FESN Bridge Structural Condition Evaluation Method Based On Multidimensional Data

Effectively assessing the health status of bridges is of great significance to the safe operation of bridges.With the wide application of bridge structure health monitoring,various types of sensors are placed on the bridge structure,with the long-term operation of the system,large-scale high-dimensional time series data are collected.The data collected by a single sensor at a certain position of the bridge is single-dimensional time series data,which only reflects the state of a local position in the bridge structure,but cannot reflect the structural health of the whole bridge.The high-dimensional time series data monitored at each position of the bridge can fully describe the overall situation of the bridge.Therefore,effectively analyzing such multi-dimensional time series data and evaluating its state is an urgent problem to be solved in bridge state assessment.Based on this,this thesis presents a bridge structure state assessment method for processing multi-dimensional data.In this thesis,a method for evaluating the health status of bridge structures based on multi-dimensional time series data is proposed.The multi-dimensional time series data of bridges are analyzed and processed by Functional Echo State Network(FESN).This method utilizes the non-linear mapping ability and storage capacity of the storage layer to accumulate the time characteristics of each input data,and maps the multi-dimensional time series data of the bridge into the high-dimensional space,so that it can be linearly separable in the new high-dimensional space.Then,the time aggregation algorithm is used to process the data mapped into the high-dimensional space and identify the multi-dimensional time series data of the bridge.At the same time,in order to learn the output-weight function matrix,a spatio-temporal aggregation algorithm based on orthogonal function basis expansion is proposed.This method not only enhances the identifiability of multi-dimensional time series data of bridges,but also takes into account the relative importance of data in different time steps by leveraging non-linear mapping capacity of a reservoir and time feature accumulation.The method proposed in this thesis firstly carries out experimental verification on the classic data set ASCE Benchmark,and the experimental results are compared with the traditional methods of SVM,KNN and NB.The comparison experiment shows that the method proposed in this thesis has a high accuracy and can effectively evaluate the health state of bridge structures.In order to further verify the effectiveness of the proposed method,this article experiments on the real bridge dataset,The results also show that the proposed method performs better than the traditional method in terms of average accuracy and stability.Through experiments,it is concluded that the method proposed in this thesis is effective and feasible in bridge structure state assessment.At the same time,the parameters of each method are adjusted and analyzed to some extent in the experiments of two classical data sets,which provides suggestions for parameter setting.Finally,the evaluation method proposed in this thesis uses the particle swarm optimization algorithm to optimize the parameters,namely PSO-FESN.Four basic parameters(N,R,SR,SD)of the method are optimized,and a large number of experiments are carried out on two classical datasets.The experimental results show that the accuracy of the optimized data sets is improved to a certain extent,which provides a theoretical basis for setting parameters on the two sets of data sets.