ML-Based Utility Pole Localization Using Ambient DAS Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for localizing utility poles using distributed acoustic sensing (DAS) are inefficient and costly, relying on manual labor to identify pole locations by analyzing DAS signal patterns generated from hammer knocks.
Innovation Solution
A machine learning model that analyzes two-dimensional spatiotemporal maps of dynamic strains on optical sensor fibers, separating low-frequency and high-frequency features using a contrastive loss function and Gaussian distribution to automatically detect pole locations without requiring domain knowledge, enabling efficient and low-workforce utility pole localization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling by human experts is used to identify pole locations, then measurement precision can be achieved, but productivity is severely reduced and costs increase
Solution Approach 1:
The patent replaces the manual mechanical process of human experts physically knocking poles and visually analyzing DAS signal patterns with an automated machine learning system. The ML model automatically processes DAS data to identify pole locations, eliminating the need for human intervention while maintaining identification accuracy and dramatically increasing productivity.
Solution Approach 2:
The system enables self-service by allowing the machine learning model to autonomously identify and label pole locations without requiring human experts. The model trains on labeled data and then independently processes new DAS data to detect poles, making the system self-sufficient and eliminating manual labor requirements.
2Reliability
If manual knocking and impacting of every pole is performed, then reliable pole location data can be collected, but loss of time and workforce requirements increase significantly
Solution Approach 1:
The patent applies preliminary action by using ambient DAS data that continuously captures acoustic events from environmental sources. Instead of performing manual knocking at the time of measurement, the system has already captured relevant acoustic signatures in the ambient data, which are then processed by the ML model to identify poles, saving significant time and effort.
Solution Approach 2:
The system implements continuity of useful action by utilizing continuous ambient DAS monitoring that constantly captures acoustic events. This continuous data collection replaces the discrete, intermittent manual knocking process, allowing the system to accumulate useful data over time and process it efficiently without requiring repeated manual interventions.
3Productivity
If automated machine learning methods are used for pole detection, then productivity increases and manual effort decreases, but device complexity increases due to signal processing requirements
Solution Approach 1:
The patent applies segmentation by dividing the complex signal processing task into distinct frequency components - specifically separating low-frequency and high-frequency features from DAS signals. The machine learning model processes these segmented frequency components independently and combines them for pole detection, making the overall system more manageable and interpretable despite the underlying complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution provides an efficient and cost-effective method for utility pole localization, automating the process and reducing manual effort, allowing for accurate detection of pole locations using ambient data from distributed fiber optic sensing systems.
Implementation Method 1
A DFOS/DAS interrogator located at one end of an optical sensor fiber remotely captures dynamic strains on the optical sensor fiber induced by acoustic events
Data Source
AI summary
Systems and methods for utility pole localization employing a DFOS/DAS interrogator located at one end of an optical sensor fiber remotely capture dynamic strains on the optical sensor fiber induced by acoustic events. A captured two-dimensional spatiotemporal map in an ambient noisy environment is analyzed by a trained machine learning model which then automatically detects an area in which a pole is located without requiring domain knowledge. Original DFOS/DAS signals are separated into pole regions and non-pole region time series for machine learning model training. A contrastive loss function measures similarities between low-frequency and high-frequency features. A Gaussian distribution is applied to the original signals to generate weighted labels to eliminate effects of label noise. The machine learning model fuses low-frequency and high-frequency features in the frequency domain for pole region classification. A contrastive loss is combined with cross entropy loss to measure a low-high frequency feature distance.