Datasets of narrow thermal hysteresis behaviour Ti-Ni-based HT-SMAs and the predicted accumulated local effects

This article refers to data derived from a research article entitled “Prediction of narrow HT-SMA thermal hysteresis behaviour using explainable machine learning” [1]. It is based on the knowledge that alloying Ti-Ni-based shape memory alloys (SMAs) with additional ternary or multicomponent elements can alter the SMAs’ characteristic transformation temperatures, including the thermal hystereses. Two datasets are reported. The first and primary dataset documents experimental Ti-Ni-based shape memory alloys’ high-transformation temperature characteristics reported in the literature. The second auxiliary dataset presented in this article was obtained following the explainable prediction of the narrow high-temperature thermal hysteresis behaviour in Ti-Ni-based high-transformation temperature SMAs (HT-SMAs). The second dataset is intended to generalise and summarise the ML prediction and visualisation of the thermal hysteresis behaviour as also observed experimentally in multiple reports elsewhere. The datasets are provided as supplementary files and the second dataset is also visualised as an intuitive marginal effects plot. We believe that these data will find applications in advancing experimental and theoretical HT-SMA research.


a b s t r a c t
This article refers to data derived from a research article entitled "Prediction of narrow HT-SMA thermal hysteresis behaviour using explainable machine learning" [1] .It is based on the knowledge that alloying Ti-Ni-based shape memory alloys (SMAs) with additional ternary or multicomponent elements can alter the SMAs' characteristic transformation temperatures, including the thermal hystereses.Two datasets are reported.The first and primary dataset documents experimental Ti-Ni-based shape memory alloys' hightransformation temperature characteristics reported in the literature.The second auxiliary dataset presented in this article was obtained following the explainable prediction of the narrow high-temperature thermal hysteresis behaviour in Ti-Ni-based high-transformation temperature SMAs (HT-SMAs).The second dataset is intended to generalise and summarise the ML prediction and visualisation of the thermal hysteresis behaviour as also observed experimentally in multiple reports elsewhere.The datasets are provided as supplementary files and the second dataset is also visualised as an intuitive marginal effects plot.We believe that these data will find applications in advancing experimental and theoretical HT-SMA research.
© 2023 The Author(s  [2] and other compatible ML packages installed] Data format The data reported in this article are in the following formats: i. Raw, tabulated and plotted Description of data collection Two datasets are presented; the main dataset comprises secondary data and the auxiliary dataset is simulated.The main dataset was collected from experimental observations reported in peer-reviewed research articles: The following data entries were captured: i. Identification of experimental Ti-Ni-based alloy, e.g.TiNi, ii.Specification of the Ti-Ni-based alloys' multicomponent elemental composition (in at.%) iii.The reported characteristic transformation temperatures, iv.The derived narrow thermal hysteresis characteristics The additional data reported are not secondary data.The data were acquired following a machine learning computation using the above-described hardware and software.In summary:

Value of the Data
The value of the data is summarised below: • The main dataset is valuable because it consolidates significant prior published data on narrow HT-SMA thermal hysteresis behaviour; • The second dataset is valuable because it improves the generalized understanding of the influence of the Ti-, Ni-, Pd-, Pt-, V-, Hf-, Zr-, Cu-, and Co-contents have on the narrow HT-SMA thermal hysteresis behaviour • Some materials used in the development of Ti-Ni-based HT-SMAs are rare-precious metals that are scarce and prohibitively expensive, researchers working on HT-SMA development can significantly benefit from the availability of these consolidated and simulated datasets.• As demonstrated in the original paper [ 1,2 ], these simulated data agree with prior published data, the datasets will be used to investigate and validate newly proposed narrow HT-SMA candidates.• Finally, these datasets are part of a case study curated to demonstrate the benefits of interpreting ML-predicted outcomes in the advanced materials and engineering field.

Objective
The work grows from a need to develop prediction models that are not only accurate but explainable from a practical metallurgical persuasion.The objective of the work was to produce results and outcomes that communicate usable knowledge to the advanced materials and metallurgical community.The data reported in this article are intended for practitioners -those researchers busy planning, conducting, or interpreting their own experimental or computational work.The consolidated main dataset will find applications in advancing experimental and theoretical HT-SMA developments while the simulated dataset will summarise our findings [1] to those readers -without having them leave their work to try and understand ML, posthoc models, or underlying 'codework'.

Data Description
The data presented in this article are divided into two components.That is, (i) the main dataset which consolidates significant prior published data on narrow HT-SMA thermal hysteresis behaviour of novel Ti-Ni-based HT-SMAs, (ii) an auxiliary dataset simulated based on the related research article findings [ 1 , 3 ], and a figure ( Fig. 1 ) summarising the auxiliary dataset.The data are included as data files.The description of these files has been provided in Table 1 .

Experimental Design and Methods
The main dataset is constructed from HT-SMA thermal hysteresis behaviour observations reported in experimental peer-reviewed research reports.It also builds upon data published by Yamabe-Mitarai [ 6 ], Frenzel, et al. [ 7 ], and Ma, Karaman, and Noebe [ 8 ].The original dataset has at least 297 alloys.After removing non Ti-Ni-based HT-SMAs, entries missing some data, and eliminating alloys with wider HT-SMA hysteresis behaviour ( > 60 °C), the constructed dataset is reduced to 190 Ti-Ni HT-SMAs.
The reported auxiliary data are simulated, and not necessarily secondary data; the data were acquired following machine learning and post-hoc ALE computations using the above-described hardware and software.

(a) An ML-based XG-Boost model was developed using the Scikit-Learn and XGBoost libraries in the Python environment. (b) The model inputs data collected from experimental reports in the public domain to predict
the narrow hysteresis behaviour ( < 60 °C) in novel Ti-Ni HT-SMAs.The data is structured as a dataframe herein each row defines an alloy's multicomponent elemental composition as taught in our previous work [ 3 ].

(c) The ML model development and implementation involves the prediction of the narrow hys-
teresis behaviour in novel Ti-Ni HT-SMA.The model follows a conventional ML-supervised regression where the accuracy of the model is assessed by comparing the ML predictions against true (previously known) values.See Fig. 2 .

Table 1
The list of all data files included.The dataset is also visualized here in the form of a main-effects type first-order ALE plot (see Fig. 1 ).The ALE plot has nine subplots; each subplot is an x-y data sample wherein the x variables are the changing content of an alloying element (in at.%) and the y variable is the effect that change has on the predicted T value.Therefore, each subplot captures the narrowing (downward) or widening (upward) trends each alloying element has on the predicted thermal hysteresis behaviour [ 5 ].Lastly, ALE plots (see Fig. 1 ) were generated and validated/interpreted according to prior experimental findings, e.g.[ 4,[6][7][8][9][10][11][12][13][14][15][16][17].

Ethics Statement
The research work did not involve human or animal subjects or the collection of data from social media or similar platforms.The use of artificial intelligence (AI) tools reported in this article, and the original study, is responsible and in line with the highest ethical standards including Australia's AI Ethics Principles [ 18 ].None of the data were collected from any social media platforms.

Fig. 1 .
Fig. 1.The first-order ALE main effects plots showing the influences of the alloys' elemental on the HT-SMA hysteresis predicted behaviour [ 4 ].

Fig. 2 .
Fig. 2. Regression scatter plot for the predicting ML model; The x-axis represents the actual T values and the y-axis the predicted T values.The dashed red line has diagonal parity and a slope of 1.
(d) The ALE post-hoc analysis technique was implemented in the Python environment using the PyALE package to analyze the marginal effects of each element's composition on the narrow hysteresis behaviour.The implementation of the ALE method produces a dataset showing how see how the ML prediction (y-axis) changes as a function of the predictor features (x-axis) i.e., Ti-, Ni-, Pd-, Pt-, V-, Hf-, Zr-, Cu-, and Co-amounts in the alloy.see Fig. 1 .(e) The dataset has nine sets of simulated samples.More specifically, the y-axis is the predicated expected change in ML predicted T value that affects the predictions of the model (i.e., predicted T) while holding the values of all the other compositions constant.
). Published by Elsevier Inc.This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ) The main dataset was collected from experimental observations reported in peer-reviewed research articles.The auxiliary dataset reported in this article was acquired following the prediction of narrow thermal hysteresis behaviour in Ti-Ni-based HT-SMAs using explainable machine learning and subsequent model-agnostic methodologies, namely accumulated local effects (ALE), and the SHapley Additive exPlanations (SHAP).A custom-built deep learning GPU workstation was used to acquire the auxiliary dataset.
The second dataset is simulated based on an ML model.The dataset has nine tables and figures: i.Each table captures the data on the effect of each alloying element -on the predicted thermal hysteresis behaviour, T ii.Each ALE plot visualizes data captured in the corresponding table ALE plots