Data on the coefficient of friction and its prediction by a machine learning model as a function of time for open-cell AlSi10Mg-Al2O3 composites with different porosity tested by pin-on-disk method

This data article presents the experimental data of the wear behavior of four types of open-cell AlSi10Mg materials and open-cell AlSi10Mg-Al2O3 composites with different pore sizes under dry sliding conditions tested by pin-on-disk method. The data include the coefficient of friction (COF) as a function of time for each material, as well as the predictions of COF using a machine learning model - Extreme Gradient Boosting. The data were generated to investigate the effect of pore size and reinforcement on the friction and wear properties of open-cell AlSi10Mg-Al2O3 composites, which are promising materials for lightweight and wear-resistant applications. The data can also be used to validate theoretical models or numerical simulations of wear mechanisms in porous materials, as well as to optimize the material design and processing parameters to enhance the wear resistance of open-cell AlSi10Mg materials. The data are available in DWF and XLSX format and can be opened by any text editor or spreadsheet software. The data article is related to an original research article entitled “Production and Tribological Characterization of Advanced Open-Cell AlSi10Mg-Al2O3 Composites”, where the details of the experimental methods, the microstructural characterization, and the analysis of the wear mechanisms are provided [1].


a b s t r a c t
This data article presents the experimental data of the wear behavior of four types of open-cell AlSi10Mg materials and open-cell AlSi10Mg-Al 2 O 3 composites with different pore sizes under dry sliding conditions tested by pin-on-disk method.The data include the coefficient of friction (COF) as a function of time for each material, as well as the predictions of COF using a machine learning model -Extreme Gradient Boosting.The data were generated to investigate the effect of pore size and reinforcement on the friction and wear properties of open-cell AlSi10Mg-Al 2 O 3 composites, which are promising materials for lightweight and wear-resistant applications.The data can also be used to validate theoretical models or numerical simulations of wear mechanisms in porous materials, as well as to optimize the material design and processing parameters to enhance the wear resistance of open-cell AlSi10Mg materials.The data are available in DWF and XLSX format and can be opened by any text editor or spreadsheet software.The data article is related to an original research article entitled "Production and Tribological Characterization of Advanced Open-Cell AlSi10Mg-Al 2 O 3 Composites", where the details of the experimental methods, the microstructural characterization, and the analysis of the wear mechanisms are provided [1].
© The data were collected from dry wear tests performed on four types of open-cell AlSi10Mg materials with different pore sizes and reinforcements.The materials were fabricated by liquid-state processing route by the replication method.The wear tests were conducted using a pin-on-disk tribometer at room temperature with linear velocity 0.5 m.s −1 , load 50 N and time of 420 s.The specimens were cylindrical pins with diameter 10 mm and length 20 mm.The counterface was a hardened steel disk with hardness 62 HRC and roughness Ra 1.4 μm.The COF was determined as the ratio of the friction force to the applied load.The COF was recorded as a function of time for each specimen during the tests.The predictions of COF were made using an AlSi10Mg-Al 2 O 3 composites can be analyzed using these data.• The performance of the machine learning model for predicting the coefficient of friction as a function of sliding time for different materials can be evaluated using these data.• The possibility of optimizing the material design and processing parameters to enhance the wear resistance of open-cell AlSi10Mg materials can be explored using these data.

Objective
The objective of this data article is to present the experimental data of the wear behavior of four types of open-cell AlSi10Mg materials with different pore sizes and reinforcements under dry sliding conditions.The data include the coefficient of friction (COF) as a function of time for each material, as well as the predictions of COF using the machine learning model -Extreme Gradient Boosting (XGBoost).The data were generated to investigate the effect of pore size and reinforcement on the friction and wear properties of open-cell AlSi10Mg materials, which are promising materials for lightweight and wear-resistant applications.The data can also be used to validate theoretical models or numerical simulations of wear mechanisms in porous materials, as well as to optimize the material design and processing parameters to enhance the wear resistance of open-cell AlSi10Mg foams.This data article is related to an original research article [1] , where the details of the experimental methods, the microstructural characterization, and the analysis of the wear mechanisms are provided.The data article adds value to the published article by making the raw and processed data publicly available and reusable for further research and development.
The data folders and files are stored organized in the in Mendeley Data repository.The repository consists of two main folders: "Pin-on-disk_data" and "Prediction".The "Pin-on-disk_data" folder has three subfolders that store the raw, processed, and average data of the COF as a function of time for each of the four tested materials, as well as the plots and the python script for the COF calculation.The "Prediction" folder has three subfolders that store the input data, the output data, and the python script for the predictions of COF vs sliding time using a XGBoost model.The output data includes the actual and predicted values of COF for two sets (test and validation) of each material, the performance metrics of the predictions, and the plots of the actual vs predicted COF as a function of time for each material.
The files stored in the Output_data subfolder are: The file stored in the Python_COF_prediction subfolder is: • "XGB-COF.py":Python script used for visualizing and storing a PNG file with the average actual COF vs the predicted as a function of time, calculating and storing in a TXT file the performance metrics of the predicted COF, and storing in a XLSX file the data used for the visualization of the average actual COF.test-set and validation-set for the four materials, which include mean squared error (MSE), root mean squared error (RMSE), coefficient of determination (R2), and mean absolute error (MAE), as shown in Table 1 .Table 2 shows the descriptive statistics of the average COF for all materials, including the mean, median, standard deviation, minimum and maximum values.

Experimental Design, Materials and Methods
This section describes the experimental design and methods used to acquire the data of the wear behavior of four types of open-cell AlSi10Mg materials and open-cell AlSi10Mg-Al 2 O 3 composites with different pore sizes under dry sliding conditions.The data include the COF as a function of time for each material, as well as the predictions of COF using the machine learning model -Extreme Gradient Boosting.

Testing Procedure
The testing procedure involved performing dry wear tests on four types of open-cell AlSi10Mg materials with different pore sizes and reinforcements under room temperature conditions.The materials were fabricated by using the replication method [2] and the squeeze casting [3] , as described in detail in the original research article [1] .The specimens were cylindrical pins with diameter 10 mm and length 20 mm.The counterface was a hardened steel disk with hardness 62 HRC and roughness Ra 1.4 μm.Acetone was used to clean the specimens prior to and following the tests.The pin-on-disk test was conducted with four different sam ples, corresponding to the four types of open-cell AlSi10Mg materials.The wear experiment was repeated three times on each sample to confirm the accuracy of the results.The chemical composition of the base alloy is presented in Table 3 .The choice of the parameters for designing the plan of the experiment was based on the following criteria: • The parameters should be relevant and significant for the research objectives.
• The parameters should be available and feasible in terms of time, cost, and equipment.
• The parameters should be compatible and interactive with each other and with the materials and composites.• The parameters should have a range and level that can cover the possible variations and scenarios of the experiment.
The chosen parameters for the experiment were: • The type, size and content of the reinforcement: Al 2 O 3 , size of particles 30 0-40 0 μm, 5 wt.%.
• The linear velocity, load and sliding time of the dry wear tests: 0.5 m.s −1 , 50 N and 420 s.
• The input data and output data for the XGBoost model: COF vs sliding time.

Testing Equipment
The testing equipment used for the dry wear tests was a Ducom Rotary Pin/Ball-on-Disk tribometer, model TR-20.The tribometer is equipped with a load cell that measures the friction force.The tribometer also has a data acquisition system that records the friction force and the sliding distance as a function of time.The device can perform the ASTM G99 standard test method for wear testing with the Pin-on-Disk method [4] .

Data Acquisition
The data acquisition system of the tribometer recorded the friction force and the sliding distance as a function of time for each test.The data were stored in DWF format and can be accessed by any software for text editing or spreadsheets.The COF was determined as the ratio of the friction force to the applied load.The COF was recorded as a function of time for each specimen during the tests.The data files include three datasets for the COF as a function of time for each of the four tested materials.Four average COF as a function of time files of the three datasets of each of the four materials.One file used for the predictions of COF vs sliding time by using a XGBoost model.The data files are in DWF (raw) and XLSX (processed) format and can be opened by any text editor or spreadsheet software.

Coefficient of Friction Calculation
The Python script "calculate-COF-Al-Al2O3.py" performs several tasks to analyze the friction data for four different materials: AC, C, AE and E. First, it imports some libraries and modules that are needed for data analysis, plotting and file handling.Then, it reads data from several Excel files that contain friction force and time measurements for each material and assigns them to variables.Next, it calculates the COF for each dataset using a formula and a constant normal force of 50 N.After that, it computes the average coefficient of friction and time for each material by taking the mean of the three datasets.It also saves the average values as new Excel files using pandas functions.Finally, it creates two plots that show the average COF versus time for each pair of materials (AC and C, AE and E) using matplotlib functions and options.It applies a style sheet, a shadow effect, labels, legends, limits and fonts to the plots to make them more appealing and informative.It shows the plots on the screen and saves them as PNG files with high resolution.

Coefficient of Friction Prediction using XGBoost
The prediction code used for this study is a python script named "XGB-COF.py".The script uses the XGBoost library to train and test a machine learning model that predicts the coefficient of friction as a function of sliding time for different materials [ 5 , 6 ].The script takes as input one of the four files named "pred_COF_AC.xlsx","pred_COF_C.xlsx","pred_COF_AE.xlsx",and "pred_COF_E.xlsx",which contain the average coefficient of friction and sliding time data for each material.The script can only process one file at a time.The script splits the data into training and testing sets with a ratio of 80:20, and performs a 5-fold cross-validation on the training set to tune the hyperparameters of the XGBoost model.The script then evaluates the performance of the model on the testing set using four metrics: MSE, RMSE, R2, MAE.The script plots and saves the average actual vs predicted coefficient of friction as a function of time for each material in four PNG files of the four materials.The script saves the performance metrics in four TXT files, and also saves the data in four XLSX files of the actual and predicted values of the COF for two different sets (test and validation) of the four materials.
The data were collected by performing dry wear tests at room temperature with linear velocity 0.5 m •s −1 , load 50 N and sliding time of 420 s.Each label of the specimen corresponds to three datasets from pin-on-disk tests with open-cell AlSi10Mg-Al 2 O 3 composite (AC; AE) and open-cell AlSi10Mg material (C; E) with different pore sizes.The labels and their corresponding pore sizes are:

Fig. 1 aFig. 1 .
Fig. 1 a presents the average COF vs time for materials AC and C with pore size 80 0-10 0 0 μm.Fig. 1 b shows the average COF vs time AE and E with pore size 10 0 0-120 0 μm.Both figures in Fig 1 were created by a python script included in the repository "calculate-wear-Al-Al2O3.py".Fig. 2 shows four plots of the average actual vs predicted COF as a function of time for the materials AC, C, AE, and E. All figures presented in Fig 2 were created by a python script included in the repository "XGB-COF.py".The XGBoost model calculated the performance metrics of the

Fig. 2 .
Fig. 2. Average actual vs predicted coefficient of friction as a function of sliding time for the following materials: (a) AC; (b) C; (a) AE; (b) E.
2023 The Author(s).Published by Elsevier Inc.This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )

1. Value of the Data •
These data are useful for understanding the wear behavior of open-cell AlSi10Mg materials and open-cell AlSi10Mg-Al 2 O 3 composites with different pore sizes under dry sliding conditions.• Researchers and engineers who are interested in developing lightweight and wear-resistant materials for various applications such as sliding contact bearings, where low friction and high load-bearing capacity are required, can benefit from these data.• The experimental results can be compared with theoretical models or numerical simulations of wear mechanisms in porous materials.• The effect of pore size and reinforcement on the friction and wear properties of open-cell

Table 1
Test-set and validation-set performance metrics.

Table 2
Descriptive statistics of the average COF (dimensionless) for all materials.

Table 3
Chemical composition of base alloy AlSi10Mg.