Real-time prediction of Poisson's ratio from drilling parameters using machine learning tools

Estimation of rocks' failure parameters from drilling data by using artificial neural network

Comprehensive and precise knowledge about rocks' mechanical properties facilitate the drilling performance optimization, and hydraulic fracturing design and reduces the risk of wellbore-related problems. This paper is concerned with the failure parameters, namely, cohesion and friction angle which are conventionally estimated using Mohr's cycles that are drawn using compressional tests on rock samples. The availability, continuity and representability, and cost of acquiring those samples are major concerns. The objective of this paper is to investigate an alternative technique to estimate these parameters from the drilling data. In this work, more than 2200 data points were used to develop and test the correlations built by the artificial neural network. Each data point comprises the failure parameters and five drilling records that are available instantaneously in drilling rigs such as rate of penetration, weight on bit, and torque. The data were grouped into three datasets, training, testing, and validation with a corresponding percentage of 60/20/20, the former two sets were utilized in the models' building while the last one was hidden as a final check afterward. The models were optimized and evaluated using the correlation coefficient (R) and average absolute percentage error (AAPE). In general, the two models yielded good fits with the actual values. The friction angle model yielded R values around 0.86 and AAPE values around 4% for the three datasets. While the model for cohesion resulted in R values around 0.89 and APPE values around 6%. The equation and the parameters of those models are reported in the paper. These results show the ability of in-situ and instantaneous rock mechanical properties estimation with good reliability and at no additional costs.

A Deep Learning Method for the Prediction of the Index Mechanical Properties and Strength Parameters of Marlstone

The index mechanical properties, strength, and stiffness parameters of rock materials (i.e., uniaxial compressive strength, c, ϕ, E, and G) are critical factors in the proper geotechnical design of rock structures. Direct procedures such as field surveys, sampling, and testing are used to estimate these properties, and are time-consuming and costly. Indirect methods have gained popularity in recent years due to their time-saving and highly accurate results, which are comparable to those obtained through direct approaches. This study presents a procedure for establishing a deep learning-based predictive model (DNN) for obtaining the geomechanical characteristics of marlstone samples that have been recovered from the South Pars region of southwest Iran. The model was implemented on a dataset resulting from the execution of numerous geotechnical tests and the evaluation of the geotechnical parameters of a total of 120 samples. The applied model was verified by using benchmark learning classifiers (e.g., Support Vector Machine, Logistic Regression, Gaussian Naïve Bayes, Multilayer Perceptron, Bernoulli Naïve Bayes, and Decision Tree), Loss Function, MAE, MSE, RMSE, and R-square. According to the results, the proposed DNN-based model led to the highest accuracy (0.95), precision (0.97), and the lowest error rate (MAE = 0.13, MSE = 0.11, and RMSE = 0.17). Moreover, in terms of R², the model was able to accurately predict the geotechnical indices (0.933 for UCS, 0.925 for E, 0.941 for G, 0.954 for c, and 0.921 for φ).

Machine learning-based prediction and inverse design of 2D metamaterial structures with tunable deformation-dependent Poisson's ratio

With the aid of recent efficient and prior knowledge-free machine learning (ML) algorithms, extraordinary mechanical properties such as negative Poisson's ratio have extensively promoted the diverse designs of metamaterials with distinctive cellular structures. However, most existing ML approaches applied to the design of metamaterials are primarily based on a single property value with the assumption that the Poisson's ratio of a material is stationary, neglecting the dynamic variability of Poisson's ratio, termed deformation-dependent Poisson's ratio, during the loading process that a metamaterial other than conventional materials may experience. This paper first proposes a crystallographic symmetry-based methodology to build 2D metamaterials with complex but patterned topological structures, and then converts them into computational models suitable for molecular dynamics simulations. Then, we employ an integrated approach, consisting of molecular dynamics simulations capable of generating and collecting a large dataset for training/validation, in addition to ML algorithms (CNN and Cycle-GAN) able to predict the dynamic characteristics of Poisson's ratio and offer the inverse design of a metamaterial structure based on a target quasi-continuous Poisson's ratio-strain curve, to eventually unravel the underlying mechanism and design principles of 2D metamaterial structures with tunable Poisson's ratio. The close match between the predefined Poisson's ratio response and that from the generated structure validates the feasibility of the proposed ML model. Owing to the high efficiency and complete independence from prior knowledge, our proposed approach offers a novel and robust technique for the prediction and inverse design of metamaterial structures with tailored deformation-dependent Poisson's ratio, an unprecedented property attractive in flexible electronics, soft robotics, and nanodevices.

Real-time prediction of formation pressure gradient while drilling

Accurate real-time pore pressure prediction is crucial especially in drilling operations technically and economically. Its prediction will save costs, time and even the right decisions can be taken before problems occur. The available correlations for pore pressure prediction depend on logging data, formation characteristics, and combination of logging and drilling parameters. The objective of this work is to apply artificial neural networks (ANN) and adaptive neuro-fuzzy inference system (ANFIS) to introduce two models to estimate the formation pressure gradient in real-time through the available drilling data. The used parameters include rate of penetration (ROP), mud flow rate (Q), standpipe pressure (SPP), and rotary speed (RS). A data set obtained from some vertical wells was utilized to develop the predictive model. A different set of data was utilized for validating the proposed artificial intelligence (AI) models. Both models forecasted the output with a good correlation coefficient (R) for training and testing. Moreover, the average absolute percentage error (AAPE) did not exceed 2.1%. For validation stage, the developed models estimated the pressure gradient with a good accuracy. This study proves the reliability of the proposed models to estimate the pressure gradient while drilling using drilling data. Moreover, an ANN-based correlation is provided and can be directly used by introducing the optimized weights and biases, whenever the drilling parameters are available, instead of running the ANN model.