Search

This document is a special issue of the journal "Materials" that focuses on the recent progress in the structure transition, processing, characterization, and applications of metallic materials. The issue includes research papers on various topics such as the development of dispersion-strengthened copper alloys, the characterization of carbide precipitation in steel, the effect of magnetic fields on aluminum bronze, and the enhancement of impact toughness in cast iron through heat treatment. The issue aims to provide comprehensive and up-to-date information for researchers, engineers, and industry experts in the field of metallic materials. [Extracted from the article]

Sensitivity analysis of influencing factors on frost resistance is carried out in this paper, and a two-stage neural network model based on grey theory and Back Propagation Neural Networks (BPNNs) is established for the sake of predicting the frost resistance of active-admixture recycled concrete quickly and accurately. Firstly, the influence degree of cement, water, sand, natural aggregate, recycled aggregate, mineral powder, fly ash, fiber and air-entraining agent on the frost resistance of active-admixture recycled-aggregate concrete was analyzed based on the grey system theory, and the primary and secondary relationships of various factors were effectively distinguished. Then, the input layer of the model was determined as cement, water, sand, recycled aggregate and air-entraining agent, and the output layer was the relative dynamic elastic modulus. A total of 120 datasets were collected from the experimental data of another author, and the relative dynamic elastic modulus was predicted using the two-stage BPNN prediction model proposed in this paper and compared with the BPNN prediction results. The results show that the proposed two-stage BPNN model, after removing less-sensitive parameters from the input layer, has better prediction accuracy and shorter run time than the BPNN model. [ABSTRACT FROM AUTHOR]

Giga-casting, a revolutionary approach for manufacturing large, single-piece car body components from aluminium, has emerged as a potential game-changer in the automotive industry. However, these large, thin-walled castings are prone to distortions during solidification and heat treatment processes. Straightening these distortions is crucial to ensure structural integrity, facilitate downstream assembly, and maintain aesthetic qualities. This paper proposes a novel method for straightening giga-cast components using a multi-pin straightening machine. The machine's versatility stems from its ability to adapt to various geometries through multiple strategically controlled straightening pins. This paper introduces the concept of a "straightening stroke decision algorithm" to achieve precise straightening and overcome the challenges of complex shapes. This algorithm determines the stroke length for each pin, combining a polynomial model representing the global stiffness of the component with a machine learning model that captures the stiffness changes arising from the current geometry. The effectiveness of the proposed approach is evaluated through comprehensive numerical experiments using finite element analyses. The straightening performance is assessed for the straightening algorithm with different machine learning models (deep neural network and XGBoost) and compared to a traditional optimisation method. The proposed surrogate models decided the straightening strokes so that the maximum remaining distortion became 0.02% of the largest dimension of each target geometry. The results of the numerical experiment showed that the proposed straightening method is suitable for straightening distortion in large thin-walled components. [ABSTRACT FROM AUTHOR]

Existing research in metasurface design was based on trial-and-error high-intensity iterations and requires deep acoustic expertise from the researcher, which severely hampered the development of the metasurface field. Using deep learning enabled the fast and accurate design of hypersurfaces. Based on this, in this paper, an integrated learning approach was first utilized to construct a model of the forward mapping relationship between the hypersurface physical structure parameters and the acoustic field, which was intended to be used for data enhancement. Then a dual-feature fusion model (DFCNN) based on a convolutional neural network was proposed, in which the first feature was the high-dimensional nonlinear features extracted using a data-driven approach, and the second feature was the physical feature information of the acoustic field mined using the model. A convolutional neural network was used for feature fusion. A genetic algorithm was used for network parameter optimization. Finally, generalization ability verification was performed to prove the validity of the network model. The results showed that 90% of the integrated learning models had an error of less than 3 dB between the real and predicted sound field data, and 93% of the DFCNN models could achieve an error of less than 5 dB in the local sound field intensity. [ABSTRACT FROM AUTHOR]

Ultra-high-performance concrete with coarse aggregates (UHPC-CA) has the advantages of high strength, strong shrinkage resistance and a lower production cost, presenting a broad application prospect in civil engineering construction. In view of the difficulty in establishing a mathematical model to accurately predict the mechanical properties of UHPC-CA, the back-propagation artificial neural network (BP-ANN) method is used to fully consider the various influential factors of the compressive strength (CS) and flexural strength (FS) of UHPC-CA in this paper. By taking the content of cement (C), silica fume (SF), slag, fly ash (FA), coarse aggregate (CA), steel fiber, the water–binder ratio (w/b), the sand rate (SR), the cement type (CT), and the curing method (CM) as input variables, and the CS and FS of UHPC-CA as output objectives, the BP-ANN model with three layers has been well-trained, validated and tested with 220 experimental data in the studies published in the literature. Four evaluating indicators including the determination coefficient (R2), the root mean square error (RMSE), the mean absolute percentage error (MAPE), and the integral absolute error (IAE) were used to evaluate the prediction accuracy of the BP-ANN model. A parametric study for the various influential factors on the CS and FS of UHPC-CA was conducted using the BP-ANN model and the corresponding influential mechanisms were analyzed. Finally, the inclusion levels for the CA, steel fiber, and the dimensionless parameters of the W/B and sand rate were recommended to obtain the optimal strength of UHPC-CA. [ABSTRACT FROM AUTHOR]

In this work, we focus on the prediction of the influence of CO2 laser parameters on the kerf properties of cut spruce wood. Laser kerf cutting is mainly characterized by the width of kerf and the width of the heat-affected zone, which depend on the laser power, cutting speed, and structure of the cut wood, represented by the number of cut annual rings. According to the measurement results and ANN prediction results, for lower values of the laser power (P) and cutting speed (v), the effect of annual rings (ARs) is non-negligible. The results of the sensitivity analysis show that the effect of v increases at higher energy density (E) values. With P in the range between 100 and 500 W, v values between 3 and 50 mm·s−1, and AR numbers between 3 and 11, the combination of P = 200 W and v = 50 mm·s−1, regardless of the AR value, leads to the best cut quality for spruce wood. In this paper, the main goal is to show how changes in the input parameters affect the characteristics of the cutting kerf and heat-affected zones for all possible input parameter values. [ABSTRACT FROM AUTHOR]

Warm mix flame retardant asphalt mixture can reduce the energy dissipation and harmful gas emissions during asphalt pavement construction, as well as mitigate the adverse effects of road fires. For this, this paper studies the design and performance of a mixture modified with a combination of warm mix agent and flame retardant, and the pavement performance and flame retardancy of the modified mixture are evaluated. Additionally, a flame retardancy prediction model based on the radial basis function (RBF) neural network model is established. On this basis, the principal components analysis (PCA) model is used to analyze the most significant evaluation indicators affecting flame retardancy, and finally, a three-dimensional finite element model is developed to analyze the effects of loading on the pavement structure. The results show that compared to virgin asphalt mixture, the modified mixture shows a reduction in mixing and compaction temperatures by approximately 12 °C. The high-temperature performance of the mixture is improved, while the low-temperature performance and moisture stability slightly decrease, but its flame retardancy is significantly enhanced. The RBF neural network model revealed that the established flame retardancy prediction model has a high accuracy, allowing for precise evaluation of the flame retardancy. Finally, the PCA model identified that the combustion time has a significant effect on the flame retardancy of the asphalt mixture, and the finite element model revealed that the displacements of the warm mix fire retardant asphalt mixture were lower than virgin asphalt mixture in all directions under the loading. [ABSTRACT FROM AUTHOR]

The quality of Ti alloy casing is crucial for the safe and stable operation of aero engines. However, the fluctuation of key process parameters during the investment casting process of titanium alloy casings has a significant influence on the volume and number of porosity defects, and this influence cannot be effectively suppressed at present. Therefore, this paper proposes a strategy to control the influence of process parameters on shrinkage volume and number. This study constructed multiple regression prediction models and neural network prediction models of porosity volume and number for a ZTC4 casing by simulating the gravity investment casting process. The results show that the multiple regression prediction model and neural network prediction model of shrinkage cavity total volume have an accuracy of over 99%. The accuracy of the neural network prediction model is higher than that of the multiple regression model, and the neural network model realizes the accurate prediction of shrinkage defect volume and defect number through pouring temperature, pouring time, and mold shell temperature. The sensitivity degree of casing defects to key process parameters, from high to low, is as follows: pouring temperature, pouring time, and mold temperature. Further optimizing the key process parameter window reduces the influence of process parameter fluctuation on the volume and number of porosity defects in casing castings. This study provides a reference for actual production control process parameters to reduce shrinkage cavity and loose defects. [ABSTRACT FROM AUTHOR]

This paper tackles the issue of moisture variation in wood-based materials, explicitly focusing on melamine-coated particleboard (hereafter referred to as melamine) and medium-density fiberboard (MDF) used in the third phase of wood industry transformation. The approach involves a comprehensive strategy for predicting moisture content variation, incorporating numerical simulation, experimental testing, and the application of artificial neural network (ANN) technology to enhance accuracy in furniture manufacturing. The developed ANN models are tailored to predict moisture content changes under specific thermal comfort conditions. Remarkably, these models demonstrate high precision, with an average error margin of only 1.40% for 8% moisture content (MC) and 2.85% for 12% MC in melamine, as well as 1.42% for 8% MC and 2.25% for 12% MC in MDF. These levels of precision surpass traditional models, emphasizing this study's novelty and practical relevance to the industrial context. The findings indicate that ANN models adapt to diverse environmental conditions, presenting a robust tool for optimizing moisture management in wood-based materials. This research contributes valuable insights for improving the reliability and efficiency of moisture content predictions in the wood industry. [ABSTRACT FROM AUTHOR]

The manufacturing processes and design of metal and alloy products can be performed over a wide range of strain rates and temperatures. To design and optimize these processes using computational mechanics tools, the selection and calibration of the constitutive models is critical. In the case of hazardous and explosive impact loads, it is not always possible to test material properties. For this purpose, this paper assesses the efficiency and the accuracy of different architectures of ANNs for the identification of the Johnson–Cook material model parameters. The implemented computational tool of an ANN-based parameter identification strategy provides adequate results in a range of strain rates required for general manufacturing and product design applications. Four ANN architectures are studied to find the most suitable configuration for a reduced amount of experimental data, particularly for cases where high-impact testing is constrained. The different ANN structures are evaluated based on the model's predictive capability, revealing that the perceptron-based network of 66 inputs and one hidden layer of 30 neurons provides the highest prediction accuracy of the effective flow stress–strain behavior of Ti64 alloy and three virtual materials. [ABSTRACT FROM AUTHOR]

Structures made of heterogeneous materials, such as composites, often require a multiscale approach when their behavior is simulated using the finite element method. By solving the boundary value problem of the macroscale model, for previously homogenized material properties, the resulting stress maps can be obtained. However, such stress results do not describe the actual behavior of the material and are often significantly different from the actual stresses in the heterogeneous microstructure. Finding high-accuracy stress results for such materials leads to time-consuming analyses in both scales. This paper focuses on the application of machine learning to multiscale analysis of structures made of composite materials, to substantially decrease the time of computations of such localization problems. The presented methodology was validated by a numerical example where a structure made of resin epoxy with randomly distributed short glass fibers was analyzed using a computational multiscale approach. Carefully prepared training data allowed artificial neural networks to learn relationships between two scales and significantly increased the efficiency of the multiscale approach. [ABSTRACT FROM AUTHOR]

In the numerical analysis of manufacturing processes of metal parts, many material properties depending on, for example, the temperature or stress state, must be taken into account. Often these data are dependent on the temperature changes over time. Strongly non-linear material property relationships are usually represented using diagrams. In numerical calculations, these diagrams are analyzed in order to take into account the coupling between the properties. An example of these types of material properties is the dependence of the kinetics of phase transformations in the solid state on the rate and history of temperature change. In literature, these data are visualized Continuous Heating Transformation (CHT) and Continuous Cooling Transformation (CCT) diagrams. Therefore, it can be concluded that time series analysis is important in numerical modeling. This analysis can also be performed using neural networks. This work presents a new approach to storing and analyzing the data contained in the discussed CCT diagrams. The application of Long-Short-Term Memory (LSTM) neural networks and their architecture to determine the correct values of phase fractions depending on the history of temperature change was analyzed. Moreover, an area of research was elements that determine what type of information should be stored by LSTM network coefficients, e.g., whether the network should store information about changes of single phase transformations, or whether it would be better to extract data from differences between several networks with similar architecture. The purpose of the studied network is strongly different from typical applications of artificial neural networks. The main goal of the network was to store information (even by overfitting the network) rather than some form of generalization that allows computation for unknown cases. Therefore, the authors primarily investigated in the ability of the layer-based LSTM network to store nonlinear time series data. The analyses presented in this paper are an extension of the issues presented in the paper entitled "Model of the Austenite Decomposition during Cooling of the Medium Carbon Steel Using LSTM Recurrent Neural Network". [ABSTRACT FROM AUTHOR]

The deep groove ball bearing is one of the most important components of the rotary motion system and is the research subject in this paper. After factory assembly, new ball bearings need to pass quality control. The conventional approach relies on measuring the vibration amplitudes for each unit and sorting them into classes according to the vibration level. In this paper, based on experimental research, models are created to predict the vibration class and analyze the dynamic behavior of new ball bearings. The models are based on artificial neural networks. A feedforward multilayer perceptron (MLP) was applied, and a backpropagation learning algorithm was used. A specific method of training groups of artificial neural networks was applied, where each network provided an answer to the input within the group, and the final answer was the mean value of the answers of all networks in the group. The models achieved a prediction accuracy of over 90%. The main aim of the research was to construct models that are able to predict the vibration class of a new ball bearing based on the geometric parameters of the bearing rings. The models are also applied to analyze the influence of surface roughness of the raceways and the internal radial clearance on bearing vibrations. [ABSTRACT FROM AUTHOR]

Bacterial-based self-healing concrete (BSHC) is a well-known healing technology which has been investigated for a few decades for its excellent crack healing capacity. Nevertheless, considered as costly and time-consuming, the healing performance (HP) of concrete with various types of bacteria can be designed and evaluated only in laboratory environments. Employing machine learning (ML) models for predicting the HP of BSHC is inspired by practical applications using concrete mechanical properties. The HP of BSHC can be predicted to save the time and cost of laboratory tests, bacteria selection and healing mechanisms adoption. In this paper, three types of BSHC, including ureolytic bacterial healing concrete (UBHC), aerobic bacterial healing concrete (ABHC) and nitrifying bacterial healing concrete (NBHC), and ML models with five kinds of algorithms consisting of the support vector regression (SVR), decision tree regression (DTR), deep neural network (DNN), gradient boosting regression (GBR) and random forest (RF) are established. Most importantly, 22 influencing factors are first employed as variables in the ML models to predict the HP of BSHC. A total of 797 sets of BSHC tests available in the open literature between 2000 and 2021 are collected to verify the ML models. The grid search algorithm (GSA) is also utilised for tuning parameters of the algorithms. Moreover, the coefficient of determination (R2) and root mean square error (RMSE) are applied to evaluate the prediction ability, including the prediction performance and accuracy of the ML models. The results exhibit that the GBR model has better prediction ability (R2GBR = 0.956, RMSEGBR = 6.756%) than other ML models. Finally, the influence of the variables on the HP is investigated by employing the sensitivity analysis in the GBR model. [ABSTRACT FROM AUTHOR]

The design of concrete mixtures is crucial in concrete technology, aiming to produce concrete that meets specific quality and performance criteria. Modern standards require not only strength but also eco-friendliness and production efficiency. Based on the Three Equation Method, conventional mix design methods involve analytical and laboratory procedures but are insufficient for contemporary concrete technology, leading to overengineering and difficulty predicting concrete properties. Machine learning-based methods offer a solution, as they have proven effective in predicting concrete compressive strength for concrete mix design. This paper scrutinises the association between the computational complexity of machine learning models and their proficiency in predicting the compressive strength of concrete. This study evaluates five deep neural network models of varying computational complexity in three series. Each model is trained and tested in three series with a vast database of concrete mix recipes and associated destructive tests. The findings suggest a positive correlation between increased computational complexity and the model's predictive ability. This correlation is evidenced by an increment in the coefficient of determination (R2) and a decrease in error metrics (mean squared error, Minkowski error, normalized squared error, root mean squared error, and sum squared error) as the complexity of the model increases. The research findings provide valuable insights for increasing the performance of concrete technical feature prediction models while acknowledging this study's limitations and suggesting potential future research directions. This research paves the way for further refinement of AI-driven methods in concrete mix design, enhancing the efficiency and precision of the concrete mix design process. [ABSTRACT FROM AUTHOR]

In this paper, laser-induced breakdown spectroscopy (LIBS) combined with a probabilistic neural network (PNN) was applied to classify engineering structural metal samples (valve stem, welding material, and base metal). Additionally, utilizing data from the plasma emission spectrum generated by laser ablation of samples with different aging times, an aging time prediction model based on a firefly optimized probabilistic neural network (FA-PNN) was established, which can effectively evaluate the service performance of structural materials. The problem of insufficient features obtained by principal component analysis (PCA) for predicting the aging time of materials is addressed by the proposal of a time-frequency feature extraction method based on short-time Fourier transform (STFT). The classification accuracy (ACC) of time-frequency features and principal component features was compared under PNN. The results indicate that, in comparison to the PCA feature extraction approach, the time-frequency feature extraction method based on STFT demonstrates higher accuracy in predicting the time of aging materials. Then, the relationship between classification accuracy (ACC) and settings of PNN was discussed. The ACC of the PNN model for both the material classification test set and the aging time test set achieved 100% with Firefly (FA) optimization algorithms. This result was also compared with the ACC of ANN, KNN, PLS-DA, and SIMCA for the aging time test set (95%, 87.5%, 85%, and 62.5%, respectively). The experimental results demonstrated that the classification model using LIBS combined with FA-PNN could realize better classification accuracy. [ABSTRACT FROM AUTHOR]

This paper is devoted to the determination of the coefficient of friction (COF) in the drawbead region in metal forming processes. As the test material, AW-5251 aluminium alloys sheets fabricated under various hardening conditions (AW-5251-O, AW-5251-H14, AW-5251-H16 and AW-5251H22) were used. The sheets were tested using a drawbead simulator with different countersample roughness and different orientations of the specimens in relation to the sheet rolling direction. A drawbead simulator was designed to model the friction conditions when the sheet metal passed through the drawbead in sheet metal forming. The experimental tests were carried out under conditions of dry friction and lubrication of the sheet metal surfaces with three lubricants: machine oil, hydraulic oil, and engine oil. Based on the results of the experimental tests, the value of the COF was determined. The Random Forest (RF) machine learning algorithm and artificial neural networks (ANNs) were used to identify the parameters affecting the COF. The R statistical package software version 4.1.0 was used for running the RF model and neural network. The relative importance of the inputs was analysed using 12 different activation functions in ANNs and nine different loss functions in the RF. Based on the experimental tests, it was concluded that the COF for samples cut along the sheet rolling direction was greater than for samples cut in the transverse direction. However, the COF's most relevant input was oil viscosity (0.59), followed by the average counter sample roughness Ra (0.30) and the yield stress Rp0.2 and strength coefficient K (0.05 and 0.06, respectively). The hard sigmoid activation function had the poorest R2 (0.25) and nRMSE (0.30). The ideal run was found after training and testing the RF model (R2 = 0.90 ± 0.028). Ra values greater than 1.1 and Rp0.2 values between 105 and 190 resulted in a decreased COF. The COF values dropped to 9–35 for viscosity and 105–190 for Rp0.2, with a gap between 110 and 130 when the oil viscosity was added. The COF was low when the oil viscosity was 9–35, and the Ra was 0.95–1.25. The interaction between K and the other inputs, which produces a relatively limited range of reduced COF values, was the least relevant. The COF was reduced by setting the Rp0.2 between 105 and 190, the Ra between 0.95 and 1.25, and the oil viscosity between 9 and 35. [ABSTRACT FROM AUTHOR]

This paper introduces a novel method for estimating the cyclic stress–strain curves of steels based on their monotonic properties and plastic strain amplitudes, utilizing artificial neural networks (ANNs). ANNs were trained on a substantial number of experimental data for steels, collected from relevant literature, and divided into subgroups according to alloying elements content (unalloyed, low-alloy, and high-alloy steels). Only monotonic properties that were proven to be relevant for the estimation of points on the stress–strain curve were used. The performance of the developed ANNs was assessed using an independent set of data, and the results were compared to experimental values, values obtained by existing empirical estimation methods, and by previously developed ANNs. The results showed that the new approach which combines relevant monotonic properties and plastic strain amplitudes as inputs to ANNs for cyclic stress–strain curve estimation is better than the previously used approach where ANNs estimate the parameters of the Ramberg–Osgood material model separately. This shows that a more favorable approach to the estimation of cyclic stress–strain behavior would be to directly estimate corresponding material curves using monotonic properties. Additionally, this may also reduce inaccuracies resulting from simplified representations of the actual material behavior inherent in the material model. [ABSTRACT FROM AUTHOR]

Replacing a specified quantity of cement with Class F fly ash contributes to sustainable development and reducing the greenhouse effect. In order to use Class F fly ash in self-compacting concrete (SCC), a prediction model that will give a satisfactory accuracy value for the compressive strength of such concrete is required. This paper considers a number of machine learning models created on a dataset of 327 experimentally tested samples in order to create an optimal predictive model. The set of input variables for all models consists of seven input variables, among which six are constituent components of SCC, and the seventh model variable represents the age of the sample. Models based on regression trees (RTs), Gaussian process regression (GPR), support vector regression (SVR) and artificial neural networks (ANNs) are considered. The accuracy of individual models and ensemble models are analyzed. The research shows that the model with the highest accuracy is an ensemble of ANNs. This accuracy expressed through the mean absolute error (MAE) and correlation coefficient (R) criteria is 4.37 MPa and 0.96, respectively. This paper also compares the accuracy of individual prediction models and determines their accuracy. Compared to theindividual ANN model, the more transparent multi-gene genetic programming (MGPP) model and the individual regression tree (RT) model have comparable or better prediction accuracy. The accuracy of the MGGP and RT models expressed through the MAE and R criteria is 5.70 MPa and 0.93, and 6.64 MPa and 0.89, respectively. [ABSTRACT FROM AUTHOR]

The production of thin-walled beams with various cross-sections is increasingly automated and digitized. This allows producing complicated cross-section shapes with a very high precision. Thus, a new opportunity has appeared to optimize these types of products. The optimized parameters are not only the lengths of the individual sections of the cross section, but also the bending angles and openings along the beam length. The simultaneous maximization of the compressive, bending and shear stiffness as well as the minimization of the production cost or the weight of the element makes the problem a multi-criteria issue. The paper proposes a complete procedure for optimizing various open sections of thin-walled beam with different openings along its length. The procedure is based on the developed algorithms for traditional and soft computing optimization as well as the original numerical homogenization method. Although the work uses the finite element method (FEM), no computational stress analyses are required, i.e., solving the system of equations, except for building a full stiffness matrix of the optimized element. The shell-to-beam homogenization procedure used is based on equivalence strain energy between the full 3D representative volume element (RVE) and its beam representation. The proposed procedure allows for quick optimization of any open sections of thin-walled beams in a few simple steps. The procedure can be easily implemented in any development environment, for instance in MATLAB, as it was done in this paper. [ABSTRACT FROM AUTHOR]

Inspired by classical graph neural networks, we discuss a novel quantum graph neural network (QGNN) model to predict the chemical and physical properties of molecules and materials. QGNNs were investigated to predict the energy gap between the highest occupied and lowest unoccupied molecular orbitals of small organic molecules. The models utilize the equivariantly diagonalizable unitary quantum graph circuit (EDU-QGC) framework to allow discrete link features and minimize quantum circuit embedding. The results show QGNNs can achieve lower test loss compared to classical models if a similar number of trainable variables are used, and converge faster in training. This paper also provides a review of classical graph neural network models for materials research and various QGNNs. [ABSTRACT FROM AUTHOR]

No study has been reported to use machine learning methods to predict the full-range test curves of circular CFST columns. In this paper, the long short-term memory (LSTM) network was introduced to calculate the axially compressive load–strain curves of the circular CFST columns according to an experiment database of limited scale. To improve the feasibility of input data for the recurrent neural network algorithm, data preprocessing methods and data configurations were discussed. The prediction results indicate that the LSTM network provides more accurate estimations compared with the artificial neural networks, random forest and support vector regression. Meanwhile, this method can be used to calculate the mechanical properties including the elastic modulus, ultimate bearing capacity, and the ductility of the columns with acceptable accuracy for engineering practice (the prediction error within 20%). For future research, it is expected that the machine learning method will be applied to predict the structural response of different members under various loading conditions. [ABSTRACT FROM AUTHOR]

In this paper, the possibility of determining the thermal, elastic and geometric characteristics of a thin TiO2 film deposited on a silicon substrate, with a thickness of 30 μm, in the frequency range of 20 to 20 kHz with neural networks were analysed. For this purpose, the geometric (thickness), thermal (thermal diffusivity, coefficient of linear expansion) and electronic parameters of substrates were known and constant in the two-layer model, while the following nano-layer thin-film parameters were changed: thickness, expansion and thermal diffusivity. Predictions of these three parameters of the thin-film were analysed separately with three neural networks. All of them together were joined by a fourth neural network. It was shown that the neural network, which analysed all three parameters at the same time, achieved the highest accuracy, so the use of networks that provide predictions for only one parameter is less reliable. The obtained results showed that the application of neural networks in determining the thermoelastic properties of a thin film on a supporting substrate enables the estimation of its characteristics with great accuracy. [ABSTRACT FROM AUTHOR]

Artificial neural networks (ANNs) have been an important approach for predicting the value of flow stress, which is dependent on temperature, strain, and strain rate. However, there is still a lack of sufficient knowledge regarding what structure of ANN should be used for predicting metal flow stress. In this paper, we train an ANN for predicting flow stress of In718 alloys at high temperatures using our experimental data, and the structure of the ANN is optimized by comparing the performance of four ANNs in predicting the flow stress of In718 alloy. It is found that, as the size of the ANN increases, the ability of the ANN to retrieve the flow stress results from a training dataset is significantly enhanced; however, the ability to predict the flow stress results absent from the training does not monotonically increase with the size of the ANN. It is concluded that the ANN with one hidden layer and four nodes possesses optimized performance for predicting the flow stress of In718 alloys in this study. The reason why there exists an optimized ANN size is discussed. When the ANN size is less than the optimized size, the prediction, especially the strain dependency, falls into underfitting and fails to predict the curve. When the ANN size is less than the optimized size, the predicted flow stress curves with the temperature, strain, and strain rate will contain non-physical fluctuations, thus reducing their prediction accuracy of extrapolation. For metals similar to the In718 alloy, ANNs with very few nodes in the hidden layer are preferred rather than the large ANNs with tens or hundreds of nodes in the hidden layers. [ABSTRACT FROM AUTHOR]

Developing accurate and interpretable models to forecast concrete's self-healing behavior is of interest to material engineers, scientists, and civil engineering contractors. Machine learning (ML) and artificial intelligence are powerful tools that allow constructing high-precision predictions, yet often considered "black box" methods due to their complexity. Those approaches are commonly used for the modeling of mechanical properties of concrete with exceptional accuracy; however, there are few studies dealing with the application of ML for the self-healing of cementitious materials. This paper proposes a pioneering study on the utilization of ML for predicting post-fire self-healing of concrete. A large database is constructed based on the literature studies. Twelve input variables are analyzed: w/c, age of concrete, amount of cement, fine aggregate, coarse aggregate, peak loading temperature, duration of peak loading temperature, cooling regime, duration of cooling, curing regime, duration of curing, and specimen volume. The output of the model is the compressive strength recovery, being one of the self-healing efficiency indicators. Four ML methods are optimized and compared based on their performance error: Support Vector Machines (SVM), Regression Trees (RT), Artificial Neural Networks (ANN), and Ensemble of Regression Trees (ET). Monte Carlo analysis is conducted to verify the stability of the selected model. All ML approaches demonstrate satisfying precision, twice as good as linear regression. The ET model is found to be the most optimal with the highest prediction accuracy and sufficient robustness. Model interpretation is performed using Partial Dependence Plots and Individual Conditional Expectation Plots. Temperature, curing regime, and amounts of aggregates are identified as the most significant predictors. [ABSTRACT FROM AUTHOR]

The fly ash-slag geopolymer is regarded as one of the new green cementitious materials that can replace cement, but it is difficult to predict its mechanical properties by conventional methods. Therefore, in the present study, the back propagation (BP) artificial neural network technique is used to predict the compressive strength of the fly ash-slag geopolymer. In this paper, data from the published literature were collected as the training set and the experimental results from laboratory experiments were used as the test set. Eight input parameters were determined, as follows: the percentage of fly ash, the percentage of slag, the water–cement ratio, the curing age, the modulus of alkali activator, the mass ratio of NaOH to Na2SiO3 and the moles of Na2O and SiO2 in the alkali activator. Three multilayer artificial neural network models were constructed using the Levenberg–Marquardt (LM), Bayesian regularization (BR) and scaled conjugate gradient (SCG) algorithms to compare the prediction accuracy of the compressive strength of the fly ash-slag geopolymer paste at different ages (3, 7, and 28 d). It was concluded that the training set error of the BR–BP neural network was the smallest. Ultimately, the hyperparameter optimization of the BR–BP neural network was carried out to compare the training set and the test set errors before and after the optimization, and the results show that the BR–BP neural network model with hyperparameter optimization had the highest prediction accuracy. [ABSTRACT FROM AUTHOR]

In engineering practice, due to the high compressibility and very low shear strength of organic soils, it is difficult to construct an embankment on organic subsoil. High variability and significant change in geotechnical parameters cause difficulties in predicting the behavior of organic soils under embankment loading. The aim of the paper was to develop empirical relationships used in the preliminary design for evaluating the settlement and undrained shear strength of organic subsoil loaded by embankment based on data obtained from four test sites. Statistical multiple regression models were developed for evaluating the settlement in time and undrained shear strength in time individually for peat and gyttja. Neural networks to predict the settlement and undrained shear strength in time for peat and gyttja simultaneously as double-layer subsoils as well as a separate neural network for peat and a separate neural network for gyttja as single-layer subsoils were also developed. The vertical stress, thickness, water content, initial undrained shear strength of peat and gyttja, and time were used as the independent variables. Artificial neural networks are characterized by greater prediction accuracy than statistical multiple regression models. Multiple regression models predict dependent variables with maximum relative errors of about 35% to about 60%, and neural networks predict output variables with maximum relative errors of about 25% to about 30%. [ABSTRACT FROM AUTHOR]

Drawbeads are used when forming drawpieces with complex shapes to equalise the flow resistance of a material around the perimeter of the drawpiece or to change the state of stress in certain regions of the drawpiece. This article presents a special drawbead simulator for determining the value of the coefficient of friction on the drawbead. The aim of this paper is the application of artificial neural networks (ANNs) to understand the effect of the most important parameters of the friction process (sample orientation in relation to the rolling direction of the steel sheets, surface roughness of the counter-samples and lubrication conditions) on the coefficient of friction. The intention was to build a database for training ANNs. The friction coefficient was determined for low-carbon steel sheets with various drawability indices: drawing quality DQ, deep-drawing quality DDQ and extra deep-drawing quality EDDQ. Equivalents of the sheets tested in EN standards are DC01 (DQ), DC03 (DDQ) and DC04 (EDDQ). The tests were carried out under the conditions of dry friction and the sheet surface was lubricated with machine oil LAN46 and hydraulic oil LHL32, commonly used in sheet metal forming. Moreover, various specimen orientations (0° and 90°) in relation to the rolling direction of the steel sheets were investigated. Moreover, a wide range of surface roughness values of the counter-samples (Ra = 0.32 μm, 0.63 μm, 1.25 μm and 2.5 μm) were also considered. In general, the value of the coefficient of friction increased with increasing surface roughness of the counter-samples. In the case of LAN46 machine oil, the effectiveness of lubrication decreased with increasing mean roughness of the counter-samples Ra = 0.32–1.25 μm. With increasing drawing quality of the sheet metal, the effectiveness of lubrication increased, but only in the range of surface roughness of the counter-samples in which Ra = 0.32–1.25 μm. This study investigated different transfer functions and training algorithms to develop the best artificial neural network structure. Backpropagation in an MLP structure was used to build the structure. In addition, the COF was calculated using a parameter-based analytical equation. Garson partitioning weight was used to calculate the relative importance (RI) effect on coefficient of friction. The Bayesian regularization backpropagation (BRB)—Trainbr training algorithm, together with the radial basis normalized—Radbasn transfer function, scored best in predicting the coefficient of friction with R2 values between 0.9318 and 0.9180 for the training and testing datasets, respectively. [ABSTRACT FROM AUTHOR]

This paper presents the development and evaluation of neural network models using a small input–output dataset to predict the thermal behavior of a high-speed motorized spindles. Different neural multi-output regression models were developed and evaluated using Keras, one of the most popular deep learning frameworks at the moment. ANN was developed and evaluated considering the following: the influence of the topology (number of hidden layers and neurons within), the learning parameter, and validation techniques. The neural network was simulated using a dataset that was completely unknown to the network. The ANN model was used for analyzing the effect of working conditions on the thermal behavior of the motorized grinder spindle. The prediction accuracy of the ANN model for the spindle thermal behavior ranged from 95% to 98%. The results show that the ANN model with small datasets can accurately predict the temperature of the spindle under different working conditions. In addition, the analysis showed a very strong effect of type coolant on spindle unit temperature, particularly for intensive cooling with water. [ABSTRACT FROM AUTHOR]

Prior to the long-term utilization of solid oxide fuel cell (SOFC), one of the most remarkable electrochemical energy conversion devices, a variety of difficult experimental validation procedures is required, so it would be time-consuming and steep to predict the applicability of these devices in the future. For numerous years, extensive efforts have been made to develop mathematical models to predict the effects of various characteristics of solid oxide fuel cells (SOFCs) components on their performance (e.g., voltage). Taking advantage of the machine learning (ML) method, however, some issues caused by assumptions and calculation costs in mathematical modeling could be alleviated. This paper presents a machine learning approach to predict the anode-supported SOFCs performance as one of the most promising types of SOFCs based on architectural and operational variables. Accordingly, a dataset was collected from a study about the effects of cell parameters on the output voltage of a Ni-YSZ anode-supported cell. Convolutional machine learning models and multilayer perceptron neural networks were implemented to predict the current-voltage dependency. The resulting neural network model could properly predict, with more than 0.998 R2 score, a mean squared error of 9.6 × 10−5, and mean absolute error of 6 × 10−3 (V). Conventional models such as the Gaussian process as one of the most powerful models exhibits a prediction accuracy of 0.996 R2 score, 10−4 mean squared, and 6 × 10−3 (V) absolute error. The results showed that the built neural network could predict the effect of cell parameters on current-voltage dependency more accurately than previous mathematical and artificial neural network models. It is noteworthy that this procedure used in this study is general and can be easily applied to other materials datasets. [ABSTRACT FROM AUTHOR]

The analysis of the basic characteristics of various research methods is highly needed to predict the residual life of the pipeline accurately, help managers understand the operational risks, and provide a reference for developing pipeline transportation and maintenance inspection plans and anti-corrosion measures. Based on a comprehensive investigation of the existing research on the residual life of the pipeline, this paper finds that the current mainstream life prediction method, based on historical statistical data, has the shortcomings of inconsistent modeling methods, inconsistent basic data, and a lack of comparative evaluation among methods. Moreover, considering the in-depth study of BP neural network modeling, grey theory modeling, time series modeling, and exponential smoothing modeling, optimal prediction models using different methods based on the same historical data are established. These optimal modeling methods are discussed, and the feasible modeling path for the accurate prediction of the pipeline's residual life is given by comparing the prediction accuracy of each model. In addition, the findings serve as a guide for developing an anti-corrosion strategy by highlighting the contribution of the prediction results of the residual life to pipeline decision-making. By comparison, it is found that the accuracy of the four prediction models is as follows: the grey theory prediction model, the exponential smoothing prediction model, the BP neural network prediction model, and the time series prediction model, from high to low, respectively. [ABSTRACT FROM AUTHOR]

Determining the quality of Ti-6Al-4V parts fabricated by selective laser melting (SLM) remains a challenge due to the high cost of SLM and the need for expertise in processes and materials. In order to understand the correspondence of the relative density of SLMed Ti-6Al-4V parts with process parameters, an optimized extreme gradient boosting (XGBoost) decision tree model was developed in the present paper using hyperparameter optimization with the GridsearchCV method. In particular, the effect of the size of the dataset for model training and testing on model prediction accuracy was examined. The results show that with the reduction in dataset size, the prediction accuracy of the proposed model decreases, but the overall accuracy can be maintained within a relatively high accuracy range, showing good agreement with the experimental results. Based on a small dataset, the prediction accuracy of the optimized XGBoost model was also compared with that of artificial neural network (ANN) and support vector regression (SVR) models, and it was found that the optimized XGBoost model has better evaluation indicators such as mean absolute error, root mean square error, and the coefficient of determination. In addition, the optimized XGBoost model can be easily extended to the prediction of mechanical properties of more metal materials manufactured by SLM processes. [ABSTRACT FROM AUTHOR]

As a key component of a high-power laser device, fused silica optics needs to bear great laser energy, and laser damage is easily generated on the optical surface. In order to improve the service life and availability of optics, it is necessary to repair the damaged optics. In this work, the repair technique of damaged, fused silica optics was studied. The neural network method was mainly used to establish the correlation between the number of small-scale damage points and the repair depth. The prediction accuracy of the model is better than 90%. Based on the neural network model, the removal depth parameters were optimized with the suppression coefficient of the damage points. The processing effect of the optimized parameters was verified by magnetorheological polishing experiments. In this paper, a repair technique based on a neural network was proposed, which avoids the low efficiency caused by processing iterations in the repair process, and can accurately what was expected. The method proposed in this work has an important reference value in the repair process of fused silica optics. [ABSTRACT FROM AUTHOR]

A considerable amount of discarded building materials are produced each year worldwide, resulting in ecosystem degradation. Self-compacting concrete (SCC) has 60–70% coarse and fine particles in its composition, so replacing this material with another waste material, such as recycled aggregate (RA), reduces the cost of SCC. This study compares novel Artificial Neural Network algorithm techniques—Levenberg–Marquardt (LM), Bayesian regularization (BR), and Scaled Conjugate Gradient Backpropagation (SCGB)—to estimate the 28-day compressive strength (f'c) of SCC with RA. A total of 515 samples were collected from various published papers, randomly splitting into training, validation, and testing with percentages of 70, 10 and 20. Two statistical indicators, correlation coefficient (R) and mean squared error (MSE), were used to assess the models; the greater the R and lower the MSE, the more accurate the algorithm. The findings demonstrate the higher accuracy of the three models. The best result is achieved by BR (R = 0.91 and MSE = 43.755), while the accuracy of LM is nearly the same (R = 0.90 and MSE = 48.14). LM processes the network in a much shorter time than BR. As a result, LM and BR are the best models in forecasting the 28 days f'c of SCC having RA. The sensitivity analysis showed that cement (28.39%) and water (23.47%) are the most critical variables for predicting the 28-day compressive strength of SCC with RA, while coarse aggregate contributes the least (9.23%). [ABSTRACT FROM AUTHOR]

In recent years, cutting edge preparation became a topic of high interest in the manufacturing industry because of the important role it plays in the performance of the cutting tool. This paper describes the use of the drag finishing DF cutting edge preparation process on the cutting tool for the broaching process. The main process parameters were manipulated and analyzed, as well as their influence on the cutting edge rounding, material remove rate MRR, and surface quality/roughness (Ra, Rz). In parallel, a repeatability and reproducibility R&R analysis and cutting edge radius re prediction were performed using machine learning by an artificial neural network ANN. The results achieved indicate that the influencing factors on re, MRR, and roughness, in order of importance, are drag depth, drag time, mixing percentage, and grain size, respectively. The reproducibility accuracy of re is reliable compared to traditional processes, such as brushing and blasting. The prediction accuracy of the re of preparation with ANN is observed in the low training and prediction errors 1.22% and 0.77%, respectively, evidencing the effectiveness of the algorithm. Finally, it is demonstrated that the DF has reliable feasibility in the application of edge preparation on broaching tools under controlled conditions. [ABSTRACT FROM AUTHOR]

Stencil printing is the most crucial process in reflow soldering for the mass assembly of electronic circuits. This paper investigates different machine learning-based methods to predict the essential process characteristics of stencil printing: the area, thickness, and volume of deposited solder paste. The training dataset was obtained experimentally by varying the printing speed (from 20 to 120 mm/s), the size (area ratio from 0.35 to 1.7) of stencil apertures, and the particle size (characterized by a log-normal distribution) in the solder paste. Various machine learning-based methods were assessed; ANFIS–adaptive neuro-fuzzy inference systems; ANN artificial neural networks (with different learning methods); boosted trees, regression trees, SVM–support vector machines. Each method was optimized and fine-tuned with hyperparameter optimization, and the overfitting phenomenon was also prevented with cross-validation. The regression tree was the best performing approach for modelling the stencil printing, while ANN with the Bayesian regularization learning method was only slightly worse. The presented methodology for fine-tuning, parameter optimization, and the comparison of different machine learning-based methods can easily be adapted to any application field in electronics manufacturing. [ABSTRACT FROM AUTHOR]

In this paper, eight variables of cement, blast furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate and flow are used as network input and slump is used as network output to construct a back-propagation (BP) neural network. On this basis, the learning rate, momentum factor, number of hidden nodes and number of iterations are used as hyperparameters to construct 2-layer and 3-layer neural networks respectively. Finally, the response surface method (RSM) is used to optimize the parameters of the network model obtained previously. The results show that the network model with parameters obtained by the response surface method (RSM) has a better coefficient of determination for the test set than the model before optimization, and the optimized model has higher prediction accuracy. At the same time, the model is used to evaluate the influencing factors of each variable on slump. The results show that flow, water, coarse aggregate and fine aggregate are the four main influencing factors, and the maximum influencing factor of flow is 0.875. This also provides a new idea for quickly and effectively adjusting the parameters of the neural network model to improve the prediction accuracy of concrete slump. [ABSTRACT FROM AUTHOR]

Concrete is an economical and efficient material for attenuating radiation. The potential of concrete in attenuating radiation is attributed to its density, which in turn depends on the mix design of concrete. This paper presents the findings of a study conducted to evaluate the radiation attenuation with varying water-cement ratio (w/c), thickness, density, and compressive strength of concrete. Three different types of concrete, i.e., normal concrete, barite, and magnetite containing concrete, were prepared to investigate this study. The radiation attenuation was calculated by studying the dose absorbed by the concrete and the linear attenuation coefficient. Additionally, artificial neural network (ANN) and gene expression programming (GEP) models were developed for predicting the radiation shielding capacity of concrete. A correlation coefficient (R), mean absolute error (MAE), and root mean square error (RMSE) were calculated as 0.999, 1.474 mGy, 2.154 mGy and 0.994, 5.07 mGy, 5.772 mGy for the training and validation sets of the ANN model, respectively. Similarly, for the GEP model, these values were recorded as 0.981, 13.17 mGy, and 20.20 mGy for the training set, whereas the validation data yielded R = 0.985, MAE = 12.2 mGy, and RMSE = 14.96 mGy. The statistical evaluation reflects that the developed models manifested close agreement between experimental and predicted results. In comparison, the ANN model surpassed the accuracy of the GEP models, yielding the highest R and the lowest MAE and RMSE. The parametric and sensitivity analysis revealed the thickness and density of concrete as the most influential parameters in contributing towards radiation shielding. The mathematical equation derived from the GEP models signifies its importance such that the equation can be easily used for future prediction of radiation shielding of high-density concrete. [ABSTRACT FROM AUTHOR]

The dynamic development of new technologies enables the optimal computer technique choice to improve the required quality in today's manufacturing industries. One of the methods of improving the determining process is machine learning. This paper compares different intelligent system methods to identify the tool wear during the turning of gray cast-iron EN-GJL-250 using carbide cutting inserts. During these studies, the experimental investigation was conducted with three various cutting speeds vc (216, 314, and 433 m/min) and the exact value of depth of cut ap and federate f. Furthermore, based on the vibration acceleration signals, appropriate measures were developed that were correlated with the tool condition. In this work, machine learning methods were used to predict tool condition; therefore, two tool classes were proposed, namely usable and unsuitable, and tool corner wear VBc = 0.3 mm was assumed as a wear criterium. The diagnostic measures based on acceleration vibration signals were selected as input to the models. Additionally, the assessment of significant features in the division into usable and unsuitable class was caried out. Finally, this study evaluated chosen methods (classification and regression tree, induced fuzzy rules, and artificial neural network) and selected the most effective model. [ABSTRACT FROM AUTHOR]

The indentation test is a popular method for the investigation of the mechanical properties of materials. The technique, which combines traditional indentation tests with mapping the shape of the imprint, provides more data describing the material parameters. In this paper, such methodology is employed for estimating the selected material parameters described by Ramberg–Osgood's law, i.e., Young's modulus, the yield point, and the material hardening exponent. Two combined identification methods were used: the P-A procedure, in which the material parameters are identified on the basis of the coordinates of the indentation curves, and the P-C procedure, which uses the coordinates describing the imprint profile. The inverse problem was solved by neural networks. The results of numerical indentation tests—pairs of coordinates describing the indentation curves and imprint profiles—were used as input data for the networks. In order to reduce the size of the input vector, a simple and effective method of approximating the branches of the curves was proposed. In the Results Section, we show the performance of the approximation as a data reduction mechanism on a synthetic dataset. The sparse model generated by the presented approach is also shown to efficiently reconstruct the data while minimizing error in the prediction of the mentioned material parameters. Our approach appeared to consistently provide better performance on the testing datasets with considerably easier computation than the principal component analysis compression results available in the literature. [ABSTRACT FROM AUTHOR]

Single point incremental forming (SPIF) is one of the most promising technologies for the manufacturing of sheet metal prototypes and parts in small quantities. Similar to other forming processes, the design of the SPIF process is a demanding task. Nowadays, the design process is usually performed using numerical simulations and virtual models. The modelling of the SPIF process faces several challenges, including extremely long computational times caused by long tool paths and the complexity of the problem. Path determination is also a demanding task. This paper presents a finite element (FE) analysis of an incrementally formed truncated pyramid compared to experimental validation. Focus was placed on a possible simplification of the FE process modelling and its impact on the reliability of the results obtained, especially on the geometric accuracy of the part and bottom pillowing effect. The FE modelling of SPIF process was performed with the software ABAQUS, while the experiment was performed on a conventional milling machine. Low-carbon steel DC04 was used. The results confirm that by implementing mass scaling and/or time scaling, the required calculation time can be significantly reduced without substantially affecting the pillowing accuracy. An innovative artificial neural network (ANN) approach was selected to find the optimal values of mesh size and mass scaling in term of minimal bottom pillowing error. However, care should be taken when increasing the element size, as it has a significant impact on the pillow effect at the bottom of the formed part. In the range of selected mass scaling and element size, the smallest geometrical error regarding the experimental part was obtained by mass scaling of 19.01 and tool velocity of 16.49 m/s at the mesh size of 1 × 1 mm. The obtained results enable significant reduction of the computational time and can be applied in the future for other incrementally formed shapes as well. [ABSTRACT FROM AUTHOR]

A material-tailored special concrete composite that uses a synthetic fiber to make the concrete ductile and imposes strain-hardening characteristics with eco-friendly ingredients is known as an "engineered geopolymer composite (EGC)". Mix design of special concrete is always tedious, particularly without standards. Researchers used several artificial intelligence tools to analyze and design the special concrete. This paper attempts to design the material EGC through an artificial neural network with a cross-validation technique to achieve the desired compressive and tensile strength. A database was formulated with seven mix-design influencing factors collected from the literature. The five best artificial neural network (ANN) models were trained and analyzed. A gradient descent momentum and adaptive learning rate backpropagation (GDX)–based ANN was developed to cross-validate those five best models. Upon regression analysis, ANN [2:16:16:7] model performed best, with 74% accuracy, whereas ANN [2:16:25:7] performed best in cross-validation, with 80% accuracy. The best individual outputs were "tacked-together" from the best five ANN models and were also analyzed, achieving accuracy up to 88%. It is suggested that when these seven mix-design influencing factors are involved, then ANN [2:16:25:7] can be used to predict the mix which can be cross-verified with GDX-ANN [7:14:2] to ensure accuracy and, due to the few mix trials required, help design the SHGC with lower costs, less time, and fewer materials. [ABSTRACT FROM AUTHOR]

Cone Penetration Testing (CPT) is a quick survey in situ method through which soil parameters are not determined directly, but have to be estimated using derived relations between required soil parameter and soil resistance at the testing probe. Boundary conditions affect the reliability of the estimated soil parameters, therefore controlled laboratory conditions were applied to the intended CPT procedure analysis. Density, pycnometry, oedometer and direct shear tests of kinetic sand were performed to prove its usability as a reference testing material for further CPT laboratory analysis. The results of testing the kinetic sand are presented in this paper. Executed tests proved the kinetic sand as a reliable material in terms of the homogeneity and consistency of its physical and mechanical parameters. The material is utilizable as a substitution of cohesive sandy soils in physical modeling without the negative impact of the consistency-dependent behavior of fine-grained soils. However, some differences in parameters with respect to the natural soils should be taken into account. Neural network theory and numerical approach will be applied to the intended CPT laboratory analysis under controlled boundary conditions using kinetic sand to evaluate its potential for the determination of soil parameters. [ABSTRACT FROM AUTHOR]

This paper shows the surface quality results after finishing milling of AZ91D and AZ31 magnesium alloys. The study was performed for variable technological parameters: cutting speed, feed per tooth, axial depth of cut and radial depth of cut. The tools used in the study were two carbide cutters with a different tool cutting edge helix angle. The measurement of the research results presented the surface roughness parameters was made on the lateral faces and the end faces of the specimens. Statistical analysis and simulations using artificial neural networks were carried out with the Statistica software. The normality of the distribution was examined, and the hypotheses of the equality of mean values and variance were verified. For the AZ91D magnesium alloy on the lateral and the end faces (Ra, Rz parameters), simulations were carried out. Two types of ANN were used: MLP (Multi-layered perceptron) and RBF (Radial Basis Function). To increase the machining stability and to obtain a high surface finish, the more suitable tool for finishing milling is the tool with a helix angle of λs = 20°. Artificial neural networks have been shown to be a good tool for predicting surface roughness parameters of magnesium alloys after finishing milling. [ABSTRACT FROM AUTHOR]

Fiber-reinforced polymer (FRP) has several benefits, in addition to excellent tensile strength and low self-weight, including corrosion resistance, high durability, and easy construction, making it among the most optimum options for concrete structure restoration. The bond behavior of the FRP-concrete (FRPC) interface, on the other hand, is extremely intricate, making the bond strength challenging to estimate. As a result, a robust modeling framework is necessary. In this paper, data-driven hybrid models are developed by combining state-of-the-art population-based algorithms (bald eagle search (BES), dynamic fitness distance balance-manta ray foraging optimization (dFDB-MRFO), RUNge Kutta optimizer (RUN)) and artificial neural networks (ANN) named "BES-ANN", "dFDB-MRFO -ANN", and "RUN-ANN" to estimate the FRPC interfacial-bond strength accurately. The efficacy of these models in predicting bond strength is examined using an extensive database of 969 experimental samples. Compared to the BES-ANN and dFDB-MRFO models, the RUN-ANN model better estimates the interfacial-bond strength. In addition, the SHapley Additive Explanations (SHAP) approach is used to help interpret the best model and examine how the features influence the model's outcome. Among the studied hybrid models, the RUN-ANN algorithm is the most accurate model with the highest coefficient of determination (R2 = 92%), least mean absolute error (0.078), and least coefficient of variation (18.6%). The RUN-ANN algorithm also outperformed mechanics-based models. Based on SHAP and sensitivity analysis method, the FRP bond length and width contribute more to the final prediction results. [ABSTRACT FROM AUTHOR]

The complex and non-linear nature of material properties evolution during 3D printing continues to make experimental optimization of Fused Deposition Modeling (FDM) costly, thus entailing the development of mathematical predictive models. This paper proposes a two-stage methodology based on coupling limited data experiments with black-box AI modeling and then performing heuristic optimization, to enhance the viscoelastic properties of FDM processed acrylonitrile butadiene styrene (ABS). The effect of selected process parameters (including nozzle temperature, layer height, raster orientation and deposition speed) as well as their combinative effects are also studied. Specifically, in the first step, a Taguchi orthogonal array was employed to design the Dynamic Mechanical Analysis (DMA) experiments with a minimal number of runs, while considering different working conditions (temperatures) of the final prints. The significance of process parameters was measured using Lenth's statistical method. Combinative effects of FDM parameters were noted to be highly nonlinear and complex. Next, artificial neural networks were trained to predict both the storage and loss moduli of the 3D printed samples, and consequently, the process parameters were optimized via Particle Swarm Optimization (PSO). The optimized process of the prints showed overall a closer behavior to that of the parent (unprocessed) ABS, when compared to the unoptimized set-up. [ABSTRACT FROM AUTHOR]

The shear transfer mechanism of steel fiber reinforced concrete (SFRC) beams without stirrups is still not well understood. This is demonstrated herein by examining the accuracy of typical empirical formulas for 488 SFRC beam test records compiled from the literature. To steer clear of these cognitive limitations, this study turned to artificial intelligence (AI) models. A gray relational analysis (GRA) was first conducted to evaluate the importance of different parameters for the problem at hand. The outcomes indicate that the shear capacity depends heavily on the material properties of concrete, the amount of longitudinal reinforcement, the attributes of steel fibers, and the geometrical and loading characteristics of SFRC beams. After this, AI models, including back-propagation artificial neural network, random forest and multi-gene genetic programming, were developed to capture the shear strength of SFRC beams without stirrups. The findings unequivocally show that the AI models predict the shear strength more accurately than do the empirical formulas. A parametric analysis was performed using the established AI model to investigate the effects of the main influential factors (determined by GRA) on the shear capacity. Overall, this paper provides an accurate, instantaneous and meaningful approach for evaluating the shear capacity of SFRC beams containing no stirrups. [ABSTRACT FROM AUTHOR]

This paper mainly proposes two kinds of artificial neural network (ANN) models for predicting the plastic anisotropy properties of sheet metal using spherical indentation test, which minimizes measurement time, costs, and simplifies the process of obtaining the anisotropy properties than the conventional tensile test. The proposed ANN models for predicting anisotropic properties can replace the traditional complex dimensionless analysis. Moreover, this paper is not limited to the prediction of yield strength anisotropy but also further accurately predicts the Lankford coefficient in different orientations. We newly construct an FE spherical indentation model, which is suitable for sheet metal in consideration of actual compliance. To obtain a large dataset for training the ANN, the constructed FE model is utilized to simulate pure and alloyed engineering metals with one thousand elastoplastic parameter conditions. We suggest the specific variables of the residual indentation mark as input parameters, also with the indentation load–depth curve. The profile of the residual indentation, including the height and length in different orientations, are used to analyze the anisotropic properties of the material. Experimental validations have been conducted with three different sheet alloys, TRIP1180 steel, zinc alloy, and aluminum alloy 6063-T6, comparing the proposed ANN model and the uniaxial tensile test. In addition, machine vision was used to efficiently analyze the residual indentation marks and automatically measure the indentation profiles in different orientations. The proposed ANN model exhibits remarkable performance in the prediction of the flow curves and Lankford coefficient of different orientations. [ABSTRACT FROM AUTHOR]

The motivation of the presented paper is the desire to create a universal tool to analyse the process of austenite decomposition during the cooling process of various steel grades. The presented analysis concerns the application of Recurrent Artificial Neural Networks (RANN) of the Long Short-Term Memory (LSTM) type for the analysis of the transition path of the cooling curve. This type of network was selected due to its ability to predict events in time sequences. The proposed generalisation allows for the determination of the austenite transformation during the continuous cooling process for various cooling curves. As training data for the neural network, values determined from the macroscopic model based on the analysis of Continuous Cooling Transformation (CCT) diagrams were used. All relations and analyses used to build training/testing or validation sets are presented in the paper. The modelling with the use of LSTM network gives the possibility to determine the incremental changes of phase transformation (in a given time step) with the assumed changes of temperature resulting from the considered cooling rate. [ABSTRACT FROM AUTHOR]

The presented study investigates the effects of weight percentages of boron carbide reinforcement on the wear properties of aluminum alloy composites. Composites were fabricated via ball milling and the hot extrusion process. During the fabrication of composites, B4C content was varied (0, 5, and 10 wt.%), as well as milling time (0, 10, and 20 h). Microstructural observations with SEM microscopy showed that with an increase in milling time, the distribution of B4C particles is more homogeneous without agglomerates, and that an increase in wt.% of B4C results in a more uniform distribution with distinct grain boundaries. Taguchi and ANOVA analyses are applied in order to investigate how parameters like particle content of B4C, normal load, and milling time affect the wear properties of AA2024-based composites. The ANOVA results showed that the most influential parameters on wear loss and coefficient of friction were the content of B4C with 51.35% and the normal load with 45.54%, respectively. An artificial neural network was applied for the prediction of wear loss and the coefficient of friction. Two separate networks were developed, both having an architecture of 3-10-1 and a tansig activation function. By comparing the predicted values with the experimental data, it was demonstrated that the well-trained feed-forward-back propagation ANN model is a powerful tool for predicting the wear behavior of Al2024-B4C composites. The developed models can be used for predicting the properties of Al2024-B4C composite powders produced with different reinforcement ratios and milling times. [ABSTRACT FROM AUTHOR]