Thermocouple Grid Dataset for Data-Driven Sensor Signal Reconstruction in Brake Tribometer Experiments

Résumé fourni par la source

This dataset contains preprocessed thermocouple measurements from brake tribometer experiments for data-driven sensor signal reconstruction. The measurements originate from an embedded thermocouple grid used to investigate the temperature field during repeated braking events. The dataset was prepared to support the reconstruction of biased thermocouple signals caused by sensor misplacement and the reconstruction of artificially masked missing values in temperature time series. The dataset contains 211 brake-event sequences with temperature signals from 12 thermocouples, denoted as TC0 to TC11. The signals were sampled at 10 Hz. The individual brake-event sequences contain between 510 and 532 time steps and were zero-padded to a common sequence length of 532 time steps. The data are provided together with a mask, sequence identifiers, and feature names. The data array contains the thermocouple measurements, while the mask indicates which samples are valid and can be used for model training and evaluation. The sequence identifiers assign each sample or time step to its corresponding brake-event sequence. The feature names specify the thermocouple channel associated with each signal dimension. The dataset is intended for supervised learning experiments on sensor signal reconstruction. For biased-sensor reconstruction, correctly installed thermocouple channels can be used to reconstruct affected channels. For missing-value reconstruction, artificial masks can be applied to complete temperature sequences to evaluate reconstruction methods under controlled missing-value fractions and missing-patch configurations. A load_data.py script is included to facilitate easy loading and inspection of the dataset. The dataset is associated with the study “Transformer-Based Reconstruction of Biased and Missing Sensor Signals”. File contents: data: Preprocessed thermocouple measurement data. mask: Boolean or binary mask indicating valid samples. mask = 1 indicates a valid sample, and mask = 0 indicates padding or invalid samples. seq_ids: Sequence identifiers assigning each sample or time step to a brake-event sequence. feature_names: Names of the thermocouple channels. The included channels are TC0, TC1, TC2, TC3, TC4, TC5, TC6, TC7, TC8, TC9, TC10, and TC11.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.