Wrocław University of Science and Technology

The spectral shift as a function of length in Rayleigh-based OFDR for temperature and humidity measurement

The dataset contains spectral shift measurements as a function of optical fiber length acquired using a Rayleigh scattering-based optical frequency domain reflectometry (OFDR) distributed optical fiber sensor. Measurements were performed using SM 1500 (7.8/125) optical fibers and acquired with a LUNA OBR 4600 reflectometer in a controlled climate chamber (Binder KMF 115). The investigated sensor consisted of experimentally prepared bare and polyimide-coated optical fiber sections placed side by side to enable simultaneous temperature and relative humidity sensing.

Measurements were performed for temperatures ranging from 30°C to 80°C with a 10°C step and relative humidity ranging from 20% to 80% with a 10% step. The dataset includes spectral shift values along the fiber length for all measured environmental conditions.

The dataset was used for training and evaluation of machine learning models, including Multilayer Perceptrons (MLP) and Convolutional Neural Networks (CNN), for automated prediction of temperature and relative humidity in distributed optical fiber sensors based on Rayleigh scattering.

This repository contains the complete framework used for data preprocessing and further machine learning (ML)-based analysis..

ML analysis

Data preprocessing

To perform data preprocessing:

1. Create a directory named dane and place all measurement files inside it.
2. Open the PreprocessData.ipynb notebook and run all cells.
3. After execution, the following directories should be created:

• data_preprocessed – containing training data,
• validation_data_preprocessed – containing testing/validation data.

Notebooks

• The models directory contains the tested machine learning models.
• The visualize_data.ipynb notebook allows preliminary visualization of the processed data.
• The calculate_statistics.ipynb notebook generates a JSON file containing experiment statistics and converts the JSON results into CSV format, which can be imported into Excel for creating tables and summaries.

Scripts

• data_preprocessor.py performs preprocessing of raw measurement data.
• import_data.py contains functions for importing preprocessed data into dataframe format.
• data_selection.py contains functions for: reducing the number of measurement points along the optical fiber, averaging or non-averaging data selection, train-test splitting based on environmental states and a defined ratio, k-fold cross-validation based on measurement states.