“In order to exploit the full potential of these electric aircraft, you need an intelligent control system that improves their robustness and especially their resilience against a variety of faults,” says Soon-Jo Chung, Bren professor of control and dynamic systems at Caltech and Senior Research Scientist at JPL, which Caltech manages for NASA. “We have developed such a fault-tolerant system, crucial for safety-critical autonomous systems, and it introduces the idea of virtual sensors for detection of any fault using machine learning and control methods adaptive. »
Chung’s group used a model of his autonomous flying ambulance to test his NFFT control method. Image credit: Caltech
Multiple rotors mean many possible points of failure
Engineers build these hybrid electric planes with multiple propellers, or rotors, in part for redundancy purposes: If one rotor fails, there are enough working motors remaining to stay in flight. However, to reduce the energy required to make flights between urban areas (say 10 or 20 miles), the craft also needs fixed wings. However, having both rotors and wings creates many possible points of failure in each aircraft. And that leaves engineers wondering how best to detect when something is wrong with part of the vehicle.
Engineers could include sensors for each rotor, but even that wouldn’t be enough, Chung says. For example, an airplane with nine rotors would need more than nine sensors, since each rotor might need one sensor to detect a failure in the rotor structure, another to notice if its engine stops working , and yet another to alert in the event of a signal wiring problem. occurs. “You could potentially have a distributed system of highly redundant sensors,” Chung says, but that would be expensive, difficult to manage, and would increase the weight of the aircraft. The sensors themselves could also fail.
With NFFT, Chung’s group proposed an alternative and innovative approach. Building on previous efforts, the team developed a deep learning method that not only reacts to strong winds but also detects, on the fly, when the aircraft has suffered an onboard failure. The system includes a neural network pre-trained on real flight data, then learns and adapts in real time based on a limited number of changing parameters, including an estimate of the operating efficiency of each rotor of the plane at some point. time.
“This does not require any additional sensors or hardware for fault detection and identification,” says Chung. “We simply observe the behavior of the plane: its attitude and position over time. If the aircraft deviates from its desired position from point A to point B, NFFT can detect that something is wrong and use the information it has to compensate for that error.
And the correction occurs extremely quickly: in less than a second. “As you fly the plane, you can really feel the difference that the NFFT makes in maintaining the controllability of the plane when an engine fails,” says scientist Matthew Anderson, author of the paper and pilot who helped carry out the flight tests. “The real-time control overhaul makes it seem like nothing has changed, even if one of your engines just stopped working. »
Overview of virtual sensors
The NFFT method relies on real-time control signals and algorithms to detect where a fault is located. So Chung says it can provide any type of vehicle with essentially free virtual sensors to detect problems. The team has primarily tested the control method on aerial vehicles it is developing, including the Autonomous Flying Ambulance, a hybrid electric vehicle designed to quickly transport injured or sick people to hospitals. But Chung’s group has tested a similar, fault-tolerant control method on land vehicles and plans to apply NFFT to boats.
Written by Kimm Fesenmaier
Source: Caltech
Originally published in The European Times.
source link eu news
