Semiotic Labs has developed a predictive maintenance system that can identify equipment failures up to five months before they happen. It’s like telling the future. Brian Dixon dug deeper.
Although predictive maintenance is not in itself a new concept, the SAM4 system developed by Netherlands-based Semiotic Labs employs a somewhat unique approach to detecting maintenance issues associated with AC motors and other items of rotating equipment, such as pumps, fans, compressors and conveyors. While most systems in use today use vibration analysis to identify tell-tale signs of potential problems, SAM4, explains co-founder Simon Jagers, instead uses artificial intelligence (AI) and machine learning (ML) to learn and then monitor an asset’s electrical waveforms for any deviations or abnormalities indicative of future failure.
One obvious benefit of this is that the SAM4 system negates the need to install vibration sensors on the assets in question, something that can be highly problematic if such items are situated in a hard-to-reach or hazardous locations as can often be the case onboard a vessel. Moreover, as the SAM4 system focusses on electrical waveforms, its sensors can be quickly and simply installed inside the appropriate motor control cabinet thus enhancing both safety and convenience as well as potential scalability.
Once such installation has been completed, the SAM4 system embarks on an automated learning process to establish the normal operating profile of the target asset’s waveforms. This, Jagers explains, typically takes about two weeks to perform. “We then start to monitor [the asset] 24/7, analysing the data and detecting failures up to five months before downtime happens,” he says.
As well as allowing plenty of time to source replacement components, such lengthy advanced warning is a clear boon to users from the maritime sector given that a ship may well be away at sea for a considerable period. Importantly, the AI and ML employed also imbue the SAM4 system with a high degree of accuracy and reliability.
“The combination of electrical waveform data and automated powerful analytics [means SAM4] can detect well over 90% of failures whereas traditional solutions typically detect 50-70%,” Jagers says, noting that SAM4 users are thus likely to experience four times fewer failures than might otherwise be the case with alternative systems.
“We take a purely data-driven approach. We let the data tell our algorithms how to respond, which makes it a lot more accurate and a lot more scalable because we can deploy and start learning automatically,” he continues. As such, the SAM4 system requires “zero manual handling on the client’s side” once in operation.
Furthermore, by employing a data-driven rather than a model-based approach, the system can also provide users with additional analytics, such as those pertaining to the energy efficiency of an individual asset or wider process. This in turn can help a user fine-tune their operations while, in the context of overall digitalisation, also help in the development of a more accurate digital twin of a vessel or facility.
Founded in 2015, Semiotic Labs currently services a range of big-name customers from a variety of different industries, including Vopak, Nouryon and ArcelorMittal among others. Moreover, it is presently working with a large maritime operator with a view to developing a standalone version of the SAM4 system with very limited bandwidth requirements that should be ready for a fleet-wide deployment and subsequent wider commercial roll-out within the next 18 months.
