Royal Navy eyeing cost savings with new failure forecasting neural network

The Royal Navy is looking to achieve cost savings with a new health management and risk forecasting tool which would use data science and analytical modelling to predict potential defects or failures.

The two-stage recurrent neural network is being developed by a company called decisionLab and is being sponsored by Joint Forces Command and funded by the MOD’s Defence Innovation Fund through the Defence and Security Accelerator’s (DASA) fast track ‘Revolutionise the human information relationship for Defence’ competition.

Originally developed for the civilian aviation market, this research funding has enabled the product to be redeveloped to be suitable for the military market.

It’s hoped it will in time provide an insight into the future, allowing maintenance engineers to view the status of their systems and the predicted health of that system a day, week, or even a fortnight in advance.

The Royal Navy has invested £150,000 in the development of this neural network for exploitation on-board a Type 45 destroyer, and pull through onto the Type 26 – if proven successful.

At present, decisionLab is training their neural network on 1.8 billion lines of Type 45 platform management system data. Each day the system gets smarter and more capable, and under current development plans this system will be installed onboard HMS DIAMOND for a trial in the summer.

It will allow the user to validate system assumptions and help contextualise events to further train and improve the model.