IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0709787
(2007-02-23)
|
등록번호 |
US-7440857
(2008-10-21)
|
우선권정보 |
FR-06 09992(2006-11-15) |
발명자
/ 주소 |
|
출원인 / 주소 |
|
대리인 / 주소 |
|
인용정보 |
피인용 횟수 :
1 인용 특허 :
4 |
초록
▼
The invention relates to a method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ANN), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade; the network (ANN) is a supervised co
The invention relates to a method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ANN), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade; the network (ANN) is a supervised competitive learning network (SSON, SCLN, SSOM) having an input to which vibration spectral data measured on the rotorcraft is applied, the network outputting data representative of which rotor blade presents a defect or an adjustment error or data representative of no defect, and where appropriate data representative of the type of defect that has been detected.
대표청구항
▼
What is claimed is: 1. A method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ANN), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade, wherein the network (ANN) is a superv
What is claimed is: 1. A method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ANN), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade, wherein the network (ANN) is a supervised competitive learning network having an input to which vibration spectral data measured on the rotorcraft is applied, the network outputting data representative of which rotor blade presents a defect or an adjustment error or data representative of no defect, and where appropriate data representative of the type of defect that has been detected in which a learning algorithm is used of the supervised self-organizing network type, and the output space is one of the group consisting of i) a square mesh, ii) a hexagonal mesh, and iii) an irregular mesh. 2. A method according to claim 1, in which the data applied to the input of the network is a sequence of vibration modulus and phase data corresponding to harmonics of the rotation frequency of the rotor. 3. A method according to claim 1, in which the supervised competitive learning network includes a layer of competitive neurons given synaptic weights, the number of said weights being equal to the dimension of the spectral data vectors used. 4. A method according to claim 3, in which the number of competitive neurons is greater than the dimension of the spectral data vectors. 5. A method according to claim 3, in which, from the competitive neurons, an elected neuron is determined with which the associated synaptic weight vector is the closest to a spectral data vector presented as input, in particular by calculating a Euclidean distance between said vectors. 6. A method according to claim 1, in which, in order to supervise learning, data is added to each spectral data vector used for network learning, said added data being representative of the membership of the vector in question to a class corresponding to a determined type of defect or adjustment error. 7. A method according to claim 1, in which a supervised vector quantization algorithm (LVQ) is used for network learning, in particular an algorithm selected from the following algorithms: LVQ1, OLVQ1, LVQ2, LVQ2.1, LVQ3, MLVQ3, GLVQ, DSLVQ, RLVQ, GRLVQ. 8. A method according to claim 1, in which the dimension of the output space of the network is equal to two. 9. A method according to claim 1, in which the dimension of the output space of the network is equal to three. 10. A method according to claim 1, to supervise learning, a partitioning of the output space into subspaces is defined and the weights of the competitive neurons are modified as a function of their membership to one or another of these subspaces. 11. A method according to claim 10, in which the partitioning is substantially regular, radial, and centered, and the number of subspaces is equal to the number of classes of defect or adjustment error plus one. 12. A method according to claim 10, in which, in order to define the partitioning, the coordinates in the output space are determined for a setpoint neuron associated with a class of defect or adjustment error. 13. A method according to claim 10, in which, to supervise learning, a neighborhood function is centered on the neuron corresponding to the barycenter in the output space of the elected neuron and of the setpoint neuron corresponding to the class of the current spectral data vector. 14. A method according to claim 10, in which, to supervise learning, competition is restricted to the neurons of the subspace corresponding to the class of the current spectrum data vector. 15. A method according to claim 1, in which a plurality of networks are used connected in series, and including at least one supervised competitive learning network. 16. A method according to claim 1, in which a plurality of networks are used connected in parallel and including at least one supervised competitive learning network. 17. A method according to claim 1, in which a plurality of redundant networks are used including at least one supervised competitive learning network. 18. A method according to claim 1, in which: a supervised first competitive learning network is used to determine data identifying at least one defective or out-of-adjustment blade; and a non-supervised second competitive learning network is used to determine data identifying at least one defect or adjustment error of the defective or out-of-adjustment blade. 19. A method according to claim 1, in which an irregular mesh is used in the output space. 20. A method according to claim 19, in which a mesh is used presenting coordinates defined in the set of real numbers. 21. A system for detecting and identifying a defect or adjustment error of a rotorcraft rotor, the system comprising: a member for reading a data medium and organized to read data of measurements taken on the rotorcraft; a database containing reference vibratory signature data for the rotorcraft; a device for processing the measurement data from the time domain to the frequency domain, which device is connected to the read member in order to receive the measurement data therefrom and to output vibratory signatures for analysis; and a calculation member connected to the database and to the processing device and programmed to perform the operations of a method according to claim 1. 22. A method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ANN), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade, wherein the network (ANN) is a supervised competitive learning network having an input to which vibration spectral data measured on the rotorcraft is applied, the network outputting data representative of which rotor blade presents a defect or an adjustment error or data representative of no defect, and where appropriate data representative of the type of defect that has been detected in which a learning algorithm is used of the supervised self-organizing network type, and in which a square mesh is used in the output space. 23. A method of detecting and identifying a defect or an adjustment error of a rotorcraft rotor using an artificial neural network (ANN), the rotor having a plurality of blades and a plurality of adjustment members associated with each blade, wherein the network (ANN) is a supervised competitive learning network having an input to which vibration spectral data measured on the rotorcraft is applied, the network outputting data representative of which rotor blade presents a defect or an adjustment error or data representative of no defect, and where appropriate data representative of the type of defect that has been detected in which a learning algorithm is used of the supervised self-organizing network type, and in which a hexagonal mesh is used in the output space.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.