2020-12-29T14:14:58.000-0500

Simcenter Amesim Automation Connect
Simcenter System Simulation Client for Git
Simcenter Amesim
Teamcenter Share

The aim of this FAQ is to explain what are the differences between the standard distributive pipe models and the CFD-1D ones giving some advantages and drawbacks of each approach.

The aim of this FAQ is to list the main differences between the distributive and CFD-1D pipe submodels without going too deep into the equations used behind each submodel.

If you want to have deep explanations on the equations used, please refer to the Simcenter Amesim Hydraulic manual under the lines section.

__Standard distributive pipe models:__

This approach uses ODEs (Ordinary Differential Equations) integrated by the Simcenter Amesim solver over the time. This type of pipe model can compute the pressure and the flow rate at different locations along the pipe. The number of nodes is variable and set by the user.

To be accurate, the user should select the proper number of internal nodes as a function of the excitation bandwidth and the natural frequencies of the network.

__CFD-1D pipe models:__

This approach uses 1D Navier-Stokes equations (continuity and motion equations) and they are solved by a local solver with a Partial Differential Equations solving method (Lax-Wendroff scheme). This numerical scheme is applied locally inside the pipe. This means that the line is no longer solved by the standard ODE/DAE solver of Simcenter Amesim.

This advanced line models were specifically developed to compute wave effects with a high level of accuracy, no matter what the excitation bandwidth is.

__Advantages and drawbacks of using standard distributive pipe models:__

The standard distributive pipe models since they are using the standard Simcenter Amesim solver they are compatible with all analyzing tools available in Simcenter Amesim such as linear analysis studies and stabilizing runs.

The problem with this pipe is that when you have too many nodes, you will be adding extra state variables to your system and the computational time will increase a lot, especially when the f.d.f (frequency depending friction) option is activated.

Another drawback of these ODE methods is that the frequency content is truncated in such a way that the highest eigenvalues of the line model are often excited in a non-physical manner. Usually, distributive lines should rather be used when the pipe natural frequencies are higher than the excitation bandwidth.

__Advantages and drawbacks of using CFD-1D pipe models:__

The CFD-1D pipe models have their own time-step, which is almost a fixed step one. These pipe models are solved by a different solver than the rest of the model and there is some kind of internal co-simulation between the ODE/DAE standard Simcenter Amesim solver and the CFD-1D one. This implies that:

Moreover, the frequency dependent friction option that introduces a lot of states in the standard distributive approach is already implemented in the CFD-1D lines for no overhead in CPU time. And since this pipe models are thought to be used to study dynamic phenomena inside the pipes (e.g. water hammer effect) there is a special plot feature only available in this type of pipe models called 1D-Plot, which allows you to animate the variable along the pipe and the simulation time (see related article for more information).

So, if you want to model a log pipe with many nodes and you want to study the dynamic phenomena inside the pipe without need to perform a linear analysis study, then this pipe model will increase your simulation time and will give you access to an additional 1D-Plot for analyzing your results.

If you want to have deep explanations on the equations used, please refer to the Simcenter Amesim Hydraulic manual under the lines section.

This approach uses ODEs (Ordinary Differential Equations) integrated by the Simcenter Amesim solver over the time. This type of pipe model can compute the pressure and the flow rate at different locations along the pipe. The number of nodes is variable and set by the user.

To be accurate, the user should select the proper number of internal nodes as a function of the excitation bandwidth and the natural frequencies of the network.

This approach uses 1D Navier-Stokes equations (continuity and motion equations) and they are solved by a local solver with a Partial Differential Equations solving method (Lax-Wendroff scheme). This numerical scheme is applied locally inside the pipe. This means that the line is no longer solved by the standard ODE/DAE solver of Simcenter Amesim.

This advanced line models were specifically developed to compute wave effects with a high level of accuracy, no matter what the excitation bandwidth is.

The standard distributive pipe models since they are using the standard Simcenter Amesim solver they are compatible with all analyzing tools available in Simcenter Amesim such as linear analysis studies and stabilizing runs.

The problem with this pipe is that when you have too many nodes, you will be adding extra state variables to your system and the computational time will increase a lot, especially when the f.d.f (frequency depending friction) option is activated.

Another drawback of these ODE methods is that the frequency content is truncated in such a way that the highest eigenvalues of the line model are often excited in a non-physical manner. Usually, distributive lines should rather be used when the pipe natural frequencies are higher than the excitation bandwidth.

The CFD-1D pipe models have their own time-step, which is almost a fixed step one. These pipe models are solved by a different solver than the rest of the model and there is some kind of internal co-simulation between the ODE/DAE standard Simcenter Amesim solver and the CFD-1D one. This implies that:

- Since the CFD-1D lines are solved by a different solver, they do not introduce extra state variables, which means that adding a lot of nodes will not increase too much the CPU time.
- Also because of the different solvers there are some discontinuities in this internal co-simulation and in some cases, it can introduce some slowdown in the simulation, specially when doing a steady-state studies.
- CFD-1D lines cannot be linearized and stabilizing runs cannot be performed. The reason here is also due the different solvers, the linear analysis features are performed by the standard ODE/DAE Simcenter Amesim solver and in this case since we use a different solvers, the linearization process ignores the CFD-1D pipe models.

Moreover, the frequency dependent friction option that introduces a lot of states in the standard distributive approach is already implemented in the CFD-1D lines for no overhead in CPU time. And since this pipe models are thought to be used to study dynamic phenomena inside the pipes (e.g. water hammer effect) there is a special plot feature only available in this type of pipe models called 1D-Plot, which allows you to animate the variable along the pipe and the simulation time (see related article for more information).

So, if you want to model a log pipe with many nodes and you want to study the dynamic phenomena inside the pipe without need to perform a linear analysis study, then this pipe model will increase your simulation time and will give you access to an additional 1D-Plot for analyzing your results.