The clogging of porous media with solid particle suspension flow is modeled using two empirical parameters of filtration coefficient (7) and formation damage coefficient (/3). These parameters are typically determined through coreflood tests. This study employs machine learning techniques to predict 7 and /3 using experimental data from open literature. The prediction of /3 is based on critical porosity fraction (gamma) data and a power law equation relating /3 and gamma. Collected data were randomly partitioned into training (80 %) and testing (20 %) subsets. Four regression algorithms were employed, treating 7 or gamma as the target variable, with injection velocity (um), particle concentration (Cin), and ratio of mean pore size (Dpore) to mean particle size (Dp) as features. The extreme gradient boosting (XGBoost) algorithm showed the best performance. The feature Cin had the highest influence on 7 and gamma, revealing a significant finding previously overlooked. Postmortem analyses revealed qualitative consistencies in 7 results, supporting the existence of critical velocities. Furthermore, 7 results showed a power law relation between 7 and all three features used. An equation was formulated to estimate 7 as a function of these three features. A direct prediction of /3 using these features was established by applying the XGBoost model to predict gamma and then employing an existing power law relationship between /3 and gamma. This study demonstrated that machine learning offers an alternative approach for predicting 7 and /3, which is particularly useful for initial evaluations of clogging potentials and identification of experimental conditions to focus on.

In slurry shield tunneling, the stability of tunnel face is closely related to the filter cake. The cutting of the cutterhead has negative impact on the formation of filter cake. This study focuses on the formation time of dynamic filter cake considering the filtration effect and rotation of cutterhead. Filtration effect is the key factor for slurry infiltration. A multilayer slurry infiltration experiment system is designed to investigate the variation of filtrate rheological property in infiltration process. Slurry mass concentration CL, soil permeability coefficient k, the particle diameter ratio between soil equivalent grain size and representative diameter of slurry particles d(10)/D-85 are selected as independent design variables to fit the computational formula of filtration coefficient. Based on the relative relation between the mass of deposited particles in soil pores and infiltration time, a mathematical model for calculating the formation time of dynamic filter cake is proposed by combining the formation criteria and formation rate of external filter cake. The accuracy of the proposed model is verified through existing experiment data. Analysis results show that filtration coefficient is positively correlated with slurry mass concentration, while negatively correlated with the soil permeability coefficient and the particle diameter ratio between soil and slurry. As infiltration distance increases, the adsorption capacity of soil skeleton to slurry particles gradually decreases. The formation time of external filter cake is significantly lower than internal filter cake and the ratio is approximately 3.9. Under the dynamic cutting of the cutterhead, the formation time is positively associated with the rotation speed of cutter head, while negatively with the phase angle difference between adjacent cutter arm. The formation rate of external filter cake is greater than 98% when d(10)/D-85 <= 6.1. Properly increasing the content or decreasing the diameter size of solid-phase particles in slurry can promote the formation of filter cake.