A set of hyperparameters for six ML models, the ranges used for model optimization, and the best values obtained for three classification problems.

A set of hyperparameters for six ML models, the ranges used for model optimization, and the best values obtained for three classification problems.

Source publication
Article
Full-text available
Gas–liquid flow is a significant phenomenon in various engineering applications, such as in nuclear reactors, power plants, chemical industries, and petroleum industries. The prediction of the flow patterns is of great importance for designing and analyzing the operations of two-phase pipeline systems. The traditional numerical and empirical method...

Contexts in source publication

Context 1
... best hyperparameters were recorded when the model achieved the best accuracy on the validation set during the training. The different hyperparameters of the ML models and their optimized values which were obtained during the cross-validated training and are listed in Table 3. The models were then fitted with the training data using optimized hyperparameters. ...
Context 2
... set of hyperparameters of the different ML algorithms were optimized for three classification problems with six (6), three (3), and two (2) classes in a threefold cross-validation framework. The optimized parameter values are listed in Table 3. ...
Context 3
... set of hyperparameters of the different ML algorithms were optimized for three classification problems with six (6), three (3), and two (2) classes in a threefold crossvalidation framework. The optimized parameter values are listed in Table 3. ...
Context 4
... set of hyperparameters of the different ML algorithms were optimized for classification problems with six (6), three (3), and two (2) classes in a threefold cross dation framework. The optimized parameter values are listed in Table 3. ...
Context 5
... ML models were learned through hyperparameter optimization in a threefold cross-validated training framework. The optimized algorithmic parameters are reported in Table 3. The training was performed using all features as well as the top four important features. ...

Citations

... It provides an overall measure of how often the model makes correct predictions. Mathematically, it is expressed by Equation (5) [36,37]. ...
... This metric evaluates the accuracy of the model's positive prediction results. Mathematically, it is expressed by Equation (6) [36,37]. ...
... This metric evaluates the model's ability to detect all relevant instances within the liquid gas samples dataset. Mathematically, it is expressed by Equation (7) [36,37]. , ...
Article
Full-text available
This paper introduces an optimized deep neural network (DNN) framework for an efficient gas detection system applicable across various settings. The proposed optimized DNN model addresses key issues in conventional machine learning (ML), including slow computation times, convergence issues, and poor adaptability to new data, which can result in increased prediction errors and reduced reliability. The proposed framework methodology comprises four phases: data collection, pre-processing, offline DNN training optimization, and online model testing and deployment. The training datasets are collected from seven classes of liquid beverages and environmental air samples using integrated gas sensor devices and an edge intelligence environment. The proposed DNN algorithm is trained on high-performance computing systems by fine-tuning multiple hyperparameter optimization techniques, resulting in an optimized DNN. This well-trained DNN model is validated using unseen new testing datasets in high-performance computing systems. Experimental results demonstrate that the optimized DNN can accurately recognize different beverages, achieving an impressive detection accuracy rate of 98.29%. The findings indicate that the proposed system significantly enhances gas identification capabilities and effectively addresses the slow computation and performance issues associated with traditional ML methods. This work highlights the potential of optimized DNNs to provide reliable and efficient contactless detection solutions across various industries, enhancing real-time gas detection applications.
... Haobin Chen et al. [23] introduced an intrusive robust CNN flow pattern recognition method, based on flow-induced vibration (FIV) analysis, with the accuracy rate being above 90% when using different axis data to predict flow patterns. Noor Hafsa et al. [24] compared and analyzed machine learning (ML) and deep learning (DL), concluding that extreme gradient boosting is the optimal model for predicting the two-phase flow states of inclined or horizontal pipelines. ...
Article
Full-text available
Gas–Liquid two-phase flows are a common flow in industrial production processes. Since these flows inherently consist of discrete phases, it is challenging to accurately measure the flow parameters. In this context, a novel approach is proposed that combines the pyramidal Lucas-Kanade (L–K) optical flow method with the Split Comparison (SC) model measurement method. In the proposed approach, videos of gas–liquid two-phase flows are captured using a camera, and optical flow data are acquired from the flow videos using the pyramid L–K optical flow detection method. To address the issue of data clutter in optical flow extraction, a dynamic median value screening method is introduced to optimize the corner point for optical flow calculations. Machine learning algorithms are employed for the prediction model, yielding high flow prediction accuracy in experimental tests. Results demonstrate that the gradient boosted regression (GBR) model is the most effective among the five preset models, and the optimized SC model significantly improves measurement accuracy compared to the GBR model, achieving an R2 value of 0.97, RMSE of 0.74 m3/h, MAE of 0.52 m3/h, and MAPE of 8.0%. This method offers a new approach for monitoring flows in industrial production processes such as oil and gas.
... Among these, we selected Optuna since the hyper-parameter search space can be defined more freely by the user in this tool. A total of 6 ML classifiers (Artificial Neural Networks, Support Vector Machines, K-Nearest Neighbors, Random Forest (RF), Gradient Boosting Classifier (GBC), and Logistic Regression (LR)) were considered in this work, which are also the ones typically chosen in other flow regime classification studies [59]. Optuna was set to maximize the 3-fold cross-validation accuracy score to find the best tuned model via Bayesian Optimization. ...
Article
Accurate identification of flow regimes is paramount in several industries, especially in chemical and hydrocarbon sectors. This paper describes a comprehensive data-driven workflow for flow regime identification. The workflow encompasses: i) the collection of dynamic pressure signals using an experimentally verified numerical two-phase flow model for three different flow regimes: stratified, slug and annular flow, ii) feature extraction from pressure signals using Discrete Wavelet transformation (DWT), iii) Evaluation and testing of 12 different Dimensionality Reduction (DR) techniques, iv) the application of an AutoML framework for automated Machine Learning classifier selection among K-Nearest Neighbors, Artificial Neural Networks, Support Vector Machines, Gradient Boosting, Random Forest, and Logistic Regression, with hyper-parameter tuning. Kernel Fisher Discriminant Analysis (KFDA) is the best DR technique, exhibiting superior goodness of clustering, while KNN proved to be the top classifier with an accuracy of 92.5 % and excellent repeatability. The combination of DWT, KFDA and KNN was used to produce a virtual flow regime map. The proposed workflow represents a significant step forward in automating flow regime identification and enhancing the interpretability of ML classifiers, allowing its application to opaque pipes fitted with pressure sensors for achieving flow assurance and automatic monitoring of two-phase flow in various process industries.
... The core concept of machine learning revolves around creating mathematical models that can automatically learn and adapt by learning the patterns in data or through previous experiences (Hafsa et al., 2023;Uthayasuriyan et al., 2023). These models can be trained using a diverse range of datatypes, such as images, text, or numerical values, and various techniques are employed to extract meaningful features and relationships from the data. ...
Article
In the oil and gas industry, understanding two-phase (gas-liquid) flow is pivotal, as it directly influences equipment design, quality control, and operational efficiency. Flow pattern determination is thus fundamental to industrial engineering and management. This study utilizes the Tree-based Pipeline Optimization Tool (TPOT), an Automated Machine Learning (AutoML) framework that employs genetic programming, in obtaining the best machine learning model for a provided dataset. This paper presents the design of flow pattern prediction models using the TPOT. The TPOT was applied to predict flow patterns in 2.5 cm and 5.1 cm diameter pipes, using datasets from existing literature. The datasets went through handling of imbalanced data, standardization, and one-hot encoding as data preparation techniques before being fed into TPOT. The models designed for the 2.5 cm and 5.1 cm datasets were named as FPTL_TPOT_2.5 and FPTL_TPOT_5.1, respectively. A comparative analysis of these models alongside other standard supervised machine learning models and similar state-of-the-art similar two-phase flow prediction models was carried out and the insights on the performance of these TPOT designed models were discussed. The results demonstrated that models designed with TPOT achieve remarkable accuracy, scoring 97.66% and 98.09%, for the 2.5 cm and 5.1 cm datasets respectively. Furthermore, the FPTL_TPOT_2.5 and FPTL_TPOT_5.1 models outperformed other counterpart machine learning models in terms of performance, underscoring TPOT’s effectiveness in designing machine learning models for flow pattern prediction. The findings of this research carry significant implications for enhancing efficiency and optimizing industrial processes in the oil and gas sector.
... Currently, researchers are increasingly turning to the application of machine learning (ML) and deep learning techniques to achieve more objective flow regime identification (Hafsa et al., 2023). These approaches offer the potential to predict flow patterns in a more objective and automated manner. ...
Article
Purpose: Identifying the flow regime is a prerequisite for accurately modeling two-phase flow. This paper aims to introduce a comprehensive data-driven workflow for flow regime identification. Design/methodology/approach: A numerical two-phase flow model was validated against experimental data and was used to generate dynamic pressure signals for three different flow regimes. First, four distinct methods were used for feature extraction: discrete wavelet transform (DWT), empirical mode decomposition, power spectral density and the time series analysis method. Kernel Fisher discriminant analysis (KFDA) was used to simultaneously perform dimensionality reduction and machine learning (ML) classification for each set of features. Finally, the Shapley additive explanations (SHAP) method was applied to make the workflow explainable. Findings: The results highlighted that the DWT + KFDA method exhibited the highest testing and training accuracy at 95.2% and 88.8%, respectively. Results also include a virtual flow regime map to facilitate the visualization of features in two dimension. Finally, SHAP analysis showed that minimum and maximum values extracted at the fourth and second signal decomposition levels of DWT are the best flow-distinguishing features. Practical implications: This workflow can be applied to opaque pipes fitted with pressure sensors to achieve flow assurance and automatic monitoring of two-phase flow occurring in many process industries. Originality/value: This paper presents a novel flow regime identification method by fusing dynamic pressure measurements with ML techniques. The authors’ novel DWT + KFDA method demonstrates superior performance for flow regime identification with explainability.
... In the realm of multiphase flow, AI-based tools have been widely utilized for prediction and classification purposes [14][15][16][17][18]. However, when it comes to predicting pressure losses specifically for stable emulsion transportation using ML, there is a striking absence of substantial attempts or studies in the existing literature. ...
... The hyperparameter optimization of the ML algorithms was performed using the grid search method via a 5-fold CV framework. We chose a set of important parameters for each algorithm to optimize on the training data, which is a common practice (see, for example, [18,21]). A range or a list of values was specified for each parameter during the grid search with MSE as the evaluation metric. ...
Article
Full-text available
One of the significant challenges to designing an emulsion transportation system is predicting frictional pressure losses with confidence. The state-of-the-art method for enhancing reliability in prediction is to employ artificial intelligence (AI) based on various machine learning (ML) tools. Six traditional and tree-based ML algorithms were analyzed for the prediction in the current study. A rigorous feature importance study using RFECV method and relevant statistical analysis was conducted to identify the parameters that significantly contributed to the prediction. Among 16 input variables, the fluid velocity, mass flow rate, and pipe diameter were evaluated as the top predictors to estimate the frictional pressure losses. The significance of the contributing parameters was further validated by estimation error trend analyses. A comprehensive assessment of the regression models demonstrated an ensemble of the top three regressors to excel over all other ML and theoretical models. The ensemble regressor showcased exceptional performance, as evidenced by its high R² value of 99.7 % and an AUC-ROC score of 98 %. These results were statistically significant, as there was a noticeable difference (within a 95 % confidence interval) compared to the estimations of the three base models. In terms of estimation error, the ensemble model outperformed the top base regressor by demonstrating improvements of 6.6 %, 11.1 %, and 12.75 % for the RMSE, MAE, and CV_MSE evaluation metrics, respectively. The precise and robust estimations achieved by the best regression model in this study further highlight the effectiveness of AI in the field of pipeline engineering.
... More recent studies (Ezzatabadipour et al. 2017;Arteaga-Arteaga et al. 2021;Li et al. 2014;Mask et al. 2019;Lin et al. 2020;Hafsa et al. 2023) explored the multiple artificial intelligence algorithms (support vector machine, neural network model and a regression tree, deep learning, Random Forest, Boosted tree, Xtreme gradient boosting) for the identification of flow patterns. The studies employed pipe size, inclination angle, fluid properties, flow velocities, and other experimental data as input features for these algorithms. ...
... Due to the linear relationship between the input features and a predicted variable, these friendly-use algorithms are selected to investigate the phenomena rather than deep learning techniques. The features and details of the applied algorithms are provided in the literature (Hafsa et al. 2023;Rushd et al., 2021). ...
... The best-performing algorithm (CatBoost) orders the input features from highest to lowest rank: superficial gas velocity, inclination angle, superficial liquid velocity, pipe diameter, liquid viscosity, and liquid density. This arrangement has been found inconsistent with the previous finding (Hafsa et al. 2023), in which the superficial liquid velocity is found to be the most dominant test parameter in defining two-phase flow patterns. It is important to note that Hafsa et al. (2023) performed feature importance analysis using the Extra Tree algorithm. ...
Conference Paper
The worst-case discharge during a blowout is a major concern for the oil and gas industry. Various two-phase flow patterns are established in the wellbore during a blowout incident. One of the challenges for field engineers is accurately predicting the flow pattern and, subsequently, the pressure drop along the wellbore to successfully control the well. Existing machine learning models rely on instantaneous pressure drop and liquid hold-up measurements that are not readily available in the field. This study aims to develop a novel machine-learning model to predict two-phase flow patterns in the wellbore for a wide range of inclination angles (0 − 90 degrees) and superficial gas velocities. The model also helps identify the most crucial wellbore parameter that affects the flow pattern of a two-phase flow. This study collected nearly 5000 data points with various flow pattern observations as a data bank for model formulation. The input data includes pipe diameter, gas velocity, liquid velocity, inclination angle, liquid viscosity and density, and visualized/observed flow patterns. As a first step, the observed flow patterns from different sources are displayed in well-established flow regime maps for vertical and horizontal pipes. The data set was graphically plotted in the form of a scatter matrix, followed by statistical analysis to eliminate outliers. A number of machine learning algorithms are considered to develop an accurate model. These include Support Vector Machine (SVM), Multi-layer Perceptron (MLP), Gradient Boosting algorithm, CatBoost, and Extra Tree algorithm, and the Random Forest algorithm. The predictive abilities of the models are cross compared. Because of their unique features, such as variable-importance plots, the CatBoost, Extra Tree, and Random Forest algorithms are selected and implemented in the model to determine the most crucial wellbore parameters affecting the two-phase flow pattern. The Variable-importance plot feature makes CatBoost, Extra Tree, and Random Forest the best option for investigating two-phase flow characteristics using machine learning techniques. The result showed that the CatBoost model predictions demonstrate 98% accuracy compared to measurements. Furthermore, its forecast suggests that in-situ superficial gas velocity is the most influential variable affecting flow pattern, followed by superficial liquid velocity, inclination angle, pipe diameter, and liquid viscosity. These findings could not be possible with the commonly used empirical correlations. For instance, according to previous phenomenological models, the impact of the inclination angle on the flow pattern variation is negligible at high in-situ superficial gas velocities, which contradicts the current observation. The new model requires readily available field operating parameters to predict flow patterns in the wellbore accurately. A precise forecast of flow patterns leads to accurate pressure loss calculations and worst-case discharge predictions.
Article
The article is available free of charge until September 10, 2024 at: https://authors.elsevier.com/a/1jTLR_5-2g2d7v Abstract: Two-phase liquid–gas flows are common in industries such as mining, energy, chemicals, and oil. The gamma-ray absorption technique is a non-contact method widely used to measure parameters for such flows. By analyzing signals from scintillation detectors, flow parameters can be determined and flow structures identified. This study evaluated four types of water–air flow regimes using selected computational intelligence methods. The experiments involved a water–air flow in a horizontal pipe with a 30 mm internal diameter, using two sealed Am-241 gamma ray sources and two scintillation probes type NaI(Tl). Eight features for fluid flow were extracted from the power spectral density and the cross-spectral density of the obtained measurement signals and then used as input for the classifier. Six computational intelligence methods, including k-means, a single decision tree, a support vector machine, a probabilistic neural network, a multilayer perceptron, and a radial basis function, were applied to identify the flow regime. The results showed that all of the methods provided good results of classification for the analyzed types of water–air flow.
Article
Full-text available
Knowledge of the liquid–gas flow regime is important for the proper control of many industrial processes (e.g., in the mining, nuclear, petrochemical, and environmental industries). The latest publications in this field concern the use of computational intelligence methods for flow structure recognition, which include, for example, expert systems and artificial neural networks. Generally, machine learning methods exploit various characteristics of sensors signals in the value, time, frequency, and time–frequency domain. In this work, the convolutional neural network (CNN) VGG-16 is applied for analysis of histogram images of signals obtained for water–air flow by using gamma-ray absorption. The experiments were carried out on the laboratory hydraulic installation fitted with a radiometric measurement system. The essential part of the hydraulic installation is a horizontal pipeline made of metalplex, 4.5 m long, with an internal diameter of 30 mm. The radiometric measurement set used in the investigation consists of a linear Am-241 radiation source with an energy of 59.5 keV and a scintillation detector with a NaI(Tl) crystal. In this work, four types of water–air flow regimes (plug, slug, bubble, and transitional plug–bubble) were studied. MATLAB 2022a software was used to analyze the measurement signal obtained from the detector. It was found that the CNN network correctly recognizes the flow regime in more than 90% of the cases.