WHAT IS PREDICTIVE MAINTENANCE?

PREDICTIVE MAINTENANCE - PdM

Predictive Maintenance is the use of on-line monitoring to estimate the condition of machines.

It is much more than applying condition monitoring in your maintenance strategy as you might be doing with Condition Based Maintenance (CBM) in conjunction with manual route based monitoring. Importantly, PdM sources data during normal operations, hence minimizing disruption to operations for sampling or measuring. Traditional condition monitoring comes with a human burden on data collection and analysis. PdM removes this dependence, hence removing barriers for the application of cost effective condition monitoring to a much wider range of asset criticality.

PdM extends a Condition Based Maintenance (CBM) strategy to:

  1. Cover a greater number of machines (balance of plant)

  2. Reduce preventative maintenance burden

  3. Enable the planning of opportunistic corrective maintenance

  4. Prevents unexpected failures

PdM and CBM are complementary disciplines. If you already have a CBM program then culturally your organizations understands the benefits of monitoring machines to optimize your maintenance policy. PdM takes that to the next level.

PdM utilizes a much higher level of data driven approaches than traditional condition monitoring as the data is from on-line sources opposed to irregular manual readings. The ability to be data driven enables PdM to exploit techniques form the domains or artificial intelligence and machine learning.

Read More On How It Works

WHAT ARE THE BENEFITS?

Senseye PdM, the industry leading Predictive Maintenance solution, is cloud based and automates data analysis. Also the Senseye PdM user interface is set up for operations and maintenance teams, rather than requiring specialist analysts. This approach of automating the analysis and simplifying the reporting of the resulting actions streamlines a complex task, saving time and costs, and allows maintenance teams to multiply their efforts, reacting with increased speed and accuracy.

Another benefit of automated analysis is that Senseye PdM is scalable, so it can grow with the requirements. This means that, in addition to critical machines, the wider non-critical population can now be monitored. These subsidiary machines may not be as expensive to replace, but their downtime could still be costly.

Calculate your return on investment

WHAT DO THESE WORDS AND PHRASES MEAN?

A – Anomalousness

A – Anomalousness

“Anomalies are features in the data that are unexpected or unexplained. Anomalousness is a measure of the density of anomalies, or put another way, a measure of how abnormal the data appears.”

Dr. James Loach
Chief Scientist

B – Bathtub Curve

B – Bathtub Curve

“This is a classic representation of the probability of machine failure over the lifecycle of an asset. At the beginning, failure is more likely due to manufacturing and installation errors (early failures). For the majority of an asset’s life, failure rates become constant (random failures). Towards the end, the probability of failure begins to increase as components succumb to fatigue (wear-out failures).”

Chris Esprey
Condition Monitoring Specialist

C – Condition Monitoring

C – Condition Monitoring

“Condition Monitoring (CM) is a discipline that makes use of sensor measurements with processing to determine the condition of industrial machines. CM enables many business benefits and forms the basis of maturing maintenance practices from reactive to preventative to predictive.”

Rob Russell
CTO 

D – Diagnostics

D – Diagnostics

“The examination of symptoms and syndromes to determine the nature of faults or failures.”

Dr. Simon Kampa
CEO

E – Expert Knowledge

E – Expert Knowledge

“Expert knowledge, or context, is information used to interpret data. In our domain, it describes the characteristics of assets, how they operate and how they fail. It is also used to select condition indicators.”

Dr. Samuel Park
Data Scientist

F – Failure Modes

F – Failure Modes

“Machines don’t fail in a single consistent way. There are many moving parts that degrade over time. Failure modes are a way of describing each type of failure that a machine can experience, but importantly using a consistent term or code for each failure mode. Good practice in capturing failure modes as part of maintenance work recording is an investment in the value of your contrition monitoring data, as it provides context to the data and enhances the type of analysis that can be performed.”

Rob Russell
CTO

G – Ground Truth

G – Ground Truth

“‘Ground truth’ is the name given to any domain-specific benchmark that can be used to quantify and judge the predictive capabilities of a machine learning model that has been applied to that domain. When used appropriately, ‘Ground truth’ can be used as a model selection tool.”

Dr. Samuel Park
Data Scientist

H – HUMS

H – HUMS

“Originating from aerospace, Health and Usage Monitoring Systems (HUMS) are designed to automatically monitor the health of mechanical components, as well as the usage of an airframe and its dynamic components. HUMS have been shown to enhance safety, decrease maintenance burden, increase availability and readiness and reduce operating and support costs. The HUMS concept is now making its way into other industries, though under different names.”

Rob Russell
CTO

I – Industry 4.0

I – Industry 4.0

“It’s broadly agreed that there are four industrial revolutions: The first representing a change from agricultural to urban environments, the second representing mass production, the third for digital automation and the fourth – connectedness. Industry 4.0 represents the infusion of Information Technology into every layer of machine, resulting in Operational Technology and producing massive amounts of data that enable intelligence and optimization never before possible.”

Alex Hill
Senseye USA CEO

J – Just in Time

J – Just in Time

“Just in Time (JIT) is a fundamental manufacturing methodology originating in the Japanese manufacturing sector in the 1960’s. It focuses on shorting manufacturing cycles and reducing the need for stock holding across the entire supply chain.  The benefits affect every part of the business: the cost of inventory storage, more efficient production and product quality improvements.”

Rob Russell
CTO

K – Kalman Filter

K – Kalman Filter

“An approach where you use knowledge about the system states and educated guesses about uncertainties to help you make reasonable inferences about the future evolution of the system. When it all comes together it is almost magical.”

Dr. Lee Middleton
Data Scientist

L – Lubrication

L – Lubrication

“Lubrication is the process by which the friction/wear between two surfaces is reduced. This is commonly achieved in machinery via an application of a lubricant such as oil or grease. Changes in the particulate and moisture content of the lubricant can be used to infer wear of machine components and prompt maintenance action.”

Chris Esprey
Condition Monitoring Specialist

M – Machine Learning

M – Machine Learning

“Algorithms are instructions for turning inputs into outputs. Humans can write them, but when this is hard or slow, we can delegate the task to a computer. The algorithms that computers use to write algorithms—to infer them from samples of inputs and outputs—are called machine learning algorithms. And they are what machine learning is all about.”

Dr. James Loach
Chief Scientist

N – Notifications

N – Notifications

“Notifications are a key way to direct user attention to the right thing. It’s a fine line between too much information (‘I have too much work to do!’) and too little (‘This thing is broken!’) and one that can only be understood by keeping close relationships with regular users of the product.”

Adam Poole
Product Design Lead

O – Operational Equipment Effectiveness (OEE)

O – Operational Equipment Effectiveness (OEE)

“Operational Equipment Effectiveness (OEE) is a key measure of manufacturing productivity. It shows the productive percentage of manufacturing time by combining 3 elements of machine performance: availability (A), performance (P), and output quality (Q). Achieving an OEE score of 100% means that only good parts are manufactured, as fast as possible, and with no down time.”

Lily Hristova
Product Associate

P – Prognostics

P – Prognostics

“Prognostics is the engineering discipline of predicting machine failure. There are two main approaches, model driven uses physical understanding of the machine to predict failures, whereas data driven uses historic data and machine learning. We blend the two, using our condition monitoring expertise to bolster data driven techniques.”

Dr. Daniel Reid
Chief Software Architect

Q – Quantisation Error

Q – Quantisation Error

“Quantisation error is the process by which noise is added to a signal by the sample values being constrained to a discrete set of numbers. Quantisation error can be introduced by the use of sensors which have a low resolution relative to the features of interest in the data and can lead to reduced sensitivity in condition monitoring applications.”

Graham Bruce
Condition Monitoring Specialist

R – Root Cause

R – Root Cause

“The root cause is the fundamental reason for the occurrence of a problem. The term denotes the earliest, most basic cause of the fault. Fixing a problem at its root therefore removes the cause at its deepest layer, eradicating it completely.”

Peggy Cooper-Berger
Marketing Associate

S – Scalable

S – Scalable

“Scalability means the capacity to change the scale of any given project – from one machine to tens of thousands. The amount of data that can be captured from modern machines is enormous and increasing all the time. In order to harness it, we design our systems so that we can monitor the 100’000th sensor just as accurately and as quickly as the first.”

Oliver Skånberg-Tippen
Machine Learning Engineer

T – Trend Detection

T – Trend Detection

“Unexpected trends in data from mechanical equipment indicate significant and persistent change in machine state. This is often a sign of degradation in the run up to machine failure, thus detecting them at the right time is an important capability.”

Dr. Henry Truong
Data Scientist

U – Unplanned

U – Unplanned

“From your commute ruined by a broken-down train, to your halted production line due to machine failure, unplanned downtime can be inconvenient and costly. One of condition monitoring’s aims is to find early indicators of failure so intervention can avoid catastrophic events.”

Dr. Chris Smith
Condition Monitoring Specialist

V – Vibration

V – Vibration

“Vibration Analysis is a keystone technique in the condition monitoring of rotating machinery as small changes in the structure of mechanical components (such as bearings and gears) have a large and predictable effect on the vibration response of the component – this allows for the early identification of degradation and failure of said component.”

Chris Esprey
Condition Monitoring Specialist

W – Web App

W – Web App

“Senseye’s web app allows maintenance teams to quickly and easily interpret vast quantities of data, keeping their focus on performing critical maintenance activities. We bring the most important information to the forefront, provide the tools to analyse and understand the underlying data, all in support of real world decision making.”

Matt Wilson
Frontend Lead

X – XOr

X – XOr

“The XOr, or ‘Exclusive Or’, is a classic problem in machine learning. It uses a neural network to predict the outputs of XOr logic gates given two binary inputs. It should return a false value if the two inputs are equal and true otherwise. Complex models like this always need to be changed and adapted to solve problems for real machines.”

Henry Truong
Data Scientist

Y – Yaw

Y – Yaw

“Yaw rotation is the movement of a rigid body about its vertical axis. As a key axis of movement for machines ranging from industrial robots to wind turbines, yaw is frequently a focal point for condition monitoring applications.”

Graham Bruce
Condition Monitoring Specialist

Z – Zero

Z – Zero

Zero was a bot from the hood,
who instead of moving just stood,
his measures were flustered,
so the bot was adjusted,
giving Zero no downtime for good

Dr. Lee Middleton
Data Scientist

LOOKING FOR MORE INFORMATION?

We have a comprehensive gallery of Whitepapers, Case Studies and Blog Content available where you can learn more

Ready to start your predictive maintenance project?

Then get in touch with us to find out how your organization could benefit from our cloud-based product for Predictive Maintenance 4.0.

GET IN TOUCH