Author Correction: Research on stress curve clustering algorithm of Fiber Bragg grating sensor

Yisen Lin, Ye Wang, Huichen Qu, Yiwen XiongScientific 2023 · DOI · OpenAlex

Introduction

This article serves as a formal correction to the study "Research on stress curve clustering algorithm of Fiber Bragg grating sensor," originally published in Scientific Reports. The primary purpose of this update is to provide the complete, corrected pseudocode for the core clustering algorithms and to rectify inaccuracies in the mathematical analysis of their computational complexity. These corrections are vital for the reproducibility and theoretical validation of the proposed Fiber Bragg grating (FBG) sensor data analysis.

Corrected Algorithmic Framework

The research focuses on a density-based clustering approach tailored for FBG sensors, which are critical components in structural health monitoring. The core of the methodology is the Mutual Nearest Neighbor (MNN) approach.

Mutual Nearest Neighbor (MNN) Logic

The algorithm identifies clusters based on points that mutually recognize each other as neighbors. Specifically, the intersection of neighbor sets is defined as:

$$MNN_k(s_i) = NN_k(s_i) \cap RNN_k(s_i)$$

where $NN_k$ is the $k$-nearest neighbors and $RNN_k$ is the reverse nearest neighbors. This mutuality serves as a robust filter against noise; a connection is only validated if both points "agree" on their proximity.

Algorithm 1: Overall Clustering

The corrected Algorithm 1 outlines the initialization process:

  1. Distance Matrix Calculation: Computing distances between all sensor data points.
  2. Neighborhood Identification: Identifying the $k$-neighborhood for each point.
  3. Point Classification: Distinguishing between "backbone" points (high connectivity) and "weak" points. Points with zero mutual neighbors are explicitly labeled as noise.

Algorithm 2: MakeClusterBackbones

The most significant part of the correction is the full restoration of Algorithm 2. This procedure handles the iterative formation of clusters ($C$):

  • It evaluates the intersection of mutual neighbor sets.
  • If a new backbone point shares neighbors with an existing cluster, the sets are merged.
  • This ensures that the "skeleton" of the cluster is formed by the most reliable data points before peripheral points are considered.

Revised Complexity Analysis

The correction updates the "Time complexity" section to align with the restored pseudocode. The theoretical performance is now accurately represented as follows:

  • Sorting: Sorting points by their MNN count requires $O(N \log N)$.
  • Backbone Formation (Algorithm 2): The complexity is $O(|S| \cdot |U| \cdot \log k)$, where $|S|$ is the total number of points and $|U|$ is the upper limit of generated clusters.
  • Weak Point Merging (Algorithm 3): The final stage of merging non-backbone points is $O(|W| \cdot |U| \cdot k)$, where $|W|$ represents the set of weak points.

Conclusion

By providing the full logic for MakeClusterBackbones and correcting the complexity citations, this update ensures that the MNN clustering algorithm for FBG sensors is both mathematically sound and procedurally transparent. These corrections allow researchers to accurately implement and evaluate the performance of the clustering method in structural monitoring applications.

Summary updated

BibTeX
@inproceedings{lin2023author,
  author = {Yisen Lin and Ye Wang and Huichen Qu and Yiwen Xiong},
  title = {{Author Correction: Research on stress curve clustering algorithm of Fiber Bragg grating sensor}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41598-023-42289-6},
}
Figure