Integrating Automated Simulation Workflows with 3D Visualization for Virtual Experiments in the Metaverse

Integrating automated simulation workflows with 3D visualization for virtual experiments in the Metaverse. A workflow is a chain of tools defining the flow of data between them. In many virtual experiments, researchers manually connect multiple simulation and processing tools, which scales poorly and impedes reproducibility.

Workflows solve this because they are abstract, automated, scalable, portable, shareable, and citable. They allow existing simulation codes and processing steps to be wrapped into tools that are available in an easy-to-use interface. These can then be linked together and reconfigured without manually creating bespoke automation scripts.

The tools can also be shared, reducing the requirement to learn new interfaces and allowing researchers to instead focus on configuring simulations and interpreting results. This protocol demonstrates the deployment of a locally containerized workflow manager. In this configuration, the end user interacts with a 3D visualization platform as the front end.

The link to the workflows comes in the form of a lightweight Python extension that makes API calls to the workflow manager to launch simulations and receive the results, all from inside the visualization platform. The system presented here is general by nature and can be applied in almost any application where simulation and processing steps are performed. For this case study, a simple neutronic simulation in the context of a fusion energy power plant is used.

In a fusion reactor, deuterium and tritium are fused together to produce energy and neutrons. Deuterium is a natural isotope of hydrogen, but tritium must be produced by the reactor itself using neutrons generated in the fusion events. To simulate the neutrons produced, a Monte Carlo simulation code is used, which follows the neutrons around the reactor, applying random interactions with the material as they pass through.

As a comparative measure, a key performance indicator called the tritium breeding ratio is calculated, which provides the ratio of tritium produced by neutron interactions to the amount of tritium consumed in the reaction. The simulation uses a simple CAD geometry based on the spherical tokamak shape to model the amount of tritium produced by the blanket wrapping around the plasma vessel, and outputs a dat set of the generated neutron tracks. To begin the protocol, assume the container engine and 3D visualization platform have both been installed, following the project repository to the local machine.

Launch the workflow manager server using the provided shell script, which uses the container engine to deploy a portable instance. Once the system is running, access the local web interface to create an administrative user account and log in. Navigate to the user preferences to generate a unique API key.

This key authenticates the 3D visualization platform when it calls the workflow manager. Next, configure the 3D visualization platform by navigating to the relevant file in the extensions folder. Add the API key to this configuration file.

Ensure this private key remains secure and is not committed to public repositories. Next, configure the 3D visualization platform by navigating to the extensions menu, clicking the settings icon, and adding the local path of the extension folder from the cloned repository. Finally, enable the extension.

The auto-load option could be selected so that it is automatically enabled on relaunch. To run any single tool, input files must first be uploaded to the service. Create a new history and name it.

Histories act as separate workspaces for each set of tools or workflows run. Click the Upload button. Select the required CAD geometry and configuration datasets from the example files in the repository and click Start.

The inputs will appear in the History panel on the right-hand side of the screen. Navigate to the required Monte Carlo simulation tool on the left-hand panel and click on it. Select the relevant input data sets from the current history and run the tool by clicking execute.

The outputs will change from orange, indicating that they are processing, to green, indicating that they have finished. Gray is used to highlight pending steps and red for failed steps. The first run may take longer as the tool's container image is downloaded.

Subsequent runs should be faster. Once the tool has finished running, the results of the simulation can be viewed. Click on the tritium breeding ratio dataset in the history panel to expand it.

This displays the ratio calculated by the neutronic simulation and associated statistical error. Tools are more useful when used in workflows, which allow multiple tools to be called in a single invocation and control the flow of data between tools automatically. To replicate the previous analysis pipeline, click on Workflows, then create and name a new workflow.

Import the tools for the Tool panel by clicking on them. And then add two datasets as inputs and name them. When all the tools and inputs are on the canvas, join them together by clicking and dragging to connect the output of one step to the input of the next, defining the flow of data.

Save the workflow using the icon in the top right. The workflow is run in much the same way as a single tool. Navigate to the Workflows tab in the top bar and click the run icon on the new workflow.

Select the corresponding input datasets, then click Run Workflow. This executes all the steps that were previously run manually, saving intermediate outputs to the history. The run can be verified by inspecting the output files or by displaying the tritium breeding ratio in the interface.

To run workflows from the 3D visualization platform, the output files must first be converted to a compatible 3D format. Create a workflow that mirrors the previous one but with additional steps. A track format conversion to USD and a CAD format conversion to USD via intermediate steps.

Save this workflow. Open the 3D Metaverse platform. The enabled extension should be available in the viewport.

Dock it to one side. Click Get Workflows to pull the list of available workflows from the connected instance. Select the newly created workflow with conversion steps to USD and click Get Inputs.

The inputs defined for this workflow will appear. Select the standard CAD geometry and configuration files for each. Then, click Run Workflow.

Once the workflow is completed, a notification appears in the information window of the extension. Click Refresh under the file manager to view the list of completed workflow instances. Select the current instance.

Refresh again to retrieve the file list. Choose the files to view and click Pull File. 3D files load to the main viewport and text-based files appear in the information window.

A successful simulation run will generate a tritium breeding ratio output. Because this is a stochastic process using a small sample size for speed, the expected value using the standard configuration provided in the repository should be approximately 0.76, with a variance of plus or minus 0.01. Increasing the batch count and particle count in the config input reduces this scatter at the cost of longer run times.

A successful run also produces a set of converted visualization files, the CAD geometry, and the neutron tracks, each in a format compatible with the 3D visualization platform. These can be downloaded and inspected in an external viewer or pulled directly into the 3D visualization platform. This method demonstrates how domain expert researchers can integrate simulation tools into automated workflows for 3D virtual experiments.

The approach supports the FAIR data principles, run histories capture full providence metadata, and tools are packaged in version-controlled containers, making the analysis pipeline both reproducible and shareable. While the workflow manager itself can be run locally on standard hardware, utilization of the 3D virtualization platform's full ray tracing engine requires a compatible high-performance RTX graphics card.