====== Automated Process Optimization ======
====== Automated Process Optimization ======

With Automated Process Optimization, suitable process parameters (e.g. throughput, screw speed, barrel temperature profile) can be determined automaticallyand evaluated against defined target criteria. The optimization is performed using **Particle Swarm Optimization (PSO / swarm intelligence)**.

**Opening the optimization dialog**

To open the optimization wizard, select from the main menu:

**Command path:** ''Main menu > Simulation > Optimization…''

An optimization wizard with multiple steps will start and guide you through the setup and execution of the optimization task.

===== Choose meta machines =====

In this step you define **which machine configurations** (meta machines) shall be considered for the optimization.

{{ :en_sigma151_dlg_opt_001.png?direct&1000 |Process optimization – Choose meta machines (1/8)}}

First, one or more **diameter-independent meta machines** can be selected. A meta machine defines the general screw and barrel layout including feeding positions (e.g. single-stage, two-stage, devolatilization, side feeding, etc.) and serves as a geometry-based framework for the subsequent optimization.

=== Functions ===

In the upper part of the dialog, the **Directory** field is used to select the folder from which meta machines are loaded.

**Filters** are available to narrow down the list of meta machines:

  * **Filter by usage**
  * **Filter by material**

Info: Machines are shown when at least one filter in the category fits.

=== Machine list (left) ===

On the left, the **Machines** list shows all available meta machines. Entries can be selected for optimization.

Buttons:

  * **All** – selects all entries
  * **Clear** – clears the current selection

=== Preview and meta machine information ===

On the right, a graphical preview of the selected meta machine is displayed (screw configuration / barrel representation). Below the preview, key data is shown:

  * **Number of feedings** (number of feeding locations)
  * **Length** (shown as L/D)

In addition, a **Comment** field is displayed below the meta machine preview where further meta machine information can be read.

===== Choose machine configuration and die =====

In this step you define:

  * which **machine configuration / modular system** (*.mbk) is used to generate and apply the diameter-independent meta machine (screw and barrel configuration)
  * which **die** (default definition or user-defined die at the screw tip) is applied

{{ :en_sigma151_dlg_opt_002.png?direct&1000 |Process optimization – Choose configuration (2/8)}}

=== Machine configuration ===

In the upper area, the machine modular system for the screw and barrel configuration to be optimized must be specified. The selected file is displayed as a path to the *.mbk file.

**Available functions:**

  * Load machine configuration / modular system
  * Import configuration from another process
  * **Edit...** – opens the modular system for adjustment (available screw and barrel elements)

A graphical preview of the resulting screw/machine configuration is shown below.

=== Meta machine navigation ===

If multiple meta machines were selected, you can switch between them using:

  * **Previous**
  * **Next**

This allows you to verify whether the selected modular system can represent the chosen meta machines.

=== Die ===

In the lower area, the die is defined.

Options and functions:

  * **Default** – uses the predefined default die
  * **Other** – allows definition / import of a custom die
      - import die from file or another process
      - create a new die
  * **Edit...** – open and edit die data

===== Choose process parameters =====

In this step, the **initial process parameters** for each selected meta machine are defined. These parameters form the baseline from which the optimization starts.

{{ :en_sigma151_dlg_opt_003.png?direct&1000 |Process optimization – Choose process parameters (3/8)}}

The dialog structure is aligned with the standard [[simulation:verfahrensparameter_-_allgemein|process parameter]] workflow.

Key functions:

  * Navigation between meta machines (**Previous / Next**)
  * Display of the active **Meta machine**
  * Button **apply to all** (transfers the current settings to all selected meta machines / parameter sets)

This enables structured definition of the process parameters.

===== Process calculation settings =====

In this step you define **which model calculations** are executed within a single SIGMA simulation and under which terminating conditions iterative sub-calculations are stopped.

{{ :en_sigma151_dlg_opt_004.png?direct&1000 |Process optimization – Calculation settings (4/8)}}

Further information on calculation settings is available [[simulation:berechnung|here]].

===== Define targets =====

In this step, the **optimization targets** are defined. For the given task, process variables, limits, target values, evaluation methods, and weights are specified.

{{ :en_sigma151_dlg_opt_005.png?direct&1000 |Process optimization – Define targets (5/8)}}

The overview includes the following columns for all target parameters:

  * **Parameters (target variables)**
  * **min**
  * **max**
  * **target**
  * **Evaluation method**
  * **Weight**

=== Target variables ===

  * **Throughput**
  * **Residence time**
  * **Temperature**
  * **Specific energy**
  * **Fiber end length**
  * **Pressure at screw tip**

=== Evaluation methods ===

In the **Evaluation method** column, different evaluation logics can be selected for each target variable:

  * **maximizing** (increasing evaluation; the "max" value is the optimization target)
  * **minimizing** (decreasing evaluation; the "min" value is the optimization target)
  * **target** (the value defined under "target" is the optimization target)

The first priority of the optimization is to find processes within the specified limits. Only afterwards does the algorithm attempt to reach the target values.

=== Weighting ===

The **Weight** defines how strongly a parameter contributes to the overall rating.

  * high weight: strong influence on the overall optimization rating
  * low weight: minor influence
  * weight = 0: the parameter is not considered

=== Additional evaluation options ===

  * **Use warnings**
  * **Consider melting profile**

These options allow warning conditions and melting behavior to be included in the evaluation of the results.

===== Optimization settings – PSO strategy =====

In this step, the parameters of **PSO (Particle Swarm Optimization / swarm intelligence)** are configured.

{{ :en_sigma151_dlg_opt_006.png?direct&1000 |Process optimization – Optimization settings (6/8)}}

=== PSO parameters ===

  * **Swarm size:** Number of particles (solution candidates) evaluated per iteration (larger swarm size = broader search in parameter space, higher runtime per iteration)
  * **Acceleration coefficient C1:** Weighting of the particle movement toward its **local best** (best solution found by the particle itself).
  * **Acceleration coefficient C2:** Weighting of the particle movement toward the **global best** (best solution found by the entire swarm).
  * **Inertia weight:** Weighting of the velocity component from the **previous iteration**.

**General swarm effects:**
  * higher inertia weight -> stronger preservation of direction (more exploration)
  * lower inertia weight -> stronger focusing / smoother convergence (more exploitation)

=== Optimization terminating condition ===

You can choose between:

  * **Run under minimum deviation**
  * **Reached maximum number of iterations**
  * **Number of iterations**

=== Recommendation ===

The default parameter values are suitable for most use cases and are recommended. They can be adjusted if required, e.g. for:

  * larger search spaces
  * faster coarse optimization runs
  * more robust convergence for complex target definitions

===== Calculation / Swarm optimization =====

In this step, the actual optimization is executed using the settings defined previously.

{{ :en_sigma151_dlg_opt_007.png?direct&1000 |Process optimization – Calculation / Swarm optimization (7/8)}}

=== Status and log ===

**Status**

The current calculation state is shown in the upper left area:

  * **Calculation in progress**
  * remaining time / progress estimate

**Buttons**
  * **Cancel** – stops the running optimization
  * **Perform a new calculation** – recalculates the swarm after optimization inputs have been modified (availability depends on status)
  * **Copy** – copies the log output to the clipboard

**Log window**

The log area lists the calculation steps, e.g.:

  * Calculation started
  * Initializing swarm
  * Swarm initialized
  * Starting swarm iteration

=== Particle visualization ===

In the search-space diagram, particles (individual process candidates) are visualized.

  * **x-axis:** Screw speed
  * **y-axis:** Throughput

The particle color is interpreted using the color scale shown below the diagram.

=== Color scale / barrel temperature profile ===

Below the diagram, a color scale labeled **Barrel temperature profile** is displayed (e.g. -50 °C to +50 °C). It visualizes a temperature-related parameter or a profile deviation (barrel temperature profile delta).

=== Iteration display ===

Below the plot:

  * **Iteration** (slider / value)

This allows visualization of individual iteration states.

===== Select results / Save process =====

After the optimization is completed, the results (particles / candidate processes) are displayed in a results list.

{{ :en_sigma151_dlg_opt_008.png?direct&1000 |Process optimization – Results / Configuration selection (8/8)}}

=== Particle list ===

Under **Particles**, all calculated candidate processes are listed, typically with their rating in brackets.

Functions:

  * selection of individual particles
  * multi-selection for comparison (checkboxes)
  * sorting is typically by rating (descending / best candidate on top)

=== Processing parameters ===

In the upper right area, the most important result quantities are shown in a table for the selected particles:

  * **Throughput [kg/h]**
  * **Residence time [s]**
  * **Temperature [°C]**
  * **Specific energy [kWh/kg]**
  * **Fiber end length [µm]**
  * **Pressure at screw tip [bar]**
  * **Rating [%]**

Additionally, a **Weight** row is displayed to document the active evaluation logic.

=== Process indicators ===

The middle area provides further configuration-related data per selected candidate:

  * **Configuration**
  * **Meta machine**
  * **Screw speed [1/min]**
  * **Throughput [kg/h]**
  * **Barrel temperature profile (delta) [°C]**

This allows direct comparison between candidates.

**Particles (preview):** The lower area provides an additional visualization/comparison area for the selected particles.

**Create process:** This function generates and saves a SIGMA process from the selected optimization candidate. The wizard can then be closed using **Finish**.

===== Notes =====

=== Choose target definitions carefully ===

  * Do not set limits (min/max) too narrow so that the algorithm has sufficient search space.
  * Use target values only where a true setpoint exists (e.g. temperature, pressure).
  * Set weights intentionally (e.g. high throughput weight for production optimization; higher temperature/residence-time weight for material-sensitive processing).

=== PSO parameters ===

  * Start with the **default values**.
  * A larger **swarm size** increases the chance of finding good solutions but increases runtime.
  * More **iterations** often improve convergence, especially for complex objective functions.

=== Result evaluation ===

  * Do not focus only on the highest rating; also verify process robustness.
  * Compare several particles (e.g. similar rating but different screw speed / barrel temperature profiles).
  * Check plausibility of indicators and warnings before final process creation/saving.