Bridge Surface Crack Detection Method Based On Kernel Locality Sensitive Discriminant Analysis And Distributed Strain

Crack detections are of great significance to the safe operation of bridges.The distributed optical fiber sensor(DOFS)can measure the distributed strain over bridge surface along the length of the fiber at once with a spatial resolution of a centimeter level.The strains are sensitive to cracks.Therefore,DOFS provides a potential solution for crack detection.However,DOFS-based detection methods rarely take into consideration that there exists correlation between space-adjacent strain points.It is necessary to explore such correlation in order to improve crack detection efficiency.It can provide improved solutions for DOFS-based bridge crack detection.It has important theoretical and applicable significance to carry out the approach research.A bridge surface crack detection method is designed and implemented based on kernel locality sensitive discriminant analysis(KLSDA)and the distributed strain.The effectiveness of the proposed method is verified by experiments.The crack detection problem is considered as a binary classification problem.Firstly,DOFS are used to measure the distributed strain over the bridge surface,the measurements are standardized.Secondly,the multiple measurements of each sampling point are regarded as a time series.A KLSDA algorithm is used to project the time series to a lower dimensional subspace where its distinguishable feature representations are obtained.Then,a nearest neighbor classifier is designed which takes the feature representations and outputs the crack detection results.Finally,in-laboratory simulation steel beam experiments and in-field bridge experiments are carried out.The experimental results show that the proposed method can effectively detect cracks on the bridge surfaces.The work mainly includes the following three parts:1.The acquisition and preprocessing methods on the distributed strain are studied.The DOFS system is exploited to measure the distributed strain,which are then standardized.The independently measured distributed strain are transformed and considered as a high-dimensional time series spatially distributed over the fiber sensor.2.KLSDA-based crack detection methods are studied.A KLSDA-based detection method is proposed for the structural surface cracks.The improvement of the proposed method is verified by comparing it with traditional global and local dimensionality reduction algorithms also with LSDA algorithm..3.The experimental programs are implemented on an in-lab steel beam and an in-field bridge.The experimental results illustrate that the proposed method can accurately detect cracks on the structural surfaces.