SAVA: detect uncommon machine-behaviour valves

The Self-Adjusting Valve Algorithm (SAVA) is a mathematical machine learning algorithm, specifically designed to detect uncommon machine-behaviour for valves.  The machine learning is used to generate a “fingerprint” of the common machine behaviour and aims at detecting failure and fouling by finding deviations to these fingerprinted patterns. Valves have a particularly “non-linear” behaviour that makes it difficult to interpret these deviations. The SAVA aims to provide the context of the deviation, making it comparable to other deviations in the past, enabling the user to determine if maintenance is desired or necessary.

How does it work?
The algorithm operates by using both machine learning and physical calculations. The “shape” of a valve fingerprint has been well understood and defined by engineers. The machine learning enables us to fit this behaviour within the operating window. This “stateless” valve algorithm makes adjusting fingerprints quick and easy, and the results are comparable between valves and different time intervals. The algorithm is based on a physical principle of pressure differential, flow, and the speed at which the valve opens/ closes, and thus uses the most readily available valve measurements.

What would be the output?
The modelling process provides several short- and long-term KPIs to give a better understanding of the ongoings of the valve and the overall performance of the statistical method itself. The latter being a measure of confidence of the made fingerprint. The algorithm gives an output concerning a long-term trend deviation and a short-term trend deviation to account for fouling and failing respectively. An increased deviation of the fingerprint means that the valve is more rapidly changing or that the pressure around the valve has been changing, indicating seal-leakage.

It also provides an error metric, that provides a measure of fit of the modelling method. This is a due-diligence metric, to appreciate when something is changing. We want to be able to exclude ‘issues’ due to modelling failure and can do so by looking at the error metric.

What would be the benefits?
By using our valve algorithm, you are enabled to extract more performance out of your valves without having the need to add, change, or replace the valve or its surrounding. The algorithm will further increase your competitive advantage by:

  • Improving the plant value drivers, such as Asset Utilisation, Safety Records, and Maintenance Costs. We enable you to digitally inspect your valves and turn unplanned incidents into planned maintenance!
  • Improve Plant Control.
    We provide you with remote real-time condition-based alarms, to free up our schedule and focus on important tasks.
  • Formalising Engineering knowledge.
    We enable you to digitally store any anomalies, incidents, or remarks per valve over a time frame.

Want to know more about SAVA? Send me a message via: nickel.mortel-van-de@sitech.nl

Share this page

Subscribe to our digital newsletter