^ Figure 1. View of the exhibition in Japan.

Article by Isao Nishizawa
___

IoT sensor
Figure 2. IoT sensor image.

The proper use of AI in connection with flow control devices could significantly reduce valve problems and enhance plant uptimes.

To give an example: a development currently underway at KITZ concerns a valve diagnostics system.

This system enables operators to monitor valves remotely and realize CBM (Condition Based Maintenance) by allowing AI to learn from the data and predict valve conditions and detect anomalies. The concept has been presented at a Japanese exhibition (see Figure 1) where it attracted much interest.

IoT sensor

In order to properly utilize AI, the first step is of course to have access to a reliable flow of data. Hence KITZ has, for example, developed an IoT sensor. Easily retrofitted to the actuator, this sensor can be used to provide data about the angular velocity of the closing elements in quarter-turn valves. Such data is wirelessly sent to the cloud server and managed by the cloud server.
The IoT-sensor can be easily retrofitted to the rotating shaft of the actuator for ON-OFF actuated valves. The image is shown in Figure 2. The sensor is powered by a solar panel or battery, eliminating the need for wiring, making it easy to accommodate additional sensors and valve changes. The non-explosion-proof type and the explosion-proof type are under development to accommodate a wide range of production sites from factories to plants.

Seat deformation

AI Scores and Damaged Conditions
Figure 3. Changes in AI Scores and Damaged Conditions of Actually Confirmed Ball Sheets.

An anomaly detection algorithm has been created that utilizes AI by registering data – both historical and actual – regarding the angular velocity of the valve. It should be noted that changes in the angular velocity can indicate issues inside the valve. For example, the ball seat may become damaged if a slurry component is deposited on the seat surface.

The AI algorithm maintains a score of 1.0 when the valve is normal but changes that score if the valve seat shows early signs of damage.Figure 3 shows an example of actual verification of the validity of the AI algorithm on the customer’s production line. Maintenance was performed as soon as the score became 0.2. The intervention was very timely; damage to the ball was minor, so steps could be taken to prevent major problems from occurring.
About the author
Isao Nishizawa is manager of the department within KITZ that creates new customer value by providing concrete solutions based upon AI and IoT.

At the current time, we are working to further improve diagnostic accuracy by physically checking the valve health condition when AI-data inform us that the valve is deteriorating. We also check changing point data of angular velocity data. This enables us to update the algorithm if necessary.

Previous articleAmpo supplying more than 20000 valves for Arctic LNG 2
Next articleOxford Flow announces 42% headcount growth
Valve World Publisher