Search

Slug liquid holdup (SLH) is a critical requirement for accurate pressure drop prediction during multiphase pipe flows and by extension optimal gas lift design and production optimization in wellbores. Existing empirical correlations provide inaccurate predictions because they were developed with regression analysis and data measured for limited ranges of flow conditions. Existing SLH machine learning models provide accurate predictions but are published without any equations making their use by other researchers difficult. The only existing ML model published with actual equations cannot be considered optimum because it was selected by considering artificial neural network (ANN) structures with only one hidden layer. In this study, an ANN-based model for SLH prediction with actual equations is presented. A total of 2699 data points randomly divided into 70%, 15%, and 15% for training, validation, and testing was used in constructing 71 different network structures with 1, 2, and 3 hidden layers respectively. Sensitivity analysis revealed that the optimum network structure has 3 hidden layers with 20, 5, and 15 neurons in the first, second, and third hidden layers, respectively. The optimum network structure was translated into actual equations with the aid of the weights, biases, and activation functions. Trend analysis revealed that this study's model reproduced the expected effects of inputs on SLH. Test against measured data revealed that this study's model is in agreement with measured data with coefficient of determinations of 0.9791, 0.9727, 0.9756, and 0.9776 for training, testing, validation, and entire datasets, respectively. Comparative study revealed that this study's model outperformed existing models with a relative performance factor of 0.002. The present model is presented with visible mathematical equations making its implementation by any user easy and without the need for any ML framework. Unlike existing ANN-based models developed with one hidden layered ANN structures, the present model was developed by considering ANN structures with one, two, and three hidden layers, respectively. [ABSTRACT FROM AUTHOR]

Surrogate models provide a powerful method for simplifying calculations within complex simulations. While surrogate models are broadly applied within chemical engineering, little research exists investigating the level of surrogacy's impact on a simplified process model. In this work, artificial neural networks (ANN) and Kriging models are used as surrogate models at the process, process unit, and thermodynamic levels for a CO2 amine scrubbing process. The surrogated models are evaluated against an Aspen Plus simulation for accuracy, convergence behavior, computational cost, and ability to extrapolate. The thermodynamic and process unit models can better handle discontinuous, non‐smooth behavior, and convergence issues in the surrogated truth model, but poor conditioning in the final system of equations results in a lower accuracy and convergence rate than the process level surrogate. Beyond model accuracy, availability of diverse data, intended re‐usability, and the desired outputs must be considered when selecting a level of abstraction. [ABSTRACT FROM AUTHOR]

A comparative analysis of the application of the standard, hybrid serial gray-box, and hybrid parallel gray-box artificial neural networks was carried out to predict Cr(VI) adsorption by activated carbon. The dataset was developed through batch kinetic experiments with varying operational parameters. The major reaction transport mechanism was found to be intraparticle diffusion while the major attachment mechanism was chemical bonding. Adsorption kinetics was well represented by the diffusion-chemisorption (DC) model while desorption kinetics followed the pseudo-second order model. It was discovered that the overall DC kinetic rate, KDC was inversely proportional to both the effective diffusion coefficient as well as the diffusivity and proves useful to indirectly assess diffusional effects created by changing operational parameters. The DC kinetic model was subsequently instituted into the hybrid models. The development of the hybrid neural networks was built-in with the joint effect of operating parameters of time and particle size for the standard (SANN) and parallel neural network (PANN), together with the kinetic parameters of the DC model such as overall rate, initial rate and sorption capacity for the serial neural network (DC-ANN). The comparative performance of the networks was subsequently evaluated using error functions and the Bland-Altman plot. The hybrid neural networks produced the best prediction to the target data. Sensitivity analysis revealed a substantial positive influence on the DC-ANN model due to the inclusion of the kinetic equation parameters. Consequently, the DC-ANN model required significantly less computational effort. [ABSTRACT FROM AUTHOR]

A problem related to the development of algorithms designed to find the structure of artificial neural network used for behavioural (black-box) modelling of selected dynamic processes has been addressed in this paper. The research has included four original proposals of algorithms dedicated to neural network architecture search. Algorithms have been based on well-known optimisation techniques such as evolutionary algorithms and gradient descent methods. In the presented research an artificial neural network of recurrent type has been used, whose architecture has been selected in an optimised way based on the above-mentioned algorithms. The optimality has been understood as achieving a trade-off between the size of the neural network and its accuracy in capturing the response of the mathematical model under which it has been learnt. During the optimisation, original specialised evolutionary operators have been proposed. The research involved an extended validation study based on data generated from a mathematical model of the fast processes occurring in a pressurised water nuclear reactor. • Neural black-box modelling of selected dynamic processes. • Modelling of fast processes in pressurised water nuclear reactor. • Development of algorithms designed to find structure of artificial neural networks. • Optimal neural network architecture search. • Design of original specialised evolutionary operators for optimisation purposes. [ABSTRACT FROM AUTHOR]

In this paper, a modeling technique using spline functions with finite time difference approximation is discussed for modeling moderately nonlinear digital input/output (1/0) drivers. This method takes into account both the static and the dynamic memory characteristics of the driver during modeling. Spline function with finite time difference approximation includes the previous time instances of the driver output voltage/current to capture the output dynamic characteristics of digital drivers accurately. In this paper, the speed and the accuracy of the proposed method is analyzed and compared with the radial basis function (RBF) modeling technique, for modeling different test cases. For power supply noise analysis, the proposed method has been extended to multiple ports by taking the previous time instances of the power supply voltage/current into account. The method discussed can be used to capture sensitive effects like simultaneous switching noise (SSN) and cross talk accurately when multiple drivers are switching simultaneously. A comparison study between the presented method and the transistor level driver models indicate a computational speed-up in the range of 10-40 with an error of less than 5%. For highly nonlinear drivers, a method based on recurrent artificial neural networks (RNN) is discussed. [ABSTRACT FROM AUTHOR]

In this work, the signal and noise behaviors of a microwave transistor within its operation domain (CT, V,DS, IDS, f) are modeled by the Artificial Neural Network (ANN) and Fuzzy Logic System (FLS) without using any information on the microwave circuit theory A worked example is presented where the same data is employed for both models selected from the manufacturer's data sheets. Performances of the FLS and ANN models are compared and conclusions are drawn. [ABSTRACT FROM AUTHOR]

[Display omitted] • The stability of dry AD of FW is highly dependent on the organic loading rate. • SRT and water content are critical factors affecting methane production. • The biogas production rate prediction model for dry anaerobic digestion is developed. • The prediction model is developed via recurrent neural network black-box modeling. • Recent three days data of HRT, SCOD, TVFA, TA, and FA are used for the prediction. The stability of dry anaerobic digestion (AD) of food waste (FW) as well as the resulting methane gas generation was investigated from the perspective of system dynamics. Various organic loading rates were applied to the system by modifying the water content in the FW feed and solid retention time (SRT). The excessive organic loading due to the accumulation of volatile fatty acids (VFAs) from the feed with 80% water content during the short SRT (15 and 20 d) caused system failure. In contrast, more intermediate materials, such as VFAs, was easily converted into methane at higher water contents. In addition, the biogas production rate of dry AD was effectively predicted based on SRT, soluble chemical oxygen demand, total VFA, total ammonia, and free ammonia using a recurrent neural network—the so-called "black-box" model. This implies the feasibility of applying this data-based black-box model for controlling and optimizing complex biological processes. [ABSTRACT FROM AUTHOR]