To mitigate the risk of structural failure in coastal engineering caused by soft marine soil creep, this study presents a coupled macroscopic and mesoscopic creep model of soft marine soil to predict long-term deformation behavior of the soil. First, the mesoscopic characteristics of soft marine soil (e.g., pore, particle, and morphological characteristics) under different external pressures were obtained using a scanning electron microscope. Then, both the mesoscopic and macroscopic characteristics of soil were quantified using directional probability entropy and then used as inputs to develop the model. The model predictions agree with the experimental data. In addition, the experimental results indicate linear negative correlations between porosity and pore ratio with stress—the relationships between the fractal dimension of pore distribution and probability entropy of particle orientation under stress are generally nonlinear. Further, results of sensitivity analysis indicate that the probability entropy of particle orientation is one of the most critical parameters governing long-term creep deformation behavior of soft marine soil.

The internal damage of red-bed soft rock induced by water is pervasive. The accumulation, growth, and localization of damage is a multi-scale process that can lead to significant strength loss in red-bed soft rock. Yet, research on the critical state of deterioration process considering multi-scale failure is limited due to high degree of system freedom. Renormalization group theory is an effective approach to find critical point of phase transition in a disordered system. To apply renormalization group theory in red-bed soft rocks, this article firstly analyzed their microstructures. Then, the granular unit model and stripy unit model are proposed to describe the self-similar characteristics of red-bed soft rocks. The calculation results based on renormalization group theory are consistent with the experimental results. The critical reductions of strength induced by water are 60% in light-yellow silty mudstone and 80% in grey silty mudstone. In addition, the critical state of damage propagation caused by stress is also studied and the analytical solution is derived. Results show that the renormalization group theory can effectively couple the micro damage and strength deterioration which provides guidance to the engineering.

The study of water absorption of soft rock is of great significance to the prevention engineering disasters. However, the research on the prediction method of water absorption of soft rock considering the influence of composition, porosity, and solute is insufficient. Aiming to solve the above problem, water absorption tests are carried out by synthesizing soft rocks and water solutions. Then, the prediction model of water absorption of soft rock is established through quantitative analysis of water absorption data and compared with the water absorption characteristics of natural rock to verify the reliability of the model. The results show that the changes in water content and water absorption velocity of soft rock with time obey the second-order exponential attenuation function and the linear function (double logarithmic coordinates axis), respectively. The types of cations and anions and the type and content of clay minerals have the greatest influence on the process of rock water absorption. In addition, the water absorption prediction model could better predict the water absorption process of natural rock. The research results solve the problem of insufficient research on the soft rock water absorption prediction method considering the influence of composition, porosity and solute.

The rates of chemical reactions are highly dependent on temperature, meaning that the actual geological rock mass is affected by different temperatures. Only when the temperature effect is considered can the mechanism of the influence of temperature on the interaction between water and rock be further understood. It was found that the condition of turbulent flow is more likely to promote the softening of red-bed sandstone than the conditions of laminar flow and static water in an experimental study on the softening effects of different flow patterns on red-bed sandstone. Therefore, based on a multi-functional self-circulating open channel hydraulic test system, this paper designs and completes equal volume saturated tests of red-bed sandstone at low temperature (1 °C), medium temperature (23 °C), and high temperature (45 °C) under the turbulent conditions of three equal temperature gradients. The chemical action of the circulating solution in water flow at different temperatures, the propagation of micro-cracks in rock and the changes in mechanical indexes are discussed. The influence laws and mechanisms of the different temperatures on the softening of red-bed sandstone in turbulent flow are revealed. The results show that low-temperature flow can inhibit the softening of red-bed sandstone in the range of 1–45 °C. With the increase in water flow temperature, the development degrees of micro-structures and the mechanical damage of the corresponding rock become more notable. That is, temperature affects the physical and chemical water-rock interactions and then changes the internal structure of rock, thus affecting the softening and failure processes of red-bed sandstone. The study provides a theoretical basis for the further investigation of the softening laws and mechanisms of other red layered soft rocks by temperature under turbulent conditions.

Prediction of soft soil settlement is an important research topic in the field of civil engineering, and the least square support vector machine is one of the commonly used prediction methods at present. Nonetheless, the existing LSSVM models have problems of low search efficiency in the search process and lack of global optimal solution in the search results. In order to solve this problem, based on the leave-one-out cross-validation method, the homotopy continuation method was used to optimize the LSSVM model parameters, and then the HC-LSSVM model was constructed with the goal of minimizing the sum of squares of the prediction error of the full sample retention one. Finally, the rationality and correctness of the model are verified by engineering application. The results show that the HC-LSSVM model constructed in this study can accurately predict the settlement of soft ground, which is superior to the common LSSVM model and solves the problem that the parameters of LSSVM model cannot be solved optimally. The research results provide a new method for prediction of soft soil settlement.