The cohesion and internal friction angle of loess are important macro-mechanical parameters for evaluating the safety and stability of engineering construction. Traditional laboratory measurement methods are time-consuming and difficult to conduct on-site. This study aims to compare the effectiveness of five Machine Learning (ML) methods, namely Random Forest (RF), Support Vector Machine (SVM), Back Propagation Neural Network (BPNN), BPNN optimized by Particle Swarm Optimization (PSO-BPNN) and BPNN optimized by Genetic Algorithm (GA-BPNN), in predicting the macro-mechanical properties of loess. To this end, the study collected data from 89 undisturbed loess samples and 229 remolded loess samples to construct training and testing datasets, and used three correlation analysis methods to analyze the influence of physical parameters on mechanical properties. The study found that the water content has the most significant impact on the mechanical properties of loess. In terms of prediction ability, SVM performs the best among the ML methods used, and the determination coefficient for cohesion of undisturbed loess reaches 0.857. Although the training data is limited, the prediction performance of BPNN is significantly improved after being optimized by PSO or GA. The research results show that ML provides an effective way to study the complex mechanical behavior of loess.

To quantify the influence of basic physical properties and cyclic loading conditions on the liquefaction properties of sandy soils, this study uses a combination of physical experiments and numerical simulations to investigate the liquefaction behavior of saturated sandy soils under undrained conditions and their relationship to physical property parameters and external loads. A numerical model with discrete elements was created based on cyclic triaxial tests. A numerical study and predictive analysis of liquefaction of common bulk samples were carried out in conjunction with a PSO-BP neural network prediction model. Using a multivariate analysis of variance and a random forest model, the complexity of the influence of physical parameters and external loads on soil liquefaction was investigated. Quantitative results indicate that particle size distribution, external loads and effective internal friction angle have a significant influence on the liquefaction of saturated sandy soils. In summary, the results of this study provide new insights into understanding the liquefaction behavior of sandy soils.

In the middle and lower reaches of the Yellow River in China, loess-like silty clay is prevalent. This soil type exhibits considerable variability in its compression coefficient alpha, which can lead to differential soil settlement and consequent damage to buildings and infrastructure, thereby posing safety risks. Despite its significance, research and data on this topic are still limited. This study involves comprehensive measurement and laboratory analysis of over one thousand soil samples collected on-site. It establishes a statistical distribution model for essential parameters, including water content w, wet density rho, void ratio e, saturation Sr, liquidity index IL, liquid limit WL, plastic limit WP, and plasticity index IP, and explores the probability distribution characteristics of the physical and mechanical parameters of loess-like silty clay. Machine learning prediction models, utilizing Random Forest (RF) and Deep Neural Network (DNN) algorithms, were developed based on an extensive database to forecast the compression coefficient alpha and compression modulus ES of this soil. The predictive models demonstrated higher accuracy compared to conventional methods and hold significant practical implications for the timely prediction of the mechanical and engineering characteristics of loess-like silty clay. This research provides a robust scientific foundation for engineering design, enhances understanding of the mechanical properties and engineering attributes of this special soil expanse, and reduces the high costs and time consumption associated with engineering geological surveys, as well as the subjectivity of technical and environmental constraints and data interpretation. It serves as a valuable tool for disaster prevention and prediction.

The temporal variability of microphysical parameters of pyrolysis smoke, retrieved by inverting the characteristics of aerosol scattering and extinction, has been studied. The polarization scattering phase functions and spectral extinction coefficients were measured for 65 hours in smoke aerosols produced from thermal decomposition of pine wood during low-temperature pyrolysis in the Big Aerosol Chamber (BAC) of Institute of Atmospheric Optics, Siberian Branch, Russian Academy of Sciences. The microstructure parameters (volume concentration and mean radius of particles with division into fine and coarse fractions) and the complex refractive index of pyrolysis smoke are retrieved following the developed algorithm for inverting optical measurements. The real part of the refractive index is found to be in the vicinity of n = 1.55, and the imaginary part is in the range 0.007 < kappa < 0.009; the mean radius of fine particles varies in the narrow range 0.137-0.146 mu m. During smoke aging, the particle ensemble-mean radius monotonically increased from 0.19 to 0.6 mu m mainly due to a relative increase in the content of coarse aerosol. Results of this work are important for estimation of the radiative forcing of aerosol and improvement of climate models and algorithms of remote optical sounding.