This study will deal with the essential weaknesses of traditional elastic models to predict wellbore instability

that in most cases results in expensive drilling procedures. The aim of this study is to develop an integrated

drilling geomechanics analysis using Poro-Elasto-Plastic finite elements and machine learning. The objectives

are to, integrate drilling operations using a poro-elasto-plastic Finite Element Model; integrate drilling operations

using machine learning algorithms. The analysis evolves a synthesized framework that combines a poro-

elastoplastic Finite Element Model (FEM) and a Machine Learning (ML) to allow a dynamic and precise

in advance to increase the safety of drilling, streamline operations, and minimize non-productive time, thereby

Geomechanical issues like instability of wellbore and breakouts of well-borings continue to be a leading reason

of non-productive time in the oil and gas sector and especially in intricate and profound oil reservoirs (Zhang et

al., 2022; Osaki et al 2025a). Conventional forms of analysis have difficulties in taking into account the

complexity of such problems which are time-dependent, and coupled thermal-hydraulic-mechanical processes.

However, Poro-elasto-plastic finite element modeling, which is an advanced numerical model, has proven to

have substantial potential in offering a more realistic simulation of wellbore response through modeling of plastic

can not only help to increase the accuracy of the forecast of wellbore stability, but also costs a lot less to compute,

The aim of this study is to develop an integrated drilling geomechanics analysis using Poro-Elasto-Plastic finite

Agbami Field is a great field to conduct research on integrated drilling geomechanics based on poroelasto-plastic

finite element based on machine learning. This large deep offshore field is at the central Niger Delta in Nigeria,

Element Method (FEM) (Komijani, 2024). This combined method has the capability of dynamically forecasting

wellbore failure and revising safe mud weight margin through the processing of on-the-fly drilling information

and offset well information and thus overcoming the constraints of the traditional unchanging models. The model

would directly respond to geomechanical risks in Agbami Field, which would help to reduce unproductive time,

Fig. 1: Map of Agbami Field Showing the Oil Field, Gas Field, Liquification Plant, FPSO, and African Oil Assert

Linear-elastic assumption and analytical solutions are commonly used to describe the traditional wellbore

important in the analysis of wellbore breakout. In addition, they use coupled THM processes to capture the most

important effects of redistribution of the pore pressure, and the thermal stresses caused by the temperature

difference between drilling fluid and the formation. These thermo-poroelastic effects have a major impact on

shear failure and the stable mud window and thus they play a vital role in the modern-day drilling activities

Machine Learning has become one of the effective geomechanical analysis tools. Artificial Neural Networks

(ANNs), Random Forest (RF) and Support Vector Machines (SVR) are now commonly used as ML algorithms

angle using only drilling data (Liu et al., 2024; Osaki & Oghonyon, 2025). Khan, et al., (2025) noted that, the

main advantage of ML is that it can model non-linear, complex relationships that are computational, providing

an alternative to physics-based simulation that is fast. The latest line of research development is the combination

of these two directions. In a seminal work, AlBahrani et al. (2021) suggested an integrated scheme, such that a

Learning for Geomechanics analysis starts with the collection of a multi-scale dataset. These are geophysical

well logs (e.g., sonic, density, neutron), direct downhole measurements, such as Leak-off Tests (LOT) to calibrate

the stress, and laboratory analyses on core plugs to measure the statically rock strength and elastic properties.

One of them involves empirical correlations obtained through statistical analysis of laboratory and log data to

estimate at any given time parameters such as UCS and static Youngs modulus over the entire formation interval

Representative Elementary Volume (REV) analysis were done to control the cost of computation. This is to make

sure that the simulated volume is large enough to be representative of the entire formation and yet

computationally efficient. The calibrated in-situ stresses, pore pressure and rock properties are included in the

employed a new analytical solution that Liu and Huang (2021) had developed in which plasticity is applied at

the drainage boundary and propagates to the undrained end. Another reason why Sierra/Aria lacks a plastic

material model is because we can test the coupling solution with Sierra/Arpeggio where Sierra/Aria is resolved

in pore pressure and Sierra/SM is resolved in the displacement as stated above. In the case of plasticity,

horizontal stress along with the pore pressure, the third track is showing friction angle and horizontal stress

Depends on a single modelling

technique, like the independent Finite

Element Method (FEM) for

Incorporates several approaches. For example,

Poro-Elasto-Plastic FEM and Machine

Learning (ML) algorithms are combined in one

known hybrid approach to produce an integrated

Model Integration

Employs sequential, segregated

Strives for a system that is more cohesive. The

hybrid model is intended to be an "Integrated

focusses on using well-established

physical models to forecast soil

response and pore pressure coupling

aims for real-time application and improved

predictive capabilities. The combined FEM-ML

model is appropriate for a real-time setting and