Edge Computing and Machine Learning


In recent years, the shortage of engineers at production sites has become quite severe. Yokogawa has heard the customers' concerns that a decrease in the number of skilled engineers has a significant impact on their factories especially in the field of equipment management, because the skilled engineers was able to notice anomalies such as unusual sounds, vibrations, and deformations in equipment using their five senses, thus coping with them contributed to the high availability of equipment.

To support production sites facing these challenges, Yokogawa is proceeding with development aimed at realizing anomaly detection technology equivalent to that of the skilled engineer. The elemental technologies involved are edge computing and machine learning.

Skilled engineers conduct advanced anomaly detection based on equipment data such as sound, vibration, visual feedback, and the level of electric current flowing through equipment. Compared to the physical data on the process side (temperature, flow rate, pressure), there is a massive volume of the equipment data, which volume makes it difficult to acquire them through existing networks and subsequently store in a database.

Image of engineer

Using Yokogawa's eRT3-Plus, which meets the need for flexible control even in harsh environments at production sites, as a platform for edge computing, we are developing a system that adds an on-line modeling function using machine learning to realize the anomaly detection.

Edge computing(Yokogawa's eRT3-Plus) and machine learning


In FY2016, Yokogawa pursued the verification of elemental technologies. In trials with some customers, Yokogawa confirmed that it is possible to construct an anomaly score indicating an upward trend toward the occurrence of an equipment anomaly from the condition data collected from each piece of equipment.

Sensor data

Based on this result, Yokogawa Solution Service Corporation issued a new release on October 17, 2016, announcing the start of "Anomaly Detecting Solution for Manufacturing Equipment" in Japan.

Looking to the Future: Innovation in the Work Style of Equipment Management

With a system that calculates an anomaly score from multiple sensor data, it becomes possible to quantitatively grasp the state of equipment. Additionally, by linking such indicators to maintenance records and failure records, it is possible to grasp the condition of equipment even more precisely. This can be said to be the first step toward a revolution in the work of equipment management. Yokogawa will continue to further develop technologies and support the creation of added value at production sites.

System and PC