Machine AI Testing (RnD)

Background

The factory owner and manufacturer jointly evaluate different AI-algorithms for the process industry. Combining historical data with real-time analytics to train designated AI-services.

Prerequisites

Difficulties to connect to designated machines due to complex and complicated OT-system and security concerns.

Scope

Apply multiple AI-algorithms for real-time use-cases

• Edge computing and analytics for collecting data
• Tag Management for data selection

Sandbox, testing, and comparison of AI-functions

• Self Service requirements
• Shortage of developers/programmers
• Limited time for deployments

Organizations constraints due to many stakeholders within the project

• Roles based usage of Edge software

Solution

Crosser Node deployed in a virtual machine inside the firewall:

• Connecting to OSI-Soft PI and retrieving OPC-data
• Specific data tag management. Focusing on specific tags
• Deploying data cleansing and preparation algorithms for AI testing
• Sending data to Amazon Web Services, where AI algorithm is running

Crosser Cloud:

• Customer organization set up with roles o Resources library setup