Search

Most damage-assessment strategies for dynamic systems only distinguish between undamaged and damaged conditions without recognizing the level or type of damage or considering unseen conditions. This paper proposes a novel framework for structural health monitoring (SHM) that combines supervised and unsupervised learning techniques to assess damage using a system’s structural response (e.g., the acceleration response of big infrastructures). The objective is to enhance the benefits of a supervised learning framework while addressing the challenges of working in an SHM context. The proposed framework uses a Linear Discriminant Analysis (LDA)/Probabilistic Linear Discriminant Analysis (PLDA) strategy that enables learning the distributions of known classes and the performance of probabilistic estimations on new incoming data. The methodology is developed and proposed in two versions. The first version is used in the context of controlled, conditioned monitoring or for post-damage assessment, while the second analyzes the single observational data. Both strategies are built in an automatic framework able to classify known conditions and recognize unseen damage classes, which are then used to update the classification algorithm. The proposed framework’s effectiveness is first tested considering the acceleration response of a numerically simulated 12-degree-of-freedom system. Then, the methodology’s practicality is validated further by adopting the experimental monitoring data of the benchmark study case of the Z24 bridge.

Most damage-assessment strategies for dynamic systems only distinguish between undamaged and damaged conditions without recognizing the level or type of damage or considering unseen conditions. This paper proposes a novel framework for structural health monitoring (SHM) that combines supervised and unsupervised learning techniques to assess damage using a system's structural response (e.g., the acceleration response of big infrastructures). The objective is to enhance the benefits of a supervised learning framework while addressing the challenges of working in an SHM context. The proposed framework uses a Linear Discriminant Analysis (LDA)/Probabilistic Linear Discriminant Analysis (PLDA) strategy that enables learning the distributions of known classes and the performance of probabilistic estimations on new incoming data. The methodology is developed and proposed in two versions. The first version is used in the context of controlled, conditioned monitoring or for post-damage assessment, while the second analyzes the single observational data. Both strategies are built in an automatic framework able to classify known conditions and recognize unseen damage classes, which are then used to update the classification algorithm. The proposed framework's effectiveness is first tested considering the acceleration response of a numerically simulated 12-degree-of-freedom system. Then, the methodology's practicality is validated further by adopting the experimental monitoring data of the benchmark study case of the Z24 bridge. [ABSTRACT FROM AUTHOR]

In this work we present the results of a field trial with a FBG sensor array system for the real time monitoring of railway traffic and for the structural health monitoring of both the railway track and train wheels. The test campaign is performed on the 2nd line of Milan metropolitan underground, employing more than 50 FBG sensors along 1.5 km of the rail track, where the trains are tested during daily passenger rail transport, with a roughly maximum speeds of 90 km/h. The measurements were continuatively performed for over 6 months, with a sampling frequency of about 400 Hz. The large amount of data/sensors allows a rather accurate statistical treatment of the measurement data and permits, with dedicated algorithms, the estimation of rail and wheel wear, key traffic parameters such as the number of axles, the train speed and load, and, in the next future, the detection of localized imperfections. [ABSTRACT FROM AUTHOR]

Abstract: Hilbert–Huang transform (HHT) is an innovative data-processing technique for analyzing nonstationary and nonlinear signals. A novel HHT-based method for damage detection of bridge structures under a traveling load is proposed. The technique uses a single point measurement and is able to identify the presence and the location of the damage along the beam. The measured data are processed by the HHT technique, and none a priori information is needed about the response of the undamaged structure. Damage location is revealed by direct inspection of the first instantaneous frequency, which presents a sharp crest in correspondence of the damaged section. The identification capabilities of the proposed technique are studied varying the damage locations, crack depths and velocity of the moving load. The effect of ambient noise is also taken into account. Theoretical as well as numerical results show the identification is rather accurate, results are not very sensitive to the crack depth and ambient noise, while they are sensibly affected by the damage location and by the speed of the moving load as well. Theoretical analysis identifies a characteristic load velocity interval, depending both on the first natural frequency of the bridge and the damage location, within which the HHT can be successfully applied. [Copyright &y& Elsevier]