Sometimes you just know what the answer should be.
This is the simple basis for the Expected Value test and these come in two flavours.
Strawberry and Vanilla might be more interesting than Discrete and Analogue, but we have to do the dull stuff sometimes.
Discrete Expected Values
Let’s test the idea so you know what I mean.
Question. When an aircraft is taxiing, should the wheels be:
- Don’t know
- I have skids and can hover taxi so who needs wheels?
So for most people, (b) was the answer and we can use this to check that the undercarriage switches are showing the right value on the ground.
Add to this the second question:
When an aircraft is in the cruise, should the wheels be:
- Don’t know
- I rarely cruise
And you can see how we can test both states and confirm that the discretes for undercarriage operation are working correctly.
Analogue Expected Values
The same principle applies as for discretes, except that we might be interested in looking for the average, minimum or maximum value. Here are a couple of examples to illustrate how this works.
During taxi the average Acceleration Longitudinal signal should be less than 0.1g. We use taxi because in flight the aircraft can sit with a significant pitch attitude which will resolve the earth’s gravitational field into the longitudinal signal. Also, we average over the whole taxi phase so that accelerations and decelerations correct each other.
Other examples: On landing, the maximum brake pressure will be less than 3,500 psi.
On the ground, the minimum control deflection will be between -6 and -40 deg, and the maximum will be between +6 and +40. This reflects the fact that the crew should do a full and free check, but often don’t.
On the approach, the decision height selected will be between 50ft and 350ft.
Like all these data cleansing tests, the principle is the same as an expert diligently checking the data. The only difference is that a computer can apply tests quickly and repeatably, without losing concentration and without tiring.