Pressure formulation and adaptive control of numerical algorithms for transient flow in pipe networks
Kriel, Albertus Johannes
MetadataShow full item record
Fluid flow network simulation codes are commonly used as a design and analysis tool for many engineering problems such as gas distribution networks, power plants and heat pumps. Two formulations of conservation of momentum have been widely applied in fluid flow network simulation models namely those based on static pressure and those based on total pressure. The total pressure formulations are convenient in that they eliminate the difficulties associated with the calculation of the convective terms and components such as pipe junctions are treated in a straightforward manner based on total pressure losses. However, the different formulations of total pressure for compressible and incompressible flow require different formulations of the momentum conservation equation, which is inconvenient for implementation in a generic network simulation code. In this thesis a united total pressure formulation is first derived which is valid for all fluids and therefore eliminates the inconvenience of switching between the compressible and incompressible formulations. A non-iterative method for the solution of the non-isothermal discretised equations based on the total pressure formulation is then introduced and consistency is illustrated. The method appears to be very stable for subsonic flows, while rapid steady state convergence is observed. A systematic comparison is also done with traditional static pressure based methods and the similarities and differences between the two formulations are illuminated. The different time scales involved in the simulation of transient flow in fluid networks are problematic when conventional fixed time step methods are used for time-wise integration. The time scales associated with acoustic and kinematic wave phenomena as well as storage effects can differ by orders in magnitude. This thesis also presents a simple adaptive time step algorithm which can be readily used in conjunction with all the commonly used first order methods for fluid flow networks. Two test problems are selected to demonstrate the efficiency and savings obtained with this procedure. The adaptive time step algorithm correctly selects appropriate time steps for all phenomena and significant computational savings are observed for accurate integration. In addition, a procedure is implemented which automatically selects the appropriate integration method. The resulting algorithm is a fully adaptive algorithm which switches between a fully implicit method and a semi-implicit method. Two test problems are once again used to demonstrate the efficiency and savings. The fully adaptive algorithm correctly selects appropriate methods for all phenomena and significant additional computational savings are observed.
- Engineering