2026-11-01T01:14:13Z https://oai.datacite.org/oai

doi:10.17632/f78bmhr628.1 2025-01-17T16:20:37Z ELSEVIER BL.MENDELEY

2 BL.MENDELEY 10.17632/F78BMHR628.1 Iafolla, Lorenzo Lorenzo Iafolla Santoli, Francesco Francesco Santoli 0000-0003-2493-0109 Istituto nazionale di astrofisica Carluccio, Roberto Roberto Carluccio 0000-0003-4344-0965 Istituto Nazionale di Geofisica e Vulcanologia Chiappini, Stefano Stefano Chiappini 0000-0001-9520-7618 Istituto Nazionale di Geofisica e Vulcanologia Fiorenza, Emiliano Emiliano Fiorenza 0000-0001-7634-2152 Istituto di Astrofisica e Planetologia Spaziali Lefevre, Carlo Carlo Lefevre 0000-0003-0649-9941 Istituto di Astrofisica e Planetologia Spaziali Lucente, Marco Marco Lucente 0000-0002-5359-5437 Istituto di Astrofisica e Planetologia Spaziali Pignatelli, Alessandro Alessandro Pignatelli 0000-0002-3172-2044 Istituto Nazionale di Geofisica e Vulcanologia Chiappini, Massimo Massimo Chiappini 0000-0001-7433-9435 Istituto Nazionale di Geofisica e Vulcanologia Mendeley Data 2023 Dataset Machine Learning Accelerometer Sensor Istituto nazionale di astrofisica https://ror.org/02gh4kt33 Istituto Nazionale di Geofisica e Vulcanologia https://ror.org/00qps9a02 2023-09-26T13:14:39Z 10.17632/f78bmhr628 Creative Commons Attribution 4.0 International Temperature fluctuations significantly affect the precision of high sensitive accelerometers and gravimeters. This dataset comprises acceleration measurements obtained from a high-accuracy accelerometer, with background noise levels as low as 10^-7 m/s^2/sqrt(Hz). Additionally, it includes data from 11 thermometers strategically placed both inside the accelerometer casing and on its sensitive components. Analyzing these data, can clearly show how important is to measure the temperature in multiple points for effectively compensating for its effects over the acceleration measurements. Throughout the measurements, the accelerometer remained stationary on the laboratory floor, primarily detecting micro-seismic noise which could, to a first approximation, be disregarded. However, the accelerometer's output was not constant due to temperature variations, which were deliberately induced using heating mats affixed to the accelerometer's enclosure and a 260-watt lamp. Using these devices, we could create varying temperature fields and gradients. Assuming that the accelerometer's output is solely influenced by the temperature variations, it can be predicted using only the thermometer measurements. In other words, the errors introduced by temperature can be reconstructed using the thermometer readings. To reconstruct the effects of temperature on the accelerometer readings, supervised machine learning (ML) methods can be employed, with the accelerometer output serving as the desired output and the temperatures as the inputs. The dataset is divided into two parts: one comprises data obtained solely by employing heating mats to manipulate the temperature, while the other (with the filename "resampled_2023-03-25_2023-04-14.csv") consists of data obtained exclusively through the use of the heating lamp. Consequently, one subset can be used for training, and the other for testing, clearly demonstrating the ML model's ability to generalize across different scenarios. Each record in the dataset includes the following information: date ("Date"); accelerometer outputs ("z [m/s^2]", "x [m/s^2]", "y [m/s^2]"), with the z-axis predominantly aligned with the local vertical; temperatures of the accelerometer sensors ("T z [°C]", "T x [°C]", "T y [°C]"), measured by thermometers attached to the accelerometer's sensitive elements; temperatures of the six faces of the accelerometer box ("bottom x1 side [°C]", "x sensor face [°C]", "y sensor face [°C]", "bottom x2 side [°C]", "x elect face [°C]", "y ACQ face [°C]", "top [°C]"), where the bottom face features two thermometers; temperature of the air inside the box ("inside air [°C]"); external temperature ("external [°C]"); and altitude ("Altitude [m]"), measured by a barometer and thus reflecting variations in pressure. The data from the accelerometers and the thermometers were synchronized between each other and then resampled at 1 Hz. No filtering was applied and missing data are due to technical stops. Regione Lazio https://ror.org/008hssd09