Using data from vibration sensors on the powertrain, OEMs will be able to diagnose equipment health and avoid vessel downtime
Over the next two years, Rolls-Royce Power Systems expects to develop an equipment health monitoring system. It will capture data from its four-stroke medium-speed Bergen and MTU high-speed engines that will enable remote and autonomous vessel operations.
Under a project called Echo, Rolls-Royce Power Systems plans to ‘scale up’ its current engine condition monitoring technology to the system health level. To develop the system, Rolls-Royce Power Systems is working with German shipping company FRS Ferries, marine gearbox manufacturer ZF and IT company MAC System Solutions. FRS Ferries owns and operates monohull ropax and multi-hull, high-speed ferries that transport passengers, vehicles and cargo.
Under the Echo pilot project, the entire power and propulsion train of a high-speed ferry is being monitored, explains Rolls-Royce Power Systems director application engineering and automation marine and defence systems Kevin Daffey.
“The idea is to head off major engine problems before they become problems”
“What we’re trying to do,” says Mr Daffey, “is increase the vessel’s availability and fuel consumption using data taken from the engines, the gearbox and waterjets.” A key component of the system is the vibration sensors that have been added to the powertrain to learn more about the propulsion system and powertrain’s health.
Mr Daffey unveiled details of the project during his remarks at Riviera’s webinar, How Covid-19 will transform your vessel operations over the next 24 months, supported by Rolls-Royce Power Systems and part of a multi-week series of executive webinars.
Echo has its genesis in Rolls-Royce Power Systems’ use of machine learning techniques that underpin its engine health monitoring systems.
“For many years we’ve been developing the means to diagnose engine problems by producing alarms and alerts on local operating panels, but now we’re developing the techniques to prognose problems before they affect engine operations,” says Mr Daffey. Rolls-Royce Power Systems’ ultimate goal is to provide condition-based asset management based upon how an engine is actually operated.
Using machine learning techniques, Rolls-Royce Systems trains a neural network on three months of ‘good’ engine data from a testbed, using all the sensors available on that particular engine, explains Mr Daffey. The idea is to detect and identify anomalies. These anomalies are flagged and if previously identified, an alert is triggered at a remote operating centre that provides advice on a course of action to the engine operator.
If the anomaly has not been previously identified, it is passed to a team of experts who diagnose the root cause of the problem and advise action. This knowledge is then captured and archived in a classification library in order that it might be shared across all similar engine models.
“This means that our body of knowledge is continuously updated and captured,” says Mr Daffey. The identification of the anomaly not only benefits that particular customer, but also all customers of that particular engine class.
A key component in detecting anomalies is a knock sensor inside a gas engine or a particular cylinder. The sensor monitors the high-frequency spectrum over the entire cylinder cycle.
“It can ‘hear’ events like valves opening and closing and the combustion event,” says Mr Daffey. This Neural Observer learns the signature and detects noise levels of normal and abnormal operation.
In an example during the webinar, Mr Daffey showed that during normal operation engine cycles have a sound intensity that maintains an average level. If that level jumps up, Rolls-Royce Power Systems will alert the operator, but it does not intervene. Instead, it monitors the event and if the noise begins to increase and becomes a concern it recommends a course of action and repair.
“We are trying to increase the vessel’s availability and fuel consumption using data from the engines, gearbox and waterjets”
The idea is to head off major engine issues before they become problems, keeping a vessel operating smoothly. “This just gives an example of what we can do using machine learning to diagnose, making a timely intervention into the customer’s operation,” says Mr Daffey.
Mr Daffey says Rolls-Royce Power Systems want to develop a full power train health management solution that can be linked to its shore-based services to provide advice, insights and interventions. “We can then begin to look at the whole ship performance, or even fleet performance,” he says. “We can add ship data with external real-time data from tides, currents, weather and position of the vessel.”
Using this data, Rolls-Royce Power Systems can provide insights into operations in terms of fuel consumption, optimisation, timetable adherence and even the state of a vessel’s hull fouling.
“All of these insights help the operator make critical and timely decisions about their
equipment, their ship and their fleet,” says Mr Daffey.
At the moment, Rolls-Royce Power Systems envisages making its engine health technology available across all of its four-stroke engines, with the ability to retrofit the technology on existing models. “It’s available on all of our medium-speed Bergen engines and it’s something that we’re beginning to now look at for our high-speed engine range,” says Mr Daffey.
The machine learning technique is available across a large part of Rolls-Royce Power Systems’ Bergen engine range and has been operating for about three years, building up a library of anomalies for the medium-speed engine range both on ships and also on power generation operations. “We have just started to do this on our high-speed engines,” says Mr Daffey. Rolls-Royce is training the neural network on engine test data and will begin to roll the technology to its fleet operators when it is satisfied with the system’s library of knowledge.
“It takes time because obviously you’re looking for anomalies that occur, but the bigger the fleet that we can operate it on, the more we get the anomalies and then we get that data and we’ve got that body of knowledge,” he concludes.
Using these enablers, Rolls-Royce Power Systems can make decisions from the shore for remote operations “all the way through to autonomous operations where the ships are able to make their own decisions,” he points out. This is being developed under a system called the Artificial Chief Engineer. “It’s at a technology readiness level four at the moment, but we hope this will become a reality over the next 24 months to enable more remote and autonomous vessels of the future,” says Mr Daffey. Technology readiness levels (TRLs) are a measurement system to assess the maturity level of a particular technology, with TRL 1 being the least mature and TRL 9 the most mature. TRL 4 is equivalent to the lab environment.