I worked as a software engineer for an airline, where my team worked on a predictive maintenance project. As we know, both in airline operations and GA maintenance is carried out every N flight cycles. The goal of predictive maintenance is to identify possible issues with the engine or other parts before the actual planned maintenance, using alternative data rather than just the number of flight cycles. At the airline we used data from FDRs (egt, etc), faults and warnings data, as well as past maintenance log (replaced components, etc) to forecast pre-fault conditions.
I have this idea to try to apply the same concept for GA. I want to create a cloud-based service for flight data analysis, condition monitoring, predictive maintenance and maintenance tracking for GA. The idea is to use the same service for storing, analysing and sharing flight data with the maintenance crew.
I would like to hear what is your take on this, and whether anyone is interested in beta-testing such service.
For Flying Schools it might work. For high hours individual pilots with regular flying habits it might work.
For Syndicates like the one I joined 30 years ago, I think there is too much variation in use. Deiced runways, soft farm strips, rough runways and sea beach, have different wear effects on wheel components compared to hard runway free from chemicals, or smooth firm grass.
O200 engines are run to 1.5 times manufacturers TBO.
Let me run this past the mx chair in our club. We operate four Cessnas (plus soon a DA40), so this could be of interest.
Sounds good. Do you know Savvy MX (engine analysis)?
The tricky part of modern data use and software for „basic“ GA vs airlines is the lack of telemetry/reporting.
On a modern airliner it happens that maintenance get a fault report and even send a message while we are inflight „just for info, R2PCBS XYZ fault reported, might cause this and that, [destination] mx informed, …“
They know stuff like increasing turbine temperature before us.
Airlines save a lot of unnecessary expenses using „on condition“ mx vs cycles/hours/calendar. Maybe a concept worthy for smaller planes too?
So the data quality and quantity is based on reliable users manually uploading it. Some hardware in the plane connected to the internet (or your software connected to some API that uses ADSB data to record flight data) might be a start.
Snoopy wrote:
So the data quality and quantity is based on reliable users manually uploading it. Some hardware in the plane connected to the internet (or your software connected to some API that uses ADSB data to record flight data) might be a start.
I’ll second that. Extra hardware will also be needed because the predictive monitoring facilities in a typical GA plane are direly inadequate. Even simple engine analysers aren’t very widely adopted.
Are there any products at all?
Can a EDM for instance be set up to upload data via mobile internet automatically?
I smiled about the lack of telemetry and reporting. My aircraft is worth about 6 months of my ability to save money, I don’t carry hull insurance and my maintenance costs are so small that I couldn’t tell you what they are – it’s spare change and done on condition. Think I’ll be interested? 
It’s not the lack of data that’s the problem, it’s the lack of predictive operation. For this to be even slightly believable, the engine must have a rather advanced FADEC as a start. This is needed so that the engine always is operated within some known frame. Even if it has, this will not fix the problem of too little usage, which will spark all kinds of random events.
The idea of predictive maintenance is that by monitoring all engines (as an example), this data can be used (after some analysis, usually some machine learning or similar) for each engine individually. This is done to minimize overall maintenance cost. Still, the principle that it is cheaper to replace an item before it is broken, than after, is as true as ever. Adding reliability into this, not just the cost if unreliability, you will always end up with the result that it is better to replace a unit well before it breaks down.
The thing is, we don’t want to maintain them that way. We want to use them until something breaks, then fix them. We want to use them “on condition”. The reason is that since there is no cost related to unreliability (other than grief), this will be the cheapest method, by far. What we want, is only to be reasonable sure that the aircraft will fly trouble free the next hour.
Silvaire wrote:
I smiled about the lack of telemetry and reporting. My aircraft is worth about 6 months of my ability to save money, I don’t carry hull insurance and my maintenance costs are so small that I couldn’t tell you what they are – it’s spare change and done on condition. Think I’ll be interested?
I think this is just an American thing. My maintenance costs are no where near small for my EASA aircraft. The support from GA communities you get in the US is just incomparable to what we have here. In Europe we just have.. well EuroGA :) In my home base in Germany all other owners look at me like I am some sort of crazy person to dare changing my own spark plugs. I think the need for an online community or any type of digital services is much stronger in Europe than it’s in the US simply because there are so few opportunities to learn locally. We are mostly slaves to shops.
What would be the benefit of predictive maintenance for noncommercial GA?
In commercial operations (doesn‘t matter if airlines or manufacturing environments) the core benefit of predictive maintenance is avoiding unscheduled downtime as these cause significant disruptions (e.g. canceled flights) and cost (e.g. emergency/off hr. maintenance). Second benefit in aviation (not so much in manufacturing) is a reduction of maintenance cost due to on average longer cycles „on condition“. Another benefit could theoretically be avoiding catastrophic failures in flight – but as these are extremely rare events anyways, the benefit can also only be small.
In non-commercial GA ops, most of us run the expensive parts (props, engines, etc.) „on condition“ anyways so the savings potential is limited.
The economics of unscheduled maintenance are also quite different, as the impact of an unexpected downtime is not as big – if it doesn‘t happen somewhere far away at a different field which is again a quite rare event.
Therefore with the expectation that we would need to invest some money to make it happen (e.g. sensors, connectivity, etc.) the first question would be the benefit…
