Fault tolerant control

Prof Colin Jones

The squeaking wheel doesn’t always get the grease. Sometimes it gets replaced.
– Vic Gold

I’ve looked at two topics under the heading of ‘Fault tolerance’:

The first was during my master’s thesis, in which I studied fault identification of hybrid systems and in particular hydraulic circuits. I haven’t looked at this topic for many years now, but you can find details on the work in the publications section below.

The second was an investigation into reconfigurable flight control of civilian aircraft. In particular, we studied a Boeing 747 crash that happened in Amsterdam in 1994. In this incident, two of the engines fell off and took a goodly part of the wing surface with it. It makes an interesting case study, since the pilots were able to keep the aircraft in the air for a full 10 minutes before the crash, which is a reasonable amount of time for an online fault-tolerant system to take effect.

The approach that we took in this case is (not surprisingly) based on model-predictive control. Given a re-identified model of the damaged aircraft, we were able to control a simulated damaged aircraft through its final landing maneuvers. This work was done as an early part of the GARTEUR action group AG16 (which was just getting rolling as I completed my PhD). Many more interesting links and details on their work can be found here.

This is a simulation of the MPC-based fault-tolerant controller. The three planes are as follows:

  • White: Functional airplane, standard controller
  • Red : Damaged airplane, standard controller
  • Green : Damaged airplane, fault-tolerant controller

Fault Tolerant Control Publications

[epfl_infoscience url=”https://infoscience.epfl.ch/publication-exports/2155/?ln=en”]