Optimizing XGBoost, CatBoost, and Bagging models for predicting the maximum dry density of compacted soil using grid search hyperparameter tuning

This technical article helps identify the optimum performance AI model for predicting compaction parameters of soil. A comparative study is mapped between regression analysis (RA), Gaussian process regression (GPR), decision tree (DT), support vector machine (SVM), and artificial neural networks (ANNs) approaches using 59 soil datasets. The soil dataset consists of soil properties such as gravel content, silt content, sand content, specific gravity, clay content, plasticity index, and liquid limit. The soil properties are used as input parameters to develop the AI model to predict soil optimum moisture content and maximum dry density. The RA, GPR, SVM, DT, and ANN models are designated as MLR_X, GPR_X, SVM_X, DT_X, ANN_X, where the X is OMC and MDD. The performance of MLR_OMC, GPR_OMC, SVM_OMC, DT_OMC, LMNN_OMC, and GDANN_OMC is 0.9714, 0.9867, 0.9689, 0.9832, 0.9435, and 0.9520, respectively. Similarly, the performance of MLR_MDD, GPR_MDD, SVM_MDD, DT_MDD, LMNN_MDD, and GDANN_MDD is 0.9512, 0.9854, 0.9482, 0.9199, 0.8679, and 0.9395, respectively. Based on the performance of AI models, the GPR_OMC and GPR_MDD models are identified as the optimum performance model to predict the soil maximum dry density (MDD) and optimum moisture content (OMC). The predicted OMC and MDD are compared with laboratory OMC and MDD, and it is found that the GPR_OMC and GPR_MDD model has the potential to predict soil compaction parameters.

The California Bearing Ratio (CBR) value is a crucial soil parameter in road construction and design. Obtaining representative CBR values is challenging, requiring time-consuming and expensive testing procedures. To address this issue, regression equations were developed to establish correlations between CBR and soil index properties. Laboratory tests were conducted to determine the soaked CBR, Liquid Limit (LL), Plastic Limit (PL), Plasticity Index (PI), Maximum Dry Density (MDD), and Optimum Moisture Content (OMC) of soil samples. Regression models were then created between CBR and different sets of soil index properties using Microsoft Excel 2007. Strong correlations were observed between soaked CBR, PL, PI, OMC, and MDD (R2 = 0.744); CBR, LL, PL, OMC, and MDD (R2 = 0.702); CBR, PI, OMC, and MDD (R2 = 0.643); CBR, LL, OMC, and MDD (R2 = 0.621); and CBR, OMC, and MDD (R2 = 0.602). Among all equations, the relation CBR = 0.72 PL – 1.22 PI + 2.34 OMC + 106.97 MDD – 222.46 exhibited the strongest correlation with a P-value of 0.005 and R2 of 0.744.

In geotechnical engineering and construction, the optimal moisture content (OMC) and maximum dry density (MDD) is crucial for determining the ideal conditions for soil strength and stability in infrastructure. Traditional laboratory techniques for calculating OMC and MDD are both costly and time-consuming. Machine learning offers a potential alternative for traditional empirical approaches by making it easier to create complex prediction models and algorithms that can improve the accuracy and efficacy of forecasts of compaction parameters. Machine learning-specifically Meta-heuristic optimization (MHO)-approaches are high-level problem-solving strategies that seek optimal or near-optimal solutions to difficult optimization problems, which are frequently non-linear, multi-modal, or non-differentiable. The Genetic Algorithm (GA), Generalized Population-Based Adaptive Search (GPAS), and Particle Swarm Optimization (PSO) are three powerful meta-heuristic optimization algorithms that are commonly employed to solve complex optimization issues. The goal of this project is to develop a framework that uses meta-heuristic optimization techniques to estimate OMC and MDD. Using MHO models, the study shows a substantial correlation between OMC and MDD, respectively, with significant soil factors such as specific gravity, Atterberg limits, and grain size distribution parameters. This study depicts three distinct models for the prediction of OMC and MDD named GA, PSO, and GPAS models. Among the models, the GA model demonstrated the highest accuracy in predicting OMC (R2 = 0.9999, MSE = 0.0001), while the PSO model was most effective for MDD prediction (R2 = 0.9660, MSE = 0.1871). These findings highlight the accuracy and dependability of the GA technique, which presents a viable method for precisely forecasting the MDD and OMC of soil stabilization mixtures in a range of engineering applications. Additionally, it reduces the negative effects that soil extraction and modification have on the environment.

ABSTRACTThis study investigates the application of artificial intelligence (AI) models to predict soil compaction characteristics, specifically maximum dry density (MDD) and optimum moisture content (OMC), which are critical for stabilizing construction foundations. Traditional methods for determining MDD and OMC are labor‐intensive and often influenced by factors such as soil type, plasticity, and compaction energy (E). The research employed AI models, including random forest regression (RF‐R), gradient boosting regression (GB‐R), XGBoosting regressor (XGB‐R), and multilinear regression (ML‐R), trained on a comprehensive dataset of soil properties. For the first time, compaction energy has been used as an input variable to predict soil cement lime stabilized compaction parameters. Among the models, GB‐R demonstrated the highest prediction accuracy for MDD and OMC, outperforming RF‐R, XGB‐R, and ML‐R. The performance of built‐in models has been measured by three new index performance metrics: the a20‐index, the index of scatter (IS), and the index of agreement (IA), in addition to four common metrics. Taylor diagrams confirmed the robustness of these predictions during lab testing. A sensitivity analysis revealed that MDD and OMC were most influenced by plastic limit (PL), compaction energy (E), liquid limit (LL), and plasticity index (PI). Additionally, curve‐fitting techniques were applied to model the relationship between MDD, OMC, and these key factors. The results indicated that the GB‐R model, particularly when focused on essential features, provided superior accuracy compared to traditional regression methods, offering a reliable tool for soil stabilization assessments in construction.

This study presents the hydraulic conductivity characteristics of fly ash mixed with varying percentages of montmorillonite clay. Fly ash produced by thermal power plants may contain different toxic metals and creates environmental problems like leaching and dusting; leaching of these metals may contaminate the ground water. To mitigate these problems attempts have been made to find out possible application of fly ash as safe construction materials like fill materials, liners and covers. Previous studies show scarcity of data on the hydraulic conductivity characteristics of fly ash–montmorillonite clay mixes. Nine different types of fly ash samples collected from different thermal power plants of Eastern part of India have been mixed with varying percentages (i.e., 0, 20, 30, 40, and 50 %) of montmorillonite clay on dry weight basis of mixes. Total 45 numbers of mixes have been tested. Falling head hydraulic conductivity tests have been carried out, compacted at maximum dry density (MDD) and optimum moisture content (OMC) obtained from standard Proctor (AASHO) compaction tests. The values of hydraulic conductivity for specimens of mix with 50 % montmorillonite clay decrease to the range of 10−10–10−11 m/s from the range of 10−06–10−07 m/s of fly ash alone. Effects of flow period, montmorillonite clay content, MDD, OMC, liquid limit, plasticity index, initial void ratio, initial degree of saturation and free swell index on hydraulic conductivity of fly ash–montmorillonite clay mixes have been discussed. The values of hydraulic conductivity decrease with increase of flow period, liquid limit, plasticity index, MDD, initial degree of saturation and free swell index, and with decrease in initial void ratio of mixes. Mixes of fly ash and montmorillonite clay may find potential application as liner material.

Subgrade of a pavement should be strong enough to give adequate support to the pavement and for supporting and distributing the wheel loads. The design and behaviour of a flexible pavement depends mainly on the stability of the subgrade soil, which can be increased by compacting the soil at optimum moisture content (OMC) thus achieving maximum dry density (MDD). In this study, Modified Proctor Compaction Test (MPCT) was conducted by giving 5 different number of blows per layer so as to establish a relationship of compaction energy with OMC & MDD. Also, OMC and MDD are one of the most important parameters influencing California Bearing Ratio (CBR) test. CBR is a penetration test used for evaluating the mechanical strength of subgrade soil and for determining the thickness of the pavement required. Determination of soaked and unsoaked CBR value for a soil is time consuming and a laborious process. Hence in this paper an attempt has been made to arrive at regression equations to correlate soaked and unsoaked CBR values for the silty clay (CL) soil with the compaction characteristics, so that based on OMC and MDD, CBR value of soil can be predicted thus avoiding the time consuming process of conducting CBR tests. The soil sample used in this project was a disturbed sample collected from Thiruporur District in Chennai.

This study presents a comparative framework for evaluating the predictive performance of four machine learning models namely Gradient Boosting Machine (GBM), Random Decision Forest (RDF), Non-Parametric Regression (NP), and Decision Tree (TREE), in estimating the Unconfined Compressive Strength (UCS) of nano-doped fly ash reinforced clayey soil. The key innovation lies in combining ensemble learning with sensitivity, monotonicity, and SHAP (SHapley Additive exPlanations) analyses to enhance predictive accuracy and interpretability in geotechnical applications. Using a comprehensive dataset of key variables (Curing Days, Maximum Dry Density (MDD), Optimum Moisture Content (OMC), Fly Ash, Multi-Walled Carbon Nanotubes (MWCNT), and Sodium Hexametaphosphate (SHMP)) models were trained and validated using various statistical metrics (R², MAE, MSE etc.). GBM achieved the best performance (R²: 1.000, 0.955; MAE: 0.001, 0.022; MSE: 0.000, 0.001 in training and testing respectively), consistently outperforming RDF, NP, and TREE. Taylor diagrams, REC curves, and AOC analysis further confirmed GBM's superior generalization and minimal error rates. Sensitivity analysis identified Days (0.4556), MDD (0.2458), and SHMP (0.1558) as the most influential factors, trends that were corroborated by SHAP analysis. Monotonicity analysis validated positive relationships with Days (R² = 0.9845) and MDD (R² = 0.689), while OMC and SHMP showed inverse effects. These results establish GBM as a highly accurate and interpretable model for UCS prediction, offering a powerful tool for optimizing soil stabilization in construction and geotechnical engineering.

Soil compaction involves concretion and a relative variation of physical and mechanical properties of soils. Determining laboratory compaction characteristics such as maximum dry density (MDD) and optimum moisture content (OMC) could be vital work to manage field compaction for all earth-works structures. There are 3 necessary Atterberg limits: plastic limit (PL), liquid limit (LL), and Plastic Index (PI). The most objective of this paper is to get the relationships between compaction parameters and their Atterberg limits of fine-grained soils and to create reliable correlations. For conducting this work, forty samples are collected from a borrowed space that is found at the bank upstream of Setit watercourse. The tests of soil samples were executed at the laboratory of Dam complex of the upper Atbara project. To perform this work, the Microsoft Office Excel software was exercised for the regression analysis of compaction parameters and Atterberg limits. Several trials were created to get the relationships between Atterberg limits (LL, PL, and PI) with the compaction parameters (OMC, and MDD). From the regression analysis, it's found that OMC and MDD have an excellent relationship with the LL other than the PL and PI. It had been observed that the (OMC) has an excellent correlation with (MDD) other than the remaining parameters. From this work, it's going to be suggested to use the soil compaction properties and Liquid Limits' correlations attributable to their reliable results compared with the other correlations. The result of the paper may be helpful and applicable in numerous civil engineering sectors, particularly for preliminary investigations and prefeasibility studies of various civil engineering works.

Improvements of the soils in sites are usually necessary to achieve the desired strength, compressibility and hydraulic conductivity of soils. Generally, improvement in compaction is preferred and applied in sites. The compaction curve and associated characteristics, namely maximum dry unit weight and optimum moisture content, should be determined in the laboratory by using Proctor test procedures. However, a number of Proctor tests need to be carried out for projects like earth dams, embankments and highways. In this paper, a simplified and dependable way of obtaining compaction characteristics and compaction curves at different compaction energy levels for preliminary designs is introduced. For this purpose, Proctor compaction tests were conducted with different soils and at different compaction energy levels apart from using extensive results found in the literature. It was found out that by only getting the compaction characteristics from plastic limit an approximate compaction curve can be obtained from family curves. The approximate compaction curve compares well with the tested values too. Maximum dry density predicted is 0.986 times the maximum dry density at plastic limit, whereas predicted optimum moisture content is 0.99 times the actual optimum moisture content.

Compaction tests forms one of the importantaspects in geotechnical engineering practice. These tests are time consumingand require large quantity of soil also. In this paper based on the results ofthe compaction tests carried out for different soils of varying plasticitycharacteristics at different compaction energies and on published data, it hasbeen brought that there is a good correlation between the optimum moisturecontent and plastic limit for the . In addition to this one can predict themodified compaction parameters just knowing the plastic limit of the soil.For the present investigation, threedifferent soils from North Cyprus (Tuzla, Değirmenlik and Akdeniz) and a soilfrom Turkey (highly plastic montmorillonitic clay) were chosen. These soils areheavily in use for civil engineering activities like construction of pavements,embankments and earth retaining structures. Compaction tests were carried out at threedifferent energy levels for the four soils described.. They are standardProctor test (SP), reduced modified Proctor (RMP) and modified Proctor (MP).For the standard Proctor, the compaction energy works out to be 593.7 kJ/m3.In the modified Proctor test, the compaction energy works out to be 2693.3 kJ/m3.In the reduced modified Proctor test the procedure is same as modified Proctorexcept the number of layers are three instead of five. The compaction energyworks out to be 1616 kJ/m3. [1]Based on the experimental results obtained for maximum dry density vs.optimum moisture content for the four different soils with different compactionenergy levels it has been found that irrespective of soil typeand compaction energy levels both the maximum dry density and optimum moisturecontent are linearly related with a very high correlation coefficient of R= 0.994. Results obtained from laboratory tests aswell as from literature show that the correlation between maximum dry densityand OMC for different soils, compacted for two compaction energy levels is verygood. It is thus seen that one can predict OMCknowing the plastic limit with reasonable accuracy.Having obtained OMC one can get the maximumdry density from equation(1) obtained in this study. From experimental results it has been foundthat both OMC and maximum dry densityof Proctor’s test results and that of modified Proctor’s test results ofauthors’ as well as data collected from literature correlate very well.It is seen that the correlation is highlysatisfactory. Having obtained both OMC and maximum dry density for Proctor’senergy level one can get the OMC and maximum dry density for modified Proctorcondition also.

Maximum Dry Density (MDD) and Optimum Moisture Content (OMC) are two important parameters of soil filling, which affect the soil stability and bearing capacity, and thus the reliability and durability of facilities such as highways and bridges. Therefore, it is important to make reasonable predictions of OMC and MDD. Four machine learning algorithms, namely, Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest (RF), and Extreme Gradient Boosting Tree (XGBoost), are adopted in this paper to establish MDD and OMC prediction models. After training and testing, the best models of the four algorithms are compared. The results show that, as an ensemble learning algorithm, XGBoost is the best model for predicting MDD and OMC, with an R2 of 0.9234 for OMC, and an R2 of 0.9098 for MDD. Finally, the feature importance analysis concludes that the plastic limit (PL) and the liquid limit (LL) are the two features that affect OMC and MDD the most. The prediction of soil compaction parameters using machine learning models, especially ensemble learning, can significantly reduce the amount of laboratory work and improve the efficiency of optimizing design for soil resource utilization in engineering construction.

This chapter examines the capability of Minimax Probability Machine Regression (MPMR) and Extreme Learning Machine (ELM) for prediction of Optimum Moisture Content (OMC), Maximum Dry Density (MDD) and Soaked California Bearing Ratio (CBR) of soil. These algorithms can analyse data and recognize patterns and are proved to be very useful for problems pertaining to classification and regression analysis. These regression models are used for prediction of OMC and MDD using Liquid limit (LL) and Plastic limit (PL) as input parameters. Whereas Soaked CBR is predicted using Liquid limit, Plastic limit, OMC and MDD as input parameters. The predicted values obtained from the MPMR and ELM models have been compared with that obtained from Artificial Neural Networks (ANN). The accuracy of MPMR and ELM models, their performance and their reliability with respect to ANN models has also been evaluated.

Objectives: This paper brings out the results of an experimental programme carried out at National Institute of Technology, Hamirpur to evaluate the effectiveness of using slate quarry waste for soil stabilization by studying the compaction characteristics of various proportions of slate waste in native soil, as a pre-requirement to reach at soil-slate mix for shear strength and settlement testing for use as a sub grade material for foundations of structures and pavements. Methods/Statistical Analysis: Disturbed sampling of soil was done and was classified according to Indian Standard Soil Classification System (ISSCS), using Liquid limit and Plasticity index. Slate waste was sampled and Sieve analysis was performed. Light compaction tests were performed in order to get the Maximum Dry Density and Optimum Moisture content of soil with slate mining waste in various proportions. Findings: The soil was found to have liquid limit and plastic limit of 46.0833% and 26.1570 % respectively. According to ISSCS, Soil was found to be clay of medium compressibility. Slate waste with cu and ccof9.04 and 0.67 respectively was utilized for the study. The Optimum Moisture Content (OMC) and Maximum Dry Density (MDD) was found to be 24.989% and 1.497 g/cc, 32% and 1.95g/cc for soil and slate waste respectively. Variation of OMC and MDD with slate waste (%) added was analysed. The results show that the addition of waste material has increased the OMC as well as the MDD. An increase in MDD depicts a stable soil. The increase in OMC is due to the water absorption of slate waste. Application/Improvements: These results can be further used for study of strength, settlement behaviour, and slope stability of the soil-slate waste mix. It can be utilized for economics in pavements and improvement of Sub grade of pavements and buildings.

Soil stabilization is the alteration of physicomechanical properties of soils to meet specific engineering requirements of problematic soils. Laboratory examination of soils is well recognized as appropriate for examining the engineering properties of stabilized soils; however, they are labor-intensive, time-consuming, and expensive. In this work, four artificial intelligence based models (OMC-EM, MDD-EM, UCS-EM+, and UCS-EM−) to predict the optimum moisture content (OMC), maximum dry density (MDD), and unconfined compressive strength (UCS) are developed. Experimental data covering a wide range of stabilized soils were collected from previously published works. The OMC-EM, MDD-EM, and UCS-EM− models employed seven features that describe the proportion and types of stabilized soils, Atterberg limits, and classification groups of soils. The UCS-EM+ model, besides the seven features, employs two more features describing the compaction properties (OMC and MDD). An optimizable ensemble method is used to fit the data. The model evaluation confirms that the developed three models (OMC-EM, MDD-EM, and UCS-EM+) perform reasonably well. The weak performance of UCS-EM− model validates that the features OMC and MDD have substantial significance in predicting the UCS. The performance comparison of all the developed ensemble models with the artificial neural network ones confirmed the prediction superiority of the ensemble models.

Lateritic soil is among the major construction materials in road pavement. However, obtaining lateritic soil with sufficient strength is difficult and stabilization with cement is expensive. Stabilization with cement and lime has been well reported but there is paucity of information on stabilization with XL-Bond nanochemicals. In this study, the strength characteristics of stabilized lateritic soil with XL-Bond nanochemical were investigated. Lateritic soil samples were taken from Bariga and Mushin, Lagos State, Nigeria borrowpits (A: Latitude 6.5391°, longitude 3.3849°) and (B: Latitude 6.5273°, longitude 3.3414°), respectively. The geotechnical properties which include Moisture Content (MC), Particle Size Analysis (percentage passing sieve no 200), Plastic Index (PI), Optimum Moisture Content (OMC), Maximum Dry Density (MDD), Unconfined Compressive Strength (UCS) and California Bearing Ratio [CBR] (soaked) of lateritic soils were determined at natural state. The lateritic soils were stabilized at 2, 4, 6, and 8% of XL-Bond nanochemical by dry unit weight of the soil samples. The PI, MDD, CBR (soaked and unsoaked) and UCS of stabilized laterite soil samples were determined in line with standard methods.The MC, percentage passing sieve no 200, PI, OMC, MDD, UCS and CBR (soaked and unsoaked) of soil samples A, B were (45.6%, 38.98%, 5.0%, 10.5%, 2420 g/m3, 59kPa, 12.0% and 10.0% ); (41.1%, 54.52%, 2.0%, 12.50%, 1960g/m3, 72kPa, 4.0% and 20.0%), respectively. The PI, MDD, CBR (soaked and unsoaked) and UCS of stabilized soil with XL-Bond varied from (3-7%; 2.30-2.35g/cm3, 12-23%, 23-50% and 64-83 kPa) and (3-6%; 1.44-1.50g/cm3,6-20%, 21-40% and 79-8 kPa), respectively. There is significant effect of varying percentage of XL-Bond on CBR of stabilized samples (P = 0.1004>0.05). In conclusion, stabilization of Laterite soil with XL-Bond nanochemical enhanced its compressive strength. The stabilized lateritic soil can be used as subgrade materials in highway construction.