アストライアーソフトウエア List of Products and Services

The AI-based product performance simulation technology, known as surrogate modeling technology, has recently gained attention. Among these, the PINNs (Physics-Informed Neural Networks) technology has particularly attracted interest, with NVIDIA also releasing a general-purpose module called NVIDIA Modulus. Our company is creating practical AI surrogate models using PINNs technology with NVIDIA Modulus and our proprietary PINNs module, and we have published technical documentation on this. Please download it from the catalog below. Additionally, you can try out the PINNs surrogate model from the "Astraea Software Product Demo Page" available at the link below.

• Other analyses

This is an instructional video for searching chair shapes using Aries 3D-Matching products. It utilizes 3D shape recognition AI to search for similar chair shapes from the product library.

• Other CAD related software
• simulator

This is an introduction to our AI technology that can handle 3D shapes. The AI mixes existing vehicle shapes to create new ones. By changing the ratio of the two vehicle bodies being mixed, it is possible to generate various shapes of vehicle bodies.

• Other CAD related software
• simulator

This is a simulation operation instruction video for cantilever beam components using the Aries 3D Surrogate Model product. It utilizes AI that has learned from past simulation results to instantly predict the deformation and stress of the cantilever beam model.

• Other CAD related software
• simulator

This is an instructional video on searching for bracket parts using Aries 3D-Matching products. It utilizes 3D shape recognition AI to search for similar shapes of bracket parts from the bracket parts library during the design process.

• Other CAD related software
• simulator

This is an instructional video for searching bolt parts using Aries 3D-Matching products. It utilizes 3D shape recognition AI to search for similar shapes of bolt parts from the bolt parts library during the design process.

• Other CAD related software
• simulator

3D Shape Matching (Patent No. 7190147) Aries 3D-Matching searches for parts that match the target shape and size from a registered library and outputs them in order of highest match degree. You can easily access design data registered in PLM from the matching results. This promotes the effective use of past design assets and significantly improves work efficiency. Our company has obtained a patent for this AI technology (Patent No. 7190147). Classification by Part Shape The classification function of Aries 3D-Matching focuses solely on the three-dimensional shape of parts, classifying them into specified classes and outputting class names in order of highest match degree. You can easily access data registered in PLM from the classification results. This promotes the effective use of data and greatly enhances work efficiency. Functionality as PLM Aries 3D-Matching is equipped not only with the function of searching three-dimensional shapes but also with features as a PLM product. It includes comprehensive functions such as management of classification results, graphic data, supplementary information, text search, and permission management.

• Data Search Software

The "3D Surrogate Model" is a product developed with the intention of training AI using simulation results obtained from existing CAE analysis solvers, and then using the trained AI model to conduct similar simulations. By utilizing AI for simulations, it significantly reduces computation time and, due to its reliance on simple multiplication, prevents analysis results from halting due to divergence. Additionally, it can directly learn from the mesh as an AI model, allowing features such as material properties and edge lengths to be represented. 【Features】 ■ Significant reduction in analysis execution time ■ Stability of analysis processing ■ Improved ease of use ■ Reduction in analysis resources *For more details, please refer to the PDF materials or feel free to contact us.

• 3D CAD

The "3D Matching AI PLM" can perform more accurate shape matching using 3D data, which contains more information than 2D data. If you already have an existing PLM, it is possible to link the matching results with your existing PLM. The AI model capable of 3D matching included in this product uses our patented technology (Patent No. 7190147). 【Features】 ■ Enables more accurate shape matching ■ Equipped with an AI model capable of 3D matching ■ Can link matching results with your existing PLM ■ Recognizes objects using two methods: classification and matching ■ System configuration can be flexibly tailored to customer requirements *For more details, please refer to the PDF materials or feel free to contact us.

• Data Search Software

The "3D AI Model Generator" allows you to easily create a trained 3D AI model simply by providing existing 3D data and its classification names (labels). To effectively utilize existing 3D data, it reduces the time and effort required for data collection and processing. Since the Generator handles the environment setup and implementation for building and training the AI model, specialized knowledge of AI is not required. 【Features】 ■ Easily create AI models just by preparing 3D data ■ Reduces time and effort for data collection and processing ■ No specialized knowledge of AI is required ■ Continuous enhancement of AI models can be performed ■ Ability to review the trained data and its results *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAD

This document introduces the development of "AI for retinal images." It includes information on the "performance of the developed 'AI for retinal images'," "the journey to the completion of this medical AI and future prospects," and "promotion of the developed medical AI." It is capable of determining whether "unlearned diseases" are normal or abnormal, even when the training data and test data are completely independent. 【Contents】 ■ Introduction ■ Performance of the developed "AI for retinal images" - Verification 1: Can it detect unlearned diseases? - Verification 2: Can it diagnose even when the training and test data are completely independent? - Summary of the AI's performance *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

Our company, which is engaged in AI×CAE activities, is also researching and investigating "surrogate models." As the application range of CAE expands and the pursuit of higher accuracy in simulation results becomes more prominent, the complexity of the considerations has increased, leading to issues such as excessive computation time in simulations that start from scratch. With "surrogate models," the time required for analysis execution can be significantly reduced, and utilizing pre-trained models does not require specialized knowledge, making them easy to use. 【Issues with CAE Simulation】 ■ The complexity of CAE considerations has increased, resulting in excessive computation time in simulations that start from scratch. ■ Rising costs of software and hardware, shortage of specialized engineers, and increasing education costs. ■ Maintenance costs for software are also rising. *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAD

This document explains 'Explainable AI (XAI)', which has emerged as a new challenge for AI. It covers topics such as "AI challenges," "the purpose of XAI," and "approaches to achieving XAI." The initiative for "XAI" aims to create more explainable AI while maintaining learning performance. 【Contents】 ■ AI challenges ■ Purpose of XAI ■ Relationship between the two objectives of XAI ■ Approaches to achieving XAI ■ Improvement of explainability *For more details, please refer to the PDF document or feel free to contact us.

• Software (middle, driver, security, etc.)

"Similarity judgment" refers to the process of searching for shapes similar to the input shape from a registered database. Feature quantities of the shape are extracted, and to compare them with the feature quantities of existing shapes in the database, generative models such as VAE and GAN are used for training. In many industries, such as manufacturing and construction, there is a shift towards design based on three-dimensional shapes using 3D CAD. Additionally, computer-aided technologies related to design and manufacturing, known as CAD, CAM, and CAE, are also based on three-dimensional shapes. 【Features】 - Extract feature quantities of shapes and compare them with existing shape feature quantities in the database. - Use generative models such as VAE and GAN for training. - Direct comparison using 3D shape recognition AI. - String comparison using text-based AI. *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAM

"New Shape Synthesis" refers to the automatic generation of new shapes by blending feature vector quantities extracted from a properly trained AI model, similar to similarity judgment, across multiple shapes. A network using a type of deep learning called Variational Autoencoder (VAE) is constructed to create new vehicle body data. Multiple timings are established to add feature representations to the Decoder part of the VAE system, enhancing the generation results. 【Features】 ■ Automatically generates new shapes by blending feature vector quantities extracted from a properly trained AI model across multiple shapes. ■ Constructs a network using the VAE system to create new vehicle body data. ■ The time required for shape synthesis is just a few seconds, allowing designers to explore designs as they come to mind. *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAM

"Shape classification" is the task of classifying three-dimensional shapes into predefined classes. It uses convolution techniques that minimize the loss of shape features, allowing for classification whether focusing on a part of the shape or viewing the overall shape from a broader perspective. It can be utilized not only for simple class categorization but also for other tasks. Additionally, when calculating cutter paths for NC lathes for injection molding and forging molds, it automatically identifies bolt hole shapes from the mold model and removes them from the overall shape. [Features] - Uses convolution techniques that minimize the loss of shape features - Allows classification whether focusing on a part of the shape or viewing the overall shape - Can be utilized not only for simple class categorization but also for other tasks *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAM

We would like to introduce the features and technology of the "3D AI Model" that we are researching and developing. The developed "3D Shape Recognition AI Model" can be applied to functions such as "shape classification," which recognizes the characteristics of shapes and classifies them into specified groups; "shape analogy," which compares the characteristics of multiple shapes and determines their similarities; and "shape synthesis," which mixes the characteristics of multiple shapes to generate new shapes. Please feel free to consult us when needed. 【Features】 ■ Enables the incorporation of variable network structures, allowing efficient recognition of 3D shapes by AI. ■ Applicable to functions such as shape classification, shape analogy, and shape synthesis. *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAD

In this document, we introduce the activities of Astraea Software Co., Ltd. With the technology of "3D AI" that we have been developing, it is possible to directly recognize shapes from the nodes (vertices, coordinate positions) and elements (node connection information) that make up a 3D shape. 3D shapes captured by 3D scanners can also be recognized as they are and utilized in various scenes of the design process. Our company aims to permeate advanced technologies cultivated in large corporations throughout the industrial structure, providing strong support to the entire Japanese manufacturing industry. [Contents] ■ Why is 3D AI necessary? ■ Comparison of text-based AI and 3D AI ■ Comparison of 2D AI and 3D AI ■ Activity goals: Integration of AI/CAE/Cloud *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAD

Our company is engaged in research and development of 3D shape recognition technology using AI and deep learning, aiming to make better use of 3D CAD data in the manufacturing industry. Based on new AI research results, we have successfully developed a 3D AI model capable of recognizing 3D shape data. Our goal is to integrate CAE, AI, and cloud technology, promoting technological penetration from large corporations to small and medium-sized enterprises. We will advance research and development in 3D AI, contributing to the improvement of QCD not only in the manufacturing industry but also across many other sectors. 【Business Content】 ■ Manufacturing, sales, technical support, and education of technical support software *For more details, please refer to the PDF document or feel free to contact us.

• 3D CAD