Model For Prediction Of Properties Research Articles

Microstructure-informed deep learning model for accurate prediction of multiple concrete properties

Microstructure-informed deep learning model for accurate prediction of multiple concrete properties

The influence mechanism of fractal design method on mechanical properties and corrosion behavior of sheet Gyroid porous structures formed by LPBF

In order to meet the various performance requirements of bone implants, it is necessary to design a porous structure with multi-scale pores. In this paper, a set of 2-level fractal porous structures and 1-level porous structures P80 and P85 were designed based on the sheet-Gyroid porous structure. The influence of porosity and fractal design on the forming quality, mechanical properties and corrosion resistance of the porous structures was investigated. The results showed that the porosity deviation of the 2-level porous structures was in the range of 17.88% to 20.79%, which was about twice that of the 1-level porous structures. The elastic gradient of 2-level fractal porous structures varied from 1.654GPa to 3.636GPa, the compressive offset strength ranged from 50.9MPa to 111.5MPa, and the first maximum compressive strength was in the range of 57.8MPa to 136.6MPa, all of which were lower than those of 1-level porous structures with similar porosity, which extended the design range of mechanical properties of porous structures to a certain extent, and was conducive to avoiding the stress shielding effect of implants. Based on the study of mechanical properties of 2-level fractal porous structures, the mechanical properties prediction models were constructed based on the improved G-A model. In addition, it was found that the corrosion resistance of 2-level fractal porous structures was much higher than that of 1-level porous structures, and the corrosion resistance decreased with the increase of porosity. The fractal design of TPMS structure can effectively expand the design space of porous bone implants.

A Knowledge–Data Dual‐Driven Framework for Predicting the Molecular Properties of Rechargeable Battery Electrolytes

AbstractDeveloping rechargeable batteries that operate within a wide temperature range and possess high safety has become necessary with increasing demands. Rapid and accurate assessment of the melting points (MPs), boiling points (BPs), and flash points (FPs) of electrolyte molecules is essential for expediting battery development. Herein, we introduce Knowledge‐based electrolyte Property prediction Integration (KPI), a knowledge–data dual‐driven framework for molecular property prediction of electrolytes. Initially, the KPI collects molecular structures and properties, and then automatically organizes them into structured datasets. Subsequently, interpretable machine learning further explores the structure–property relationships of molecules from a microscopic perspective. Finally, by embedding the discovered knowledge into property prediction models, the KPI achieved very low mean absolute errors of 10.4, 4.6, and 4.8 K for MP, BP, and FP predictions, respectively. The KPI reached state‐of‐the‐art results in 18 out of 20 datasets. Utilizing molecular neighbor search and high‐throughput screening, 15 and 14 promising molecules, with and without Chemical Abstracts Service Registry Number, respectively, were predicted for wide‐temperature‐range and high‐safety batteries. The KPI not only accurately predicts molecular properties and deepens the understanding of structure–property relationships but also serves as an efficient framework for integrating artificial intelligence and domain knowledge.

A Knowledge-Data Dual-Driven Framework for Predicting the Molecular Properties of Rechargeable Battery Electrolytes.

Developing rechargeable batteries that operate within a wide temperature range and possess high safety has become necessary with increasing demands. Rapid and accurate assessment of the melting points (MPs), boiling points (BPs), and flash points (FPs) of electrolyte molecules is essential for expediting battery development. Herein, we introduce Knowledge-based electrolyte Property prediction Integration (KPI), a knowledge-data dual-driven framework for molecular property prediction of electrolytes. Initially, the KPI collects molecular structures and properties, and then automatically organizes them into structured datasets. Subsequently, interpretable machine learning further explores the structure-property relationships of molecules from a microscopic perspective. Finally, by embedding the discovered knowledge into property prediction models, the KPI achieved very low mean absolute errors of 10.4, 4.6, and 4.8 K for MP, BP, and FP predictions, respectively. The KPI reached state-of-the-art results in 18 out of 20 datasets. Utilizing molecular neighbor search and high-throughput screening, 15 and 14 promising molecules, with and without Chemical Abstracts Service Registry Number, respectively, were predicted for wide-temperature-range and high-safety batteries. The KPI not only accurately predicts molecular properties and deepens the understanding of structure-property relationships but also serves as an efficient framework for integrating artificial intelligence and domain knowledge.

Advances in 3-D printing: polymers, fabrication mechanisms, mass balance models and applications

AbstractIn Science and Engineering, additive manufacturing, which is synonymously known as 3-D printing is a layer-by-layer computer-aided approach that is used to produce high precision commercial industrial products. So far, 3-D printing techniques, which translate computer-aided virtual 3-D models into physical objects, are gaining traction globally. However, the current evolution from conventional molding and machining to rapid prototyping followed by rapid additive manufacturing prompts new challenges for engineers and scientists alike. This review is particularly prompted by novel innovations in 3-D printing with respect to emerging fabrication designs and mechanisms, printable polymeric materials, and current applications. Central to all spotlights is the discussion on 3-D printing techniques discussed under 3 broad categories; (i) extrusion-based methods, (ii) powder binding technologies and (iii) photopolymerization methods and corresponding polymer design aspects, processing parameters and mass balance models. Aspects of mass-balance models addressed in this review include models of multimaterial 3-D printing of viscoelastic inks, Newtonian fluid in extrusion, the Gaussian Beam model and multiscale computational simulations for prediction of macroscale properties of various polymeric materials. This review also highlights selected applications that demonstrate how polymer-based 3-D printing is being exploited in industry, and future perspectives.

Reverse design of molecule‐process‐process networks: A case study from HEN‐ORC system

AbstractThe integrated design of the heat exchanger network (HEN) and organic Rankine cycle (ORC) system with new working fluids is a complex optimization problem. It involves navigating a vast design space across working fluid molecules, ORC processes, and networks. In this article, a new two‐stage reverse strategy is developed. The optimal HEN‐ORC configurations and operating conditions, and the thermodynamic properties of the hypothetical working fluid are identified by an equation of state (EOS) free HEN‐ORC model in the first stage. With two developed group contribution‐artificial neural network thermodynamic property prediction models, working fluid molecules are screened out in the second stage from a database containing more than 430,000 hydrofluoroolefins (HFOs). The presented method is employed in two cases, where new working fluids are found. The total annual cost of Case 1 is 12%–22% lower than the literature, and the power output of Case 2 is 5%–8% higher than the literature.

Effect of Porosity and Gradient Parameters on Compressive Mechanical Properties of Sheet Gyroid Gradient Porous Structures and Construction of Mechanical Properties Prediction Model

A series of uniform porous structures and radial stepwise and radial linear gradient structures are designed based on sheet‐gyroid to explore the influence mechanism of porosity and gradient parameters on forming quality and compressive mechanical properties. On this basis, mathematical models for predicting mechanical properties of porous structures based on uniform, discrete gradient, and linear gradient porosity distribution are established. The volume fraction deviation of uniform porous structures increases gradually with the increase of porosity. The relative density of porous structures ranges from 96.78 to 98.79%. The influence of porosity on compressive mechanical properties is investigated, and a prediction model of the equivalent elastic modulus and yield strength of porous structures is constructed based on the generalized G‐A model. The elastic modulus of the radial stepwise gradient porous structures is predicted by combining the rule of mixing. The deviation between the predicted results and the experimental results ranges from 0.21 to 2.61%. At the same time, a prediction model of the compressive mechanical properties of the radial linear gradient porous structure is constructed, and it is found that the predicted value is somewhat different from the experimental value, which is related to the synergistic strengthening effect of the gradient porous structure.

Hybrid CNC–MXene Nanolubricant for Tribological Application: Characterization, Prediction, and Optimization of Thermophysical Properties Evaluation

In the present work, hybrid Cellulose Nanocrystal–MXene (CNC–MXene) nanolubricants were prepared via a two-step method and investigated as potential heat-transfer hybrid nanofluids for the first time. CNC–MXene nanolubricants were synthesized via a two-step method by varying the weight percentage of CNC–MXene nanoparticles (ranging from 0.01 to 0.05 wt%) and characterized using Fourier-Transform Infrared Spectroscopy and TGA (Thermogravimetric Analysis). Response surface methodology (RSM) was used in conjunction with the miscellaneous design model to identify prediction models for the thermophysical properties of the hybrid CNC–MXene nanolubricant. Minitab 18 statistical analysis software and Response Surface Methodology (RSM) based on Central Composite Design (CCD) were utilized to generate an empirical mathematical model investigating the effect of concentration and temperature. The analysis of variance (ANOVA) results indicated significant contributions from the type of nanolubricant (p < 0.001) and the quadratic effect of temperature (p < 0.001), highlighting non-linear interactions that affect viscosity and thermal conductivity. The findings showed that the predicted values closely matched the experimental results, with a percentage of absolute error below 9%, confirming the reliability of the optimization models. Additionally, the models could predict more than 85% of the nanolubricant output variations, indicating high model accuracy. The optimization analysis identified optimal conditions for maximizing both dynamic viscosity and thermal conductivity. The predicted optimal values (17.0685 for dynamic viscosity and 0.3317 for thermal conductivity) were achieved at 30 °C and a 0.01% concentration, with a composite desirability of 1. The findings of the percentage of absolute error (POAE) reveal that the model can precisely predict the optimum experimental parameters. This study contributes to the growing field of advanced nanolubricants by providing insights into the synergistic effects of CNC and MXene in enhancing thermophysical properties. The developed models and optimization techniques offer valuable tools for tailoring nanolubricant formulations to specific tribological applications, potentially leading to improved efficiency and durability in various industrial settings.

XGBoost-based prediction of electrical properties for anode aluminium foil

Anode aluminum foil (AAF) shows advantages such as high electrical conductivity, high specific capacitance, and low cost, making it a high-quality electrode material for energy storage. However, the formation of AAF is a material process with multiple influencing factors interacted with each other, and it is difficult to analyze the influencing mechanism of each factor and predict the electrical properties. To solve the aforementioned problem, the process data of AAF is collected from a actual production line as the research subject firstly. The correlation coefficient between the process parameters and withstand voltage (WV) is used for feature selection of the data sets. XGBoost-based prediction model is established to predict the electrical properties of AAF, which is evaluated using R2 and root-mean-square error (RMSE). The results indicate that the process parameter X70 (Voltage of repair 3) shows the most significant effect on the WV. The magnitude of the WV shows a stepwise increase with the increase of X70. The result of prediction model is a R2 of 0.990 and a RMSE of 3.888. The XGBoost-based prediction model for the electrical properties (WV) of AAF shows excellent prediction characteristic, which contributes to the optimization of process parameters and the properties prediction.

Topological models of yttrium aluminosilicate glass based on molecular dynamics and structure characterization analysis

AbstractTopological constraint theory (TCT) and molecular dynamics simulations (MD) are utilized to study the effect of Y2O3/SiO2 substitution on the short‐ and medium‐range structures and properties in aluminosilicate glasses containing Y2O3. Experimental methods have been applied to characterize the structures of the as‐prepared glasses to verify the accuracy of MD simulations. It was found that the Y–O bond distance is around 2.38 Å and the Y average coordination number is around 5.78. The introduction of Y2O3 disrupts the Si–O and Al–O bonds in the network structure and elevates the proportion of non‐bridging oxygen. Meanwhile, the calculation results of Si–O and Al–O bond distances show Y shortens the distance between Si, Al cations and oxygen ions, improving the network structure tightness. Furthermore, the glass transition temperature (Tg), Vickers hardness (HV), and elastic modulus (E) of glass increase with the increase of Y2O3 content, while the viscosity decreases. TCT was used to analyze the relationship among glass compositions, structures, and properties, and the properties prediction models for Tg, HV, and E were established. The effectiveness of TCT prediction models was proved by the comparison between the results predicted by the models and the data reported in the literature.

Prediction Models for the Hybrid Effect of Nano Materials on Radiation Shielding Properties of Concrete Exposed to Elevated Temperatures

In modern construction, nanomaterials can be added to concrete to improve its radiation shielding properties. A prediction model for the gamma-ray radiation shielding properties, such as linear attenuation coefficient (LAC) of nanomaterial-based concrete remains necessary. This study aims to develop prediction models for LAC of nano modified concrete (NMC) using nano alumina (NA) and carbon nano tubes (CNTs) subjected to high temperatures. These models are based on linear regression and three machine learning algorithms: M5 Prime (M5P), random forest (RF), and extreme gradient boosting (XGBoost). To build the models, results of 117 concrete cubic specimens (100mm size) was utilized as a dataset with varying amounts of NA and CNTs. Relevant variables including temperature, exposure time, and NA and CNTs dosages were considered. To further validate the predicted results, various statistical metrics and K-fold cross-validation were employed to compare and validate the models' output. The results showed that the XGBoost model outperformed other models, with the highest R2 of 0.9975 and the lowest error of 0.4365%. In addition, the impact of each parameter on LAC of NMC was assessed using Shapley Additive Explanations (SHAP) analysis. Sensitivity and SHAP analyses reveal that CNTs has the highest influence on the radiation shielding properties of NMC. In comparison with the previously developed ANN model, the XGBoost and ANN models showed approximately the same prediction power for predicting LAC of NMC.

A generalized adsorption model of CO2-CH4 in shale based on the improved Langmuir model

CO2 enhanced shale gas recovery (CO2-ESGR), as one of the most promising carbon capture, utilization and storage (CCUS) technologies, can both improve the shale gas production for clean energy supply and mitigate greenhouse gas emissions for the goal of carbon neutrality. The adsorption characteristics of CO2 and CH4 in shale play a key role in the reserve assessment of shale gas and the implementation of CO2-ESGR. The problem of accurate prediction of the adsorption and competitive adsorption characteristics of CO2 and CH4 in shale is still one of the main challenges for CO2-ESGR. To bridge the gap, this work developed a generalized prediction model for the adsorption properties of CH4, CO2, and their mixed gas in shale by improving the Langmuir model, covering a large range of temperature, pressure, and total organic carbon content (TOC). The mathematical models for the key parameters of the Langmuir model, the saturated adsorption capacity Q0 and Langmuir pressure PL, were built for the single-component gas based on their functional relationship with temperature and TOC, and for the mixed gas based on their relationship with Q0 and PL of the single-component gas and CO2 content. Then an improved Langmuir model for the adsorption properties of CO2-CH4 in shale was established. The accuracy of the improved model was verified with a relative deviation below 5 % in a large range of temperature (303–363 K), pressure (0–15 MPa) and TOC (1.06 %-3.83 %) by comparing the calculated adsorption with the experimental data. This generalized adsorption model not only ensures the prediction accuracy of CO2-CH4 adsorption characteristics in shale but also expands its application convenience in a large range of temperatures, pressures and TOC, providing a certain theoretical basis for CO2-ESGR technology.

Machine learning models for predicting the compressive strength of agro-waste stabilized bricks for sustainable buildings

Excessive consumption of time and resources are the major challenges of conducting laboratory experiments to determine the mechanical properties of building/construction materials. There is a need to explore various prediction models for mechanical properties of construction materials. This study is aimed at using machine learning models to predict the compressive strength of stabilized bricks for sustainable buildings. Data for the independent variables generated from laboratory experiments were used for the compressive strength prediction. Several machine learning models were explored using Lazy Predict Python library. Extreme Gradient Boosting Regressor with the coefficient of determination (R2) score of 99.45% and root mean square error of 0.06 outperformed all the tested models. SHAP (SHapley Additive exPlanations) analysis was used to show the distribution of the impacts of the features on the predicted compressive strength. The SHAP analysis results showed that higher percentages of the stabilizers (with reference to 0%, 2%, 4%, and 6% of stabilizers) have a positive impact on the prediction of compressive strength of stabilized bricks. In addition, higher water content, lower lateritic soil contents, lower cube sizes and curing at higher values of temperature favoured the prediction of compressive strength of stabilized bricks. The proposed model can be adapted by civil engineers and researchers for the prediction of the compressive strength of agro-waste stabilized construction materials. The implication of this study will save the time and resources required by the construction professionals to obtain the compressive strength agro-waste stabilized bricks during the production of bricks for sustainable construction applications.

Construction of a prediction model for properties of wear-resistant steel using industrial data based on machine learning approach

Construction of a prediction model for properties of wear-resistant steel using industrial data based on machine learning approach

Hybrid fragment-SMILES tokenization for ADMET prediction in drug discovery

Background:Drug discovery and development is the extremely costly and time-consuming process of identifying new molecules that can interact with a biomarker target to interrupt the disease pathway of interest. In addition to binding the target, a drug candidate needs to satisfy multiple properties affecting absorption, distribution, metabolism, excretion, and toxicity (ADMET). Artificial intelligence approaches provide an opportunity to improve each step of the drug discovery and development process, in which the first question faced by us is how a molecule can be informatively represented such that the in-silico solutions are optimized.Results:This study introduces a novel hybrid SMILES-fragment tokenization method, coupled with two pre-training strategies, utilizing a Transformer-based model. We investigate the efficacy of hybrid tokenization in improving the performance of ADMET prediction tasks. Our approach leverages MTL-BERT, an encoder-only Transformer model that achieves state-of-the-art ADMET predictions, and contrasts the standard SMILES tokenization with our hybrid method across a spectrum of fragment library cutoffs.Conclusion:The findings reveal that while an excess of fragments can impede performance, using hybrid tokenization with high frequency fragments enhances results beyond the base SMILES tokenization. This advancement underscores the potential of integrating fragment- and character-level molecular features within the training of Transformer models for ADMET property prediction.

Unleashing the potential of cell painting assays for compound activities and hazards prediction.

The cell painting (CP) assay has emerged as a potent imaging-based high-throughput phenotypic profiling (HTPP) tool that provides comprehensive input data for in silico prediction of compound activities and potential hazards in drug discovery and toxicology. CP enables the rapid, multiplexed investigation of various molecular mechanisms for thousands of compounds at the single-cell level. The resulting large volumes of image data provide great opportunities but also pose challenges to image and data analysis routines as well as property prediction models. This review addresses the integration of CP-based phenotypic data together with or in substitute of structural information from compounds into machine (ML) and deep learning (DL) models to predict compound activities for various human-relevant disease endpoints and to identify the underlying modes-of-action (MoA) while avoiding unnecessary animal testing. The successful application of CP in combination with powerful ML/DL models promises further advances in understanding compound responses of cells guiding therapeutic development and risk assessment. Therefore, this review highlights the importance of unlocking the potential of CP assays when combined with molecular fingerprints for compound evaluation and discusses the current challenges that are associated with this approach.

3DReact: Geometric Deep Learning for Chemical Reactions.

Geometric deep learning models, which incorporate the relevant molecular symmetries within the neural network architecture, have considerably improved the accuracy and data efficiency of predictions of molecular properties. Building on this success, we introduce 3DReact, a geometric deep learning model to predict reaction properties from three-dimensional structures of reactants and products. We demonstrate that the invariant version of the model is sufficient for existing reaction data sets. We illustrate its competitive performance on the prediction of activation barriers on the GDB7-22-TS, Cyclo-23-TS, and Proparg-21-TS data sets in different atom-mapping regimes. We show that, compared to existing models for reaction property prediction, 3DReact offers a flexible framework that exploits atom-mapping information, if available, as well as geometries of reactants and products (in an invariant or equivariant fashion). Accordingly, it performs systematically well across different data sets, atom-mapping regimes, as well as both interpolation and extrapolation tasks.

Research and Prediction Model for Water Loss Properties in Poly(Magnesium acrylate)/ Cement Hybrid Network Composites

Research and Prediction Model for Water Loss Properties in Poly(Magnesium acrylate)/ Cement Hybrid Network Composites

Application of machine learning models for property prediction to targeted protein degraders

Machine learning (ML) systems can model quantitative structure-property relationships (QSPR) using existing experimental data and make property predictions for new molecules. With the advent of modalities such as targeted protein degraders (TPD), the applicability of QSPR models is questioned and ML usage in TPD-centric projects remains limited. Herein, ML models are developed and evaluated for TPDs’ property predictions, including passive permeability, metabolic clearance, cytochrome P450 inhibition, plasma protein binding, and lipophilicity. Interestingly, performance on TPDs is comparable to that of other modalities. Predictions for glues and heterobifunctionals often yield lower and higher errors, respectively. For permeability, CYP3A4 inhibition, and human and rat microsomal clearance, misclassification errors into high and low risk categories are lower than 4% for glues and 15% for heterobifunctionals. For all modalities, misclassification errors range from 0.8% to 8.1%. Investigated transfer learning strategies improve predictions for heterobifunctionals. This is the first comprehensive evaluation of ML for the prediction of absorption, distribution, metabolism, and excretion (ADME) and physicochemical properties of TPD molecules, including heterobifunctional and molecular glue sub-modalities. Taken together, our investigations show that ML-based QSPR models are applicable to TPDs and support ML usage for TPDs’ design, to potentially accelerate drug discovery.

3DSGIMD: An accurate and interpretable molecular property prediction method using 3D spatial graph focusing network and structure-based feature fusion

A comprehensive representation of molecular structure is essential for establishing accurate and reliable molecular property prediction models. However, fully extracting and learning intrinsic molecular structure information, especially spatial structure features, remains a challenging task, leading that many molecular property prediction models still have no enough accuracy for the real application. In this study, we developed an innovative and interpretable deep learning method, termed 3DSGIMD, which predicted the molecular properties by integrating and learning the spatial structure and substructure information of molecules at multiple levels, and generated the focusing weights by aggregating spatial and adjacency information of molecules to improve understanding of prediction results. We evaluated the model on 10 public datasets and 14 cell-based phenotypic screening datasets. Extensive experimental results indicated that 3DSGIMD achieved superior or comparable predictive performance compared with some existing models, and the individually designed components contributed significantly to the advanced performance of the model. In addition, we also provided insight into the interpretability of our model via visualizing the focusing weights and perturbation analysis, and the results showed that 3DSGIMD can pinpoint crucial local structures and bits of molecular descriptors associated with the predicted properties. In summary, 3DSGIMD is a competitive molecular property prediction method that holds the potential to aid drug design and optimization.