Smart Way for Consistent Cement Bond Evaluation and Reducing Human Bias Using Machine Learning

Abstract

Cement evaluation data acquired in oil and gas wells for confirmation of zonal isolation, channeling in cement behind casing and well integrity. All available technologies for cement evaluations are primarily measurements of acoustic parameters like amplitude of first arrival, full waveform recording of refracted wave, impedance and attenuation or there a combination. Generally, operationsPetrophysicist, petroleum engineer or service providers are responsible for evaluation of cement bond logs and propose remedial jobs if required. This cement bond log interpretation is quite subjective if performed through pure visual interpretation and its accuracy depends on objective of well, work pressure and experience of interpreter. This paper talks about how ADNOC Onshore has leveraged machine learning (ML) for interpretation of various cement bond logs from several service providers. In green fields, cement evaluation is very important in all new wells to ensure good cement quality and zonal isolation and it is also equally important in in brown fields where 1000s of wells were drilled where well integrity issues are of common occurrence. Major challenges in evaluation are inconsistency and human bias in interpretation and as a result, interpretation may vary from one interpreter to the other. Authors have tested different ML techniques (Random Forest Classification & Neural Net) and smart way of data training. Final recommendation is to use nested models instead of single model. In this technique, input measurements and required solutions will be classified and divided into different classes, a separate ML model will be built for each class and combine all the models to get final cement evaluation and recommendations. Final results include data quality flags, cement bond quality, zonal isolation, channeling, micro annuals, areal cement map (if analyzed field wise), anomaly maps (if analyzed data from offset wells at different times), and recommendation. In this project, a simple easy to use user interface has been developed to browse the cement logs and use the trained ML models to predict the cement evaluations with a click of a button. This ML based cement bond evaluation is proved to be very effective and saved 75% of efforts by operational Petrophysicists. The interpretation accuracy has been significantly improved. This method has potential to be used at rig site. These models are very cost effective with, minimum human bias, and improves consistency as well as independent of wells or type of reservoirs. These models are tested across ADNOC Onshore Wells and can be extended to any well irrespective of geographic locations. This paper discusses machine-learning approach, evaluation of various algorithms and testing results for cement evaluation log from various service providers.

Assisted Cement Log Interpretation Using Machine Learning

“…Viggen et al (2021) further improved this performance by reducing the negative combined effect of interpreter subjectivity and data complexity identified in their previous work by means of feature engineering (i.e., designing predictive features based on the raw log data) and using ML algorithms that are less susceptible to overfitting (Webb 2011). Voleti et al (2020) also reported good results with a comparable tool trained and tested on a small number of wells, estimating that it could save their company 75% of their current interpretation effort.…”

“…To help avoid getting many very short zones as seen in previous publications on automatic cement log interpretation tools (Reolon et al 2020, Viggen et al 2020, Voleti et al 2020, and Viggen et al 2021, the tool includes a post-processing component. The user can specify the minimum length of a zone (e.g., 3 m), and the tool will merge and/or replace shorter zones based on the classifier's class probability distribution.…”

Cementing Quality Prediction in the Shunbei Block Based on Genetic Algorithm and Support Vector Regression

“…In addition to this, there are other types. Deepak Kumar Voleti et al examined the efficacy of random forest classification and neural network models in cement evaluation and proposed the use of nested models that have low bias, improving interpretation accuracy and cement evaluation in onshore wells; however, the work did not address the issues of model generalizability, data dependency, or potential overfitting [15]. Ni Hongmei et al introduced a particle swarm optimization algorithm based on stochastic global optimization in neural network training.…”

A Novel Cementing Quality Evaluation Method Based on Convolutional Neural Network

“…Belozerov (2018) et al [7] used neural networks to identify reservoir locations from logging data, and Gkortsas (2019) et al [15] used support vector machines and neural networks to automatically identify ultrasonic waveform characteristics, which can predict additional information about the longitudinal wave velocity of annular materials in cased wells. Deepak Kumar Voleti et al [9] (2020) established different machine learning algorithms, such as random forest and neural network prediction based on CBL-VDL, and ultrasonic imaging data, to output the prediction results of cementing quality. Santos, L. et al [10] (2021) used the Gaussian process regression algorithm for training and generated new characteristic curves based on CBL and VDL logging data to accurately evaluate the cementing quality.…”

“…In 2015, Chen Xiangjun et al [8] proposed evaluating the cementing quality according to the method of acoustic energy, which has a high interpretation efficiency but insufficient accuracy. In 2020, Deepak Kumar Voleti et al [9] established different machine learning algorithms, such as CBL-VDL-based random forest and neural network prediction, and ultrasonic imaging data, to output the prediction results of cementing quality and achieve an automatic interpretation of cementing quality. In 2021, Santos, L. et al [10] used the Gaussian process regression algorithm for training, generated new characteristic curves according to CBL and VDL logging data, and accurately evaluated the cementing quality through the new curves.…”