning: Transfer learning builds a training set for new assets by finding
a way to modify or map data from similar assets, even like assets within a particu-
lar manufacturing batch. It still requires a small amount of data to understand the
actual transfer mechanism but users no longer need to train a new model from
scratch. “We focus in particular on transductive transfer learning where we have
some information about the difference in task,” Ardis said. “So we’re not just
trying to match distributions but very explicitly looking for the same relationships,
the same sorts of connections from a physics-based perspective that we should
expect from the original design.”

“We don’t need to build and train a model for one specific pump because we’ve
seen over 20,000 pumps before,” Yoskovitz said. “We know what cavitation looks like,
what bearing wear looks like. We’ve already built these models using a generic phys-
ics-based approach.” Of course, building a standard diagnostic model for a fixed-
speed rotating asset is quite different than for a custom machine with complex dy-
namics. Here, mapping from asset to asset is more complex. “For anomaly detection,
we build a baseline for a specific machine because the different recipes change the
behavior, the environment from one site to another changes the behavior. We build a
model for baseline machine behavior, conceptualized into operations, as well. Then
we can detect anomalies if anything goes wrong.”

Timing matters with data monitoring

Make sure that data acquisition and processing is fast enough to keep up with the
19
Getting started with AI-based predictive maintenance

industrial process and assets being monitored. This requires evaluating the data and  Back to TOC
its relationship to the physical entities.

• Define the forecast window: This refers to the time elapsed between the indica-
tor and failure. In other words, how much lead time is required to not only prevent
catastrophic failure but to enable the least disruptive, least expensive repair? A
small gearbox with a replacement on-site may only need minutes of notice for a
repair. Replacing a custom kilowatt-class motor located on the roof of a factory
might require days or even weeks of lead time both to order a new motor and to
rent a crane to move it. What data and what ML model will deliver this?

• Define the target window: Even before failure, machine performance will begin
to degrade. The model should account for this. The goal is not just to continue
making product or delivering services, the goal is to make salable product and
meet service level agreements.

• Define the feature window: If we are tracking a specific type of data logging an
average or an FFT, for example, what is the window covered by that data set?

If a predictive-maintenance solution can’t provide results quickly enough for the user
to take action in a timely fashion, then there is not a good business case to pursue it.
”A model is only as good as the data available to it at any point in time,” Ardis said .
“The ability to update the model to reflect the latest sense of continuing operation –
is it normal or trending abnormal? – is highly limited by the update rate of the data.
So it is not so much a question of how much data we are able to store as how quickly
we are refreshing the data stream to make sure that we are able to take action.” Con-
20
Getting started with AI-based predictive maintenance

versely, if the timeframe of the activity is slow, an ML-based solution will probably be  Back to TOC
overengineered (and overly expensive) for the need.

Data preparation
As with most types of computer modeling, data needs to be prepared before it can
be applied to ML. This goes beyond simply cleansing and formatting. Given the sheer
volumes involved, the raw data needs to be broken out into subsets of data that sat-
isfy a particular business question or manipulated to deliver new insights. This takes
place using a collection of techniques known as feature engineering, essential to the
success of any ML project.

“It’s actually more important than the modeling,” Genzer said. “You’re trying to find
out which sensors and which parts are the most meaningful.”

Feature engineering consists of feature selection and feature generation:

• Feature selection is the process of narrowing down the field by finding out which
columns in the database are the most relevant, which ones are more correlated to
the target variable. An installation might generate a thousand or more columns of
data but only few of those columns are necessary to identify developing defects.

• Feature generation is a process of combining columns to create more useful attri-

butes, for example by adding a pair of columns, multiplying them, etc.

Between feature selection and feature generation, it’s possible to create an optimized
set of data columns that will have the best possible ability to find meaningful models.
21
Getting started with AI-based predictive maintenance

Feature engineering is an essential step but should be approached with care. “The  Back to TOC
danger that engineers almost instinctively get into is they think they know which col-
umns are going to drive the predictive class,” Genzer said. “This biases the model.
You really want the computer to find what it thinks the columns are and have an open
mind. You need to apply a little bit of judgment because you don’t want it to find
things that are ridiculous, but we don’t want to limit the computer’s ability to find sig-
nal that you may not know exists.”

Data modeling and machine learning

The whole point of ML is to use data and algorithms to develop a model that de-
scribes the physical system in operation and can be used on new data to deliver ac-
tionable insights in an ongoing fashion. A detailed discussion of ML algorithms and
models for predictive maintenance is outside the scope of this article. Instead, we
can focus on some big picture points. Let’s start with one of the most common user
mistakes, which is to approach a project with a preconceived notion of which model
to use before the data is ever captured. This gets the process backward. The key is to
start from the data, from the business problem, and then find which algorithms best
satisfy the criteria.

Another point to bear in mind is that building a model is not a case of one and done.
Often, projects result in multiple models which then need to be evaluated to deter-
mine which model or models most effectively characterize the system.

“A lot of what we do in terms of machine learning in AI is explore a vast array of po-

tential models simultaneously,” Ardis said. “There isn’t a simple answer to, ‘What are
the models?’ because the answer is, “Everything that we can get our hands on.’ In
addition, we look at dynamism of the ability to handle them together.”
22
Getting started with AI-based predictive maintenance

It’s not necessarily effective to train five models, for example, determine some rela-  Back to TOC
tive effective weighting, then set that up as a constant model that will be used for the
foreseeable future. A more dynamic approach might provide a better solution.
Ardis said, “Can we have a setup that is actually selecting dynamically which models
to include, which to exclude, and how to rebalance and utilize them best based upon
their performance in terms of providing information over time that tracks with the real
ground truth?”

Part of the challenge lies in the fact that machines are constantly changing over time;
“normal” operation will look different today than it will next year. The training data
needs to be broad enough to show some of that progression. Model validation also
needs to test whether the model is resilient enough or whether it is overfitting to the
initial data set. “The challenge is in getting someone to understand that sometimes
it is better to have a model whose training performance is 10% lower but validates to
understand the whole general space much more effectively,” Ardis said.

Particularly for engineers and data scientists, it’s easy to get caught up in the hunt,
trying to optimize the model to increase the figure of merit by a fraction of a percent.
This type of improvement may be conceptually satisfying but in the business context,
the more relevant question is whether the change saves money or increases profitabil-
ity through higher throughput or product quality. “I think it’s a mistake to optimize the
models based on recall or precision,” Genzer said. “It’s better to optimize based on
profit or loss. Build the model such that either the savings is optimized or the profit is
maximized.”

23
Getting started with AI-based predictive maintenance

ML-based predictive maintenance deployment  Back to TOC

ML-based predictive maintenance applications can be deployed in a variety of ways.
They are most commonly deployed in cloud-based applications accessed via web
server or run as dedicated appliances on the shop floor.

Deployment doesn’t mean that the job is over, however. Anybody who has ever had a
carefully curated Pandora channel go from playing John Lee Hooker to Justin Bieber
while they are not paying attention understands the issue of model drift, or concept
drift. Best practices call for deploying the chosen model(s), then continuously build-
ing so-called challenger models based on new training sets. The challenger models
should mimic the model in deployment. If they don’t, it could mean that there is a
problem with the deployed model that requires a rebuild. It could mean that the
initial conditions may have changed, either as the (still healthy) machine changes,
environmental conditions change, or operating requirements evolve. Once again, the
situation requires a value judgment: Is there an actual problem with the model that
needs to be corrected or is the model different but the assets are still turning out
quality parts at the desired throughput?

Conclusion
ML is a powerful tool for implementing a predictive maintenance program on the
shop floor. Starting with the business case and taking the time to develop a deep
understanding of the data will help ensure quality results. In general, today’s ML tools
for predictive maintenance are designed to streamline the process of extracting in-
sights from data to improve business operations. “Machine learning isn’t magic and
shouldn’t be treated as such,” Ardis said.

24
Getting started with AI-based predictive maintenance

“You don’t need to be a data scientist to realize savings in your plant with [ML predic-  Back to TOC
tive maintenance tools],” Genzer said. “It is not rocket science. It’s math and anyone
can learn it.”

References
1. McKinsey & Company, “Smartening up with Artificial Intelligence (AI) – What’s in
it for Germany and its Industrial Sector?”

Kristin Lewotsky
Kristin Lewotsky, contributing editor, Association for Advancing Automation

25
Preventive maintenance software:
Doing preventive maintenance  Back to TOC

the right way

Preventive maintenance is a better way of preventing equipment breakdowns
before they cause problems on the plant floor.

P reventive maintenance is a better — and, perhaps more importantly, actually pos-

sible — way of preventing equipment breakdowns before they wreak havoc. It’s all
enabled by preventive maintenance software, which is a whole lot simpler to get your
hands on than an Delorean.

Let’s take a look at preventive maintenance: what it means, why it’s important and how
to do it the right way.

Preventive maintenance fundamentals

First things first: What do we mean when we say preventive maintenance?

At its core, preventive maintenance is exactly what it sounds like: a planned mainte-
nance method based on preventing equipment breakdowns or other asset manage-
ment issues before they arise. Of course, you can’t stop every problem, so preventive
maintenance is just one part of a complete maintenance management approach. You’ll
also want to make room for the other members of the maintenance family: reactive
maintenance and predictive maintenance.

Here’s how these three siblings stack up:

26
Preventive maintenance software: Doing preventive maintenance the right way

Reactive maintenance is the oldest kid in the family and represents a somewhat out-  Back to TOC
dated break/fix approach. Essentially, you’re reacting to breakdowns by scheduling a
maintenance work order or otherwise moving to get things back online. While reactive
maintenance deserves a spot in your overall maintenance management approach, it
shouldn’t be your only method because it can quickly become wasteful.

Preventive maintenance is the middle sibling. It’s a step above reactive maintenance
in the proactivity department. Think of it this way: A reactive approach would be to
help people pay for car repairs after accidents in a dangerous intersection, while a pre-
ventive approach is to put in a new stop sign — and naturally, the latter makes more
sense. Of course, it’s important to know when and where to implement this particular
solution to get the best return on investment (ROI); simply put, you want to minimize
the number of breakdowns you have to react to in the first place.

Predictive maintenance is the youngest — the “hip” one who’s up-to-date on all the
latest tech and uses frankly indecipherable slang. As the name suggests, this approach
is all about predicting equipment or asset issues and performing the right maintenance
work at the right time. With this kid on your side, you’ll see all kinds of benefits to your
asset management and maintenance operations. Just like its siblings, though, there’s
a time and place for predictive maintenance; it’s not the kind of solution you can (or
should) deploy at every turn.

The takeaway here is that preventive maintenance is a big deal — and more than wor-
thy of a place in your digital transformation journey. After all, other efforts to digitize
your plant won’t mean much if you’re still living in the stone age of maintenance man-
agement, right? Trust us: Your equipment, maintenance work teams and even your
budget will thank you. 27
Preventive maintenance software: Doing preventive maintenance the right way

How not to do preventive maintenance  Back to TOC

Now that you know what preventive maintenance is, how it compares to its siblings and why
it’s just one part of a planned maintenance program, let’s take a look at how not to do it.

That’s right: Just like most things in life, there’s a wrong way to do preventive mainte-
nance. In fact, there are a few wrong ways. To further complicate matters, there’s not
even a “right” way to compare against — a preventive maintenance program should
be designed to fit your plant, which means there’s not necessarily a rulebook to follow
for every maintenance task or facilities management question.

Luckily, it’s not too difficult to avoid significant preventive maintenance pitfalls. You just
have to know what to look for:

Time-based maintenance
In some cases, you might be right when scheduling maintenance work based on how
much time has passed since your equipment’s last checkup. However, if your entire
approach is to perform service just because time has elapsed, you’re doing preventive
maintenance the wrong way.

Generally speaking, it’s often smarter to build your maintenance schedule around a
usage measurement like mileage, cycles or rotations. Even placing a work order based
on hours of run-time is better than writing “maintenance” on your calendar every three
months “just because.”

The best approach is to have an indicator of wear and tear — something that helps you
predict upcoming equipment issues and plan your maintenance scheduling according-
28
Preventive maintenance software: Doing preventive maintenance the right way

ly. Naturally, that’s where predictive maintenance and preventive maintenance team up  Back to TOC
to help you navigate work order management, maintenance costs and more.

Manual processes
Here’s a secret: Any time manual processes are the keystone of your equipment main-
tenance, inventory management or facility management, something is probably wrong.
That’s because manual processes are outdated, inefficient and usually unnecessary. Of
course, robots aren’t stealing our jobs any time soon — but we can certainly hand off
every tedious, repetitive maintenance task that only serves as another opportunity for
human error.

More importantly, the idea of preventive maintenance is to make your life easier, your
equipment more reliable and your planned maintenance more efficient — not to add
extra work to your plate. Simply put, if manual processes are taking over your plant,
you know something’s gone sideways in your preventive maintenance program.

Tech overload
It’s tempting to over-invest in technology. After all, so many tech tools promise to be
the end of asset management troubles, work order woes and planned maintenance
frustrations — who wouldn’t want to have as many of those solutions as possible?

The truth, however, is that “tech overload” can quickly turn ROI into R-O-bye. That’s
right: Too much technology drains your budget, hurts your efficiency and can even
make it more difficult to clearly and accurately make decisions about preventive main-
tenance task distribution, downtime mitigation, spare part management and more.

29
Preventive maintenance software: Doing preventive maintenance the right way

Of course, that’s not to say that technology is the enemy; tech for tech’s sake is the  Back to TOC
enemy. Before you start adding sensors and other devices to make predictive and pre-
ventive maintenance possible, you need a foundation that will facilitate reactive, pre-
ventive and predictive maintenance as necessary.

That’s where preventive maintenance software comes in.

Preventive maintenance software saves the day

It’s easy to get intimidated by all those pitfalls, right? At this point, you might be think-
ing it would be easier to just find that ‘81 Delorean.

Lucky for you, preventive maintenance software is here to save the day.

Preventive maintenance software like a computerized maintenance management

system (CMMS) is the foundation you need to avoid those pitfalls. It may not be a
rulebook — remember, there’s only one “right” way to implement a predictive main-
tenance program, and that’s whatever works for your plant — but it does act as a
guardrail to keep you from falling into traps like over-investment in technology.

How is that possible? Well, it all comes down to the concept of digital transformation.
When you utilize the right kind of preventive maintenance software or maintenance
management software, you’re essentially restructuring your approach from the ground
up — and along the way, you’re building preventive maintenance right into your fac-
tory’s DNA. As you digitize other elements of your workflows, you’ll be able to base
every decision on what worked for preventive maintenance, thanks to that software you
chose to invest in from the very beginning. Plus, with no more preventive maintenance
30
Preventive maintenance software: Doing preventive maintenance the right way

scheduling issues or maintenance activity madness, you’ll be able to focus on other  Back to TOC
things — like asset essentials or production efficiency.

Sound too good to be true? It’s not — and here are a few preventive maintenance soft-
ware benefits that prove it:

• Transparency: To ensure that asset and maintenance activity tracking is accurate

and efficient, you need visibility into every moving piece. Preventive maintenance
software gives you this top-down view, allowing you to leverage transparency and
make every maintenance connection count via better, more informed decisions.

• Simplicity: There’s a lot of data involved with asset management, maintenance

operations and all the other things you do to keep your equipment up and running.
With preventive maintenance software, you’ll have a centralized database where you
can organize, analyze and utilize this information — even from your mobile device.

• Efficiency: If preventive maintenance is all about making your life easier, preven-
tive maintenance software is what enables that improvement. This software is de-
signed to help maintenance activities waste less time and money while improving
equipment performance and other asset management stats.

• Optimization: When should you automate a task, and when should you leave it
in the capable hands of your maintenance operations team? Where is predictive
maintenance necessary, and where would effective preventive maintenance suf-
fice? How much tech do you actually need? To answer all these questions and end
up with a well-optimized plant, you need preventive maintenance software.
31
Preventive maintenance software: Doing preventive maintenance the right way

Four methods for choosing preventive maintenance software  Back to TOC

for your plant
Now that we’ve sold you on the concept, it’s time to deliver a little bad news: Not all
preventive maintenance software is created equal. That means it’s up to you to choose
the right software for your plant.

But how do you do that? Where do you start when you’re still trying to figure out the
right time for reactive maintenance, or how much tech is too much tech?

Simple: Look for the signs of good preventive maintenance software. Here are a few to
get you started:

1. Real-time data capture

If you really want to know when a work order is necessary or what the best mainte-
nance approach is, you can’t base your decisions on yesterday’s information. You need
up-to-the-second data that will give you a complete, precise view of what’s going on
with your equipment.

For that reason, the best preventive maintenance software has the capability to cap-
ture and display data in real-time — and not just from machines. The software should
also make it easy to review user-generated data right next to equipment stats and
asset management information, giving you all the facts you need to make preventive
maintenance a breeze.

2. Automated notifications
What good is a preventive maintenance system if it’s not drawing your attention to rel-
32
Preventive maintenance software: Doing preventive maintenance the right way

evant insights? Preventive maintenance software should have automated notifications  Back to TOC
that trigger in response to specific events. Some software can even generate and send
an automatic work order depending on your specifications, cutting out the need for a
middle man and saving time, money and resources in the process.

3. Consistent visibility
Real-time data capture is an important step in being able to know which preventive main-
tenance measures are necessary. However, unless that visibility is consistent and constant,
you won’t know which solutions led to which results, or which problems merit your focused
problem solving resources — and, eventually, you’ll be relying exclusively on reactive
maintenance or manual processes again. That’s why it’s crucial for preventive maintenance
software to include data analysis capabilities, enabling you to learn from your decisions
and prevent even more equipment failures or asset downtime in the future.

4. Improvement framework
One way to think about preventive maintenance is “the creation of new best prac-
tices.” Of course, that doesn’t mean much if you’re unable to standardize your new
approach across the entire plant — so your preventive maintenance software should
allow you to share insights, solutions and best practices with everyone. It should also
include a framework for future improvements; after all, even a best practice might be
strengthened once your teams get ahold of it.

Trent Maw
Trent Maw, senior director of marketing, L2L (Leading2Lean).

33
Natural Gas SPOTLIGHT: How well have you optimized your
compressor?

Compressor “Accurately measure the mass flow of your fuel to avoid

environmental fines and properly account for allocation.”

Application Application Challenges

Optimizer • Turndown

• Mass flow needed

• Limited straight run

• High Vibration

Ideal Products

• FT4A

— Custom
hydrocarbon
gas mix
Natural Gas Compressor Application Optimizer

Blowdown/Emergency Venting  Back to TOC

“Maintain accurate emissions data even during a
blowdown event”

Application Challenges

• Wide range of flow rates

• Pressure changes

• Low flow cutoff needed

• Changing gas composition

• High Vibration

Ideal Products

• FT4A

— Velocities up to 65000 sfpm

• FT4X

— Datalogger with daily totals, remote mount optional

35
Natural Gas Compressor Application Optimizer

Packing Vent  Back to TOC

“Accurately measure leakage to predict mainte-
nance needs.”

Application Challenges

• Low flow

• Low pressure

• Small pipe sizes

• High Vibration

Ideal Products

• FT4A

— Custom hydrogen
carbon mix

— Non-cantilevered sensor
(for high vibration service)

36
Natural Gas Compressor Application Optimizer

 Back to TOC

https://www.foxthermal.com/contact-us/
https://www.foxthermal.com/contact-us/
GET EXPERT HELP
https://www.foxthermal.com/contact-us/
https://www.foxthermal.com/contact-us/
https://www.foxthermal.com/contact-us/
https://www.foxthermal.com/contact-us/

37
How to avoid COVID-19-related
schedule delays using Lean  Back to TOC

If implemented effectively, Lean planning practices can help meet critical

production start dates and make up for delays caused by the pandemic

A s a result of the COVID-19 pandemic, many manufacturing companies placed

their design and construction projects on hold. Many of these projects were
critical to meeting production targets and already had aggressive construction sched-
ules before the pandemic began to impact the global economy. As the manufacturing
industry begins to ramp up again, owners are looking for creative solutions during
design and construction to make up for lost time.

Effectively implementing Lean practices during initial planning can help identify de-
sign, procurement and construction strategies for bringing facilities online in a com-
pressed timeframe. Here are several key considerations and insights.

Develop an effective plan

Historically, manufacturing leaders have preferred the traditional design/bid/build
model, which takes a linear, step-by-step approach to project delivery and uses the
critical path method of scheduling. While this model typically provides the most con-
trol and the least amount of risk to the owner, the linear nature of this method makes it
the longest in terms of schedule duration.

Based on Lean concepts, pull planning focuses on the desired end state and works
backwards to identify activities that can be done in advance or completed simultane-
ously with the goal of meeting that end state. Pull planning is a collaborative process
38
How to avoid COVID-19-related schedule delays using Lean

 Back to TOC

that requires participation from all project stake- Rather than traditional design/bid/build,
holders, and as a result, helps build trust and con- alternative delivery methods may be
necessary to meet start of production
sensus — elements critical to a successful project. deadlines. Courtesy: SSOE Group
Done correctly, an effective pull plan will drive the
development of an efficient and feasible project delivery method.

Schedule acceleration versus procurement protocols

A pull planning session may identify opportunities for significant schedule savings.
39
How to avoid COVID-19-related schedule delays using Lean

However, taking full advantage of these opportunities may require owners to approach  Back to TOC
the procurement process in a less than traditional way.

Consider an open book procurement approach where the general contractor is trans-
parent with the owner’s procurement team on bids received for key trades or project
components. This enables owners to maintain competitive bidding protocols while
being open to potential schedule savings alternatives. Since open book procurement
is a relatively new concept, consider sharing previous project examples or case studies
to educate procurement staff on how they can remain involved and retain control over
the procurement process.

Rethink corporate standards

Many institutional and manufacturing owners have established supplier networks and
longstanding purchasing agreements that support operations and maintenance. There
is a reason that light fixtures and roof membranes are typically the same across mul-
tiple locations. You can streamline maintenance stock, purchase equipment and parts
in bulk, reduce cost per unit and simplify overall maintenance programs.

However, constraints such as these can pose challenges when trying to accelerate proj-
ect schedules. Should speed to market be a priority business driver, owners are encour-
aged to be open to non-standardization and consider alternative products and equip-
ment that still meet their requirements but might be available at reduced lead times.

For example, following the pandemic, standard switchgear for a manufacturing plant
may take 26 weeks or more to be delivered, while a switchboard that provides a simi-
lar function can be delivered in half the time. Additionally, a manufacturing plant may
40
How to avoid COVID-19-related schedule delays using Lean

require more than 400 light fixtures. Placing an order of this size may take a supplier  Back to TOC
weeks to produce and deliver to the site.

By dividing the order among multiple suppliers or considering alternate light fixtures
that still meet the specification but are a different model, the fixtures may be delivered
and installed before the larger order is even shipped. This can apply to HVAC systems
and other building components as well. It is important to note that, when considering
alternate solutions, careful review is required by the installing contractor, engineer and
owner. Electrical and mechanical systems require constant maintenance, so the facility
lifecycle risks and rewards must be evaluated by all parties.

In another example, a greenfield automotive assembly plant includes heavy-duty

cranes that require numerous large columns to support. Traditionally, these columns
are steel. However, due to steel shortages, steel fabricators could not meet the proj-
ect’s aggressive schedule.

During a planning session to address the issue, the design team proposed using pre-
cast concrete columns rather than steel. Although a deviation from the client’s stan-
dard, the alternative solution was accepted, and the precast columns were fabricated
and installed eight weeks sooner than the steel would have been delivered.

Jumpstart the construction process

An effective pull planning session may also identify opportunities to break up projects
into smaller scopes of work. These “pull ahead” packages can enable an owner to start
construction activities before the bulk of design is complete, enabling a significantly
compressed overall project schedule.
41
How to avoid COVID-19-related schedule delays using Lean

For example, take a greenfield manufacturing plant. By completing the sitework design  Back to TOC
and awarding an early site/civil package, the project can begin construction before
other building elements are designed. Foundations and steel can be completed sepa-
rately, enabling steel mill orders to be placed before the bulk of the mechanical/elec-
trical design is complete.

Large mechanical and electrical equipment can require up to six months lead time,
so specifying and ordering substations, switchgear, chillers, air handling units (AHUs)
and so on early places them in the manu facturing queue sooner. Finally, the balance
of the architectural, mechanical and electrical systems can be designed after con-
struction has started, allowing the owner additional time to refine their own process
equipment design and subsequent utility requirements, giving more time for the en-
gineering details for the production lines to be refined without impacting the overall
construction schedule.

By breaking the project up into smaller packages rather than completing the entire
project design in a single comprehensive package, owners can achieve significant
schedule savings.

The importance of owner involvement

Every project will face challenges as it moves forward. How the project team addresses
those challenges can make or break a project schedule. The owner’s project leadership
sets the tone from the very beginning and should facilitate collaboration and create an
environment of trust and respect of each participant to ensure issues do not become
exercises in finger pointing or devolve into blame game scenarios.

42
How to avoid COVID-19-related schedule delays using Lean

By being open to innovative ideas and encouraging all team members to participate  Back to TOC
and contribute to solutions, the owner establishes an effective project culture. In ad-
dition to being open minded, a good owner’s lead (or owner’s representative) should
have the ability to actively listen, effectively communicate and most importantly, be
empowered and be comfortable making decisions.

Accelerating project execution after COVID-19

Many manufacturing organizations use the same project delivery method over and
over because “that is how it has always been done.” As a result, it is crucial for pro-
fessional services consultants and contractors to make clear to manufacturers that
the traditional design/bid/ build approach typically comes with the longest duration,
and that an alternative delivery method may be necessary to meet start-of-production
deadlines.

In many cases, pandemic-related project delays have provided the necessary catalyst
for manufacturers to investigate alternative approaches. If implemented effectively,
these Lean practices can help manufacturers meet critical production start dates and
make up for delays caused by the pandemic.

Brandon Darroch and Brian Arend

Brandon Darroch, PMP, is southeast division manager and senior associate at SSOE
Group, a global project delivery firm for architecture, engineering and construction
management. He has 15 years of experience leading projects for both general con-
tractors and design firms. Brian Arend, PE, LEED AP is electrical power department
manager and senior associate at SSOE Group. He has 17 years of industry experience
and has worked across many different manufacturing industries as well as automotive.
43
Six sigma principles could
improve additive manufacturing  Back to TOC

quality control
Utilizing six sigma principles, team of university researchers and standards
groups are working on finding ways to set standards for quality control when
it comes to additive manufacturing.

A dditive manufacturing offers design flexibility and expanded functionality, but the
quality and process can drastically differ across production machines. With ap-
plications in aerospace, health care and automotive industries with potential for mass
customization, additive manufacturing needs quality management.

To address this concern, Hui Yang, a professor of industrial engineering at Penn State,
and a team of researchers from Penn State, University of Nebraska—Lincoln and the
National Institute of Standards and Technology (NIST) proposed the design, devel-
opment and implementation of a new data-driven methodology for quality control in
additive manufacturing.

“Like an ecosystem, we have people working in isolated efforts in different areas of

additive manufacturing, and systems engineers can help connect the dots to pro-
vide a framework for quality management,” Yang said. “Quality is indispensable,
and if we design a system-level framework of quality management from the start,
then we have higher quality and better productivity at less cost. Ultimately, every-
one wants to do high-precision, high-end manufacturing, but if quality suffers at
any step during production, you lose the competitive advantage needed for the
44
Six sigma principles could improve additive manufacturing quality control

global market. Leveraging data to control and ensure high quality products helps  Back to TOC
keep that advantage.”

How six sigma helps additive manufacturing

The team worked together to analyze various academic papers to deduce a six sigma
framework of quality control for additive manufacturing, which lead to their proposed
systems engineering approach.

The method hinges on six sigma, a popular approach that uses data-driven tactics to
eliminate defects, drive profits and improve quality of products. Through their detailed
analysis, the team suggested that this five-step approach of defining, measuring, ana-
lyzing, improving and controlling can further the quality management when applied to
additive manufacturing.

“Via the research we analyzed, we identified the critical challenges of additive man-
ufacturing and where quality standards are lacking,” Yang said. “For each step in the
process, you need to identify the sticking points, which is where methods such as ma-
chine learning can come into play and help show an engineer or designer how to con-
trol the process to avoid defects.”

Tim Simpson, interim head of the School of Engineering Design, Technology, and
Professional Programs; Paul Morrow Professor in Engineering Design and Manufactur-
ing; and professor of mechanical engineering and industrial engineering, explained
how such defects can become massive liabilities when considered in the context of
mass-produced products.

45
Six sigma principles could improve additive manufacturing quality control

“If your goal is to use additive manufacturing to make parts for a car or a plane, then  Back to TOC
that part better not fail,” Simpson said. He also noted the cost of failed parts can add
up — a failed metal build, he said, could “easily cost 10 to 20 thousand dollars and
require multiple iterations along the way.”

By looking for the quality gaps in the standards for mass-produced parts, Simpson said
their proposed methodology is critical to ensure quality production with additive man-
ufacturing for both high volume and custom products.

“Quality control processes and methods are established for mass production, where
you make hundreds to millions of things,” Simpson said. “Additive manufacturing
enables customization, and the current quality control methods and accepted practic-
es do not readily apply when you are only making one or a few of an item. We have to
think differently to ensure highly quality parts.”

Miranda Buckheit
Miranda Buckheit, Penn State University.

46
 Maintenance

Thank you for visiting the Maintenance eBook!

Content Archive

2021 Winter Edition If you have any questions or feedback about the contents
in this eBook, please contact CFE Media at
2021 Fall Edition
[email protected]
2021 Summer Edition

2021 Spring Edition We would love to hear from you!