Neural network evaluation of pile foundation damage recognition for high-pile piers based on cost expectation
There have been many achievements in the theoretical research on the application of neural networks to structural damage detection.However,due to the lack of accuracy and completeness of basic data,the generalization ability of networks was limited,and the traditional network evaluation model was often unable to evaluate whether such neural networks have practical application value,thus the practical application of this technique in the field of structural damage identification in port engineering was limited.In this paper,the cost expectation of using neural network damage identification results in engineering is proposed.This cost expectation considered the economic loss of further testing caused by misdiagnosis in practical engineering applications,as well as the significant risk of structural instability and failure caused by missed diagnosis.In addition,a new damage identification network evaluation method based on cost expectationfor high pile wharf is established.This evaluation method took the missed diagnosis rate and misdiagnosis rate of damage identification as the evaluation index,which improved the traditional method of evaluating the feasibility of actual network engineering only from the perspective of recall rate.
damage identificationneural network evaluationcost expectationshigh pile wharftraining set
万方数据
维普
