DFOS Data Segmentation for Temperature and Traffic Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing distributed fiber optic sensing (DFOS) systems face challenges in accurately estimating real-time temperature changes and vehicle trajectories along roads, particularly due to mixed temperature and traffic patterns, dynamic environmental conditions, and the difficulty in obtaining labeled data.
Innovation Solution
The proposed solution employs a generalized framework that uses 2D DFOS data to pre-train a masked autoencoder without requiring labeled data. This framework divides 2D data into grids, applies image distortion techniques for data augmentation, and connects the autoencoder to an estimation network for accurate temperature and traffic pattern recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If classic time delay estimation models are used for temperature estimation, then temperature changes can be detected, but accumulated errors occur due to multiple-step estimation
Solution Approach 1:
The patent segments the temperature estimation problem into multiple spatial zones along the fiber optic cable. Each zone is independently analyzed using the autoencoder model to identify local temperature changes and vehicle trajectories. This segmentation allows the system to process complex DFOS data in manageable sections, reducing error propagation compared to global multiple-step estimation while maintaining comprehensive monitoring coverage.
Solution Approach 2:
The patent introduces a masked autoencoder as an intermediary computational model between the raw DFOS data and the final temperature estimation. This autoencoder learns intrinsic features of temperature and traffic patterns from unlabeled data, serving as a mediator that transforms complex optical signal variations into meaningful temperature and trajectory information in a single step, thereby avoiding the error accumulation inherent in traditional multi-step estimation chains.
2Measurement precision
If supervised learning is used for pattern recognition, then accurate temperature and traffic patterns can be learned, but large amounts of labeled data are required
Solution Approach 1:
The patent applies preliminary unsupervised pre-training of the autoencoder on large amounts of unlabeled DFOS data before fine-tuning with少量 labeled data. This preliminary action allows the model to learn intrinsic features of temperature and traffic patterns from the data distribution itself, establishing a strong feature representation that requires minimal labeled data for final supervised adaptation, thus resolving the contradiction between accuracy and data quantity requirements.
Solution Approach 2:
The masked autoencoder performs self-service by learning to reconstruct and understand the DFOS data structure without external labels. The model automatically identifies intrinsic patterns in temperature changes and vehicle trajectories through self-supervised learning mechanisms, generating its own learning signals from the data's inherent structure. This self-service capability eliminates the need for extensive manual labeling while maintaining high pattern recognition accuracy.
3Ease of operation
If traditional DFOS processing is used, then simple temperature monitoring is achieved, but complex external factors like strain and vibration cannot be distinguished
Solution Approach 1:
The patent implements a universal autoencoder model that simultaneously handles multiple types of DFOS signals including temperature variations, strain effects, and vibration patterns. The model is designed to process diverse environmental factors through a unified architecture that learns to distinguish between different physical phenomena affecting the optical fiber. This multi-functional approach maintains processing simplicity while achieving robustness against complex external factors, as the single model adapts to various signal types rather than requiring separate processing chains.
Solution Approach 2:
The patent utilizes parameter changes in the DFOS signal characteristics to distinguish between different external factors. The autoencoder learns to identify specific parameter patterns associated with temperature changes versus strain or vibration effects by analyzing variations in optical signal properties. Through this parameter-based differentiation, the system maintains operational simplicity while achieving adaptability to complex environmental conditions, as the model automatically adjusts its interpretation based on the observed signal parameter variations.
Data Source
AI summary
Disclosed are systems, methods, and structures that provide more accurate temperature measurements and/or derived measurements using distributed fiber optic sensing (DFOS) systems and methods. DFOS systems and methods according to aspects of the present disclosure employ distributed fiber optic sensing that determines real-time temperature changes and vehicle trajectories from two-dimensional (2D) DFOS data with very few labeled data. The 2D data is first divided into multiple grids and then pre-processed with image distortion methods to enrich diversity of temperature change patterns. The transformed grids are used to pre-train a masked autoencoder, which advantageously does not require labels. The encoder of the autoencoder learns intrinsic features of temperature and traffic patterns, which are later connected to an estimation network to solve downstream tasks trained on a small set of labeled data.