Prediction Of Capacity Research Articles

Enhancing FRP-concrete interface bearing capacity prediction with explainable machine learning: A feature engineering approach and SHAP analysis

This study introduces a novel approach to predict the shear bearing capacity of FRP-concrete interfaces using explainable machine learning. Eight algorithms are employed: three standalone models (Artificial Neural Network, Support Vector Regression, and Decision Tree) and five ensemble learning models (Bagging, Random Forest, Adaptive Boosting, Gradient Boosting, and Extreme Gradient Boosting). Four scenarios with varying input features, including engineered features inspired by mechanics-based bearing capacity equations, are examined. Notably, the inclusion of engineered features such as the stiffness of the FRP strip (Kf) significantly enhanced prediction accuracy and efficiency, although the width correction coefficient (bf/bc) did not yield significant benefits, contrary to findings in mechanics-based models. The Extreme Gradient Boosting (XGBoost) algorithm emerged as the top-performing model, achieving an impressive R-square value of 0.949. In terms of explainability, the study utilized the Shapley Additive Explanation (SHAP) technique to comprehensively elucidate the significance, dependency, and interaction effects of features within the best-performing model. Remarkable, approximately 75 % of the total mean absolute SHAP values were attributed to Kf and bf of the FRP, highlighting their pivotal role in the prediction process. This detailed explanation of the model using SHAP instills trust in the predictions and facilitates its practical implementation in real-world scenarios.

Flexural performance of functionally graded UHPFRC-NSC beams

This paper presents an experimental and numerical study on the flexural performance of Functionally Graded Reinforced Concrete (FGRC) beams using Ultra-High Performance Fiber Reinforced Concrete (UHPFRC) with Normal-Strength Concrete (NSC). The experimental results were used to examine the applicability of the codes’ predictions of flexural capacity adopted in ACI 318-2019, EC-2, and ECP-203-2020. The flexural design recommendations of AFGC-2013, JSCE-2008, and KCI-2012 were also examined. The experimental program involved ten simply supported beams with identical cross-sectional dimensions and two different tensile reinforcement ratios (ρl = 1.38 % and 5.68 %). FGRC beams consisting of UHPFRC (with fcu = 174.3 MPa) and NSC (with fcu = 45.8 MPa) were tested under four-point bending. The layer of UHPFRC was positioned either in the utmost compression region or in both the compression and tension regions. The experimental results revealed that the FGRC beams exhibited strength and ductility comparable to those of pure UHPFRC beams but at a significantly reduced cost. When UHPFRC layers constituted 66.7 % of beam thickness, the FGRC beams were capable of carrying 97.4 % of the load sustained by a full UHPFRC beam, while for UHPFRC layers constituted 40 % of beam thickness the FGRC beam carried 77.5 % of that of a full UHPFRC beam. Increasing ρl% from 1.38 % to 5.68 % in the tested specimens led to a considerable enhancement in the flexural capacity. The sectional analysis conducted in accordance with the recommendations of AFGC-2013, JSCE-2008, and KCI-2012 showed precise predictions of the flexural strength in both UHPFRC and FGRC beams, while the flexural strength calculated using the equations adopted in the studied codes gave conservative estimations. The proposed numerical model displayed a high level of agreement with the experimental findings of the studied FGRC beams.

Dynamic modeling and defrost optimization for air source heat pumps: A deep learning and autoregression approach

Air source heat pumps (ASHPs) are celebrated for their exceptional energy efficiency and their capacity to curtail greenhouse gas emissions. However, a critical challenge in ASHPs research is the effective and accurate management of the defrosting process to preserve peak performance. The development of robust and precise simulation models is crucial yet complex for formulating and assessing defrosting strategies. To address this, our study investigates the integration of deep learning models with feature encoding to establish dynamic models for ASHPs. We evaluated various deep learning architectures, achieving a Mean Absolute Percentage Error (MAPE) of 3.5% in heating capacity predictions. Moreover, our analysis reveals that feature encoding substantially lowers the MAPE by 12%, thereby enhancing model accuracy. We employed a comprehensive suite of evaluation metrics to benchmark the performance across models, finding that the Temporal Convolutional Network (TCN) marginally surpasses others, including Transformer, LSTM, and LightGBM models. Additionally, we utilized transfer learning to boost the generalization capabilities models across different outdoor conditions (temperature and humidity), facilitating effective few-shot training. Finally, our study introduces a novel defrosting control strategy based on Heating Capacity Degradation (HCD). Using dynamic modeling, we determined optimal control thresholds (89%-94%) that ensure maximum heating efficiency during frosting conditions. This strategy improves the average heating capacity by 101-278 W compared to conventional defrosting methods. The insights from this research not only advance defrosting control but also serve as a valuable reference for dynamic modeling and control optimization in the heat pump sector. The proposed control method enhances ASHPs performance and energy efficiency, supporting efforts towards achieving carbon neutrality.

Polyurethane Foam and Algae-Based Activated Carbon Biocomposites for Oil Spill Remediation.

This study investigates the incorporation of algae-based activated carbon into polyurethane foam to improve a biocomposite for gasoil sorption. The biocomposites were thoroughly analyzed using various techniques to examine the properties of both the blank foam and the algae activated carbon foam with a carbon content of 4.41 mass% and particle diameter of 500 µm. These techniques included Scanning Electron Microscopy (SEM), thermogravimetric analysis (TGA), and density analysis. The TGA analysis revealed that the biocomposites had an impact on the onset temperature (Tonset) of the foams. Higher concentrations of the biocomposites resulted in a decrease in Tonset from approximately 310 °C in the blank foam (PUF0) to 300 °C in the composite (PUF3B). The final residue percentage also decreased from around 20% in PUF0 to 10% in PUF3B. Density analysis showed that the apparent density of the foam increased from 0.016 g/cm3 in the blank foam to 0.020 g/cm3 in the biocomposite (PUF3B), while the real density slightly decreased from 0.092 g/cm3 to 0.076 g/cm3, indicating a reduction in overall porosity from 82.5% to 74.4%. All foams that were modified showed an increase in their ability to absorb gasoil in a PUF/gasoil/water system. The optimized biocomposite (PUF1B), with 1.14 mass% of 500 µm algae carbon, displayed the highest sorption capacity, starting at approximately 50 g/g at 1.5 h and increasing to 53 g/g over 72 h. The analysis of adsorption kinetics revealed that by utilizing adsorption isotherms, particularly the Langmuir isotherm, a more accurate fit to the data was achieved. This allowed for the prediction of the maximum gasoil adsorption capacity. This study aims to further develop, analyze, and utilize biocomposites made from algae-based activated carbon and polyurethane. These materials offer a sustainable and environmentally friendly approach to cleaning up oil spills.

Predicting the Compression Capacity of Screw Piles in Sand Using Machine Learning Trained on Finite Element Analysis

Screw piles (often referred to as helical piles) are widely used to resist axial and lateral loads as deep foundations. Multi-helix piles experience complex interactions between the plates which depend on the soil properties, pile stiffness, helix diameter, and the number of helix plates among other factors. Design methods for these piles are typically highly empirical and there remains significant uncertainty around calculating the compression capacity. In this study, a database of 1667 3D finite element analyses was developed to better understand the effect of different inputs on the compression capacity of screw piles in clean sands. Following development of the numerical database, various machine learning methods such as linear regression, neural networks, support vector machines, and Gaussian process regression (GPR) models were trained and tested on the database in order to develop a prediction tool for the pile compression capacity. GPR models, trained on the numerical data, provided excellent predictions of the screw pile compression capacity. The test dataset root mean square error (RMSE) of 29 kN from the GPR model was almost an order of magnitude better than the RMSE of 225 kN from a traditional theoretical approach, highlighting the potential of machine learning methods for predicting the compression capacity of screw piles in homogenous sands.

Multi-scale prediction of remaining useful life of lithium-ion batteries based on variational mode decomposition and integrated machine learning

Accurate and reliable prediction of the remaining useful life (RUL) of lithium-ion batteries (LIB) is very important for the safety of power systems. To solve the nonlinear and time-varying problems of LIB aging trajectories, an RUL prediction method based on variational mode decomposition (VMD) and integrated machine learning is proposed. First, the VMD algorithm is used to mine the internal information in LIB data to capture the long-term downward trend and capacity regeneration hidden in the capacity degradation. After that, the Elman neural network optimized by the gray wolf algorithm and Gaussian process regression were used to model and predict the intrinsic mode function (IMF) and residual sequence respectively. Next, the predicted IMF component and residual sequence are fused to obtain the LIB aging trajectory, and the capacity prediction error is corrected through the bidirectional long short-term memory network. Finally, the LIB data set is used as the test data of the algorithm. The results show that the proposed method has better robustness and nonlinear tracking ability, the mean absolute error is controlled within 1.5 %, and the RUL prediction error is within 3 cycles.

Capacity prediction of lithium-ion batteries based on ensemble empirical mode decomposition and hybrid machine learning

Capacity prediction of lithium-ion batteries based on ensemble empirical mode decomposition and hybrid machine learning

An interpretable capacity prediction method for lithium-ion battery considering environmental interference

Predicting the capacity of lithium-ion battery (LIB) plays a crucial role in ensuring the safe operation of LIBs and prolonging their lifespan. However, LIBs are easily affected by environmental interference, which may impact the precision of predictions. Furthermore, interpretability in the process of predicting LIB capacity is also important for users to understand the model, identify issues, and make decisions. In this study, an interpretable method considering environmental interference (IM-EI) for predicting LIB capacity is introduced. Spearman correlation coefficients, interpretability principles, belief rule base (BRB), and interpretability constraints are used to improve the prediction precision and interpretability of IM-EI. Dynamic attribute reliability is introduced to minimize the effect of environmental interference. The experimental results show that IM-EI model has good interpretability and high precision compared to the other models. Under interference conditions, the model still has good precision and robustness.

Machine-learning assisted analysis on the seismic performance of steel reinforced concrete composite columns

This paper presents the data-driven analysis on the seismic performance of steel reinforced concrete (SRC) composite columns by using machine learning (ML) algorithms. A total of 248 test data of SRC columns subjected to the combined axial compressive force and low reversed cyclic horizontal force was collected from the published literatures to form the database. Since the seismic action triggers the SRC columns to suffer from different failure mode and hence exhibit quite different load bearing mechanism, six ML algorithms were employed to facilitate the failure mode classification and bearing capacity prediction. It was found that the Random Forest (RF) model predicts the failure mode most accurately, while XGBoost model delivers the best estimation of bearing capacity. To further evaluate the feasibility of ML models and unveil the interaction between different variables, interpretability analysis of ML models was carried out. In comparison with the conventional empirical judgement and theoretical derivation, ML model delivers enhanced accuracy and robustness to predict the seismic performance of SRC columns, and hence it could be used as a promising alternative for the preliminary design and analysis of SRC columns.

Fiber-based model for rectangular/square concrete-filled steel tubular columns under arbitrary end moment ratios considering member imperfection

This paper develops a fiber-based model for rectangular/square concrete-filled steel tube (CFST) columns under arbitrary end moments considering the member imperfection. The accuracy of the proposed model regarding the prediction of capacity, deflection curves and full range curves is verified using a collected test database. The verified model is further used to investigate the behaviour of rectangular/square CFST columns under arbitrary end moments. It was found that the varying trends of strength and deformation capacity of rectangular/square CFST columns under arbitrary end moments are opposite while increasing the end moment ratio β, L/r ratio and fc, whereas those are the same if increasing b/t ratio and fy. Higher fy and/or smaller b/t ratio may be adopted in order to achieve larger strength and deformation capacity simultaneously. The location of maximum deflection at peak state hkp/L is only significantly affected by end moment ratio β, whereas the location of critical section at peak state hcp/L is significantly affected by end moment ratio β, L/r ratio and yield strength fy. Moreover, current design codes, i.e., EC4, AISC 360 and GB50936–2014, were found to poorly predict the resistance to axial load and bending moment of rectangular/square CFST columns under arbitrary end moments.

Using meta-heuristic algorithms with multi-layer perceptron for prediction of ultimate bearing capacity

ABSTRACT This study presents an innovative approach to predicting the Ultimate Bearing Capacity (Qu) in civil engineering, integrating a Multi-layer Perceptron (MLP) neural network using the Artificial Rabbits Optimization (ARO) and Improved Manta-Ray Foraging Optimizer (IMRFO) methods. The MLP effectively captures complex, non-linear relationships within geotechnical datasets, while the IMRFO and ARO algorithms fine-tune the MLP’s parameters for optimal performance. Validation results demonstrate the efficacy of this method, with the MLP+IMRFO (MLIM) model achieving an R2 value of 0.996 and a MARE of 0.113, highlighting its potential for enhancing predictive accuracy in civil engineering applications. This combined methodology offers a state-of-the-art solution for Qu forecasting, contributing to safer and more efficient construction practices.

The Impact of Forest Management Inventory Factors on the Ecological Service Value of Forest Water Conservation Based on Machine Learning Algorithms

Based on forest management inventory data, this study applies machine learning algorithms to explore the relationships between forest water conservation capacity and forest management inventory factors, thus providing more extensive insights into forest water conservation services. By integrating the InVEST model and machine learning algorithms, this study identifies the key factors related to water conservation services based on forest management inventory factors and investigates the differences in and accuracy of forest water conservation models using the random forest algorithm. The results are as follows: (1) The determination coefficients (R2) of the three machine learning models range from 0.508 to 0.869, with root mean square errors (RMSEs) ranging from 28.380 to 69.339. The performance of these models is generally satisfactory, with the random forest algorithm showing superior results. (2) By leveraging the advantages of the three machine learning algorithms in handling categorical data, this study analyzes the contributions of forest management inventory factors, revealing the impact mechanisms of forest-type water conservation services. (3) The integration of machine learning algorithms allows for better processing of the scale and correlation of independent variables, providing more objective information on the main controlling factors of forest water conservation. (4) Predictions of water conservation capacity using machine learning are consistent with that of the InVEST model. The water conservation per unit area shows a variation trend as follows: slow-growing broadleaf forests > shrub forests > middle-growing broadleaf forests > cunninghamia lanceolata forests > fast-growing broadleaf forests > pine forests > bamboo forests. (5) Since this study considers only the factors available in the forest management inventory, which does not encompass all relevant influencing factors, it is difficult to fully address the complexities of how forest water conservation services interact with forest structure. Therefore, further research is needed to investigate the intrinsic mechanisms underlying the interactions between water conservation and forest management inventory factors.

Optimized CatBoost-Based Soft-Computing Models for Prediction of the Ultimate Bearing Capacity of T-Shaped Footings Subjected to Eccentric Load

Optimized CatBoost-Based Soft-Computing Models for Prediction of the Ultimate Bearing Capacity of T-Shaped Footings Subjected to Eccentric Load

Predicting the Adhesive Layer Thickness in Hybrid Joints Involving Pre-Tensioned Bolts.

While most academic studies focus on the properties of cured joints, this research addresses the manufacturing process of hybrid joints in their uncured state. Hybrid joints that combine adhesive bonding with pre-tensioned bolts exhibit superior mechanical performance compared to exclusively bonded or bolted joints. However, the adhesive flow during manufacturing in hybrid joints often results in a nonuniform adhesive thickness, where obtaining an exact thickness is crucial for accurate load capacity predictions. This paper presents experiments involving three different adhesives, providing precise measurements of the adhesive layer thickness distribution, which served as a reference when evaluating and validating the subsequent numerical predictions. The numerical predictions were performed using computational fluid dynamics (CFD) to model the flow behavior of the adhesives during the bonding process and their interactions with the metal substrates. The CFD predictions of the adhesive layer thickness showed good agreement with the experimental data, with the relative differences between the average experimental and numerical thickness values ranging from 4.07% to 27.1%. The results were most accurate for the adhesive with sand particles, whose particles remained intact, ensuring that the adhesive's rheology remained unchanged. The results highlight the importance of the rheological behavior of the adhesive in the final distribution of the adhesive layer thickness, thereby expanding the understanding of these joints.

Unveiling the Origins of “Memory Effects” in Thick Electrodes through 2D XRF-XANES Mapping and Modeling

Making electrodes much thicker than the current commercial lithium-ion batteries is a promising approach to further increase the energy density of rechargeable batteries by reducing the weight/volume fraction of inactive components at the battery cell level. However, thick electrodes suffer inferior rate performance as non-uniform reaction develops within the electrode due to the increased ionic and electronic transport distances, which reduces the capacity utilization. In this work, we demonstrate yet another challenge faced by the application of thick electrodes (>100 µm), that is, they exhibit pronounced “memory effect” that is not seen in their thin electrode counterpart. While “memory effect” was initially observed in nickel-cadmium and nickel metal-hybrid batteries, which experience capacity reduction upon repeated shallow discharging, here we use the terminology to refer to the phenomenon that the (dis)charging performance of a lithium-ion battery depends on its cycling history. An example is illustrated in Figure 1, which shows the 2C discharge capacity of a 150μm-thick LiFePO4 electrode in a half cell when it starts from an initial state of charge (SoC) at 50%. Our experiment reveals that the discharge capacity has a strong dependence on how the electrode was first brought to the initial state. When it is charged to 50% SoC at 1.5C, the subsequent discharge capacity is 137% larger compared to when it is discharged to the same SoC at 1.5C. Similar phenomenon is also observed in another mainstream cathode NMC622. We show that such memory effect is caused by the reaction polarization present in the thick electrodes. To this end, 2D X-ray fluorescence (XRF) spectroscopy and Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS) were employed to characterize the reaction distribution within the electrodes, and its correlation to the (dis)charging kinetics and OCV relaxation were investigated by electrochemical testing and pseudo-2D battery simulation. Our study highlights the issues created by the memory effect on the capacity prediction and SoC determination, which need to be addressed to develop a better battery management system for thick-electrode-based battery cells and packs. Figure 1

A new prediction model for residual deformation capacity of corroded steel bars

Corrosion of reinforcing steel bars (rebar) is one of the primary factors leading to the performance degradation of reinforced concrete (RC) structures. It is essential to take into consideration the deteriorating effects of corrosion when it comes to assessing the performance of existing RC structures. However, due to the uncertainties and randomness associated with the degree and distribution of corrosion, accurately assessing the residual deformation capacity of corroded rebar is a significant challenge. This paper presents a comprehensive theoretical model for the prediction of the residual deformation capacity of corroded rebar. Taking into account the mechanical characteristics of the steel material, three classes of non-uniform corrosion in rebar are firstly defined on a quantitative basis to characterize the deformation distribution patterns under tension. Based on the concept of deformation accumulation, a simplified theoretical model is proposed for evaluating the residual deformation capacity of corroded rebar, incorporating both the non-uniformity of corrosion and the sample (or gauge) length; thus, a length scale is built into the nominal strain measure to accommodate the use of different sample or gauge lengths. Numerical tensile tests on 500 randomly generated samples of rebar with varying corrosion levels are conducted, and the parameters of the theoretical model for the three corrosion classes are determined based on a regression analysis of the tensile test results. Comparisons with actual experimental results and existing degradation models demonstrate that the proposed theoretical model provides a more accurate estimation for the deformation capacity of corroded rebar. The model offers a reliable reference for accurately assessing the tensile performance of corroded rebar in subsequent engineering applications.

Initial Engineering Studies of Battery Capacity Prediction Power of "Electric Ship of FTK UNSADA" Container Concept from Jakarta to Ibu Kota Nusantara

Initial Engineering Studies of Battery Capacity Prediction Power of "Electric Ship of FTK UNSADA" Container Concept from Jakarta to Ibu Kota Nusantara

Axial compressive mechanical behavior of a CFRP-confined composite member composed of bamboo strips and a thin-walled steel tube

To utilize fully the excellent mechanical properties of bamboo, thin-walled steel, and carbon fiber-reinforced polymer (CFRP), a CFRP-confined composite member composed of bamboo strips and a thin-walled steel tube (CBSC) was proposed to design a bamboo member with high-bearing capacity and ductility. The axial compressive tests were conducted on 26 CBSC members. Based on the validated finite-element (FE) model, a parametric analysis was performed to investigate the effects of different parameters on the axial compressive mechanical behaviors of CBSC members. Finally, the predictions of the ultimate axial compressive bearing capacity derived from the unified theory of concrete-filled steel tube columns agreed well with the test and FE results. The results showed that the failure mode of short CBSC members was strength failure, including material damage in the middle of the member, and end-crushing failure. However, the slender CBSC members exhibited typical global buckling failure. CBSC members can fully utilize the mechanical characteristics of the three materials to exert their combined effects. In terms of improving the bearing capacity, the recommended number of CFRP layers for all CBSC members is one except for the short CFRP-confined bamboo strip, thin-walled, circular steel-tube composite members, which is equal to two. The critical slenderness ratios of the CBCSC specimens and CBSSC specimens for distinguishing strength failure and global buckling are 19 and 36, respectively. The proposed predicted equations can accurately estimate the ultimate bearing capacity of CBSC members under axial compression.

Behaviour of steel-fibre-reinforced geopolymer concrete-filled square steel tubular stub column under axial compression

Resource utilisation of industrial waste has become a crucial issue in construction aiming for sustainable development. Research on geopolymer concrete-filled square steel tubular (GCFST) structures is limited. Moreover, methods for calculating the bearing capacity of steel-fibre-reinforced GCFST (SFRGCFST) structures require further investigation. This study presents an experimental and analytical investigation of the axial compression performance of SFRGCFST stub columns. Twenty-four specimens were tested to assess the effect of steel tube thickness, steel fibre (SF) volume fraction, SF aspect ratio, and concrete strength grade on failure modes, load–deformation behaviour, load–strain response, ultimate resistance, and ductility of SFRGCFST columns. The results indicated that the failure modes of SFRGCFST columns are influenced by the coefficient index, which varies with the test parameters. When the coefficient index exceeded 0.88, the failure mode of the specimens transitioned from shearing failure to waist-drum-shaped failure. The average ultimate bearing capacity of the SFRGCFST columns increased by 35.13 % compared with that of the plain concrete-filled steel tubes (CFSTs). When the SF volume fraction reached 0.9 %, the ultimate bearing capacity of the SFRGCFST columns increased by 21.85 % compared with that of steel-fibre-reinforced concrete-filled steel tube (SFRCFST) columns. The incorporation of SF significantly improved the deformation capacity and ductility of the SFRGCFST columns and moderated the slope of the post-peak load–displacement and load–strain curves, slowing the descent of the curve. Utilising the multi-axial strength criterion and mechanical equilibrium equation for square CFSTs, a formula for calculating the ultimate bearing capacity of SFRGCFST stub columns was developed. This formula enables accurate prediction of the axial compressive load capacity of SFRGCFST stub columns and provides a basis for revising relevant standards.

Interpretable Capacity Prediction of RC Bridges Based on Genetic Algorithm-enhanced Support Vector Machine Learning

Interpretable Capacity Prediction of RC Bridges Based on Genetic Algorithm-enhanced Support Vector Machine Learning