Mechanical strength estimation of ultrafine-grained magnesium implant by neural-based predictive machine learning

doi:10.1016/j.matlet.2021.130627

Materials Letters

Volume 305, 15 December 2021, 130627

https://doi.org/10.1016/j.matlet.2021.130627 Get rights and content

Highlights

•
The relation between SPD and the mechanical behavior of the biodegradable Mg was modeled.
•
ANFIS and SVM were employed to estimate σ_YS, σ_UTS, and δ.
•
GEP and GP were utilized to further verify the estimation capability of neural-based predictive ML.
•
The GEP model possessed the lowest error values and it was more accurate than the ANFIS and SVM models.
•
Both ANFIS and SVM-based models achieved high accuracy for predicting mechanical properties of the UFG Mg.

Abstract

The relation between severe plastic deformation (SPD) and the mechanical behavior of the biodegradable magnesium (Mg) implants is not clearly understood yet. Thus, the present study aims to provide, for the first time, a framework for modeling the mechanical features of the ultrafine-grained (UFG) biodegradable Mg-based implant. First, an adaptive neuro-fuzzy inference system (ANFIS) and support vector machine (SVM) were employed to determine relationships between SPD parameters, including the kind of metal forming process, the number of the pass, and temperature of the procedure based on the restricted training dataset. Second, gene expression programming (GEP) and genetic programming (GP) were then used to further verify the estimation capability of neural-based predictive machine learning techniques. Comparison of estimation results with real data confirmed that both ANFIS and SVM-based models had high accuracy for predicting the mechanical behavior of UFG Mg alloys for fracture fixation and orthopedic implants.

Graphical abstract

Introduction

Mg-based alloys are remarkably lightweight metals with a density that is close to that of natural bone and much less than that of conventional Ti-based implants [1]. They have devoted considerable attention owing to their similar mechanical response to those of bone [2], nevertheless, the widespread applications of these biodegradable alloys are still hindered because of their extreme degradation levels and the resultant loss in mechanical strength.

More recently, numerous reports have investigated the potential role of SPD methods in fabricating UFG Mg alloys for fracture fixation and orthopedic implants. Besides, different mathematical modeling approaches were developed, however, some drawbacks should be addressed through new manufacturing strategies and modeling methods [3]. In view of the assorted parameters of the SPD procedures with nonlinear interlinkages, a precise estimation of the material properties dictates plenty of complex interactions that are not commercially viable [4]. Therefore, mechanical properties modeling of the orthopedic implants, for instance, UFG Mg alloys provides the basis for analyzing and adjusting the biomechanical behavior under machine learning (ML) perspective [5].

Despite the importance of developing biomechanical findings based on the ML algorithms, most previous reports have concentrated only on the manufacturing processes and descriptions of the processed alloys [6]. Therefore, this work gives for the first time, a comparative framework between neural-based models (ANFIS and SVM) and genetic-based approaches (GEP and GEP), to forecast the mechanical properties of the severely deformed biodegradable Mg alloy using limited datasets. It should be noted that this is a newborn field of research with lots of challenges and open questions which will be further developed in future work to address accuracy, scalability, and process efficiency.

Section snippets

Material and methods

The ZK60 Mg alloy was acquired as an extruded bar, followed by machining into the desired shape (cylindrical). The details of the manufacturing process are given elsewhere [4]. Fig. 1a-d shows a schematic illustration of the PTCAP tools and process, the experimental output, as well as the dimension of the tensile test specimens. Here, ANFIS was employed and adjusted membership functions and the relevant factors towards the target data sets using the MATLAB software [7].

The ANFIS structure was

Results and discussions

In the present work, the parameters of ANFIS and SVM were learned from the training dataset, where the split was adjusted to 50/50 for training and testing [11]. The mechanical features of the severely deformed biodegradable ZK60 Mg alloy extracted from our previous findings [4] compared to the literature values [12], [13], [14], [15], [16], are tabulated in Table S1 (Supplementary data 1). From this table, ZK60-3 processed alloy exhibited higher mechanical strength than those of the

Conclusions

Four different genetic- and neural-based models were developed to predict the mechanical properties of the severely deformed biodegradable ZK60 Mg alloy. The predictions from the ANFIS and SVM models showed a satisfactory level of forecasting of the mechanical response of the UFG Mg-based implant, with an |R| higher than 0.8 for all simulations. By comparing the genetic-based models with neural-based models, it was found that the GEP model possessed the lowest error values and it was more

CRediT authorship contribution statement

Maohua Li: Writing - original draft, Investigation. Mohsen Mesbah: Conceptualization, Methodology. Alireza Fallahpour: Software, Validation. Bahman Nasiri-Tabrizi: Methodology, Data curation, Writing - original draft. Baoyu Liu: Supervision, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (16)

M.S. Dargusch et al.
Improved biodegradable magnesium alloys through advanced solidification processing
Scr. Mater.
(2020)
G. Faraji et al.
Parallel tubular channel angular pressing (PTCAP) as a new severe plastic deformation method for cylindrical tubes
Mater. Lett.
(2012)
C. Prakash et al.
Characterization of indigenously coated biodegradable magnesium alloy primed through novel additive manufacturing assisted investment casting
Mater. Lett.
(2020)
J.-D. Wu et al.
An expert system of price forecasting for used cars using adaptive neuro-fuzzy inference
Expert Syst. Appl.
(2009)
E. Mostaed et al.
Microstructure, texture evolution, mechanical properties and corrosion behavior of ECAP processed ZK60 magnesium alloy for biodegradable applications
J. Mech. Behav. Biomed.
(2014)
F.-D. Dumitru et al.
ZK60 alloy processed by ECAP: Microstructural, physical and mechanical characterization
Mater. Sci. Eng. A
(2014)
Y. He et al.
Microstructure and mechanical properties of ZK60 alloy processed by two-step equal channel angular pressing
J. Alloys Compd.
(2010)
D. Orlov et al.
Improvement of mechanical properties of magnesium alloy ZK60 by integrated extrusion and equal channel angular pressing
Acta Mater.
(2011)

There are more references available in the full text version of this article.

Cited by (9)

Ultrafine-grained Mg alloy: Preparation, properties, design strategy
2024, Journal of Materials Research and Technology
As the lightest metal structural materials, Mg alloys have great application potential in transportation, aerospace, and biomedicine. However, the further application is hindered by their low strength, plasticity, and poor corrosion resistance. The preparation of Mg alloys with ultra-fine grain (UFG) structure is regarded as a good solution to improve their performance, and has been reported by a large number of literatures. The purpose of this paper is to summarize the development of fabrication processes of UFG Mg alloys and the effects of UFG structure on the mechanical properties and corrosion resistance. In essence, this review aims to elucidate the evolution and current state of UFG structure in Mg alloys. Additionally, it seeks to expound upon the strategies for designing and preparing these structures and shed light on the future development trends. The ultimate goal is to advance the utilization of UFG Mg alloys in both structural and functional materials.
Interpretable machine learning-based analysis of mechanical properties of extruded Mg-Al-Zn-Mn-Ca-Y alloys
2023, Journal of Alloys and Compounds
In this study, the mechanical properties of as-extruded Mg-Al-Zn-Mn-Ca-Y alloys were quantitatively investigated with respect to alloying elements, extrusion temperature, microstructure and texture through interpretable machine learning (IML). To overcome the lack of data, two methods were devised to augment the existing dataset by 39 times using the mean and standard deviation of the measured data. Artificial neural networks predicted room-temperature tensile properties with an accuracy ranging from 0.842 to 0.997 based on R² using 12 predictors for a total of 1179 data points. Shapley additive explanation identified that Al and Mn are the key determinants for strength and elongation, respectively. Partial dependence plots investigated the interaction of all features to understand the quantitative correlation between features. This IML approach revealed that texture, solid solution and secondary particles are related to the main strengthening mechanism of as-extruded Mg alloys. These results can provide insights into the utilization of IML approach to predict material properties and describe key variables for designing lightweight structural metals.
Application of logistic regression, support vector machine and random forest on the effects of titanium dioxide nanoparticles using macroalgae in treatment of certain risk factors associated with kidney injuries
2023, Environmental Research
Citation Excerpt :
Numerous health-promoting properties of marine algae, such as anti-inflammatory, antibacterial, and anticancer properties, have been discovered via the evaluation of their bioactivities in vitro and in vivo (Salamati et al., 2020; Kamyab et al., 2018b; Cheah et al., 2018). Furthermore, marine algae are a rich source of prebiotics, which are complex carbohydrates (Lopez-Santamarina et al., 2020; Mosleh-Shirazi et al., 2022; Kamyab et al., 2014c). Additionally, marine algae are a good source of lipids, dietary fiber, omega-3 fatty acids, minerals, proteins, essential amino acids, polysaccharides, and vitamins A, B, C, and E. Studies showing the anti-inflammatory and antioxidant actions of sulfated polysaccharides (PLS) have led to the conclusion that these compounds are extremely advantageous physiologically (Bilal and Iqbal, 2019; Tavakkoli et al., 2022b; Kamyab et al., 2016b; Putri et al., 2011).
The use of titanium dioxide (TiO₂) nanoparticles in many biological and technical domains is on the rise. There hasn't been much research on the toxicity of titanium dioxide nanoparticles in biological systems, despite their ubiquitous usage. In the current investigation, samples were exposed to various dosages of TiO₂ nanoparticles for 4 days, 1 month, and 2 months following treatment. ICP-AES was used to dose TiO₂ into the tissues, and the results showed that the kidney had a significant TiO₂ buildup. On the other hand, apoptosis of renal tubular cells is one of the most frequent cellular processes contributing to kidney disease (KD). Nevertheless, the impact of macroalgal seaweed extract on KD remains undetermined. In this work, machine learning (ML) approaches have been applied to develop prediction algorithms for acute kidney injury (AKI) by use of titanium dioxide and macroalgae in hospitalized patients. Fifty patients with (AKI) and 50 patients (non-AKI group) have been admitted and considered. Regarding demographic data, and laboratory test data as input parameters, support vector machine (SVM), and random forest (RF) are utilized to build models of AKI prediction and compared to the predictive performance of logistic regression (LR). Due to its strong antioxidant and anti-inflammatory powers, the current research ruled out the potential of using G. oblongata red macro algae as a source for a variety of products for medicinal uses. Despite a high and fast processing of algorithms, logistic regression showed lower overfitting in comparison to SVM, and Random Forest. The dataset is subjected to algorithms, and the categorization of potential risk variables yields the best results. AKI samples showed significant organ defects than non-AKI ones. Multivariate LR indicated that lymphocyte, and myoglobin (MB) ≥ 1000 ng/ml were independent risk parameters for AKI samples. Also, GCS score (95% CI 1.4–8.3 P = 0.014) were the risk parameters for 60-day mortality in samples with AKI. Also, 90-day mortality in AKI patients was significantly high (P < 0.0001). In compared to the control group, there were no appreciable changes in the kidney/body weight ratio or body weight increases. Total thiol levels in kidney homogenate significantly decreased, and histopathological analysis confirmed these biochemical alterations. According to the results, oral TiO₂ NP treatment may cause kidney damage in experimental samples.
Machine learning-based design of biodegradable Mg alloys for load-bearing implants
2023, Materials and Design
Citation Excerpt :
In the field of Mg, both corrosion current density and compressive yield strength of Mg-Ca-Zn alloys were predicted by ANN with an architecture of 4-hidden nodes [17]. Mechanical strength of ultrafine-grained ZK60 implant was estimated by neural-based models such as adaptive neuro-fuzzy inference system, radial basis function and multilayer perceptron [7,8]. It was also found through feed forward neural networks with back propagation algorithm that CO2 and the composition of the buffer system played a crucial role in Mg corrosion [25].
For load-bearing applications, biodegradable Mg alloys require high strength and slow degradation rates to support bone regeneration. This study proposes a design guide of Mg-Zn-Mn-Sr-Ca (ZMJX) alloys for load-bearing Mg implants using machine learning. To this end, it quantitatively investigates the correlation between 4 alloying elements with content of 0–3 wt%, ultimate compressive strength (UCS) and in vitro corrosion rate (CR) in ZMJX alloys. Cascade-forward neural networks predict UCS and CR with high accuracy of over 0.95 for a total of 840 data points. Random forest regression identifies Zn as a major determinant of UCS and CR. Based on this, three chemical compositions are recommended with improved compressive strength and in vitro corrosion resistance by well-verified neural network models. The proposed Mg alloys have UCS of 244–305 MPa and CR of 0.31–0.83 mm/y according to the change of the Zn content. These results can not only provide deep insights into ZMJX alloys, but also recommend a compositional window for load-bearing Mg implants.
Application of machine learning in key properties of medical metal materials
2024, Chinese Journal of Tissue Engineering Research
Prediction of High-Temperature Creep Life of Austenitic Heat-Resistant Steels Based on Data Fusion
2023, Metals

View all citing articles on Scopus

View full text

Mechanical strength estimation of ultrafine-grained magnesium implant by neural-based predictive machine learning

Highlights

Abstract

Graphical abstract

Introduction

Section snippets

Material and methods

Results and discussions

Conclusions

CRediT authorship contribution statement

Declaration of Competing Interest

Scr. Mater.

Mater. Lett.

Mater. Lett.

Expert Syst. Appl.

J. Mech. Behav. Biomed.

Mater. Sci. Eng. A

J. Alloys Compd.

Acta Mater.