How to transform manufacturing four ways with AI

Automate 2024: Microsoft and OPC Foundation AI insights

• Artificial intelligence (AI) integration can help manufacturing with the labor shortage, as explained by Microsoft and OPC Foundation at Automate 2024.
• AI can help domain experts with programming, interacting with and troubleshooting machines and automation.
• OPC Foundation formed a working group on AI.

Artificial intelligence (AI) can help with the manufacturing labor shortage in many ways, including helping with engineering, data analysis, programming or updating control code and machine troubleshooting, explained Holger Kenn, chairperson, board of directors, OPC Foundation and director of business strategy, AI, data and emerging technology, Microsoft Corp., at Automate 2024 by A3, the Association for Advancing Automation, in Chicago (Figure 1). Kenn made the comments during the “AI in the Machine – AI and OPC UA in Industrial Manufacturing and Robotics,” talking about what now is possible with AI in manufacturing and what may be possible very soon.

Figure 1: Holger Kenn, chairperson, board of directors, OPC Foundation and director of business strategy, AI, data and emerging technology, Microsoft Corp., at Automate 2024 by A3, the Association for Advancing Automation, in Chicago, spoke about “AI in the Machine – AI and OPC UA in Industrial Manufacturing and Robotics.” He talked about what now is possible with AI in manufacturing and what may be possible very soon. Courtesy: Mark T. Hoske, Control Engineering

AI and interconnectivity can help with the manufacturing labor shortage

Manufacturing has a labor shortage. OPC Foundation provides interoperability standards to add meaning to manufacturing data, Kenn said. Many organizations are involved in OPC Unified Architecture (OPC UA) collaboration (Figure 2). Figure 3 shows resulting OPC Foundation companion specifications. Semantic information helps interoperability and is human and machine readable. A Github OPC Foundation specification repository contains UANodeSets and other normative files. https://github.com/OPCFoundation/UA-Nodeset

Figure 2: Many organizations are involved in OPC Unified Architecture (OPC UA) collaboration, explained Holger Kenn, chairperson, board of directors, OPC Foundation and director of business strategy, AI, data and emerging technology, Microsoft Corp., at Automate 2024 by A3, the Association for Advancing Automation, in Chicago. Courtesy: Mark T. Hoske, Control Engineering

AI can rescue old code, translate to modern languages

AI in engineering uses large language models to read and apply the OPC UA specification. Write code in C# for interacting with OPC server. Microsoft Copilot software generates code for standard programming languages, explains and documents existing code, writes code in domain-specific languages and converts code from old domain-specific languages to modern languages, Kenn said.

“If we have a curated set of training data, building large language models can be trained properly to take the legacy code forward. It still will need attention, but it provides a starting point.”

Manufacturing requires more because there’s a need for data and systems specification. Generating machine-readable specification from device documentation. Conversion between existing device specification standards, code-generation for device interaction based on specification. While it can be done manually, the return on investment improves by using AI, he suggested.

Figure 3: OPC Foundation companion specifications involve collaboration for many industry organization, said Holger Kenn, chairperson, board of directors, OPC Foundation and director of business strategy, AI, data and emerging technology, Microsoft Corp., at Automate 2024 by A3, the Association for Advancing Automation, in Chicago. Courtesy: Mark T. Hoske, Control Engineering

AI provides domain experts with programming superpowers

An example AI application would be using AI for natural language interaction with large amounts of data. Users don’t need domain-specific query languages and data structures; they can just ask questions, Kenn said. A prompt can be used to generate a query. Results are added to prompt, and the answer is generated based on prompt and retrieved information. Queries can be performed with embedding models and vector databases (AI) or with conventional database queries, such as SQL. This provides a superpower for domain experts because there’s no need for a data science team to build custom analytics.

AI as the next-generation user interface

AI used in factory can be the next-generation user interface, Kenn said. Data analysis and document retrieval can be combined with AI. Natural language documentation and maintenance records are indexed through embedding models and vector databases. Historical and current data from systems can be retrieved from data lakes via query generation. Large language models can combine query results from documentation and data to give natural language answers. Questions don’t need to be typed. Speech recognitions can turn spoken voice to text questions. A possible application would be speech-based fault investigation and maintenance procedures for conversation-based root-cause analysis.

AI can enable operators to help with servicing machines more quickly and effectively.

Another example Kenn described was to help collaborative robots. Perception-based AI identifies objects and locations. National language instruction and interoperability can work with existing automation systems to product flexible manufacturing based on mobile and humanoid robots. An operator can tell the robot to “Pack those boxes.”

Figure 4: An OPC Foundation AI working group, announced in April, will look at AI technologies in the engineering phase, using OPC UA and OT servers as data sources for AI systems, according to Holger Kenn, chairperson, board of directors, OPC Foundation and director of business strategy, AI, data and emerging technology, Microsoft Corp., at Automate 2024 by A3, the Association for Advancing Automation, in Chicago. Courtesy: Mark T. Hoske, Control Engineering

AI working group examines what’s possible for AI in manufacturing

An OPC Foundation AI working group announced in April looks at AI technologies in engineering phase, using OPC UA and OT servers as data sources for AI systems (Figure 4). It uses AI-generated configuration files for industrial connectivity software leveraging OPC UA. Use AI-based interfaces can be used for the operation of automation systems. This can help expand the domain of automation and increase the efficiency of engineers. Using AI, robotics can be applied for lower cost, higher variability applications in smaller manufacturing facilities.

Questions and answers about AI on the plant floor

While answering audience questions, Kenn said 65 members are interested in the OPC Foundation AI working group, including major cloud vendors. Domain-specific language models might result. Participation is also interesting for large system integrators involved in OT, IT and AI.

The idea is for AI to improve plant-floor capabilities and an interface for operations might be an extension.

AI can be used in edge devices and platforms, just as other applications are being applied, Kenn said. Process industry applications, such as for remote locations may want to use AI locally when satellite connection fails. AI containers could run on edge devices. OPC Foundation will stay out of the cloud versus edge debate, welcoming machine and cloud-based proponents to the working group, Kenn said.

Mark T. Hoske is editor-in-chief, Control Engineering, WTWH Media, mhoske@wtwhmedia.com.

KEYWORDS

Automate 2024, AI for plant-floor applications, AI manufacturing applications

CONSIDER THIS

Have you found the best application to apply your first, or next, robot?

Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our WTWH Media editorial team and getting the recognition you and your company deserve. Click here to start this process.