What value does AI bring?
Easy AI for achievement-backed industrial automation
Yokogawa has already accumulated know-how by solving the problems of more than 60 factories and plants using AI.
Yokogawa developed its AI based on the achievements, experience, and knowledge it has cultivated for years, and now we are incorporating AI into the popular SMARTDAC+ Series Paperless Recorder GX/GP, Data Logger GM and Data Logging Software GA10.
Additionally, our new e-RT3 Plus series F3RP70 CPU module features industrial-strength environmental worthiness and support for the Python generic programming language, making AI industrial use and development easier than ever.
These three products aim to make AI easy to use for predicting equipment malfunction, product quality, and other factors to help solve our customer’s problems.
All the achievements we have solved and the merits of introducing AI products
→ AI Product Solution Book download
Reasons for proposing adding AI technology at sites
- Three ways that AI can offer value in plants.
- Results from YOKOGAWA AI solutions
- Group of products that come standard with AI and contribute to fast AI development
- Plant control with AI and reinforcement learning
Three ways that AI can offer value in plants.
We’ve been working on solving a variety of customer issues. Of the issues that couldn’t be solved with conventional analysis methods, we were able to solve 50 by applying AI. We learned that in plants AI can offer value in the following three ways.
Abnormal Sign Detection
Predicts various equipment failures and plant shutdowns. Enables you to perform maintenance before an abnormality occurs, take countermeasures in advance, and improve your uptime.
Root Cause Analysis
Ascertains the causes of previous problems, and identify the cause of reduced quality and changes in consumed power. By identifying the areas of focus when improving and implementing quality measures, you can improve product quality.
Assigns quality indexes, predicts quality before testing, and predicts changes in quality. Get a handle on quality without destructive testing and other processes that take time to test, and reduce costs.
Results from YOKOGAWA AI solutions
These are some of the analysis results from AI solutions that YOKOGAWA has offered.
We have been solving customers’ issues in many industries including oil refineries and chemical plants using our AI.
Abnormal Sign Detection
-Quickly detects Anomality in equipment-
A shaft breakage occurred in a reducer that monitored trends with a wireless vibration sensor (Sushi Sensor). The graph below shows acceleration, speed, and surface temperature data, and the results of AI analysis based on those data. The moving average of health index calculated by AI moved from the normal to abnormal range three weeks before a failure. You can see that it captured the “Anomality” that were signs of abnormality earlier than the sensor data trend changes. By combining the Sushi Sensor with the AI, undesired state can be notiﬁed based on the equipment's vibration and surface temperature data, which enables you to predictively detect equipment abnormality and to take prior actions for equipment maintenance.
Results from YOKOGAWA AI solutions (Excerpt)
|Abnormal Sign Detection
|Predict deterioration of sensors installed in a waste water pipeline
|Predict operating conditions of a furnace by using process data
|Predict cavitation in pipes by using pressure-related data
|Root Cause Analysis
|Identify the causes of reduced efficiency of cooling compressors in area A of a plant by using process data
|Hot spring control
|Identify the causes of changes in hot water distribution by using operating data
|Paper & Pulp
|Identify the relationship between a beating machine’s power and product quality
|Electronic parts manufacturing equipment
|Identify damaged parts in an assembly process by using sensor data
|Industrial material continuous production equipment
|Predict indicators between quality measurement indicators by using 18 types of data
|Product quality values
|Estimate product quality values by using production data
Food & drug
|Medical product manufacturing equipment
|Predict the quality of completed product by using 10 types of manufacturing data
The above is only a small part.
See the AI Product Solution Book for other achievements.
Utilizing our analysis experience and technology, we can offer a lineup of easy-to-use AI products.
In addition, we have prepared a video of an example of optimal plant control using AI reinforcement learning.
→ Please refer to Plant control by AI
The dream of an AI that can control your plant is becoming a reality.
-Plant control with AI and reinforcement learning-
Yokogawa has developed the technology to use AI to directly control equipment through reinforcement learning.
Reinforcement learning is one type of machine learning in which AI learns from its own trial and error. The key feature of Yokogawa’s reinforcement learning technology making it so practical is that it can learn with a small number of trials.
Issue with a complex process
- Indicated values are destabilized due to disturbances, resulting in loss of raw materials.
- Skilled workers have difficulty passing on control adjustment technology, requiring time and labor.
- Destabilization of process and decrease in quality due to unfamiliar control methods.
Features of Yokogawa Electric’s reinforcement learning
- The AI teaches itself, and doesn’t require large amounts of teaching data.
- Learns with a small number of trials.
- Optimized with reinforcement learning technology specialized for the control field.
In a World First Yokogawa uses AI to Autonomously Control a Chemical Plant for 35 Consecutive Days
Using Yokogawa Electric's autonomous control AI, factory processes that could not be made autonomous using conventional control methods such as PID control were controlled over a long period of time while maintaining safety and quality.
For more information, please visit >> this news
ENEOS Materials Corporation has been in operation for over a year and the system has been officially implemented.
Reinforcement learning technology FKDPP is installed in a small general purpose controller.
For details, please see >> Industrial AI Platform
AI control test equipment by a Three Tank Level Control System
Real-world sites often require complex control mechanisms.
Here we show a experimental example of that, in which we apply reinforcement learning technology to control a Three Tank Level Control System.
In under 4 hours of learning, it found the ideal control method to eliminate the overshoots that occurred previously.
Optimized with reinforcement learning AI specialized for the control field
Supports Autonomous control AI
Industrial AI Platform e-RT3 Plus
This industrial AI platform supports general-purpose operating s ystems such as Linux Ubuntu and supports C and Python, which are essential for AI application development. Multi-CPU support, a wealth I/O for high scalability, and high environmental performance make this platform ideal for industrial development.
A wealth of I/O supported
A wealth of I/O and communication supported by e-RT3. You can easily get startup with the I/O access libraries accessible from Python. You can quickly develop data collection, AI applications, and more.
OSS to develop application more efficiently
You can boot most popular Linux standard distributions.
Therefore, it supports many open source software.
Sample boot able Linux (Ubuntu) image files are provided to assist in system construction. You can easily transfer your own AI applications and use network file sharing software or PC-less SCADA software.
The e-RT3 Plus CPU can be mounted next to the sequence CPU mounted in the base module. Control can be performed by the high-speedsequence CPU, while communication and AI applications can be performed by the e-RT3 Plus.
Holds up in hot, harsh environments. The e-RT3 has a fan-less design, modules that can withstand temperatures from 0°C to 55°C, and installs in factories, plants, or inside a box outdoors.
Python is a programming language widely used for AI development.
By seamlessly connecting on-site equipment and host systems, it is possible to develop applications in both IT and OT fields.
e-RT3 Plus is certified by both Microsoft Azure Edge and Amazon AWS IoT Greengrass.
By connecting an e-RT3 Plus developed application to one of the certified clouds, secure communications can be established and a new cloud based solution can be created at the customer’s production site.
High-Accurate AI using Python for industrial application in harsh environment
Reduce development time
How about challenge the issues you have given up on?
We have added a new lineup of AI control learning services using autonomous control AI (reinforcement learning AI, algorithm Factorial Kernel Dynamic Policy Programming, FKDPP)*1 for the industrial AI platform e-RT3 Plus.
By packaging the controller and services, it is possible for customers to tackle AI control themselves. Our reinforcement learning AI has the advantage of being able to learn with fewer trials than general reinforcement learning, making it extremely practical.
*1 This is a reinforcement learning AI algorithm jointly developed by Yokogawa Electric and Nara Institute of Science and Technology (NAIST), and was recognized for the first time in the world as a "reinforcement learning technology that can be used in plant" at an IEEE international conference.
Coexistence of productivity and energy savings
Application examples：heating furnace controls
- It is difficult to control PID at low temperatures with a heater of large output, resulting in overshooting and hunting.
- Wasted energy is consumed and the load beyond the set value is applied, causing deterioration and failure of the furnace.
- Settling time becomes longer, resulting in lower equipment operating rates.
- Autonomous AI control was confirmed to suppress overshoot and reduce settling time by about 65%.
[Comparison of autotuned PID control and AI control]
Utilize autonomous control AI without specialized knowledge
Thanks to simplification of the AI model creation process, even non-AI experts can create an autonomous control AI model and install it on an e-RT3 edge controller.
Uploading the equipment operating data automatically sets the identification conditions and creates a simulator.
Generated control models are displayed graphically for review. Select the AI control model most suitable for your purpose.
Equipment/Quality Easy Predictive Detection/ Future Pen
SMARTDAC+ GX/GP/GM Series
Paperless Recorders and Data Logger
What is the GX/GP/GM Series?
The GX/GP/GM series can be used to acquire, display, and record data such as temperature, voltage, current, flow, and pressure in various industry production and development sites.
The GX/GP series is a panel mount or portable paperless recorder that provides intuitive touch panel operation plus a highly flexible modular.
The GM series is a versatile and scalable data logger. Because you can add or remove modules even after installation, it of fers excellent maintainability.
Want to build an AI predictive detection system in the field
Equipment/Quality Easy Predictive Detection New!
You can easily build equipment/quality predictive detection systems. By quantifying deterioration of equipment, you can optimize parts replacement and maintenance period, improve downtime by predicting abnormal conditions in equipment, and automate work by quantifying product quality.
Equipment/Quality Predictive Monitoring
- Quantifies the degr ee of deterior ation in equipment and quality, and monitors trends
- Notifies you of abnormal signs on site
- Optimizing costs by performing predictive maintenance rather than preventive maintenance
* Certain restrictions apply with Equipment/Quality Easy Predictive Detection. See the general specifications for details.
GX/GP AI Functions
Using acquired data to predict future data, draw predicted future waveforms along with real-time data on the trend monitor. The future waveforms allow you to identify and deal with likely problems.
You can set future alarms against future measured data from the future pen. Future alarm information is shown in a future alarm summary. When a future alarm is generated, external output or email give prior notice.
Future Pen Demonstration for Predictive Detection
Examples of building recorder-centric AI solutions
Easily add AI to a recorder (GX/GP future pen, future alarm)
Process signals such as temperature and voltage can be displayed from past to present, and future waveforms by AI function. (GX/GP)
Accurate AI analysis in real time on a PC and displaying determination results on site. (GX/GP + GA10)
GX/GP collects the process signal and displays the measured values.
Send real-time data to PC, the AI analysis function of the software GA10 displays the result of AI judgment such as uncomfortable feeling detection on the GX/GP on the site side.
Ask us about the highly accurate AI consultant service.
Analyze on an embedded device with accurate AI and display determination results (GX/GP + e-RT3 Plus)
GX/GP collects the process signal and displays the measured values.
Collects process signals with GX/GP, displays measured values, transmits data in real time to the Python-compatible AI platform e-RT3 Plus, and deploys the judgment results by the AI algorithm developed by the customer on the GX/GP screen can do.
Ask us about the highly accurate AI consultant service.
AI on a stand-alone computer detects predictive abnormality sign of equipment
with Anomaly Detection,
AI analysis data logging software GA10 + Sushi Sensor
In recent years the need for maintenance is growing as equipment ages. Equipment conditions are mainly ascertained by human workers through inspections such as operator rounds inspection. However, we face the problems of declining birthrates leading to labor shortages, and the retirement of skilled workers leading to skill shortages. Also, the measured results obtained through an operator rounds inspections are not quantified, and often cannot be utilized effectively. We urgently need to create more efficient equipment maintenance mechanisms.
Automatically detects the “anomalies” that are signs of abnormalities
GA10 + Sushi Sensor are equipment predictive maintenance solution that is just like an AI operator, detecting the "anomalies" that are signs of abnormality, and automatically notifying the user.
You can dispatch worker to equipment that needs maintenance at just the right time, this reduces equipment inspection man-hours, lets you quickly discover abnormal signs, and prevents unanticipated equipment shutdowns. You realize more efficient equipment maintenance and increased plant availability.
*Sushi sensor: Up to 1000 units can be registered. (Gate Sushi function; option)
Predict the future with AI New
Use acquired data to predict future data, and display predicted future waveforms along with real time data on the trend screen. You can also set alarms against predicted data to detect problems in advance when they occur in the future so you can deal with them beforehand.
GA10 Data Logging Software - AI notifies you of anomalies
What is GA10?
GA10 is PC-based data logging software that connects various devices installed in factories and premises (such as Sushi Sensor, recorders, and data loggers) through an Ethernet network, and performs monitoring and recording.
What is Sushi Sensor?
The Sushi Sensor* is a wireless solution for IIoT that includes sensors for detecting equipment conditions. It supports low power, wide-area LoRaWAN communication standard for wireless network.
*Sushi Sensor is to be released to other countries in order. For details, please click “more information”
Realize predictive maintenance and avoid future crises with the usual operability
Realizing predictive maintenance with trend monitoring
- Monitoring of large amounts of data
- Digitizing on-site know-how
Efficient equipment maintenance
- Digitize equipment health conditions and monitor trends
- Reduce the cost of operator rounds inspections
- Share and transfer field operator's knowledge and experience of equipment maintenance
- Prevent sudden unexpected equipment failures
Quickly capture the signs of equipment abnormalities
- Capture signs, even with equipment for which it is difficult to apply threshold-based abnormality determination
- Based on a variety of data, determine signs of equipment abnormalities
Easily detect signs of equipment abnormalities with simple settings
- Simple system: Sushi Sensor, LoRaWAN gateway, and GA10
- Just specify a data period of normal conditions, and start automated analysis
- Notifies you when anomalies are detected
GA10 is PC-based data logging software for monitoring and recording data by connecting to devices distributed in a factory. Now with an AI-based Anomaly Detection function, the AI notifies you of anomalies before equipment abnormalities occur.
AI development is not that difficult! This tutorial video shows you how to develop AI application on e-RT3 F3RP70 by running AI sample program available on Partner Portal.
In a World First, Yokogawa’s Autonomous Control AI Is Officially Adopted for Use at an ENEOS Materials Chemical Plant
– One year of stable operation demonstrates this next-generation control technology can decrease environmental impact, achieve stable quality, and transform operations –
- Optimizes control to improve productivity and save energy -
- Targeting remote, autonomous operations for the process industries
In a World First, Yokogawa and JSR Use AI to Autonomously Control a Chemical Plant for 35 Consecutive Days
- Putting into practical use a next-generation control technology that takes into account quality, yield, energy saving, and sudden disturbances -
Looking for more information on our people, technology and solutions?Contact Us