A long time ago I explained that our analysis of flight data is based upon the concept of a Key Point Value, where an aspect of a flight is reduced to a single number. For example, the airspeed at takeoff or the maximum angle of bank during a flight. These are analogue measurements, that is, they vary continuously across a range of values and have a characteristic distribution, the shape depending upon the measurement itself.
In this blog I will look at two cases where the obvious thing is to produce a binary On/Off answer, but we still use an analogue KPV to provide measurement of the flight characteristics.
A customer asked for an event which would trigger if the aircraft was “Not Established on ILS Approach between 500 ft – 200 ft AAL”. OK, so they knew the name of the event and had a clear understanding of how this would map onto their Standard Operating Procedures (SOPs) but the implication was that the KPV would be a binary output of whether the ILS was established within this height range or not.
The drawback of a binary measurement like this, while it would be possible to generate, is that it provides no insight into what is happening. No view of the iceberg below the surface. We therefore produced a KPV which represents the highest altitude at which the ILS stops being established, and set a threshold for the event at 500ft. In this way, approaches that were only established down to, say, 750ft, would trigger an event while those that were established down to 350ft would not.
Boeing Hard Landing
A recent amendment to the Boeing Maintenance Manual (MM) for some types specifies when certain landing gear inspections have to be performed. The criteria is given on a chart which brings together:
- Normal acceleration on landing
- Normal acceleration sample rate
- Roll attitude in the second before landing
- Gross weight at touchdown
Here is the chart we have to implement:
Again, the obvious output is an Inspection Required / No Inspection Required status, but this is not how KPVs are designed.
From the chart, we can compute the load factor for which an inspection is required, and we simply subtract this from the normal acceleration at landing. The difference is then negative if no inspection is required, and positive if an inspection is needed. Setting a maintenance event threshold of zero makes the event trigger when an inspection is needed.
The advantage of this technique is that maintainers can judge how closely the aircraft is operating to the inspection threshold. It also opens the potential for more advanced analysis. For example, maintenance could identify the airfields most likely to see a hard landing then make sure that the tools required to carry out the inspection are available there.
I would normally provide example charts, but these two cases are still completing their test phase before being put into service, and so no live data is available yet. Here’s what a distribution with a threshold looks like, with infrequent exceedances. I’m sure you get the idea!