Missing Structural Health Monitoring Data Recovery Based on Bayesian Matrix Factorization

The exposure of bridge health-monitoring systems to extreme conditions often results in missing data, which constrains the health monitoring system from working. Therefore, there is an urgent need for efficient data cleaning method. With development big and machine-learning techniques, several methods missing-data recovery have emerged. However, optimization-based may experience overfitting demand extensive tuning parameters, trained models still substantial errors when applied unseen datasets. Furthermore, many can only process a single sensor at time, so spatiotemporal dependence among different sensors cannot be extracted recover data. Monitoring multiple organized form matrix. matrix factorization appropriate way handle To this end, hierarchical probabilistic model formulated under fully Bayesian framework by incorporating sparsity-inducing prior over factors. modeled reconstruct achieve recovery. Through experiments using continuous in-service bridge, proposed method shows good performance effect on preset rank also investigated. show that higher accuracy with same case