Effective Machine-Learning Models for Rock Mass Deformation Modulus Estimation Based on Rock Mass Classification Systems

Engineered Science, Год журнала: 2024, Номер unknown

Опубликована: Янв. 1, 2024

The rock mass deformation modulus (RMDM) plays a crucial role in dam and tunnel design.This study introduces advanced machine-learning (ML) models to predict RMDM using rating (RMR) the Q-system at Khersan-2 site southwestern Iran.Through analysis of exploratory boreholes, engineering geological properties samples, Q, RMR, RMDM, strength index (GSI), Hoek-Brown, shear constants were determined.Subsequently, seven effective ML models, namely random forest, multilayer perceptron backpropagation artificial neural network, Gaussian process regression, Knearest neighbor, simple multiple linear non-linear regression approaches, utilized estimate RMDM.Based on classification systems, was rated as having good RMR Q categories.A new empirical relationship with high accuracy established between RMR89.Furthermore, demonstrated strong correlation supported by statistical analysis.The results showed relative superiority compared ones.The employed techniques displayed remarkable estimating achieving coefficient determination (R 2 ) greater than 97%.Notably, squared exponential kernel function stood out most approach, yielding outstanding performance predicting an impressive R =0.99 RMSE=0.01 all other investigated methods.

Язык: Английский

---

Estimation of static Young’s modulus of sandstone types: effective machine learning and statistical models

Earth Science Informatics, Год журнала: 2024, Номер 17(5), С. 4339 - 4359

Опубликована: Июль 4, 2024

Abstract The elastic modulus is one of the important parameters for analyzing stability engineering projects, especially dam sites. In current study, effect physical properties, quartz, fragment, and feldspar percentages, dynamic Young’s (DYM) on static (SYM) various types sandstones was assessed. These investigations were conducted through simple multivariate regression, support vector adaptive neuro-fuzzy inference system, backpropagation multilayer perceptron. XRD thin section results showed that studied samples classified as arenite, litharenite, feldspathic litharenite. low resistance arenite type mainly due to presence sulfate cement, clay minerals, high porosity, carbonate fragments in this type. Examining fracture patterns these different ranges at values resistance, pattern shear type, which changes multiple extension with increasing compressive strength. Among influencing factors, percentage quartz has greatest SYM. A comparison methods' performance based CPM error estimating SYM revealed SVR (R 2 = 0.98, RMSE 0.11GPa, + 1.84) outperformed other methods terms accuracy. average difference between predicted using intelligent measured value less than 0.05% indicates efficiency used

Язык: Английский

---

A Comparative Study of Ensemble Learning Techniques and Mathematical Models for Rigorous Modeling of Solution Gas/Oil Ratio

SPE Journal, Год журнала: 2025, Номер unknown, С. 1 - 26

Опубликована: Фев. 1, 2025

Summary The solution gas/oil ratio (Rs) represents the quantity of gas dissolved in oil under reservoir conditions. It is a vital parameter petroleum engineering, defining content available during production. While many experimental techniques exist for measuring this ratio, they often require considerable time and resources. Thus, mathematical intelligent models are essential accurate determination. A total 720 data points from diverse geographical regions were collected published studies research, using gas-specific gravity, temperature, bubblepoint pressure, API gravity as inputs, with output. Statistical physical analyses assessed impact parameters on revealing that temperature does not always decrease gas. Beyond specific point, known inversion higher temperatures enhance solubility. set was split, 80% allocated training 20% testing. accuracy Al-Marhoun model, originally established 160 sets Middle East, evaluated test data, which produced root mean square error (RMSE) 468.79 scf/STB. recalibration coefficients 576 differential evolution (DE) algorithm led to formulation New Model 1. By incorporating effect 2 developed. Testing results showed 1 achieved an RMSE 100.97 scf/STB, while reached 105.1 both showing better compared previous models, including model. Subsequently, machine learning applied, multilayer group method handling (GMDH), voting regressor (VR), extra trees (ET), histogram-based gradient boosting regression (HGBR), extreme (XGBoost), categorical features support (CatBoost) modeling process. Notably, such ET, HGBR, XGBoost, CatBoost effectively captured data. performance statistical visual analyses. HGBR model outperformed all others, achieving 0.0044 scf/STB value 73.03 demonstrating its clear superiority among considered models.

Язык: Английский

---

Effect of textural and physical properties of the carbonate rocks on dynamic elastic modulus: application of statistical and intelligent methods

Physics and Chemistry of the Earth Parts A/B/C, Год журнала: 2025, Номер unknown, С. 103964 - 103964

Опубликована: Май 1, 2025

Язык: Английский

---

Estimating Rock Mechanical Properties Using Statistical and Intelligent Methods Based on Physical, Acoustic, and Hardness Data

Geotechnical and Geological Engineering, Год журнала: 2025, Номер 43(5)

Опубликована: Июнь 1, 2025

Язык: Английский

---

Detection and prediction of slope stability in unsaturated finite slopes using interpretable machine learning

Modeling Earth Systems and Environment, Год журнала: 2025, Номер 11(4)

Опубликована: Июнь 4, 2025

Язык: Английский

---

Developing some models to predict the uniaxial compressive strength of various sedimentary rocks (Case studies: large dam site and mine in Southeast China)

Case Studies in Construction Materials, Год журнала: 2024, Номер 21, С. e03817 - e03817

Опубликована: Окт. 6, 2024

Язык: Английский

---

A Novel Inversion Method for Permeability Coefficients of Concrete Face Rockfill Dam Based on Sobol-IDBO-SVR Fusion Surrogate Model

Mathematics, Год журнала: 2024, Номер 12(7), С. 1066 - 1066

Опубликована: Апрель 2, 2024

The accurate and efficient inversion of permeability coefficients is significant for the scientific assessment seepage safety in concrete face rockfill dams. In addressing optimization challenge with few samples, multiple parameters, strong nonlinearity, this paper proposes a novel intelligent method based on Sobol-IDBO-SVR fusion surrogate model. Firstly, Sobol sequence sampling introduced to extract high-quality combined samples coefficients, equivalent continuum model utilized forward simulation obtain theoretical hydraulic heads at monitoring points. Subsequently, support vector regression used establish complex mapping relationship between heads, convergence performance dung beetle algorithm effectively enhanced by fusing strategies. On basis, we successfully achieve precise driven multi-intelligence technologies. engineering application results show that determined can reasonably reflect characteristics dam. maximum relative error measured values each point only 0.63%, indicating accuracy meets requirements. proposed study may also provide beneficial reference similar parameter problems projects such as bridges, embankments, pumping stations.

Язык: Английский

---

Prediction of Rock's Brittleness and Dynamic Properties Utilizing Effective Artificial Intelligence Approaches

Periodica Polytechnica Civil Engineering, Год журнала: 2024, Номер unknown

Опубликована: Май 16, 2024

This research aims to determine the brittleness index (BI) and engineering properties of limestone specimens. In addition, this study evaluates effect moisture on developed models predict BI shear wave velocity (Vs), based point load (Is50), dry saturated tensile strength (Ts-d Ts-s), porosity. Gaussian process regression (GPR), multilayer feed-forward neural network (MFFNN), multiple linear (MLR) predictive were utilized. Microscopic examination specimens revealed that calcite is predominant mineral. It was observed samples with higher content exhibited greater while having lower The results obtained from MLR analysis demonstrated it possible accurately forecast (BI), as well velocities (Vs-d Vs-s) at specific sites under investigation. showed Vs prediction in conditions (Vs-d) less accurate compared (Vs-s). Conversely, relationships for estimating accuracy. all model assumptions using indicated could be reliably However, MFFNN GPR methods found more conservative these properties. Moreover, identified best transfer function training algorithm predicting BI. evaluation metrics, such R2 RMSE precision MLR.

Язык: Английский

---

The effect of mineralogical, mechanical, physical, and dynamic properties on rock brittleness using statistical and soft computing methods

Earth Science Informatics, Год журнала: 2024, Номер 18(1)

Опубликована: Дек. 17, 2024

Язык: Английский

---