Vol. 55, Issue 3, pp. 437-448 (2025)

Abstract

To obtain the state information of multiple targets in a fiber optic sensing network, we improve the target state recognition probability. A target state recognition algorithm based on data mining is proposed. An orthogonally distributed FBG sensing network is designed based on the characteristics of fiber optic sensing. The clustering weights and clustering degree for multi-target state classification are derived. A target state recognition model based on data mining is constructed. The time interval between peak-to-peak tip positions in an aliased signal containing a faster-moving target is narrower. The more targets are aliased, the greater the time width of the overall echo response. The average test speeds for target A, target B, and target C are 0.99 m/s, 2.86 m/s, and 4.87 m/s, with relative errors of 1.3%, 4.7%, and 2.5%, respectively. The experimental comparison of target recognition probabilities before and after data mining shows that the recognition rate is significantly improved after applying this algorithm. It is especially effective for overlapping signals of state-similar targets. Moreover, this algorithm also suppresses the system’s false judgment rate to a certain extent.