CMT: AI not yet intelligent enough to diagnose engine problems

Engine performance optimization still needs human-in-the-loop involvement, says CMT managing director David Fuhlbrügge

AI might be everywhere but Glueckstadt, Germany-headquartered Condition Monitoring Technologies GmbH ( CMT) is cautioning against placing sole reliance on artificial intelligence in ship condition monitoring. It says that human expertise remains essential to ensure safety and accuracy in maritime operations.

Founded in 2023, CMT is a leading developer of diagnostic systems for engine performance and fuel quality monitoring. Its solutions provide real-time analysis of wear metals, combustion parameters, and fuel contamination – critical indicators of machinery condition. The company’s portable and onboard systems are widely used to help prevent failures, optimise maintenance, and improve vessel efficiency

As the shipping industry rapidly adopts AI-driven systems for machinery testing and diagnostics, CMT acknowledges the growing potential of these technologies to process vast quantities of data and assist with condition monitoring tasks. However, the company insists that human engineers must remain “in the loop” to validate, interpret and act upon technical data, particularly aboard increasingly complex and automated vessels.

“AI has a role to play, certainly – especially when it comes to analyzing huge data sets generated during operations,” said CMT’s managing director David Fuhlbrügge. “But there are critical limitations in relying on technology alone. A sensor can tell you a pressure reading or temperature value. It cannot smell burning oil, feel excessive vibration or recognise an unusual sound in the engine room. That’s where human intuition, experience and judgement come in.”

The company currently does not embed AI in its own monitoring devices, relying instead on sophisticated algorithms to provide reliable data interpretation.

CMT suggests that in the future, AI and sensors will be relied upon to flag issues remotely. But, it says, these alerts will still require expert human evaluation, often from shore-based engineers in contact with onboard or visiting crews.

CMT envisions a likely shift towards a hybrid model where monitoring is continuous during voyages, with mobile maintenance teams dispatched to address problems in port.

“Ultimately, we anticipate a setup similar to today’s engine manufacturer service models,” Fuhlbrügge continued. “Sensors might identify a fault mid-voyage, and a flying repair team would meet the vessel at the next port. But without someone qualified to interpret those readings correctly, there’s a serious risk of either false alarms or overlooked faults.”

The concern is compounded by the technical and financial burden of deploying high numbers of reliable sensors across all areas of a ship, especially if these systems themselves become points of failure. AI still lacks the ability to emulate the “gut feeling” that seasoned engineers develop through years of experience, a critical quality when diagnosing nuanced mechanical issues.

“The shipboard engineer is effectively a multi-sensory detector,” Fuhlbrügge said. “They notice smells, vibrations, small changes in behaviour, things no current AI or sensor suite can reliably do. That kind of holistic insight is still uniquely human, and indispensable.”

CMT stresses that rather than seeking to replace engineers, AI should be used to augment their abilities, enhancing maritime safety and efficiency through collaboration between people and machines.

“Seasoned engineers see AI as a useful extension of their own insight. But younger professionals, by contrast, are often more enthusiastic about the prospect of fully automating decision-making, even to the point of removing human involvement. This is as much a cultural shift as it is a technological one,” said Fuhlbrügge.

Acknowledging the long-term potential of machine learning to replicate more sophisticated aspects of human reasoning, CMT warns that significant obstacles remain. Chief among them is the need for massive, diverse, and highly contextual datasets to train such systems effectively, as well as the enormous energy requirements to power advanced neural networks – challenges that are yet to be resolved.

In the meantime, the company calls on industry stakeholders to adopt a balanced and pragmatic approach.

“Let’s not be blinded by the promise of the full autonomous ship. Human engineers are not a relic of the past,” says Fuhlbrügge. “They’re the best safeguard we have for a safe and reliable future at sea.”