Abstract Summary
The use of fibre reinforced polymer (FRP) in civil construction applications has gained considerable popularity worldwide as suitable method for strengthening of existing concrete structures. However, the implementation of methods able to give a reliable prediction about the health of this type of structures is needed since the usual failure modes occur in a sudden and brittle way. The electromechanical impedance (EMI)-based structural health monitoring technique, based on the piezoelectric or electromechanical coupling effects, has been applied successfully for these structures. By analyzing the electric impedance variation of piezoelectric sensors installed on the structure in different frequency intervals along the monitoring process the structural damage can be investigated. However, EMI technique is very sensitive not only to structural incipient damages, but also to temperature fluctuations, that may adversely cause false alarms in real-life monitoring applications. On the other hand, FBGs are intrinsically sensitive not only to the strain but also to temperature and, thus, the combination of FBG and PZT technologies in combination with suitable data processing techniques might provide a suitable methodology to identify incipient damage in the presence of temperature variations. Machine learning (ML) approaches, such as clustering techniques, are an interesting alternative to detect hidden patterns from monitored data and, therefore, to separate the baseline from any anomalies experienced by the structure, either due to mechanical deterioration or change of temperature. This kind of techniques might identify complex patterns related to different stages of the tested structure. In this paper, an integrated electromechanical impedance approach with clustering machine learning techniques and the use of fibre Bragg grating (FBG) temperature and strain sensors for identification of concrete damage under varied temperatures is presented. Effectiveness of the proposed approach was verified on a concrete specimen strengthened with FRP NSM technique and subjected to different levels of sustained load for various time periods and at different temperatures. Strain and temperature were continuously measured using FBG sensors. In the same way, after each sustained loading period or important temperature variation, an EMI test was performed with different PZT sensors located at different areas of the beam. Even although the method has been applied in RC beams strengthened with NSM-FRP technique, it is an effective and original contribution to any SHM system based on EMI. It allows to address in a direct way, with low computational complexity, temperature variations and mechanical deterioration without using any reference baseline signatures measured at different temperatures such as it is done in the usual temperature variation compensation techniques.
