Quick analysis of residual stress and distortion in cast aluminum components
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06N-003/10
G06N-003/08
G06F-017/10
C22F-001/04
G06F-017/50
출원번호
US-0295404
(2014-06-04)
등록번호
US-9489620
(2016-11-08)
발명자
/ 주소
Wang, Qigui
Wang, Yucong
Gao, Zhiqiang
Quan, Zhibin
출원인 / 주소
GM Global Technology Operations, LLC
인용정보
피인용 횟수 :
2인용 특허 :
8
초록▼
A computer-implemented system and method of rapidly predicting at least one of residual stress and distortion of a quenched aluminum casting. Input data corresponding to at least one of topological features, geometrical features and quenching process parameters associated with the casting is operate
A computer-implemented system and method of rapidly predicting at least one of residual stress and distortion of a quenched aluminum casting. Input data corresponding to at least one of topological features, geometrical features and quenching process parameters associated with the casting is operated upon by the computer that is configured as a neural network to determine output data corresponding to at least one of the residual stress and distortion based on the input data. The neural network is trained to determine the validity of at least one of the input data and output data and to retrain the network when an error threshold is exceeded. Thereby, residual stresses and distortion in the quenched aluminum castings can be predicted using the embodiments in a tiny fraction of the time required by conventional finite-element based approaches.
대표청구항▼
1. A computer-implemented method of rapidly predicting at least one of residual stress and distortion of a quenched aluminum casting, said method comprising: receiving into said computer input data corresponding to at least one of topological features, geometrical features and quenching process para
1. A computer-implemented method of rapidly predicting at least one of residual stress and distortion of a quenched aluminum casting, said method comprising: receiving into said computer input data corresponding to at least one of topological features, geometrical features and quenching process parameters associated with said casting; andoperating said computer as a neural network to determine output data corresponding to at least one of said residual stress and distortion based on said input data, said operating configured to train said network to determine the validity of at least one of said input data and output data and to retrain said network when an error threshold is exceeded. 2. The method of claim 1, wherein said input data corresponding to at least one of topological features, geometrical features and quenching process parameters associated with said casting comprises input data corresponding to each of said topological features, geometrical features and quenching process parameters. 3. The method of claim 2, wherein said geometrical features include at least the Gaussian curvature that is determined by the formula: k(vi)=3×{2π-∑vj,vk∈n(vi)⋀eij=ejk=eki=1θ(vi,vj,vk)}∑vj,vk∈n(vi)⋀eij=ejk=eki=1A(vi,vj,vk). 4. The method of claim 3, wherein said geometrical features comprise at least a maximum dihedral angle that is calculated using the formula: θ(vi)=maxvj∈n(vj){θ(ei,j)}. 5. The method of claim 2, wherein said quenching process parameters comprises node temperature changes that take place during a quench of said casting. 6. The method of claim 2, wherein said topological features include at least a set of nearest neighbor nodes that are determined by a breadth-first-search using the following function: N(vi)=n( . . . n(n(vi))). 7. The method of claim 2, wherein said input data is received in nodal form based on a mesh simulation of said casting. 8. The method of claim 1, wherein said network is defined by at least one input layer, at least one hidden layer and at least one output layer such that training of said network is achieved by weighting values calculated by at least one of said hidden layer and said output layer within said network. 9. The method of claim 8, wherein weighting values may be changed by said network during said operating. 10. The method of claim 9, wherein at least one of said input layer and said output layer performs a linear processing step on respective data received therein. 11. The method of claim 8, wherein said hidden layer performs a nonlinear processing step on data received therein. 12. The method of claim 1, wherein said rapidly predicting comprises outputting indicia of at least one of residual stress and distortion of said casting in substantially real-time. 13. The method of claim 12, wherein said substantially real-time is no more than about ten minutes. 14. A neural network system to provide substantially real-time prediction of at least one of a residual stress and distortion of a quenched aluminum casting, the system comprising: an input configured to receive data relating to topological features, geometrical features and quenching process parameters associated with said casting;an information output configured to convey data relating to at least one of the residual stress and distortion of said casting predicted by the system;a processing unit; anda computer-readable medium comprising a computer-readable program code embodied therein, said computer-readable medium cooperative with said input, output and processing unit to operate as an artificial neural network to provide said substantially real-time prediction. 15. The system of claim 14, wherein said program code is configured to operate over at least one input layer, at least one hidden layer and at least one output layer in said network such that training of said network is achieved by weighting values calculated by at least one of said hidden layer and said output layer within said network. 16. The system of claim 14, wherein said geometrical features include at least the Gaussian curvature that is determined by the formula: k(vi)=3×{2π-∑vj,vk∈n(vi)⋀eij=ejk=eki=1θ(vi,vj,vk)}∑vj,vk∈n(vi)⋀eij=ejk=eki=1A(vi,vj,vk). 17. The system of claim 16, wherein said geometrical features comprise at least a maximum dihedral angle that is calculated using the formula: θ(vi)=maxvj∈n(vi){θ(ei,j)}. 18. The system of claim 14, wherein said quenching process parameters comprises node temperature changes that take place during a quench of said casting. 19. The system of claim 14, wherein said topological features include at least a set of nearest neighbor nodes that are determined by a breadth-first-search using the function N(vi)=n( . . . n(n(vi))). 20. The system of claim 14, wherein said input data is received in nodal form based on a mesh simulation of said casting.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (8)
Sullivan Michael Scott ; Brost Ronald David ; Chen Yaobin ; Eberhart Russell Carley, Method and apparatus for determining battery state-of-charge using neural network architecture.
Moore-Ede Martin C. ; Trutschel Udo E. ; Guttkuhn Rainer ; Heitmann Anneke M., Method of and apparatus for evaluation and mitigation of microsleep events.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.