Predictive modelling for contact angle of liquid metals and oxide ceramics by comparing Gaussian process regression with other machine learning methods

Wettability has a major effect on the performance of the corrosion of ceramic refractory under normal operating conditions. Contact angle measurement is available to characterize the wettability of liquid metals and oxide ceramics. Therefore, it is necessary to develop a contact angle prediction model with generalizability. This work emphasizes on developing a model for predicting the contact angle of a liquid metal with a solid oxide and analyzes the influence of factors affecting the contact angle when contact angle is predicted. In this paper, six contact angle prediction models are developed based on machine learning methods and contact angle data from the previous literature. The comparison between six contact angle prediction models evidences that the gaussian process regression (GPR) model has the best prediction accuracy and reaching 96%. Furthermore, the comparative results indicate that when surface energy of metal, surface energy of oxide, formation free energy of oxide, and bandgap energy of oxide are ignored respectively, the prediction accuracy of the model decreases by 4%, 3%, 1% and 1% respectively.     

Predicting oxidation damage of ultra high-temperature carbide ceramics in extreme environments using machine learning

Determining the oxidation resistance of UHTC carbides in extreme environments is challenging theoretically and experimentally due to the high dimensional complexity of influencing variables and intricate testing setups. Herein we demonstrate the use of machine learning (ML) models trained with experimental literature data to predict the oxide thickness of UHTC carbides exposed to air based on composition, mean grain size, relative densification, holding time, and temperature. A multi-dimensional database with 76 occurrences is created containing experimental results of Hf, Zr, and Ta carbides plus additives. The preprocessed database is then used to train ML models to predict their oxidation behavior. The trained model predicts the oxidation damage in the form of an average oxide thickness in UHTC carbides with a Mean Absolute Error (MAE) of ±65.45 μm for samples in the testing set that developed thicknesses up to 1000 μm. The model successfully predicted oxidation damage for a recession rate lower than 60 μm/min. It is noticed that the ensemble method MAE is increased to ±134.34 μm while forecasting the oxidation of samples with a recession rate higher than the threshold. The unprecedented approach is a novel way to predict the damage through the oxidation of carbide compounds before processing for a smarter design with room for improvement.     

Designing optical glasses by machine learning coupled with a genetic algorithm

Engineering new glass compositions have experienced a sturdy tendency to move forward from (educated) trial-and-error to data- and simulation-driven strategies. In this work, we developed a computer program that combines data-driven predictive models (in this case, neural networks) with a genetic algorithm to design glass compositions with desired combinations of properties. First, we induced predictive models for the glass transition temperature (T_g) using a dataset of 45,302 compositions with 39 different chemical elements, and for the refractive index (n_d) using a dataset of 41,225 compositions with 38 different chemical elements. Then, we searched for relevant glass compositions using a genetic algorithm informed by a design trend of glasses having high n_d (1.7 or more) and low T_g (500 °C or less). Two candidate compositions suggested by the combined algorithms were selected and produced in the laboratory. These compositions are significantly different from those in the datasets used to induce the predictive models, showing that the used method is indeed capable of exploration. Both glasses met the constraints of the work, which supports the proposed framework. Therefore, this new tool can be immediately used for accelerating the design of new glasses. These results are a stepping stone in the pathway of machine learning-guided design of novel glasses.     

Ultra-high temperature ceramics melting temperature prediction via machine learning

Melting temperature has great influence on the high temperature properties and working temperature limits of ultra-high temperature ceramics (UHTCs) In order to bypass the challenge in the measurement of ultra-high melting points, this paper proposed a novel method to predict UHTCs melting temperature via machine learning. A dataset including more than ten thousand melting temperature data has been established, which covers 8 elements and most of the known non-oxide UHTCs. We built up an element to ceramic system framework by back propagation artificial neural network (BPANN) with the accuracy approaching to 90% and the correlation coefficients approaching to 0.95. Our work provides a probability to get the high accuracy melting temperature of UHTCs, and a more convenient way to develop novel materials with higher working temperature. The given case of melting temperature prediction of Hf-C-N ceramics proves the generality of the artificial neural network (ANN). An inter-validation of melting temperature prediction using our network with materials thermodynamics and density functional theory (DFT) has been demonstrated, indicating that our network is of powerful prediction ability.     

Predicting oxidation damage in ultra high-temperature borides: A machine learning approach

Ultra-high temperature (UHT) borides are ceramics materials with melting points above 3000 °C for structural applications in extreme environments. However, at temperatures exceeding 1600 °C and under oxidizing conditions, the material suffers from detrimental degradation. Optimized design and performance of diboride materials under such extreme conditions requires filling the missing composition-microstructure-oxidation gap. This study proposes a computational data-driven framework to connect the processing and microstructure of Ultra-high temperature borides with the oxidation damage. Random Forest Regressor (RFR) model is adopted to forecast the oxide scale thickness developed after oxidation testing based on processing variables and microstructural features. The model trained on a dataset consisting of 107 samples of experimental data extracted from the literature aims to predict oxidation damage. With proper data manipulation and fine model tuning, the predictor could forecast the oxide scale thickness of UHT diborides with a Mean Absolute Error of 37.45 μm and an R-square of 0.83. This model could be used as a high-throughput scheme to design and test new UHT diborides materials computationally. A model with larger composition capabilities could also be developed in the future as more experimental data become available.     

Prediction of bending strength of Si3N4 using machine learning

The bending strength of silicon nitride (Si₃N₄) plays a vital role in its application and is influenced by various process factors. Current experimental methods for investigating Si₃N₄ ceramics exhibiting low efficiency and high cost are incapable of systematically analysing the effect of process factors on the bending strength of Si₃N₄ ceramics and quantitatively predicting the optimum process parameters. In this study, machine learning (ML) approaches based on extreme gradient boosting (XGBoost) were applied to predict and analyse the bending strength of Si₃N₄ ceramics. Because the classification model of XGBoost is easily interpretable, the factors affecting the bending strength could be quantitatively evaluated. The current model can provide a suitable order of adding sintering additives to obtain excellent bending strength in Si₃N₄ ceramics. Although this study focuses on the bending strength of Si₃N₄ ceramics, the new approach reported herein is applicable for the in silico design and analysis of other ceramic materials.     

Additives distribution and electrical properties in zinc oxide varistor prepared by a wet chemical method

Cobalt-, praseodymium-added zinc oxide varistor was prepared through a wet chemical method followed by sintering with or without calcination. Changes in grain size, compact density, additives distribution, and voltagecurrent/ capacitance-voltage relations were investigated for the characterization of the samples sintered at temperatures from 1473 to 1573 K without calcination or with calcination at 773 K for 2 h. The electrical properties were compared with nhose of samples prepared by two types of ball mill methods. The wet chemical method provided almost the same additives-distribution profile and less impurities in comparison with the ball grinding method carried out for 10–100 h. The donor concentration and the potential-barrier height for the samples were evaluated by Double Schottky Barrier Model. Addition of small amount of both cobalt and praseodymium in preparation by the wet chemical method was effective for a better nonlinearity relation between voltage and current, which has potential for a smaller sized varistor.     

Prediction of bio-oil yield during pyrolysis of lignocellulosic biomass using machine learning algorithms

This work aims to implement and use machine learning algorithms to predict the yield of bio-oil during the pyrolysis of lignocellulosic biomass based on the physicochemical properties and composition of the biomass feed and pyrolysis conditions. The biomass pyrolysis process is influenced by different process parameters, such as pyrolysis temperature, heating rate, composition of biomass, and purge gas flow rate. The inter-relation between the yield of different pyrolysis products and process parameters can be well predicted by using different machine learning algorithms. In this study, different machine learning algorithms, namely, multi-linear regression, gradient boosting, random forest, and decision tree, have been trained on the dataset and the models are compared to identify the optimum method for the determination of bio-oil yield prediction model. Analysis of the results showed the gradient boosting method to possess a regression score of 0.97 and 0.89 for the training and testing sets with root-mean-squared error (RMSE) values of 1.19 and 2.39, respectively, and overcome the problem of overfitting. Therefore, the present study provides an approach to train a generalized machine learning model, which can be employed on large datasets while avoiding the error of overfitting.     

A new estimation algorithm of physical properties based on a group contribution and support vector machine

There are two ways to evaluate the properties of unknown chemical compounds. One is by traditional approaches, which measure the desired data from the experiments and the other is by predicting them in the theoretical approaches using a kind of prediction model. The latter are considered to be more effective because they are less time consuming and cost efficient, and there is less risk in conducting the experiments. Besides, it is inconvenient to conduct experiments to obtain experimental data, especially for new materials or high molecular substances. Several methods using regression model and neural network for predicting the physical properties have been suggested so far. However, the existing methods have many problems in terms of accuracy and applicability. Therefore, an improved method for predicting the properties is needed. A new method for predicting the physical property was proposed to predict 15 physical properties for the chemicals which consist of C, H, N, O, S and Halogens. This method was based on the group contribution method that was oriented from the assumption that each fragment of a molecule contributes a certain amount to the value of its physical property. In order to improve the accuracy of the prediction of the physical properties and the applicability, we extended the database, significantly modifying the existing group contribution methods, and then established a new method for predicting the physical properties using support vector machine (SVM) which is a statistical theory that has never been used for predicting the physical properties. The SVM-based approach can develop nonlinear structure property correlations more accurately and easily in comparison with other conventional approaches. The results from the new estimation method are found to be more reliable, accurate and applicable. The newly proposed method can play a crucial role in the estimation of new compounds in terms of the expense and time.     

Machine learning of materials design and state prediction for lithium ion batteries

With the widespread use of lithium ion batteries in portable electronics and electric vehicles,further improvements in the performance of lithium ion battery materials and accurate prediction of battery state are of increasing interest to battery researchers.Machine learning,one of the core technologies of artificial intelligence,is rapidly changing many fields with its ability to learn from historical data and solve complex tasks,and it has emerged as a new technique for solving current research problems in the field of lithium ion batteries.This review begins with the introduction of the conceptual framework of machine learning and the general process of its application,then reviews some of the progress made by machine learning in both improving battery materials design and accurate prediction of battery state,and finally points out the current application problems of machine learning and future research directions.It is believed that the use of machine learning will further promote the large-scale application and improve-ment of lithium-ion batteries.     

Physical and structural properties of Ge-rich chalcogenide glass sandwiched by GeS crystalline layers

Ge-rich glass-ceramics sandwiched by GeS crystalline layers were fabricated through 10?h thermal treatments at different temperatures. Surface crystallization is evidenced by XRD investigation of glass-ceramic samples polished by different times. SEM observation shows that the thickness of the crystalline layer is about 100?µm for the sample thermally treated at 395?°C for 10?h. The physical properties, including transmission spectra, density, Vickers hardness, and thermal expansion coefficient, were characterized and discussed with the evolution of GeS crystalline layers. This work not only establishes the correlation between microstructure and physical properties of chalcogenide glass-ceramics sandwiched by GeS, but also provides important evidence of structural similarity for understanding the network structure of Ge-S chalcogenide glasses.     

Predicting nepheline precipitation in waste glasses using ternary submixture model and machine learning

Nepheline precipitation in nuclear waste glasses during vitrification can be detrimental due to the negative effect on chemical durability often associated with its formation. Developing models to accurately predict nepheline precipitation from compositions is important for increasing waste loading since existing models can be overly conservative. In this study, an expanded dataset of 955 glasses, including 352 high-level waste glasses, was compiled from literature data. Previously developed submixture models were refitted using the new dataset, where a misclassification rate of 7.8% was achieved. In addition, nine machine learning (ML) algorithms (k-nearest neighbor, Gaussian process regression, artificial neural network, support vector machine, decision tree, etc.) were applied to evaluate their ability to predict nepheline precipitation from glass compositions. Model accuracy, precision, recall/sensitivity, and F1 scores were systemically compared between different ML algorithms and modeling protocols. Model prediction with an accuracy of ~0.9 (misclassification rate of ~10%) was observed for different algorithms under certain protocols. This study evaluated various ML models to predict nepheline precipitation in waste glasses, highlighting the importance of data preparation and modeling protocol, and their effect on model stability and reproducibility. The results provide insights into applying ML to predict glass properties and suggest areas for future research on modeling nepheline precipitation.     

Application of machine learning to process simulation of n-pentane cracking to produce ethylene and propene

Modeling light olefin production was one of the main concerns in chemical engineering field. In this paper, machine learning model based on artificial neural networks(ANN) was established to describe the effects of temperature and catalyst on ethylene and propene formation in n-pentane cracking. The establishment procedure included data pretreatment, model design, training process and testing process, and the mean square error(MSE) and regression coefficient(R~2) indexes were employed to evaluate model performance. It was found that the learning algorithm and ANN topology affected the calculation accuracy. GD24223, CGB2423, and LM24223 models were established by optimally matching the learning algorithm with ANN topology, and achieved excellent calculation accuracy. Furthermore, the stability of GD24223, CGB2423 and LM24223 models was investigated by gradually decreasing training data and simultaneously transforming data distribution. Compared with GD24223 and LM24223 models, CGB2423 model was more stable against the variations of training data, and the MSE values were always maintained at the magnitude of 10~(-3)–10~(-4), confirming its applicability for simulating light olefin production in n-pentane cracking.     

Next generation pure component property estimation models: With and without machine learning techniques

Physiochemical properties of pure components serve as the basis for the design and simulation of chemical products and processes. Models based on the molecular structural information of chemicals for the following 25 pure component properties are presented in this work: (critical-) temperature, pressure, volume, acentric factor; (normal-) boiling point, melting point, auto-ignition temperature; flash point; (standard-) enthalpy of formation, Gibbs energy of formation, enthalpy of fusion, enthalpy of vaporization, liquid molar volume; (environmental-) (lethal dose-) LC50 and LD50, photo-chemical oxidation potential, bioconcentration factor, permissible exposure limit; (physicochemical-) acid dissociation constant, water-solubility, octanol–water partition coefficient, Hildebrandt solubility parameter, Hansen solubility parameters. Utilizing functional groups for molecular representation, two parallel property estimation models where the group contributions for each property are regressed through traditional regression techniques and machine learning techniques are presented. Both techniques use an a priori data analysis before regression of model parameters. A dataset with more than 24,000 chemicals for the 25 pure component properties has been utilized for the development of the two sets of property models. The efficacy of the developed models and their use are highlighted together with a discussion on the overall performance, application range, and predictive capabilities with implications to product and/or process engineering problem solutions.     

基于多元时序驱动的ELM全流程故障预测及其应用

多元时序驱动建模方法是复杂系统故障预测和系统状态评估的一种有效途径, 其中人工神经网络作为一种数据驱动的处理非线性问题的有效建模工具, 近年来在处理多元时序建模这个问题上得到了较广泛的关注。从全流程的角度出发, 首先, 运用k-近邻互信息方法对多元时序变量进行降维与相关性计算, 从而选择特征变量;其次, 提出了一种改进的趋势分析方法对系统的状态进行实时监测, 并对系统运行状态进行有效细分;最后, 针对系统潜在故障阶段, 应用极限学习机(extreme learning machine, ELM)神经网络方法对其进行故障预测。通过对青霉素发酵过程(penicillin fermentation process)进行仿真实验, 结果验证了所提方法的有效性。     

Effect of boron oxide on mechanical and thermal properties of bioactive glass coatings for biomedical applications

Bioactive glass coatings can improve the osteo integration of metallic implants with the host tissue, thereby increasing their lifespan and overall success rate. However, complex composition-structure-property relations in phosphosilicate-based bioactive glasses make experimental determination of these relations and related composition design of bioactive coatings challenging. By applying molecular dynamics (MD)-based atomistic simulations with recently developed effective potentials, this work addresses the challenge by using a material genome approach to obtain the composition and structure effects on various key properties for bioactive coating applications. A series of potential bioactive glass compositions were studied and the composition effects on the mechanical and thermal properties that are critical to these bioactive glasses as a coating to metallic implants were calculated. Particularly, by varying the level of B₂O₃ to SiO₂ substitutions, the effect of composition on various key properties was elucidated. It was found that by using cation in a 1 to 1 ratio (BO_3/2 to SiO₂) instead of the commonly used substitutions (B₂O₃ to SiO₂), the composition effect can be more clearly expressed and, hence, recommended in future composition designs. Together with careful structural analysis, the origin of property changes can be elucidated. The atomistic computer simulation-based approach is, thus, an effective way to guide future bioactive glass designs for bioactive coatings and other applications.     

Interpreting the optical properties of oxide glasses with machine learning and Shapely additive explanations

Due to their excellent optical properties, glasses are used for various applications ranging from smartphone screens to telescopes. Developing compositions with tailored Abbe number (V_d) and refractive index at 587.6 nm (n_d), two crucial optical properties, is a major challenge. To this extent, machine learning (ML) approaches have been successfully used to develop composition–property models. However, these models are essentially black boxes in nature and suffer from the lack of interpretability. In this paper, we demonstrate the use of ML models to predict the composition-dependent variations of V_d and n_d. Further, using Shapely additive explanations (SHAP), we interpret the ML models to identify the contribution of each of the input components toward target prediction. We observe that glass formers such as SiO₂, B₂O₃, and P₂O₅ and intermediates such as TiO₂, PbO, and Bi₂O₃ play a significant role in controlling the optical properties. Interestingly, components contributing toward increasing the n_d are found to decrease the V_d and vice versa. Finally, we develop the Abbe diagram, using the ML models, allowing accelerated discovery of new glasses for optical properties beyond the experimental pareto front. Overall, employing explainable ML, we predict and interpret the compositional control on the optical properties of oxide glasses.