Data for: Simultaneous separation of travel-dependent drag and added-mass coefficients for underwater manipulators
Description
This dataset contains processed research data supporting the manuscript “Simultaneous separation of travel-dependent drag and added-mass coefficients for underwater manipulators.” The study tests whether the transient drag coefficient Cd and equivalent added-mass coefficient Cm can be represented primarily as functions of nondimensional travel S/d and separated simultaneously from sectional hydrodynamic-torque data. The torque data were obtained from three-dimensional unsteady CFD simulations of a 4-m cylindrical manipulator link undergoing twelve standard accelerate–cruise–decelerate motions with different angular velocities and accelerations. The files contain smoothed Fluent total-torque histories and Tecplot ten-segment and total-torque histories. These data were processed by binned least-squares coefficient separation and physically constrained piecewise-cubic spline fitting. The coefficient-identification data include combined least-squares samples, separate Cd and Cm scatter-point datasets, spline control points, and densely sampled Cd(S/d) and Cm(S/d) curves. Twelve leave-one-out datasets are provided, each containing samples, control points, and spline curves obtained after excluding one standard case, together with an error summary. The package also contains processed time-series data for seven realistic quintic-polynomial motions covering slew angles of 15–120 degrees and durations of 5–10 s. The files provide time, S/d, angular displacement, angular velocity, angular acceleration, CFD torque, spline-model torque, and error statistics. Results for an optimal constant-coefficient baseline, curve-inflection source tables, and selected auxiliary motion workbooks are also included. The data can be used to reconstruct the identified coefficient curves, inspect sectional and total torque histories, compare the spline model with CFD, assess sensitivity to individual training cases, and recalculate the reported errors. They show systematic variation of both coefficients with S/d and support evaluation of the model’s predictive behavior. File contents, variables, units, provenance, folder structure, and checksums are documented in the README and manifest files. This repository contains processed data only. Water-tank experimental data, the main cone, orbital, two-link, and attached-cylinder spatial-motion datasets, and mesh-convergence source data are not included. Source code, Fluent UDFs, and patent-related implementation files are unavailable because the method is associated with a pending patent application. Raw CFD fields, meshes, and large Fluent case/data files are excluded because of their storage requirements.
Files
Steps to reproduce
This repository contains the processed data required to inspect and reproduce the principal data-based results reported in the associated manuscript. The complete source code and raw CFD files are not included in the current dataset. 1. Refer to the README and manifest files for the folder structure, file descriptions, variable definitions, and units. 2. To reproduce the travel-dependent coefficient curves, use the supplied control-point files or the densely sampled lookup-table files for Cd(S/d) and Cm(S/d). The curves can be plotted directly from the lookup tables or reconstructed using piecewise-cubic spline interpolation through the supplied control points, following the procedure described in the manuscript. 3. To reproduce the standard-motion torque comparisons, use the processed time-series files for the twelve standard cases. These files contain the CFD torque data and the corresponding quantities required for comparison with the travel-dependent coefficient model. 4. Recalculate the error metrics using RMSE = sqrt(mean((tau_model - tau_CFD)^2)) and relative RMSE = RMSE / sqrt(mean(tau_CFD^2)). Multiply the relative RMSE by 100 to obtain the percentage value. 5. To reproduce the leave-one-out cross-validation results, use the supplied holdout-specific coefficient curves, control points, torque data, and summary files. For each holdout case, compare the model prediction with the corresponding CFD torque and calculate the error using the same definitions given above. 6. The processed CSV and XLSX files can be inspected and plotted using Python, MATLAB, spreadsheet software, or equivalent numerical tools. No proprietary software is required to visualize the supplied processed data or recalculate the reported error metrics. The complete coefficient-identification workflow and mathematical formulation are described in the associated manuscript. The source code, Fluent UDFs, and patent-related implementation files are temporarily withheld because the associated patent application is currently undergoing substantive examination. Their public release may be considered after completion of the patent examination, subject to applicable intellectual-property requirements. Raw CFD flow fields, computational meshes, and large Fluent case/data files are not included because of their substantial storage requirements.
Institutions
- University of Science and Technology of ChinaAnhui, Hefei