eBike measurements for fatigue monitoring using acceleration differences
Documentation of the data | A thorough documentation of this dataset, including experimental setup, postprocessing and data access is provided in the description of the data and in the textfile dataset_description.txt in the dataset. | |
Countries to which the data refer | GERMANY | |
References to related material | http://dx.doi.org/10.25532/OPARA-189 | |
Description of the data | EXPERIMENTAL SETUP The sensor setup includes 5 acceleration sensors and 5 strain gauges, whose data is sampled at a frequency of 1200 Hz. The individual sensors are positioned as shown in the pictures provided in additional_information/eBike_sensor_locations_left.png and eBike_sensor_locations_right.png. Strain gauges 1 and 2 are positioned on the left side of the rear part of the frame, connected to the wheel suspension and acceleration sensors 3 and 4 are positioned such that their integration provides information related to bending of this structure. On the right side of the frame, the deformation of strain gauges 4 and 5 is similarly related to acceleration sensors 3 and 5. The estimation of strains at strain gauge 3 is less straight forward, since it is located near the stiff saddle post region of the frame. Acceleration sensors 1 and 2 provide additional information regarding the general dynamic maneuvers of the eBike, since they are aligned with and perpendicular to the direction of travel. DATA POST-PROCESSING Roughly 30 s at the start and end of each measurement file were removed, since the rider is only configuring the measurement system and getting onto or off the eBike in this time. Additionally, the acceleration data was numerically integrated to displacement time series, which required using a high pass filtered at 2 Hz before and after every integration step to avoid signal drifts. The corresponding strain data is also filtered in identical manner to ensure corresponding input and output data. In this dataset, data in the directory "raw/" includes unfiltered, unintegrated data, while data in the "processed/" directory includes filtered acceleration and strain signal, as well as displacement data resulting from integration. The "processed/" data is organized as input signals (_in.json), which include accelerations and displacements, and output signals (_out.json), which include strain data. The input signals also include FRF-model predictions of the output data, which were created using all displacement input sensor data. The FRF-model prediction data is not required for the virtual sensing task, but can either be used as an additional input to parameterize hybrid approaches or be used as a reference benchmark solution to the problem. STRUCTURE The dataset is divided into unlabeled (_UNL_) and labeled (_LAB_) measurements. Unlabeled measurements is obtained from general eBike usage, where only a very general categorization of the file content is provided in the file names. Here, "Heide" refers to measurement rides conducted in the Dresdner Heide, "city_park" refers to measurements from the Grosser Garten in Dresden and "bike_way" refers to rides conducted on bike ways, while "city_mixed" contains data stemming from a combination of streets and bike ways in Dresden, where the underground is mostly even, but can also include sections of cobblestone. Since the monitored bike contains an electric motor, which was mostly used in medium and high support mode, the velocity in the unlabeled data lies mostly in the range of 20 - 25 km/h, which might be slighly less on uneven underground. Short reference for file names: [t]_[l]_[c]_[n]_[d].[e] - [t] training assignment, either to training or testing data - [l] label information, UNL: unlabeled, LAB: labeled - [c] content, short and very general description - [n] counting number among files of the corresponding content - [d] designation, showing if processed data is assigned to input or output - [e] ending, _cn_X.float32 includes the binary data of channel X, .json includes human readible metadata Labeled measurements include short, specific riding scenarios, for which precise maneuver information is available. Riding speed was varied in steps of 5 km/h, controlled manually within an approximate 1.5 km/h range of the target speed given in the file name. Data was collected from two different types of underground, named "even" for very even asphalt underground and "cobble" for very uneven cobblestone underground. The additional_information directory includes images of the sensor equipped bike. DATA ACCESS Measurements in this dataset are stored using a binary file format, where data from each individual measurement channel is stored in separate files as a sequence of 32-bit floats. For each such file, a corresponding json-file contains relevant metadata, such as the number of channels, sampling rate, channel names, the physical unit and the nominal values of the channel. Except for the channel related file ending, json-files are named identically to the binary files. The dataset also contains a Python script "binary_to_numpy.py", which provides an interface for data access. Export to binary NumPy files (.npy) or human-readable text files (.csv) is exemplarily shown, though the script can be easily adapted to user-specific requirements. A data loading example is provided at the end of the Python script. This script was tested on a system with the following package versions, though newer versions should also work: Python Version: 3.8.3 NumPy Version: 1.19 ACKNOWLEDGMENTS The authors gratefully acknowledge funding from the German Federal Ministry for Digital and Transport within the mFUND research initiative, Funding code: 19FS2012A. This research is also funded by the European Regional Development Fund (ERDF) and co-financed by tax funds based on the budget approved by the members of the Saxon State Parliament. | |
Type of the data | Dataset | |
Total size of the dataset | 1666783430 | |
Author | Heindel, Leonhard | |
Author | Hantschke, Peter | |
Author | Kästner, Markus | |
Upload date | 2025-06-26T13:07:29Z | |
Publication date | 2025-06-26T13:07:29Z | |
Data of data creation | 2025-02 | |
Publication date | 2025-06-26 | |
Abstract of the dataset | In fatigue monitoring, the aim is to approximate the fatigue damage accumulation in a system, for example to schedule maintenance intelligently. Virtual sensing can be deployed to obtain the required information more efficiently by estimating them from readily available sensor data. This dataset contains acceleration and strain measurements from a sensor equipped eBike, designed to develop such methods. Using the acceleration data, the aim is to estimate the fatigue damage accumulation at the strain gauge positions. Here, the sensors are positioned such that differences between multiple acceleration sensors provide information that is related to the deformation of the eBike frame, so that a connection to strain gauge data can be established by computing differences between integrated displacement data. The dataset includes data intended for training and testing of data-driven modeling approaches, which is obtained from regular eBike usage. Additional labeled data enables the analysis of specific riding maneuvers. | |
Public reference to this page | https://opara.zih.tu-dresden.de/handle/123456789/1556 | |
Public reference to this page | https://doi.org/10.25532/OPARA-884 | |
Publisher | Technische Universität Dresden | |
Licence | Attribution 4.0 International | en |
URI of the licence text | http://creativecommons.org/licenses/by/4.0/ | |
Specification of the discipline(s) | 4 | |
Title of the dataset | eBike measurements for fatigue monitoring using acceleration differences | |
Research instruments | eBike of brand Nuvelos | |
Research instruments | acceleration sensors | |
Research instruments | strain gauges | |
Underlying research object | predictive maintenance | |
Software | Python | |
Project abstract | The objectives of the project are the development of fundamental digital methods for monitoring and increasing the reliability of highly integrated mechatronic systems that can be transferred to other engineering problems. The methods are to be developed within the framework of the project using the electric bicycle as an example, always with a view to the transferability and utilization of the research results to other vehicles with electric drives. These methods are a prerequisite for new business models of system providers that link product, application and service. | |
Public project website(s) | https://tu-dresden.de/ing/maschinenwesen/ifkm/nefm/forschung/projekte/epredict | |
Project title | ePredict |
Collections
