A comparative study of a traditional localization algorithm and a deep learning model for radioactive particle tracking application.

Radioactive particle tracking is a nuclear technique that tracks a sealed radioactive particle inside a volume through a mathematical location algorithm, which is widely applied in many fields such as chemical and civil engineering in hydrodynamics flows. It is possible to reconstruct the trajectory of the radioactive particle using a traditional mathematical algorithm or artificial intelligence methods. In this paper, the traditional algorithm is based on solving a minimization problem between the simulated events and a calibration dataset, and it was written using C++ language. The artificial intelligence method is represented by a deep neural network, in which hyperparameters were defined using a Python optimization library called Optuna. This paper aims to compare the potentiality of both methods to evaluate the accuracy of the radioactive particle tracking technique. This study proposes a simplified model of a concrete mixer, six NaI(Tl) detectors, and a137Cs sealed radioactive particle. The simulated measurement geometry and the dataset (3615 patterns) were developed using the MCNPX code, which is a mathematical code based on the Monte Carlo Method. The results show a mean absolute percentage error (MAPE) of 20.81%, 10.33%, and 16.84% for x, y and z coordinates, respectively, for the traditional algorithm. For the deep neural network, MAPE is 6.87%, 2.70%, and 22.79% respectively for x, y and z coordinates. In addition, an investigation is carried out to analyze whether the size of the calibration dataset influences the performance of both methods.

Prediction of fluids volume fraction and barium sulfate scale in a multiphase system using gamma radiation and deep neural network.

In the oil industry, during the production of oil and gas, barium sulfate (BaSO4) scale may occur on the inner walls of the pipelines leading to the reduction of the internal diameter, making the fluids' passage difficult and complicating the calculation of the fluids volume fraction. This paper presents a methodology to predict volume fraction of fluids and BaSO4 scale thickness from obtaining spectra of two NaI(Tl) detectors that record the transmitted and scattered beams of gamma-rays. Theoretical models for a multiphase annular flow regime (gas-saltwater-oil-scale) were developed using MCNP6 code, which is a mathematical code based on the Monte Carlo method. The simulated data was used to train a deep neural network (DNN) to predict the volume fraction of gas, saltwater and oil, and the concentric scale thickness. A Python optimization library called Optuna was used for the hyperparameters search to design the DNN architecture. The methodology presented great results, especially for scale thickness prediction. Although the results for saltwater did not reach the same level, it was still possible to predict approximately 70% of the patterns up to 10% relative error. This achievement indicates the possibility to calculate the volume fraction of fluids and the concentric scale thickness in the offshore oil industry using gamma densitometry and deep learning models.

Application of deep neural network and gamma-ray scattering in eccentric scale calculation regardless of the fluids volume fraction inside a pipeline.

Scale formation is one of the major problems in the oil industry as it can accumulate on the surface of the pipelines, which could even fully block the fluids' passage. It was developed a methodology to detect and quantify the maximum thickness of eccentric scale inside pipelines using nuclear techniques and an artificial neural network. The measurement procedure is based on gamma-ray scattering using NaI(Tl) detectors and a137Cs radiation source that emits gamma-rays of 662 keV. The simulations considered an annular flow regime composed of barium sulfate scale, oil, saltwater and gas, and three percentages of these fluids were used. In the present investigation, a study of detectors configuration was carried out to improve the measurement geometry and the simulations were made using the MCNP6 code, which is a mathematical code based on the Monte Carlo method. The counts registered in the detectors were used as input data to train a deep neural network (DNN) that uses rectifier activation functions instead of the usually sigmoid-based ones. In addition, a hyperparameters search was made using open software to develop the final DNN architecture. Results showed that the best detector configuration was able to predict 100% of the patterns with the maximum relative error of 5%. Moreover, the achieved mean absolute percentage error was 0.42% and the regression coefficient was 0.99996 for all data. The results are promising and encourage the use of DNN to calculate inorganic scale regardless of the fluids volume fraction inside pipelines.

Application of radioactive particle tracking and an artificial neural network to calculating the flow rate in a two-phase (oil-water) stratified flow regime.

A multiphase flow is defined as the transport of two or more fluids with different properties flowing together inside a pipeline. After offshore oil production, it is necessary to control the amount of transported fluids based on flow rate measurements. Therefore, in this study, we developed a simulation method for predicting the volume fraction and calculating the superficial velocity for a two-phase flow based on radioactive particle tracking, which involves using a sealed radiation source inside the pipeline in order to obtain volume fraction measurements. The test section for the multiphase flow comprised oil and saltwater under a stratified flow regime, with a polyvinyl chloride pipe, four NaI(Tl) detectors, and a137Cs radioactive particle that emitted gamma-rays at 662 keV. Simulations were conducted using the MCNP6 code, which is a mathematical code based on the Monte Carlo method. Volume fraction predictions were obtained using a multilayer perceptron neural network with a backpropagation algorithm. The novel feature of this method is the combination of radioactive particle tracking with an artificial neural network in order to predict volume fractions in multiphase flows. The results showed that 91.65% of the predicted patterns were within 5% of the relative error. In addition, the time delay was determined using the cross-correlation function to obtain the superficial velocity in three different volume fractions, which allowed each phase flow rate to be calculated in these cases.

Determination of eccentric deposition thickness on offshore horizontal pipes by gamma-ray densitometry and artificial intelligence technique.

The extraction of oil is accompanied by water and sediments that, mixed with the oil, cause the formation of scale depositions in the pipelines walls promoting the reduction of the inner diameter of the pipes, making it difficult for the fluids to pass through interest. In this sense, there is a need to control the formation of these depositions to evaluate preventive and corrective measures regarding the waste management of these materials, as well as the optimization of oil extraction and transport processes. Noninvasive techniques such as gamma transmission and scattering can support the determination of the thickness of these deposits in pipes. This paper presents a novel methodology for prediction of scale with eccentric deposition in pipes used in the offshore oil industry and its approach is based on the principles of gamma densitometry and deep artificial neural networks (DNNs). To determine deposition thicknesses, a detection system has been developed that utilizes a 1 mm narrow beam geometry of collimation aperture comprising a source of 137Cs and three properly positioned 2″×2″ NaI(Tl) detectors around the system, pipe-scale-fluid. Crude oil was considered in the study, as well as eccentric deposits formed by barium sulfate, BaSO4. The theoretical models adopted a static flow regime and were developed using the MCNPX mathematical code and, secondly, used for the training and testing of the developed DNN model, a 7-layers deep rectifier neural network (DRNN). In addition, the hyperparameters of the DRNN were defined using a Baysian optimization method and its performance was validated via 10 experiments based on the K-Fold cross-validation technique. Following the proposed methodology, the DRNN was able to achieve, for the test sets (untrained samples), an average mean absolute error of 0.01734, mean absolute relative error of 0.29803% and R2 Score of 0.9998813 for the scale thickness prediction and an average accuracy of 100% for the scale position prediction. Therefore, the results show that the 7-layers DRNN presents good generalization capacity and is able to predict scale thickness with great precision, regardless of its position inside the tube.

Prediction of volume fractions in three-phase flows using nuclear technique and artificial neural network.

This work presents methodology based on nuclear technique and artificial neural network for volume fraction predictions in annular, stratified and homogeneous oil-water-gas regimes. Using principles of gamma-ray absorption and scattering together with an appropriate geometry, comprised of three detectors and a dual-energy gamma-ray source, it was possible to obtain data, which could be adequately correlated to the volume fractions of each phase by means of neural network. The MCNP-X code was used in order to provide the training data for the network.