Osvaldo N. Oliveira Jr.*, Jose F. Rodrigues Jr., Maria Cristina F. de Oliveira**

*São Carlos Institute of Physics, University of São Paulo (USP), CP 369, 13560-970 São Carlos, SP, Brazil

**Institute of Mathematics and Computer Science, University of São Paulo (USP), CP 668

13560-970 São Carlos, SP, Brazil

As we approach the end of the second decade of the 21st century, we may envisage a completely different paradigm for generating knowledge, which might become a reality in a few decades. According to this new paradigm, artificial systems (machines!) will be able to generate knowledge; that is to say, for the first time in history, knowledge would be created without human intervention. This prediction is based upon developments in machine learning following decades of intense research, which have achieved innovative leaps in recent years. Most notably, relevant progress in prediction schemes, classification methods, and advanced modelling have made it possible for machines to outperform humans in various intellectually-demanding tasks.

Very stringent requirements must be met before this paradigm becomes a reality, of which perhaps the most important ones are associated with challenges in the so-called Big Data methodologies. The first step towards having autonomous machines capable of generating knowledge is the fostering of data-intensive discovery. In fact, the challenges in storing, managing, sharing and mining massive amounts of data are far from trivial. Several layers of resources and tools are required, which include enormous storage, data security, information management, and, overall, efficacious machine intelligence.

These new concepts can obviously be applied to any field, and materials science is no exception. Any technology based upon complex devices or hardware relies on materials science. With the large body of data detected by a diversity of computerized sensors during experimental testing, it is essential that automatic tools be designed so that Big Data translates into useful knowledge. Many efforts are being made around the world; therefore, it is time to analyse what has been accomplished in a collection of selected papers dealing with machine learning applied to materials science.

In this article selection, the reader will find compelling examples from different areas of materials science. Just to give a flavour of what has been compiled, we mention the diversity of materials and applications in a sample of interesting papers. From the development of concrete for civil construction, as presented in the paper Machine learning in concrete strength simulations: Multi-nation data analytics, by Chou et al., to the design of lithium-ion batteries, as explained in Application of machine learning methods for the prediction of crystal system of cathode materials in lithium-ion batteries, by Shandiz and Gauvin. Common to these papers is evidence that machine learning does enhance human capability in predicting the properties of materials. This is also beautifully illustrated in the proposal Material synthesis and design from first principle calculations and machine learning, by Takahashi and Tanaka, who use a database of materials and their properties to “teach” machine learning methods to  predict new materials with desirable traits; a similar approach is adopted by Khan, Shamsi and Choi in the contribution entitled Correlating dynamical mechanical properties with temperature and clay composition of polymer-clay nanocomposites, which relies on support vectors and artificial neural networks to identify non-linear correlations between temperature and composition and mechanical properties in polymer-clay nanocomposites.

The achievements highlighted in this selection include work heavily- focused on materials; as observed in power generation, a topic touched upon in Multi-Model Ensemble for day ahead prediction of photovoltaic power generation, by Pierro et al.; or in the case of medicine, as described in the paper Human breath-print identification by E-nose, using information-theoretic feature selection prior to classification, by Wang  et al., who demonstrate how the breath of a patient has the potential to enable the detection of clinical conditions based on machine learning classification. Two papers demonstrate that computer vision might be improved by knowledge coming from machinery, as in the work A computer vision approach for automated analysis and classification of microstructural image data, by DeCost and Holm; and in the contribution Image driven machine learning methods for microstructure recognition, by Chowdhury et al. In these two later papers, the task of identifying the types of microstructures is performed by employing numeric visual features used to feed machine-driven methods.

We trust the novel ideas, concepts and results compiled in this issue will inspire further investigation on how to enable machines to “see” and infer beyond the ability of humans, which is highly constrained by limited perception, memory and reasoning. Such investigations are essential to pave the way towards better and more affordable solutions to critical problems affecting the quality of everyday human life.


Article Selection

Application of neural networks to predict the elevated temperature flow behavior of a low alloy steel
Lin Y.C., Zhang J., Zhong J.
Computational Materials Science, 2008    

Optimization of process conditions in casting aluminum matrix composites via interconnection of artificial neurons and progressive solutions
Shabani M.O., Mazahery A   
Ceramics International, 2012

Stochastic molecular descriptors for polymers. 4. Study of complex mixtures with topological indices of mass spectra spiral and star networks: The blood proteome case
Cruz-Monteagudo M., Munteanu C.R., Borges F., Cordeiro M.N.D.S., Uriarte E., Chou K.-C., González-Díaz H.   
Polymer, 2008

Estimation of exposed temperature for fire-damaged concrete using support vector machine
Chen B.-T., Chang T.-P., Shih J.-Y., Wang J.-J.                 
Computational Materials Science, 2009

Correlating dynamical mechanical properties with temperature and clay composition of polymer-clay nanocomposites          
Khan A., Shamsi M.H., Choi T.-S.                 
Computational Materials Science, 2009

Lattice constant prediction of orthorhombic ABO3 perovskites using support vector machines
Javed S.G., Khan A., Majid A., Mirza A.M., Bashir J.                    
Computational Materials Science, 2007

Two semi-empirical approaches for the prediction of oxide ionic conductivities in ABO3 perovskites
Xu L., Wencong L., Chunrong P., Qiang S., Jin G.
Computational Materials Science, 2009

A computer vision approach for automated analysis and classification of microstructural image data
Decost B.L., Holm E.A
Computational Materials Science, 2015

Evaluation of machine learning interpolation techniques for prediction of physical properties
Bélisle E., Huang Z., Le Digabel S., Gheribi A.E.                
Computational Materials Science, 2015        

Data mining our way to the next generation of thermoelectrics
Sparks T.D., Gaultois M.W., Oliynyk A., Brgoch J., Meredig B.
Scripta Materialia, 2016        

An implementation of artificial neural-network potentials for atomistic materials simulations: Performance for TiO2
Artrith N., Urban A.   
Computational Materials Science, 2016        

Material synthesis and design from first principle calculations and machine learning
Takahashi K., Tanaka Y.
Computational Materials Science, 2016

PV power forecast using a nonparametric PV model
Almeida M.P., Perpiñán O., Narvarte L.                   
Solar Energy, 2015

Machine learning in concrete strength simulations: Multi-nation data analytics
Chou J.-S., Tsai C.-F., Pham A.-D., Lu Y.-H.           
Construction and Building Materials, 2014

Human breath-print identification by E-nose, using information-theoretic feature selection prior to classification
Wang X.R., Lizier J.T., Berna A.Z., Bravo F.G., Trowell S.C.                      
Sensors and Actuators, B: Chemical, 2015

CaPrM: Carbonation prediction model for reinforced concrete using machine learning methods
Taffese W.Z., Sistonen E., Puttonen J.                       
Construction and Building Materials, 2015

Image driven machine learning methods for microstructure recognition
Chowdhury A., Kautz E., Yener B., Lewis D.
Computational Materials Science, 2016

Multi-Model Ensemble for day ahead prediction of photovoltaic power generation
Pierro M., Bucci F., De Felice M., Maggioni E., Moser D., Perotto A., Spada F., Cornaro C.      
Solar Energy, 2016

Non-metallic coating thickness prediction using artificial neural network and support vector machine with time resolved thermography
Wang H., Hsieh S.-J., Peng B., Zhou X.                    
Infrared Physics and Technology, 2016        

Multi-scale parallel temperature error processing for dual-mass MEMS gyroscope
Shen C., Li J., Zhang X., Tang J., Cao H., Liu J.                  
Sensors and Actuators, A: Physical, 2016     

Application of machine learning methods for the prediction of crystal system of cathode materials in lithium-ion batteries   
Attarian Shandiz M., Gauvin R.        
Computational Materials Science, 2016

A comparison of stress in cracked fibrous tissue specimens with varied crack location, loading, and orientation using finite element analysis
Peloquin J.M., Elliott D.M.               
Journal of the Mechanical Behavior of Biomedical Materials, 2016

Automated voxelization of 3D atom probe data through kernel density estimation
Srinivasan S., Kaluskar K., Dumpala S., Broderick S., Rajan K.                 
Ultramicroscopy, 2015          

Evaluation of the splitting tensile strength in plain and steel fiber-reinforced concrete based on the compressive strength
Behnood A., Verian K.P., Modiri Gharehveran M.               
Construction and Building Materials, 2015