Artificial Intelligence

Machine Automation Controller

NX701-Z□00 / NY5□2-Z□00

Machine learning and

Optimization on the edge
Get your machine data
out of the cloud
Data-driven decisions help extract the most value from your machines
Manufacturing is rapidly advancing, and the world is facing a shrinking labor force and a shortage of skilled workers. Omron is helping to
build a factory of the future where people and machines work in harmony by using AI and Machine Learning technologies at the Edge
level and migrating tacit knowledge – such as the intuition of experienced operators – into explicit knowledge.

Optimize your overall equipment effectiveness

Reduce equipment stops Maximize equipment performance Ensure product quality

Host connection
(optional)
AI controller
Machine learning model
High-speed time-series AI engine Sysmac Library for AI controllers
database

AI Predictive AI Machine AI Quality-Yield

Maintenance Library Optimization Library Improvement Library

Control engine

Collection, Utilization ,
sensing control
Input device Output device

How are you utilizing your machine data?

The artificial intelligence machine automation controller (AI Controller) integrates state-of-the-art machine learning functionality
into an Edge level industrial controller, allowing you to leverage machine level information in real time to prolong equipment life
and improve product quality. The controller learns the data patterns of nominal machine behavior without being explicitly pro-
grammed, so that anomalies can be detected and acted upon immediately.

Integrated AI and Controls processing in real time

Data monitoring frame Data monitoring frame Data monitoring frame

AI processing for frame AI processing for frame

AI AI
AI processing

Nominal behavior Anomaly detected

Control Normal operation within controls program

processing Immediate response to anomaly

Leverage the expanding Sysmac Library available from Omron, based on data science and research, avoiding costly development
of in-house analytics and optimization capabilities, as well as custom solutions.
 Artificial Intelligence Machine Automation Controller

Find the economic optimum point

for machine maintenance

Make data-driven maintenance decisions

Each machine is unique and critical to production. The AI Controller learns from each machine to help move
from scheduled and reactive maintenance toward status-based maintenance when it is truly necessary.

Doing maintenance and replacing components

Current : Machine
too late or too early generate losses
Scheduled and reactive maintenance status
Maintenance after failure

Skilled engineers perform maintenance

Failure
based on their intuition and experience
regularly or after failure has occurred
( time-based maintenance). Component a
b c
d
Time

Time-based
maintenance

With AI Controller :
Status-based maintenance
Just-in-time maintenance and
replacement minimize losses
Machine
AI Controller monitors machine status status
using machine data.
Maintenance is performed based on
Failure
machine status when the data shows
Error
leading indicators through Anomaly
Detection. Component a Status-based
b maintenance
c
d
Time

Benefits expected from status-based maintenance

1. Minimized downtime reduces production losses

2. Just-in-time maintenance reduces reactive costs
3. Replacing components only when necessary reduces stock of components
4. Error locations can be identified following leading indicators
5. Maintenance work can be standardized and skilled engineers can create new value

4
 Machine
status
Predictive maintenance procedure using AI Machine
status
Machine
status Usual behavior defined in learning model
Machine
status Usual behavior defined in learning model
Machine
status Usual behavior defined in learning model
Generating a learning model Usual behavior defined in learning model Time

Step1 A learning model including a threshold value is

generated from current machine data.
Usual behavior defined in learning model Time
Time
Time
(Nominal machine behavior is learned.) Time
Machine
status
Machine
status
Machine
status
Machine
Monitoring the machine status
Machine
status

Step2 Machine data is monitored based

on the learning model.
If the specific features exceed the
Time
Time
Time
threshold value, a notification is issued. Time
Time
Machine
status
Machine
status
Machine
status
Machine
Setting a new threshold value status

3
Machine
The machine status is checked. status
Step Time
If no error is found, Time
a new threshold value is set. Time
Time
Time
Machine
status Perform maintenance when needed
Machine
status Perform maintenance when needed
Maintenance performed Machine
Error
status Perform maintenance when needed

4
Machine
Error
status Perform maintenance when needed
Programmatically determine what maintenance is
Step Machine
Error
status Perform maintenance when needed
required, including component replacement. Error Time
Error Time
Time
Time
Time
Generating a learning model Machine
status Usual behavior defined in new learning model
Machine
with new components status Usual behavior defined in new learning model

5
Machine
status Usual behavior defined in new learning model
Step A new learning model including the threshold value is Machine
status Usual behavior defined in new learning model
generated based on the previous error level after Machine
Usual behavior defined in new learning model
status
components are replaced. Repeating these steps Time
makes status-based maintenance more reliable. Time
Time
Time
Time

5
 Artificial Intelligence Machine Automation Controller

Detect machine anomalies quickly and accurately

By integrating the machine learning engine within the machine controller, the AI Controller enables the highest speed and security of
data processing. Anomalies can be accurately detected in milliseconds.

Detection comparison between AI and conventional method

( time-series data such as voltage and current)

Nominal machine behavior Anomalous machine behavior

Threshold detection Pattern detection Outlier detection using feature

by program by human eye values by AI
Measured
Measured
Measured Measured
Measured
Measured Measured
Measured
Measured Measured
Measured
Measured
value
value
value value
value
value value
value
value value
value
value

Detectable
Detectable
Detectable Detectable
Detectable
Detectable Detectable
Detectable
Detectable

Not
NotNot Detectable
Detectable
Detectable Detectable
Detectable
Detectable
detectable
detectable
detectable

Threshold
Threshold
Threshold
value
value
value

Not
NotNot Not
NotNot
detectable
detectable
detectable detectable
detectable
detectable Detectable
Detectable
Detectable

Monitored
Monitored
Monitored Monitored
Monitored
Monitored Monitored
Monitored
Monitored Monitored
Monitored
Monitored
frame
frame
frame frame
frame
frame frame
frame
frame frame
frame
frame
Time
Time
Time Time
Time
Time Time
Time
Time Time
Time
Time

Program Human eye AI

Cannot detect changes that occur Cannot detect minute changes that Can detect minute changes that the
below the threshold value the human eye cannot distinguish program and human eye cannot
distinguish

Functions to detect quickly and accurately

Measured
value Variable a
High-speed Time Series ● Periodic sampling
Database Function as fast as 125 μs (NX7)
or 500 μs (NY5)
Collection and storage of time-series data are
fully synchronized with the control cycle. The
periodically sampled data is used to understand 0 Time
machine behavior, enabling creation of accurate Measured
value Variable b ● I/Ovariables can be
learning models and judgment. Moreover, data
sampled synchronously,
can optionally be saved to external Host, feeding
with less than 1 µs jitter
enterprise level IoT systems and advanced
analysis.

0 Time

6
Machine learning for anomaly detection
Data collection Data analysis Data utilization
Time-series data collection, feature value creation Mining, machine learning Real-time monitoring and action

Features values are generated from data Features with highest correlation to The machine learning model is
gathered from the real-world machine in anomalies are extracted. A machine transferred to the AI controller.
production learning model is generated from the Machine status is
analysis results monitored Standard
in real deviation
time.

Feature values Standard deviation < Θ1

Nominal machine behavior
Maximum = a
Minimum = b Skewness < Θ2
Average = c
Standard
Feature values
deviation =d Normal Average < Θ3
Skewness = e class
Maximum = a
Kurtosis = f
Minimum = b
Normal
Average = c class
Standard
deviation = d
Skewness = e Normal Abnormal
Machine learning class class
Kurtosis = f
model Standard deviation

Anomalous machine behavior

Feature values
Skewness Average
Maximum = g
Minimum = h
Average = i
Standard Skewness Average Blue : Learning data indicating nominal behavior
Feature values
deviation =j
Skewness = k Green : Judged as nominal behavior
Maximum = g
Kurtosis = l Red : Judged as anomalous behavior
Minimum = h Blue : Learning data indicating nominal behavior
Average = i
Standard
Light blue : Threshold value
deviation = j
Skewness = k
Kurtosis = l

Feature value 3

Ultra-high-speed ● Ultra-high-speed AI engine can calculate

in several milliseconds
Machine Learning engine
The AI Machine Learning engine provides both ● One machine learning model can
speed and accuracy. Omron developed this discriminate multiple operating modes
unique technology based on the Isolation Forest
machine learning engine, which is ideal for real- ● Up to 16 feature dimensions
time processing, and fine-tuned it to increase
detection accuracy. The algorithm is applicable
to multimodal data and can be used for high-mix
production lines where two or more operating
modes are required. Feature Feature
value 2 value 1
(Example of three dimensions)

7
 Artificial Intelligence Machine Automation Controller

The AI Predictive Maintenance Library

Pre-made machine learning models

to detect anomalous machine behavior
The AI Predictive Maintenance Library, a collection of Omron developed Function Blocks specific to the AI Controller,
calculates optimal feature values to judge real-world machine mechanism behavior.

Feature value 3

AI Predictive
Maintenance Library
Nominal behavior
AI

Anomalous behavior
Three feature values to Feature Feature
value 2 value 1
judge behavior
(Example of three dimensions)

Robustness minimizes effects of environmental changes

Through research and data science, Omron machine learning models have been designed to minimize the effects of environmental
change using specific feature values, helping to stabilize predictive maintenance activities.

Average value and other Omron-developed

general feature values feature values 10℃ 25℃ 40℃
Measured Measured
value Failure value Failure

Error

Threshold
Error Threshold
value set value set
at 40°C at 40°C

Time Time

Normal state just after start-up is Normal value at 10°C is Unstable state just after Stable detection even in
incorrectly detected as error incorrectly detected as error start-up is ignored environments with significant
temperature changes
※The above results were obtained under Omron’s test conditions. The same results are not guaranteed for all conditions.

8
System configuration

AI Controller Software utilities Optimal web server connection

for additional analysis
Configuration tool Visualization tool

Web server

AI Viewer
AI Operator
[ NX Series ]
Install the AI controller software
utilities on the host computer
・AI Operator ・AI Viewer
NA
Machine interface

Firewall

Proxy server,
DNS server, etc.

[ NY Series ] AI controller
AI Controller software utilities are ・Time Series Database Function
pre-installed in Windows on the ・Feature Value/Machine Learning Function
NY Series ・WebAPI Connection Function
・AI Operator
・AI Viewer
AI Predictive Maintenance Library
NY NX AI
AI Controller AI Controller

EtherCAT slaves

9
 Artificial Intelligence Machine Automation Controller

Ordering Information
NX-series AI Controller
Specifications
Current (Power )
Product Name Model
Program capacity Memory capacity for variables Number of motion axes consumption

4 MB : 256 NX701-Z700
Retained during power
NX701 interruption 40 W
CPU Units 80MB ( including SD Memory
with AI function 256 MB : Card and End Cover)
Not retained during power
interruption 128 NX701-Z600

NY-series AI Controller
Specifications
Product Name Operating Number of RAM memory Interface Model
CPU type Storage size
system motion axes ( non-ECC type ) option

64 NY512-Z500-1XX214T1X

32 RS-232C NY512-Z400-1XX214T1X
Windows
16 NY512-Z300-1XX214T1X
Industrial Box PC Embedded Intel®Core™ 128GB×2 SSD
16GB
with AI function Standard i7-4700EQ iMLC/pSLC
64 NY512-Z500-1XX214T2X
7 - 64bit
32 DVI-D NY512-Z400-1XX214T2X

16 NY512-Z300-1XX214T2X

64 NY532-Z500-112214T10

32 RS-232C NY532-Z400-112214T10
Windows
16 NY532-Z300-112214T10
Industrial Panel PC Embedded Intel®Core™ 128GB×2 SSD
16GB
with AI function Standard i7-4700EQ iMLC/pSLC
64 NY532-Z500-112214T20
7 - 64bit
32 DVI-D NY532-Z400-112214T20

16 NY532-Z300-112214T20

10
AI Controller Services (Americas)
Services are required to allow customers to get best results from Sysmac AI Controller solution. These services are scheduled when the controls machinery is sufficiently ready
to produce and establish a normal baseline. Startup Service will include any necessary AI Controller software utilities, licenses, and AI Predictive Maintenance Library files
required for the application and expected user outcomes.

Service Part Numbers Service Description

Required multi-day Service includes group training on concepts, technology, and hands-on work to start data
AIC-STARTUPSUPPORT collection, feature extraction, model creation, AI Controller Predictive Maintenance Library installation, and
utilization within machine controls program.
Optional Service for re-training additional members after initial Startup Service, as well as assistance with
AIC-RELEARNINGSUPPORT feature extraction, model creation, and threshold setting with the AI Controller software utilities provided
during Startup.

11
OMRON AUTOMATION AMERICAS HEADQUARTERS • Chicago, IL USA • 847.843.7900 • 800.556.6766 • www.omron247.com

OMRON CANADA, INC. • HEAD OFFICE OMRON ELETRÔNICA DO BRASIL LTDA • HEAD OFFICE
Toronto, ON, Canada • 416.286.6465 • 866.986.6766 • www.omron247.com São Paulo, SP, Brasil • 55.11.2101.6300 • www.omron.com.br

OMRON ELECTRONICS DE MEXICO • HEAD OFFICE OMRON ARGENTINA • SALES OFFICE

México DF • 52.55.59.01.43.00 • 01-800-226-6766 • [email protected] Cono Sur • 54.11.4783.5300

OMRON ELECTRONICS DE MEXICO • SALES OFFICE OTHER OMRON LATIN AMERICA SALES
Apodaca, N.L. • 52.81.11.56.99.20 • 01-800-226-6766 • [email protected] 54.11.4783.5300

Authorized Distributor: Controllers & I/O

• Machine Automation Controllers (MAC) • Motion Controllers
• Programmable Logic Controllers (PLC) • Temperature Controllers • Remote I/O
Robotics
• Industrial Robots • Mobile Robots
Operator Interfaces
• Human Machine Interface (HMI)
Motion & Drives
• Machine Automation Controllers (MAC) • Motion Controllers • Servo Systems
• Frequency Inverters
Vision, Measurement & Identification
• Vision Sensors & Systems • Measurement Sensors • Auto Identification Systems
Sensing
• Photoelectric Sensors • Fiber-Optic Sensors • Proximity Sensors
• Rotary Encoders • Ultrasonic Sensors
Safety
• Safety Light Curtains • Safety Laser Scanners • Programmable Safety Systems
• Safety Mats and Edges • Safety Door Switches • Emergency Stop Devices
• Safety Switches & Operator Controls • Safety Monitoring/Force-guided Relays
Control Components
• Power Supplies • Timers • Counters • Programmable Relays
• Digital Panel Meters • Monitoring Products
Switches & Relays
• Limit Switches • Pushbutton Switches • Electromechanical Relays
• Solid State Relays
Software
• Programming & Configuration • Runtime

R92I-E-01 Note: Specifications are subject to change. © 2018 Omron. All Rights Reserved. Printed in U.S.A.

Printed on recycled paper.