Nucleation and growth of bubbles in porous media are important problems encountered in processes such as pressure depletion and boiling. Experiments and numerical simulations are studied in micromodel geometries to understand their basic aspects. Experiments of bubble growth by pressure depletion are carried out in 2-0 etched-glass micromodels and in Hele-Shaw cells. Nucleation of bubbles and subsequent growth of gas clusters are visualized. Contrary to the bulk or Hele-Shaw cells, gas clusters in the micromodel have irregular and ramified shapes and share many of the features of an external invasion process (such as percolation during drainage). A pore network numerical model developed simulates the growth of multiple gas clusters under various conditions. It is based on the solution of the convection-diffusion equation and accounts for capillary and viscous forces, which play an important role in determining growth patterns. Numerical simulation agrees well with the experimental results.
The nonsynchronous motion of a rigid rotor-squeeze film damper system was investigated. This research classifies the phenomenon of the system with perfectly preloaded centering springs, reveals its existence conditions, and analyzes the system’s topological structures associated with the phenomenon. When the centering springs of the system do not perfectly balance the gravity load, a unidirectional gravitational residual is resulted. This research also studied the effect of the gravitational residual on the system’s behavior and the degree of stability for the various topological structures of the corresponding autonomous system. It shows that the residual improves the performance of the system because it either suppresses the nonsynchronous orbit of the system or changes it into various subharmonic orbits. This research shows that though the squeeze film damper system is inherently stable, it can have some undesirable nonsynchronous behavior for a wide range of system parameters.
A modified approach for pattern recognition from derivative plots by using neural network methodology is proposed. Instead of training only one single comprehensive neural net to recognize all possible patterns, multiple neural nets are employed in which each neural net is trained to recognize patterns for a specific conceptual reservoir model. This improves the full representation for each model. At the same time the proposed approach assists in recognition of various models with similar responses. A hybrid approach is suggested where the output from the proposed neural network can be cross correlated with models suggested from other data sources. Introduction Pressure transient test data, prior to any quantitative interpretation, must be examined on a qualitative basis. The purpose of qualitative examination is the selection of the most related conceptual reservoir model. Conceptual models from other data bases (geophysical and geological surveys and production test) should also be consulted for proper model selection. Prior to the introduction of derivative plots, the first generation ISDP's (Industry Standard Diagnostic Plots) consisted of linear, semilog and log-log plots. Ershaghi and Woodbury discussed the shortcomings of the first generation plots. The second generation diagnostic plots based on the derivative of pressure vs. time has been more diagnostic than the first generation ones. While considerable efforts in years past have focused on the formulation of various conceptual models and the prediction of the shapes of corresponding diagnostic plots, no major comparative work has been published to cross correlate various models with similar responses. Because of multitude of models and the sensitivity of responses to parametric control, a computer based technique was deemed necessary to perform the categorization. The rule based systems such as the work presented by Stewart and Du and Allain and Horne were found cumbersome for examination of tens of thousands of expected patterns tested in this study. We focused on the application of neural networks. Neural networks have provided the means for a new brand of creativity in problem solution. The use of neural networks to emulate intelligence is on the premise that much learning can be characterized as statistical. The purpose of this study was to examine the similarity of responses among non-similar conceptual models using neural networks, to examine the robustness of a single comprehensive neural network and to propose improved structures for application of neural networks to well test modeling. REVIEW OF NEURAL NETWORK APPLICATION The application of neural networks to identify well test interpretation model was introduced by Al-Kaabi and Lee and some of the shortcomings of this method was discussed by Juniardi and Ershaghi. Detail description of the neural network approach can be found in these two references. A brief summary is given here to pinpoint some of the related issues. Back-propagation neural networks are used for pattern identification in well test analysis. The structure of such a neural network consists of nodes arranged in three layers: an input layer i, a hidden layer j and an output layer k. Nodes in the hidden and output layers are connected by links to the previous layer. The links provide the channels for propagation of input from layer i to j and from j to k. A weight factor is generated for each link.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.