Drainage flow of a viscous compressible gas from a semi-sealed narrow conduit is a pore-scale model for studying the fundamental flow physics of fluid recovery from a porous reservoir without using fluid injection. Thermal effect has been routinely neglected for these flows in the traditional petroleum engineering literature. Since the motion is entirely driven by volumetric expansion, temperature change always accompanies the density change. This thesis examines such thermal effects on the drainage flow.

Thermal drainage flow is first studied by simultaneously solving the linearized continuity, momentum and energy equations for adiabatic walls. It is shown that even in the absence of an imposed temperature drop, gas expansion induces a transient temperature decrease inside the channel, which slows down the drainage process compared to the isothermal model and Lighthill’s model. For a given density drop, gas drains out faster as the initial-to-final temperature ratio increases; and the transient density can undershoot the final equilibrium value. A parametric study is then carried out to explore the influence of various thermal boundary conditions on drainage flow. It is found that as the wall transitions from adiabatic to isothermal condition, the excess density changes from a plane wave solution to a non-plane wave solution and the drainage rate increases. It is shown that when the exit is also cooled and the wall is non-adiabatic, the total recovered fluid mass exceeds the amount based on the isothermal theory which is determined by the initial and final density difference alone. Finally, a full numerical simulation is conducted to mimic the channel-reservoir system using the finite volume method. The Ghost-Cell Navier-Stokes Characteristic Boundary Condition technique is applied at the far end of the truncated reservoir, which is an open boundary. The results confirm the conclusions of the linear theory.

This study investigates the relation between credit supply competition among banks and their clients’ conditional accounting conservatism (i.e., asymmetric timely loss recognition). The Interstate Banking and Branching Efficiency Act (IBBEA) of 1994 permits banks and bank holding companies to expand their business across state lines, introducing a positive shock to credit supply competition in the banking industry. The increase in credit supply competition weakens banks’ bargaining power in the negotiation process, which in turn may weaken their ability to demand conservative financial reporting from borrowers. Consistent with this prediction, results show that firms report less conservatively after the IBBEA is passed in their headquartered states. The effect of the IBBEA on conditional conservatism is particularly stronger for firms in states with a greater increase in competition among banks, firms whose operations are more concentrated in their headquarter states, firms with greater financial constraints, and firms subject to less external monitoring. Robustness tests confirm that the observed decline in conditional conservatism is causally related to the passage of IBBEA. Overall, this study highlights the impact of credit supply competition on financial reporting practices.

Templates are wildly used in Web sites development. Finding the template for a given set of Web pages could be very important and useful for many applications like Web page classification and monitoring content and structure changes of Web pages. In this thesis, two novel sequence-based Web page template detection algorithms are presented. Different from tree mapping algorithms which are based on tree edit distance, sequence-based template detection algorithms operate on the Prüfer/Consolidated Prüfer sequences of trees. Since there are one-to-one correspondences between Prüfer/Consolidated Prüfer sequences and trees, sequence-based template detection algorithms identify the template by finding a common subsequence between to Prüfer/Consolidated Prüfer sequences. This subsequence should be a sequential representation of a common subtree of input trees. Experiments on real-world web pages showed that our approaches detect templates effectively and efficiently.