Advanced

1. Presentation

Advanced Preferences Screen
1	The Preferences dialog box is opened from the supervisor.
2	Fifth tab is Advanced Preferences.
3	Logger level: INFO, WARN, DEBUG Logger level represents the level of information given by log files provided by FluxMotor®.
4	Setting memory allocation: Yes / No. In case of answer “Yes” the maximum memory allocation for applications can be defined by user.
5	Maximum memory allocation for applications. It is recommended to keep default values as far as possible. In case of a trouble consult our support team. These fields are in write mode in advanced FluxMotor® user level.
6	Buttons to enable the Flux solver distribution mode. By default, this option is set to “No”. And to define the maximum number of cores used by the distribution (See note below)
7	Maximum allocated memory for secondary Flux2D and FluxSkew software when the Flux solver distribution mode is selected. Note: Ony static allocated memory can be used. The dynamic memory allocation is not available.
8	Button to close the Preferences dialog box without applying the changes.
9	Button to apply changes and close the Preferences dialog box.

Enable Flux distribution mode

Introduction

This section gives information about the establishment of a parametric distribution of the FluxMotor solver, “Flux”, on a single machine. Distributed computing allows the user to save computation time. For example, a test “Scalar Maps” may be automatically distributed if the Flux distribution mode is enabled. In this case, several Flux projects will run at the same time to solve all the required test configurations.

The main parameter of a distributed computation is the “Maximum number of cores used in distribution mode” (i.e., the number of running Flux in parallel).

Principles

The “Flux distribution mode” allows the user to save computation time by parallelizing the finite element (FE) solving of a test.

Indeed, FE solving is done by parallelizing several independent configurations of a FE problem (such as the value of the stator current) instead of running them sequentially.

A primary Flux project is launched and controls all the other secondaries projects (distribution). The primary project oversees the gathering of all the results obtained during the solving process by all the sub-projects Flux as shown in the figure below:

Flux distribution mode - Flux process operating principle

When the Flux distribution mode is activated, FluxMotor automatically uses it and manages the number of cores used to reduce, as far as possible, the computation time.

Depending on the number of computation steps, FluxMotor will adapt the number of cores to have at least 10 steps to solve per core.

In other words, if you select a maximum number of 5 usable cores and solve a test with 25 computation steps ("Characterization - Model - Maps" test with an Id, Iq grid of 5x5), FluxMotor will only distribute the resolution over 2 cores, which is the best configuration according to the FE project to solve.

An example of the process is shown in the following image for the test “Characterization - Model - Maps” of SMPM IR 3PH:

Flux distribution mode - FluxMotor process operating principle SMPM IR (OR) 3PH - “Characterization - Model - Maps”

The following graph shows an example of the solving time versus the number of cores used for a Flux project with 2525 computation steps to solve:

Computation time evolution in function of the number – FE Flux project example with 2525 steps to solve

Note: Note in FluxMotor:

• An FE Flux project will be considered of interest for distribution if it contains more than 20 computation steps to solve.
• From the user's point of view, an FE Flux project’s computation time includes the creation of the FE project, which can take from 45 seconds to several minutes, depending on the case.
• In most cases, FluxMotor tests involve the sequential solving of several FE Flux projects, and some of these FE Flux projects have no interest in being distributed (less than 20 computation steps).

According to the previous points, the trend presented in the graph above is therefore not exact in the context of a FluxMotor tests, but it gives a general trend, especially for some tests like “Characterization - Model - Maps” (SM PM and SM RSM) or “Characterization - Model - Scalar Maps” (IM SQ).

How to set up a parametric distribution on a single machine

In FluxMotor, the Flux distribution mode (number of secondary Flux in parallel) must be set in the Preferences in the section Advanced. It’s just needed to set “Flux distribution mode” to “Yes” and to select the “Maximum number of cores” you want to use.

Warning:

When using the distribution mode, the primary Flux can be used either with dynamic or static memory allocation. However, it is highly recommended to use the dynamic memory allocation.

When using the distribution mode, only the static memory can be used for the secondary Flux. The dynamic memory allocation is not available.

Flux distribution mode
1	Button to enable the Flux solver distribution mode. By default, this option is set to “No”.
2	Button to select the maximum number of cores used by the distribution
3	Button to set the maximum allocated memory for the primary Flux (Dynamic allocated memory is recommended)
4	Button to set the maximum allocated memory for the secondary Flux (Only static allocated memory can be selected)

Warning - Beta version mode

Sometime, the non-entirely qualified feature like tests for instance, are nevertheless available for testing. In this case, a "Beta version" banner warns the user to be careful when analyzing the results.

For example, in the current version, evaluation of NVH for induction machines with squirrel cage is concerned by this mode.