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.     

Predicting and interpreting oxide glass properties by machine learning using large datasets

With the advent of powerful computer simulation techniques, it is time to move from the widely used knowledge-guided empirical methods to approaches driven by data science, mainly machine learning algorithms. We investigated the predictive performance of three machine learning algorithms for six different glass properties. For such, we used an extensive dataset of about 150,000 oxide glasses, which was segmented into smaller datasets for each property investigated. Using the decision tree induction, k-nearest neighbors, and random forest algorithms, selected from a previous study of six algorithms, we induced predictive models for glass transition temperature, liquidus temperature, elastic modulus, thermal expansion coefficient, refractive index, and Abbe number. Moreover, each model was induced with default and tuned hyperparameter values. We demonstrate that, apart from the elastic modulus (which had the smallest training dataset), the induced predictive models for the other five properties yield a comparable uncertainty to the usual data spread. However, for glasses with extremely low or high values of these properties, the prediction uncertainty is significantly higher. Finally, as expected, glasses containing chemical elements that are poorly represented in the training set yielded higher prediction errors. The method developed here calls attention to the success and possible pitfalls of machine learning algorithms. The analysis of the SHAP values indicated the key elements that increase or decrease the value of the modeled properties. It also estimated the maximum possible increase or decrease. Insights gained by this analysis can help empirical compositional tuning and computer-aided inverse design of glass formulations.     

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.     

Modeling of a reaction network and its optimization by genetic algorithm

Continuous endeavors are going on in many research works to find out the strategy to mathematically model and optimize complex reaction networks in order to maximize the main product and at the same time keeping the reactor dimensions within some acceptable limits. The aim of this work is to provide with a strategy for efficient modeling and optimization of reaction networks for reaction controlled processes. Genetic algorithm (GA) has been used for optimizing complex search spaces with multiple optima. Formation of styrene monomer from the ethylbenzene dehydrogenation, with several by-products in a fixed bed reactor, is taken as an example for this study. Two activation energies are found to be the best in term of maximizing styrene productivity.     

Electrical and optical properties of a polyblend electrolyte

A potassium ion conducting polyblend electrolyte based on polyvinyl pyrrolidone (PVP)+polyvinyl alcohol (PVA) complexed with KBrO₃ was prepared using solution cast technique. The electrical conductivity increased with increasing dopant concentration. Optical absorption studies were made in the wavelength range (200-600 nm) on pure (PVP+PVA) and KBrO₃ doped (PVP+PVA) films. The absorption edge was observed at 5.13 eV for undoped (PVP+PVA) while it ranged from 4.88 to 5.0 eV for differently KBrO₃-doped samples. The direct band gaps for undoped and KBrO₃ doped (PVP+PVA) films were found to be, respectively, 5.05 and 4.95, 4.86 and 4.90 eV while the indirect band gaps were 5.03 and 4.88, 4.79 and 4.83 eV, respectively. The absorption edge and the band gaps moved towards lower energies as the dopant concentration was increased up to 20 wt% of the dopant. For further increase in dopant concentration these values started increasing again. This is explained in terms of formation of charge transfer complexes between the dopant and the host matrix. The thermal properties of these films were investigated with differential scanning calorimetry (DSC). The variation in film morphology is examined by scanning electron microscopic examination.     

Prediction of thermodynamic properties of polyvinylpyrrolidone solutions by artificial neural networks coupled with genetic algorithm

In the present study, the artificial neural networks coupled with the genetic algorithm (ANN–GA) models were used to predict the thermodynamic properties of polyvinylpyrrolidone (PVP) solutions in water and ethanol at various temperatures, mass fractions, and molecular weights of polymer. The genetic algorithm (GA) was used to find the best weights and biases of the network and improve the performance of ANNs. The proposed model was composed of three input variables including the temperature of the solution, the mass fraction, and molecular weight of the polymer. Density, viscosity, and surface tension of PVP solutions with various molecular weights (10,000, 25,000, and 40,000) in water and ethanol have been measured in the temperature range 20–55°C and various mass fractions of polymer. The ANN–GA models were trained by the experimental datasets and the prediction of density, surface tension, and viscosity of PVP solutions was performed using these models. The predicted values were compared with the experimental ones and the mean absolute relative error was less than 0.5% for the density and surface tension and about 3% for the viscosity of solutions.     

Optimization of short-time gasoline blending scheduling problem with a DNA based hybrid genetic algorithm

Gasoline blending is a key process in the petroleum refinery industry posed as a nonlinear optimization problem with heavily nonlinear constraints. This paper presents a DNA based hybrid genetic algorithm (DNA-HGA) to optimize such nonlinear optimization problems. In the proposed algorithm, potential solutions are represented with nucleotide bases. Based on the complementary properties of nucleotide bases, operators inspired by DNA are applied to improve the global searching ability of GA for efficiently locating the feasible domains. After the feasible region is obtained, the sequential quadratic programming (SQP) is implemented to improve the solution. The hybrid approach is tested on a set of constrained nonlinear optimization problems taken from the literature and compared with other approaches. The computation results validate the effectiveness of the proposed algorithm. The recipes of a short-time gasoline blending problem are optimized by the hybrid algorithm, and the comparison results show that the profit of the products is largely improved while achieving more satisfactory quality indicators in both certainty and uncertainty environment.     

遗传-分布估计算法求解化工生产中一类带多工序的异构并行机调度问题

针对化工生产中广泛存在的一类带多工序的异构并行机调度问题,即部分产品需多工序加工,同时不同产品间带序相关设置时间的异构并行机调度问题（heterogeneous parallel machine scheduling problem with multiple operations and sequence-dependent setup times, HPMSP_MOSST）,提出了一种遗传-分布估计算法（genetic algorithm-estimation of distribution algorithm, GA-EDA）,用于优化最早完工时间（makespan）。首先,提出了一种基于GA的概率模型训练机制,用来提高概率模型在算法进化初期的信息积累量,进而提高搜索的效率;其次,设计了一种有效的GA与EDA混合策略,使得算法的全局探索和局部开发能力得到合理平衡。计算机模拟验证了GA-EDA的有效性和鲁棒性。     

Polyamides obtained by direct polycondesation of 4‐[4‐[9‐[4‐(4‐aminophenoxy)‐3‐methyl‐phenyl] fluoren‐9‐YL]‐2‐methyl‐phenoxy]aniline with dicarboxylic acids based on a diphenyl‐silane moiety

Polyamides (PAs) containing fluorene, oxyether, and diphenyl‐silane moieties in the repeating unit were synthesized in > 85% yield by direct polycondesation between a diamine and four dicarboxylic acids. Alternatively, one PA was synthesized from an acid dichloride. The diamine 4‐[4‐[9‐[4‐(4‐aminophenoxy)‐3‐methyl‐phenyl]fluoren‐9‐yl]‐2‐methyl‐phenoxy]aniline ( 3 ) was obtained from the corresponding dinitro compound, which was synthesized by nucleophilic aromatic halogen displacement from p‐chloronitrobenzene and 9,9‐bis (4‐hydroxy‐3‐methyl‐phenyl)fluorene ( 1 ). Monomers and polymers were characterized by FTIR and ¹H, ¹³C, and ²⁹Si‐NMR spectroscopy and the results were in agreement with the proposed structures. PAs showed inherent viscosity values between 0.14 and 0.43 dL/g, indicative of low molecular weight species, probably of oligomeric nature. The glass transition temperature (T_g) values were observed in the 188–211°C range by DSC analysis. Thermal decomposition temperature (TDT_10%) values were above 400°C due to the presence of the aromatic rings in the diamine. All PAs showed good transparency in the visible region (>88% at 400 nm) due to the incorporation of the fluorene moiety. © 2010 Wiley Periodicals, Inc. J Appl Polym Sci, 2011