An unsupervised clustering-based sectionalized displacement reconstruction method for smart geogrids integrated with fiber Bragg grating sensors

•An clustering-based displacement reconstruction method for the geogrids with FBG sensors was proposed.•Two novel descriptors were designed as features for clustering.•The method was validated at three levels: numerical simulations, laboratory experiments and model testing. It has been found that cumulative measurement errors adversely affect the accurate reconstructions of the displacements of smart geogrids integrated with fiber Bragg grating sensors. The correction coefficients tend to vary with the deformation forms of the geogrids, which hinder the accurate self-corrections of the smart geogrids. Therefore, in order to address these issues and increase the displacement measurement accuracys, a sectionalized displacement reconstruction method for the smart geogrids was proposed. Two novel feature descriptors, which were termed as the displacement variation direction and logarithm of the strain absolute value, were designed. This was followed by an unsupervised clustering algorithm, which was designed to process the two designed features and categorize the smart geogrids into different sections. Each of the sections were corrected using the corresponding correction coefficients, which were optimized using a differential evolution algorithm. Then, simulations of smart geogrids with different lengths and different deformation patterns were performed for the purpose of validating the performance of the proposed method. It was determined that the mean of mean absolute errors (MAEs) of the simulation cases had been decreased from 18.6 mm to 2.2 mm. In addition, the proposed method was validated using laboratory experiments and model testing. The experimental results indicated that for the three different performed experiments, the proposed method had successfully reduced the mean MAEs from 12 mm to 5 mm. Therefore, smart geogrids using sectionalized displacement reconstruction methods were confirmed to be promising new developments in the field of practical civil engineering applications.