3D Matching AI PLM

What is the “Aries” series?

The company name of Astraea Software Co., Ltd. is derived from Astraea, the goddess of justice in Greek mythology.

Based on scientific management methods, we will make unselfish management decisions in accordance with justice and fairness, grow the company, remember the altruistic spirit, and pursue the well-being of all customers, society and employees. ..

Based on that idea, we named our product series Aries, Aries, which is one of the 12 constellations.

What is the “3D Matching AI PLM”?

In the manufacturing industry of automobiles, electric appliances, etc., 3D CAD systems are widely used, and design data is stored as 3D shapes. However, the current situation is that there is no AI model for handling 3D data.

Based on the latest CNN research results, Astrier Software has succeeded in convolving 3D shape data for the first time in the world.

With this technology, it is possible to perform more accurate shape matching with 3D data, which has more information than 2D.

In order to effectively use the result of matching in 3D, it is necessary to link the result with the existing information. As a PLM product for that purpose, we have developed and released a 3D matching AI PLM equipped with an AI model capable of 3D matching.

However, if the customer has an existing PLM, it is also possible to link the matching result with the existing PLM owned by the customer.

In addition, the 3D matching AI model installed in the 3D matching AI PLM is the AI ​​model created by the 3D AI model Generator.

Classification (filter) and matching by 3D AI model

3D Matching AI We will explain how the 3D AI model installed in PLM recognizes the object to be recognized.

The 3D AI model recognizes objects in two ways: classification and matching.

1. Classification means that the AI ​​model that learned the 3D data labeled for each class determines which class the specified object belongs to and the certainty that it belongs to that class.

2. Matching means that the AI ​​model matches the specified object and the learned 3D data one by one, including the shape and dimensions, and quantifies the difference.

About class hierarchy

The 3D AI model recognizes objects in two ways: classification and matching.
In order to classify, it is necessary to prepare the classified learning data as training data in the 3D AI model.
As shown in the figure below, it is necessary to prepare the learning data for each labeled class such as Round Head Minus and Flat Head Minus.

If you prepare the learning data of the folder structure as shown below, you can create a 3D AI model with our 3D AI model Generator.

What is classification (filter)?

Classification means that the AI ​​model that learned the 3D data labeled for each class determines which class the specified object belongs to and the certainty that it belongs to that class.

With classification (filter) What is matching?

Matching with classification is matching including shape and dimensions fo reach bolt in the specified class.

Normally, matching will be performed within the upper class of the classification result.

Since the classification result is Round Head Minus, matching is performed including the shape and dimensions in that class.

What is matching without classification (filter)?

Matching without classification is matching that includes the shape and dimensions of all bolts without classifying in advance.

We perform matching including shape and dimensions with all bolts.。
You can find another class of bolts with a high degree of matching in shape and dimensions.

3D matching AI PLM system configuration

I will explain the system configuration of 3D matching AI PLM.

The system configuration of 3D matching AI PLM can be flexibly configured according to the customer’s request.

Mainly, it depends on whether the customer does not have or has an existing PLM, so I will explain the system configuration for each of these two cases.

When using our PLM

  • The 3D data to be verified is classified and matched.
  • From the result, it will be linked with the information managed in the DB.

To create a 3D AI model, use our 3D AI model Generator.

When using your company’s PLM

  • Create an AI model for classification and matching using the 3D AI model Generator.
  • Provides a Web API that uses the created AI model.
  • Call Web API, classify and match.
  • Please link the result with the existing system owned by your company.

To create a 3D AI model, use our 3D AI model Generator.

Data structure of 3D matching AI PLM

I will explain the data structure of 3D matching AI PLM.

If you do not have an existing PLM, you will need to use our PLM.

In that case, it is necessary to prepare the data to be linked with the classification and matching result of 3D matching AI. I will explain the structure of the data using Bolt as an example.

About classification information of classification

Decide the items to classify Bolt.

Here, it will be classified by the shape of the bolt head and the bolt hole.

Information on parts by class, parts, and size

Determine the hierarchy that manages Bolt. Decide which table to link with the classification result and matching result.

Here, the classification result is linked with the class master, and the matching result is linked with the size-based parts master.

You can also use only the matching results without using the classification results.

Supplementary information on parts by class, parts, and size

You can also manage if you have information such as class, basic parts, CAD information attached to parts by size, and CAE simulation results.

PLM information management function

3D matching AI PLM has two main functions.

There are two functions, a classification / matching function for identifying objects and an information management function for managing information to be managed.

First, I will explain about PLM information management.

Class information search

You can search for classes created by combining the contents set as classification information.
Here, the class is composed of the combination of the shape of the bolt head and the screw hole.

You can refine, sort, and paging class search results.

Class details screen

You can check the detailed information of the class on the screen that transitions from the class information search result and the classification result.

You can also manage images, 3D data, and attached information.

You can go down one level in the class.
Here you can follow the links to the parts below the class.

Parts detail screen

You can check the detailed information of the parts on the class details screen and the screen that transitions from the classification / matching results.

You can also manage images, 3D data, and attached information.

You can go down one level of the part.
Here you can follow the link to the size part details screen, which is one size below the part.

Parts details screen by size

You can check the detailed information of parts by size on the part details screen and the screen that transitions from the classification / matching results.

You can also manage images, 3D data, and attached information.

Supplementary information

In addition to basic information, supplementary information can also be managed on the size-specific parts details screen.

You can also manage the Web information of the supplier’s company and the save destination of the CAD file.
You can also open the supplier’s web screen or download a CAD file.

You can also view the CAE simulation results.

3D Viewer

On the parts details screen by size, you can not only view but also operate 3D data with the viewer on the browser.

You can also move, rotate, and zoom in / out objects.

You can also display axes, wireframes, etc.

Classification / matching function

We will explain the classification and matching functions that are the main functions of 3D matching AI PLM.

About selection of identification target

In the product prototype, the Sample data prepared by us and the data scanned by the scanner were misused as identification targets.

Normally, the user uploads the 3D data to be identified and specifies the identification target.

If you select Sample data or Scan data from 3D matching in the menu, each data will be displayed in a list.
From these, select the one you want to classify and match.

3D Viewer

The selected object is displayed in 3D Viewer and various operations can be performed.

After selecting the target of classification / matching, the next method of classification / matching is selected.





The selected object will be displayed in 3D Viewer.

You can move, rotate, and zoom in / out objects.

In addition, you can switch the camera and switch the display of the axis, wire frame, etc.

You can also enlarge the display on the pop-up screen

Classification / matching method specification

〇Classification by bolt head
・Specify whether to classify before matching.

〇Number of displayed classification results
・It can be specified only when there is a 3D bolt head filter.
・Specify the top number of classification results to be displayed.

〇Number of matching results displayed
・Specify how many top matching results are displayed.

Matching with classification

Whether or not to classify by bolt head specifies whether to classify first.

If yes, classify first as shown in the left figure above, and match within that class.

If you have a 3D bolt head filter, you can specify the number of classification results to be displayed and the number of matching results to be displayed.

In the figure on the right above, since 3 is set for both, the top 3 class information with high classification accuracy and the top 3 matching bolts for each class are displayed.

No matching for classification

Whether or not to classify by bolt head specifies whether to classify first.

If there is none, class classification is not performed as shown in the left figure above, and matching is performed for all objects.

Without the 3D volt head filter, you can only specify the number of matching results to display.

In the figure on the right above, since the number of matching results displayed is set to 4, matching is performed within all objects and the top 4 bolts are displayed.

Cooperation with PLM information management function Class information

If there is a 3D bolt head filter, as a result, the specified number of class information will be displayed.
A link to the class details screen will be displayed and you can transition.

Cooperation with PLM information management function Parts information by size



As a result, the specified number of parts information by size is displayed regardless of the presence or absence of the 3D bolt head filter.

The parts information by size can be changed by displaying the link to the parts details screen and the parts details by size screen.

Also, since it is displayed in 3D Viewer, you can move, rotate, enlarge / reduce, and display in wireframe

Other

Bracket version of 3D matching AI PLM

We’ve focused on examples of handling bolts in 3D matching AI PLM, but you can also target other things like brackets.

Therefore, it is possible to support a flexible system configuration according to the customer’s environment.

Below is the main screen of the bracket version of 3D Matching AI PLM.

Matching and information management can be performed in the same way as bolts.

Also, if it is difficult to classify due to its shape like a bracket, you can only perform matching without classification. Also, in that case, class information registration is not required.

About the cost of 3D matching AI PLM

3D matching AI PLM can support flexible system configurations according to the customer’s environment. It is possible to flexibly respond to requests such as wanting to make a small start to reduce initial investment.

If you are interested in 3D Matching AI PLM with 3D Matching AI, please feel free to ask any questions.

We apologize for the inconvenience, but thank you for your inquiry.