Satavia

Contact: Dr. Adam Durant, CEO; adam.durant@satavia.com

Proposed with Microsoft

Predictive aircraft maintenance

Local company Satavia helps airlines to schedule engine maintenance based on the amount of exposure the components have had to air pollution, dust, volcanic eruptions and other factors. They have large data sets which could be used to train predictive models that might be added to the Microsoft Cortana Intelligence Solution Template Playbook for predictive maintenance in aerospace. You will need to deliver a data ingestion architecture for a range of global data, and also demonstrate an aircraft maintenance scheduling application that applies the results.

Feedback:

Unfortunately, I don’t think that second year undergraduates have any significant technical knowledge of time-series analysis.

They should be familiar with basic statistical regression, but will not have much understanding of machine learning techniques until third year. I don’t believe that we teach much in the way of frequency domain analysis or autocorrelation methods in the Computer Science degree - these are topics that would be more typical of the signal processing courses taught in Electrical Engineering.

I’m afraid I can’t advise on which features of Cortana would be most relevant to your problem, as I have not used the product myself. However, Google suggests that the “Cortana Intelligence Gallery” might contain relevant stuff: https://gallery.cortanaintelligence.com/Experiment/Time-Series-Forecasting-8

If I understand correctly, this example simply uses Azure to host an R script, but it should be relatively trivial for students to follow these instructions (if they found them!)

So to make this into an interesting design challenge, we would need to give some thought to technical infrastructure, delivery mechanisms etc, that are appropriate for your user base. It’s unlikely that any members of the team would have any prior knowledge of the aircraft maintenance industry, so we would need to come up with a relatively trivial use case that is sufficiently representative of the business issues to be interesting.

Original suggestion:

Environmental factors in the atmosphere, like dust, ice, sulphur, and volcanic, accelerate wear of aircraft components. Modern aircraft generate large volumes (several GB) of data per flight and digital predictive analytics technology has great potential to support asset health monitoring, and disrupt traditional maintenance and flight planning processes. The solution combines high spatio-temporal environmental factor analysis with machine-learning to provide advanced risk-based decision-making capability.

Data driven model predictions (in this case Satavia’s environmental factor analyses) inevitably contain some level of uncertainty, due to the reliance on a finite (typically small) sample of direct observational measurement data, or coarse resolution input meteorological data used to drive the NWP model. The project will demonstrate advanced environmental data analytics capability through the application of artificial intelligence (AI) techniques such as reinforcement learning (RL), to develop a framework for sequential decision-making under uncertainty. The objectives will include:

1. Improve the skill of the NWP-based model using training data acquired from a local in situ measurement network; 2. Correlate aircraft environmental factor exposure to component wear rates and proxy data provided by engine health management; 3. Develop autonomous decision-making capability to adjust aircraft flight plans and engine maintenance plans in near-real-time.

Aircraft original equipment manufacturers (OEM), operators, and maintenance repair organisations now increasingly share maintenance liability through ‘power-by-the-hour’ services for the lifetime of the aircraft. Unscheduled maintenance costs the industry billions of dollars each year (e.g., in 2016 sulphurous air pollution cost Rolls-Royce >£65M) and causes disruption to airline operator flight schedules. Customers currently implement highly conservative maintenance schedules at pre-defined maintenance frequency. Predicting when to do the maintenance based on environmental exposure could make huge savings for the industry. Satavia’s data-as-a-service enables aircraft operators to modify aircraft flight plans and scheduling to extend engine lifetime, while engine manufacturers can proactively adjust maintenance plans to minimise unscheduled maintenance. Satavia is currently working on a series of ‘proof-of-value’ projects with a UK-based aircraft engine manufacturer.