Salgado William Luna, Dam Roos Sophia de Freitas, Desterro Filipe Santana Moreira do, Cruz Bianca Lamarca da, Silva Ademir Xavier da, Salgado César Marques
Divisão de Radiofármacos (DIRA), Instituto de Engenharia Nuclear - (IEN), Rua Hélio de Almeida 75, 21941-906, Cidade Universitária, RJ, Brazil.
Divisão de Radiofármacos (DIRA), Instituto de Engenharia Nuclear - (IEN), Rua Hélio de Almeida 75, 21941-906, Cidade Universitária, RJ, Brazil; Programa de Engenharia Nuclear - (PEN/COPPE), Universidade Federal do Rio de Janeiro - (UFRJ), Avenida Horácio de Macedo 2030, G - 206, 21941-914, Cidade Universitária, RJ, Brazil.
Appl Radiat Isot. 2023 Oct;200:110973. doi: 10.1016/j.apradiso.2023.110973. Epub 2023 Aug 9.
To continuously monitor information about the transport of fluids by sequential batches in polyduct, found in the petrochemical industry, it is necessary to manage the mixing zone - transmix - that occurs when two fluids are being transported. This scenario demonstrates the need to estimate the interface region and the purity of the fluids in this region to improve the management of the pipeline and, thus, reduce associated costs. This study presents a measurement system based on the dual-modality gamma densitometry technique in combination with a deep neural network with seven hidden layers to predict the purity level of four different fluids (Gasoline, Glycerol, Kerosene and Oil Fuel) in the transmix. The detection geometry is composed of aCs radioactive source (emitting gamma rays of 661.657 keV) and two NaI(Tl) scintillator detectors to record the transmitted and scattered photons. The study was performed by computer simulations using the MCNP6 code, and the information recorded in the detectors was used as input data for training and evaluating the deep neural network. The proposed intelligent measurement system is able to predict the purity level of fluids with errors with mean squared error values below 1.4 and mean absolute percentage error values below 5.73% for all analyzed data.
为了持续监测石油化工行业中多管道内连续批次流体输送的相关信息，有必要对两种流体输送时出现的混合区域——过渡混合区进行管理。这种情况表明，需要估计界面区域以及该区域内流体的纯度，以改善管道管理，从而降低相关成本。本研究提出了一种基于双模态伽马密度测量技术的测量系统，该系统结合了一个具有七个隐藏层的深度神经网络，用于预测过渡混合区中四种不同流体（汽油、甘油、煤油和燃油）的纯度水平。检测几何结构由一个Cs放射源（发射能量为661.657 keV的伽马射线）和两个NaI(Tl)闪烁探测器组成，用于记录透射光子和散射光子。该研究通过使用MCNP6代码进行计算机模拟完成，探测器记录的信息被用作训练和评估深度神经网络的输入数据。对于所有分析数据，所提出的智能测量系统能够预测流体的纯度水平，其误差的均方误差值低于1.4，平均绝对百分比误差值低于5.73%。
