Multicopter online rotor fault diagnosis system
A multicopter rotor fault diagnosis system includes a multicopter, a plurality of local sensors, and a controller. The multicopter includes a body, a plurality of booms secured to the body, and a plurality of rotors. Each boom has a proximal end secured to the body and extends radially outward to a distal end. Each rotor is secured to the distal end of a respective boom. Each sensor is secured to a respective boom and is positioned a distance from the body. The local sensors are configured to measure out-of-plane strain of the booms and to continuously generate boom strain signals. The controller is in communication with the body and the local sensors. The controller is configured to receive the continuously generated boom strain signals and to perform a pattern recognition algorithm to simultaneously detect and identify faults associated with the rotors.
1 . A rotor fault diagnosis system for a multicopter that comprises a plurality of booms and a plurality of rotors, each of which are secured to a distal end of a respective one of the plurality of booms, the system comprising:
a plurality of local sensors, each of which are secured to a respective one of the plurality of booms; the plurality of local sensors are configured to measure out-of-plane strain of the plurality of booms and to continuously generate boom strain signals; and
a controller in communication with the plurality of local sensors; the controller is configured to receive the continuously generated boom strain signals and perform a pattern recognition algorithm to simultaneously detect and identify faults associated with the plurality of rotors.
2 . The system of claim 1 , wherein the multicopter further comprises a body to which a proximal end of each of the plurality of booms is secured and from which each of the plurality of booms extends; and
wherein each of the plurality of local sensors are secured to the proximal end of the respective one of the plurality of booms and positioned adjacent the body and at a distance from the body.
3 . The system of claim 2 , wherein the controller is a component of an onboard flight computer secured to the body of the multicopter.
4 . The system of claim 1 , wherein each of the plurality of local sensors is a strain gauge.
5 . The system of claim 1 , wherein the pattern recognition algorithm comprises a perceptron configured to receive as input a mean value of the boom strain signals over a duration and to output one of a plurality of linearly separable classes associated with a flight state of the multicopter.
6 . The system of claim 5 , wherein the duration is 1 second and the mean value of the boom strain signals is recalculated every 0.1 seconds based on the continuously generated boom strain signals.
7 . The system of claim 5 , wherein the plurality of linearly separable classes comprises healthy flight with wind gusts and with turbulence, healthy flight with gusts and without turbulence, healthy flight with turbulence and without wind gusts, healthy flight without turbulence and without wind gusts, and faults for each of the plurality of rotors.
8 . The system of claim 1 , wherein if the pattern recognition algorithm detects and identifies a fault associated with any one of the plurality of rotors, the controller is configured to perform a linear regression algorithm to quantify the fault of the rotor.
9 . The system of claim 8 , wherein the linear regression algorithm is configured to receive as input the out-of-plane strain measured by the local sensor of the boom associated with the faulty rotor divided by a mean out-of-plane strain of the plurality of booms and to output a predicted magnitude of degradation of the faulty rotor and a confidence interval of the predicted magnitude.
10 . The system of claim 1 , wherein if the pattern recognition algorithm detects and identifies a fault associated with any one of the plurality of rotors, the controller is configured to generate and communicate at least one corrective control signal to at least one of the non-faulty rotors to return the multicopter to a healthy flight state.
11 . A method of diagnosing rotor fault in a multicopter that comprises a plurality of booms and a plurality of rotors, each of which are secured to a distal end of a respective one of the plurality of booms, the method comprising:
continuously measuring, via a plurality of local sensors, out-of-plane strain of the plurality of booms to continuously generate boom strain signals, each of the plurality of local sensors are secured to a respective one of the plurality of booms;
receiving, at a controller in communication with the plurality of local sensors, the continuously generated boom strain signals; and
performing, via the controller, a pattern recognition algorithm to simultaneously detect and identify faults associated with the plurality of rotors.
12 . The method of claim 11 , wherein the multicopter further comprises a body to which a proximal end of each of the plurality of booms is secured and from which each of the plurality of booms extends; and wherein each of the plurality of local sensors are secured to the proximal end of the respective one of the plurality of booms and positioned adjacent the body and at a distance from the body.
13 . The method of claim 12 , wherein the controller is a component of an onboard flight computer secured to the body of the multicopter.
14 . The method of claim 11 , wherein each of the plurality of local sensors is a strain gauge.
15 . The method of claim 11 , wherein the step of performing the pattern recognition algorithm comprises:
calculating an initial mean value of the boom strain signals received over an initial 1 second duration;
inputting the mean value of the boom strain signals into a perceptron trained using data associated with a plurality of multicopter flight states and rotor fault scenarios; and
outputting one of a plurality of linearly separable classes associated with a flight state of the multicopter.
16 . The method of claim 15 , wherein if the perceptron outputs a linearly separable class associated with a healthy flight state of the multicopter, the step of performing the pattern recognition algorithm further comprises:
calculating an updated mean value of the boom strain signals received over an updated 1 second duration shifted 0.1 seconds forward from the initial 1 second duration; and
repeating the inputting and outputting steps to determine an updated flight state of the multicopter based on the updated mean value of the boom strain signals.
17 . The method of claim 15 , wherein the plurality of linearly separable classes comprises healthy flight with wind gusts and with turbulence, healthy flight with gusts and without turbulence, healthy flight with turbulence and without wind gusts, healthy flight without turbulence and without wind gusts, and faults for each of the plurality of rotors.
18 . The method of claim 11 , wherein if the pattern recognition algorithm detects and identifies a fault associated with any one of the plurality of rotors, further comprising:
performing, via the controller, a linear regression algorithm to quantify the fault of the rotor.
19 . The method of claim 17 , wherein the step of performing the linear regression algorithm comprises:
inputting the out-of-plane strain measured by the local sensor of the boom associated with the faulty rotor divided by a mean out-of-plane strain of the plurality of booms into a linear regression model trained using data associated with a plurality of multicopter flight states and rotor fault levels; and
outputting a predicted magnitude of degradation of the faulty rotor and a confidence level of the predicted magnitude.
20 . The method of claim 11 , wherein if the pattern recognition algorithm detects and identifies a fault associated with any one of the plurality of rotors, further comprising:
generating, via the controller, at least one corrective control signal; and
communicating the at least one corrective control signal to at least one of the non-faulty rotors to return the multicopter to a healthy flight state.
21 . A multicopter, comprising:
a plurality of booms;
a plurality of rotors, each of which are secured to a distal end of a respective one of the plurality of booms; and
a rotor fault diagnosis system, comprising:
a plurality of local sensors, each of which are secured to a respective one of the plurality of booms; the plurality of local sensors are configured to measure out-of-plane strain of the plurality of booms and to continuously generate boom strain signals; and
a controller in communication with the plurality of local sensors; the controller is configured to receive the continuously generated boom strain signals and perform a pattern recognition algorithm to simultaneously detect and identify faults associated with the plurality of rotors.
22 . The multicopter of claim 21 , further comprising a body to which a proximal end of each of the plurality of booms is secured and from which each of the plurality of booms extends; and
wherein each of the plurality of local sensors are secured to the proximal end of the respective one of the plurality of booms and positioned adjacent the body and at a distance from the body.
23 . The multicopter of claim 21 , wherein each of the plurality of local sensors is a strain gauge.
24 . The multicopter of claim 21 , wherein the pattern recognition algorithm comprises a perceptron configured to receive as input a mean value of the boom strain signals over a duration and to output one of a plurality of linearly separable classes associated with a flight state of the multicopter.
25 . The multicopter of claim 24 , wherein the plurality of linearly separable classes comprises healthy flight with wind gusts and with turbulence, healthy flight with gusts and without turbulence, healthy flight with turbulence and without wind gusts, healthy flight without turbulence and without wind gusts, and faults for each of the plurality of rotors.
26 . The multicopter of claim 21 , wherein if the pattern recognition algorithm detects and identifies a fault associated with any one of the plurality of rotors, the controller is configured to perform a linear regression algorithm to quantify the fault of the rotor.
27 . The multicopter of claim 26 , wherein the linear regression algorithm is configured to receive as input the out-of-plane strain measured by the local sensor of the boom associated with the faulty rotor divided by a mean out-of-plane strain of the plurality of booms and to output a predicted magnitude of degradation of the faulty rotor and a confidence interval of the predicted magnitude.
28 . The multicopter of claim 21 , wherein if the pattern recognition algorithm detects and identifies a fault associated with any one of the plurality of rotors, the controller is configured to generate and communicate at least one corrective control signal to at least one of the non-faulty rotors to return the multicopter to a healthy flight state.
29 . The multicopter of claim 21 , wherein the controller is a component of an onboard flight computer secured to the body of the multicopter.