Sparse regularization for traffic load monitoring using bridge response measurements

Abstract Traffic load monitoring (TLM) is one of important issues in bridge structural health monitoring (SHM), but there still exist such problems as lack of accuracy and efficiency for the existing methods. In this study, a sparse regularization approach is proposed for TLM based on analytical model and redundant dictionary. Firstly, an unknown moving traffic load is deemed as a combination of static and time-varying components so that a redundant dictionary can be established to independently express them. The static component is expressed by a basis vector whose elements are identical, and the time-varying one by wavelet functions for their good multi-resolution analysis characteristics. Then, the TLM problem is converted to determine a coefficient vector of dictionary, and the l 1 -norm regularization technique is adopted to obtain a sparse solution to the coefficient vector. Finally, a series of experimental studies on a hollow steel beam bridge under crossing a moving model car are conducted in laboratory to assess the effectiveness of the proposed method. Furthermore, comparative studies are carried out for assessing the effect of different measurement parameters, such as moving car speeds, car weights, strain and acceleration response data, redundant dictionaries as well as selection of regularization parameters, on the proposed method. The illustrated TLM results show that the dictionary used for TLM in this study can independently distinguish the static and time-varying components of moving traffic loads. The proposed method can effectively identify the total weight of moving traffic loads with a higher accuracy, which provides a great potential for monitoring moving vehicle loads on bridges.