What is the “Aries” series?
The name of Astraea Software is derived from Astraea, the goddess of justice in Greek mythology.
We will pursue the happiness of all employees, customers and society by making selfless management decisions based on scientific management methods and in accordance with justice and fairness, growing the company and remembering the altruistic spirit.
Based on this idea, we have named our product line after one of the 12 signs of the zodiac, Aries.
Product Creation Background
Current simulations use very complex logic to calculate 3D shapes using inverse matrix and iterative calculations. As a result, there are many problems such as time consuming calculations, divergent results, and difficult settings that can only be performed by dedicated personnel.
Many simulations are likely to be routine in nature. Routine tasks are the areas where AI excels. However, in order to handle 3D with AI, there was a problem that convolution was difficult due to the indefinite number of nodes and edges connecting them in 3D data.
Based on the latest CNN research results, Astraea Software has succeeded in convolution of 3D shape data for the first time in the world.
With this technology, we have been providing products for 3D shape classification and matching. This technology has been newly appropriated to the field of simulation, allowing us to offer new products as surrogate models.
In addition, the simulatable AI model in the 3D surrogate model is created by the 3D AI model Generator.
CAE analysis solver issues.
CAE analysis solvers have the following issues.
Aries 3D Surrogate-Model will be an AI-based product to address the above issues.
What is the Aries 3D Surrogate-Model?
3D Surrogate-Model is a product developed with the intention of training AI on the simulation results of existing CAE analysis solvers, and then using the trained AI model to perform the same simulation on AI.
Design data for routine tasks trained by AI can be simulated using Aries 3D Surrogate-Model to meet the issues of CAE analysis solvers.
Simulations of new conditions that the AI has not been trained for can be done with the CAE analysis solver, but the results can be used to additionally train the AI to simulate them using the Aries 3D Surrogate-Model.
Features of the Aries 3D Surrogate-Model
The features of the Aries 3D Surrogate-Model are as follows.
Using AI models for simulation
We are using AI for simulation. Since most simulations are routine tasks, we developed our simulation product with the idea that AI can take care of this part of the work, and people can increase the ratio of their work to other creative tasks.
Calculation speed is extremely fast.
AI is much faster than CAE analysis solvers in terms of simulation calculation speed.
The factors behind this are listed below.
The following is a comparison of calculation execution times between our CAE analysis solver and the Aries 3D Surrogate-Model.
Stable calculation process.
CAE analysis solvers perform high-load calculations such as complex inverse calculations, and depending on the boundary conditions of the simulation, the simulation may be interrupted by divergence. AI, on the other hand, is a simple multiplicative analogy, so it does not interrupt the simulation due to divergence.
Less computational resources
CAE analysis solvers perform high-load calculations such as complex inverse calculations, which may use a large amount of computing resources depending on the size of the simulation and boundary conditions. The consumption of computational resources increases in proportion to the complexity of the simulation. However, since AI is a simple multiplication analogy, the consumption of computational resources does not increase in proportion to the complexity of the simulation.
AI training methods
This section describes the scheme of how AI analogizes the simulation results of 3D shapes.
The surrogate model is achieved by analogizing the simulation with the idea of a graph neural network (GNN), which considers a FE mesh model as a graph.
An FE mesh can be considered as a graph.
Graph nodes = mesh nodes
Graph edges = element edges
Graph node encodes loads, constraints, current node state (nodal stresses, deformations).
Graph edge encodes properties similar to spring properties such as material, edge length.
We can train graph neural network to predict the next state of a graph given its current state and boundary conditions.
The configuration of the Aries 3D Surrogate-Model
About providing AI models for each simulation
Because AI is an analogous calculation, AI models are used to limit the content and target of the analysis, such as the boundary conditions of the simulation. Therefore, in order to support multiple types of simulations, Aires 3D Surrogate-Model is provided in the following configuration.
Aries 3D-Surrogate Model with PLM
We sell a packaged product called Aries 3D-PLM, which is a web application for 3D shape classification and matching. We offer the Aries 3D-Surrogate Model as one of its functions. If you do not need PLM, you can use only Aries 3D-Surrogate Model as a web application.
You can enter boundary conditions, etc. on the Web, run simulations, and check the results on the Web. Some of the functions of the CAE analysis solver are also available on the Web. It is also possible to download the simulation results.
Aries 3D-Surrogate Model with API
It is also possible to provide an API that only executes the simulation. However, you will need to modify your own tools.
Also, if it is difficult to modify the tool you are using, please contact us and we will be happy to help you.
The cost of installing the Aries 3D Surrogate-Model
The Aries 3D Surrogate-Model can be flexibly configured to suit your environment. We can flexibly respond to requests such as starting small to reduce the initial investment.
If you are interested in the Aries 3D Surrogate-Model, please feel free to ask any questions you may have.
We are sorry for the inconvenience, but please click here to contact us.