Generalized Structural Analysis Algorithm with Learning Capacity
George Lee Principal Investigator
MetadataShow full item record
This proposal presents a novel idea to integrate conventional finite element methods with neural networks to achieve a fast, time- saving, and highly accurate computational scheme for dynamic analysis of complex structures. A one-year proof-of-concept study will be conducted. By presenting a network with different complicated structures, a computational system will be developed which combines a finite element program and a neural net that can handle structures with a stiffness matrix up to 500 x 500 under static or dynamic loading. The above model, if proven successful, has a potential to revolutionize many concepts in structural/earthquake engineering. Different neural net modules - each of which is trained to perform certain tasks such as analysis, design, code provisions, etc. - may be developed, and these modules will be more reliable than and can out perform human engineers.