Trained Artificial Intelligence (AI) Model - Intergraph Smart Reference Data Standard Database - Version 12.2 - Help - Hexagon PPM

Standard Database for Intergraph Smart Reference Data Help

Language
English
Product
Intergraph Smart Reference Data Standard Database
Search by Category
Help
Standard Database for Smart Reference Data Version
12.2

You can map Smart 3D data to an existing SRD data by using the SRD Import functionality to import Smart 3D data to a defined staging area. Using configurations, you can either create or select an existing configuration to import Smart 3D data to SRD. The configuration is used to import Smart 3D data into a project and then map the imported data with the existing project of SRD data. As part of auto mapping enhancement, SRD 10.1 has implemented AI based configurations.

AI Configuration

The Artificial Intelligence (AI) configuration suggests the Commodity Codes or Codelists during the auto map procedure. An AI configuration must be trained until it reaches the required accuracy level so that the configuration can suggest accurate mappings. At any given time, you can have only one configuration for the given staging area and mapping type.

You can also import or export the mapped Commodity Codes or Codelists.

When you start training the configuration, the SRD application initially displays a filtered list of recommended Commodity Codes or Codelists in the staging area. Using the filtered list, you must train the configuration by mapping the Commodity Codes or Codelists with table details based on the selected mapping type. If you are training the configuration for the first time, the accuracy of the configuration starts at zero. Based on the number of mappings you perform, the configuration is trained, and the accuracy increases. You can stop training the configuration at any level of accuracy and resume it later.

When you are training the configuration, a bulb icon available next to the AI Configuration button blinks continuously indicating that the configuration is in Training mode. If the Configure Auto Mapping feature is not trained at all, the bulb will be OFF, indicating that the configuration is in Inactive mode.

After you finish training the configuration, you must change the mode of configuration from Training to Production. Once the configuration is in Production mode, based on the mappings performed while training the configuration, the software suggests the Commodity Codes in production. You can change the mode of configuration from Production to Training at any time to train the configuration again.

To re-train the configuration from scratch, you can reset it. When you reset the configuration, all the trainings that were done previously to train the configuration are discarded.

For more information, see AI Configuration.

SDB Trained Artificial Intelligence (AI) Model for description mapping

In the SDB 12.2 release, SDB delivers a trained AI model to provide recommendation for commodity code mapping in Smart 3D import job.

The SDB 12.2 is delivered with Smart 3D Ten specification equaling commodity codes in SDB. An AI Model is created with the mappings between S3D Ten specification commodity codes and SDB commodity codes (based on AI sampling recommendations in the training model). This trained AI Model can be directly used to get the recommendation for commodity code mapping of S3D Ten specification.

Work process to generate AI Model

  1. Create import configuration and staging area to load the required Codelist sheets (PipingCommodityType, WasherType, BoltType, GasketType and NutType) and Ten specification and part class workbooks.

  2. Update the required AI configuration settings to enable running the Sampling.exe. This file is used to create a filtered list of sample mappings from the source data.

    In training mode, mapping is done based on Sampling.exe recommendations. Additional mappings are done in production mode based on AI model recommendations provided by the trained AI model.

  3. Export the AI annotation file after mapping all the commodity codes in production mode to deliver as part of SDB.

Work process to reuse SDB delivered trained AI Model

During the S3D specification import process, go to Configure Auto Mapping function and directly import the AI annotation file to get the recommendation for commodity code mapping. See the below line drawing for work process.

Reuse SDB trained model

For more information, see Import section of Procedure for AI Configuration.

The Trained AI Model is delivered as part of SDB 12.2 and is found at: <SDB Installation Location>\SDB-SRD AI Annotation File for Smart3D Import\ AIMappings_commoditycode_SDB\ AIMappings_commoditycode_SDB12.2.xlsx.