AI standards help accelerate digitalization of smart manufacturing

One of the key drivers of the digital transformation is smart manufacturing

By Wael Diab, Chair ISO/IEC JTC 1/ SC42 for Artificial intelligence

Various analysts put the market estimate at several hundreds of billions US dollars with double digit CAGR growth. Much of this growth is being fuelled by the digitalization of the sector as emerging IT technologies such AI, big data and analytics are improving efficiencies and unlocking insights that would otherwise be unattainable.

Smart manufacturing AI standards help ensure AI systems in smart manufacturing are trustworthy

Together, IEC and ISO develop international standards for information and communication technologies through a joint technical committee (ISO/IEC JTC 1). One of the committees (SC 42) covers the entire ecosystem for artificial intelligence (AI).

Earlier this year, IEC and ISO endorsed a joint definition of smart manufacturing:

“Manufacturing that improves its performance aspects with integrated and intelligent use of processes and resources in cyber, physical and human spheres to create and deliver products and services, which also collaborates with other domains within enterprises’ value chains.”

In the definition, performance aspects include agility, efficiency, safety, security, sustainability or any other performance indicators identified by the enterprise, while other enterprise domains can include engineering, logistics, marketing, procurement, sales or any other domains named by the enterprise.

The definition also provides a few insights into smart manufacturing. These include:

  • An encompassing approach that combines cyber, physical and human aspects
  • A focus on improving both performance as well as the creation of both products and services through intelligence

The link to AI and enabling IT technologies

The focus on intelligence is fundamental to smart manufacturing. The key component is insights, whether providing them on improving operational efficiencies in manufacturing or for making intelligent decisions on what or where to manufacture. To provide those insights, IT systems are used to look at the large amounts of data that are coming from the manufacturing domain. This focus on data is the link to emerging IT applications such as AI.

Through applying big data and AI techniques, IT systems can take data analytics to the next level. For example, in machine learning based AI systems, the algorithms at the heart of AI can be used to predict when maintenance is needed dynamically, monitor and provide recommendations to improve quality, provide guidance on root cause analysis, improve yields and much more. AI not only enables these analytics, but by looking and learning from the data, the insights delivered can be tailored to the application and context it is being used in. 

Ultimately these efficiencies can lead to cost reductions in the manufacturing process and improved production times.

The need for AI standards

Standards are essential to removing barriers to deployment, addressing concerns and ultimately accelerating adoption.

Horizontal AI standards, such as those being developed by SC 42, enable smart manufacturing in a number of ways:

  • Terminology and foundational frameworks: As smart manufacturing brings together a diverse set of interests that include information technology experts and operational technology experts, a common language and a framework for the use of AI machine learning are important to the successful architecture and deployment of next generation smart manufacturing systems. SC 42 is working in two standards in this area namely ISO/IEC 22989 and ISO/IEC 23053.
  • Trustworthy AI: Key to the successful deployment of AI systems in smart manufacturing is to ensure that the system is trustworthy. To that effect, SC 42 is developing projects in this area that are applicable: 

         - AI domain overview of trustworthiness, bias and robustness of neural networks: These projects aim to introduce the topic and some of the AI context specific concepts about trustworthiness, bias and robustness. In the area of bias, the project also addresses AI- aided decision making.

         - Risk management: This project builds on the generic ISO 31000 standard for risk management in the AI domain. The document provides guidelines on the management of risk during the development and deployment of the AI system.  The purpose is to establish trust in the system by addressing issues such as this, by design and during the operation of the system in accordance with the goals of the deployment.

  • Ethics and societal concerns: The ability of AI systems to learn and make decisions brings about a host of ethical considerations. For instance, ensuring that AI enabled systems in smart manufacturing are safe. Moreover, when dealing with data and developing insights, the systems should only consider data within the application provenance and not look at data that would otherwise be unavailable to a human (commonly referred to as eavesdropping). To address these issues SC 42 is developing a project that maps such high-level concerns and looks at these across its technical programme of work. For example, ethical concerns are being collected for the various AI use cases.
  • Application guidance and use cases: One of the primary goals of SC 42 is to provide guidance to application committees within IEC, ISO and JTC 1. To-date, SC 42 has collected over 85 use cases, which include smart manufacturing and is actively working with committees and organizations looking at this domain.
  • Governance implications of AI: When an AI system is deployed in an organization, questions may arise by non-technical executives managing and deciding on the deployment of such systems. By collaborating with the committee covering IT service management and IT governance (SC 40), through a joint working group, SC 42 is developing a standard that would aid in answering some of these questions.

Additional standards work in the area of big data and analytics are of relevance to smart manufacturing deployments. As large data sets are collected over the life of a smart facility and across different facilities, big data techniques for processing the information and deriving analytics may also be applied. SC 42 has published and is developing some standards which include foundational projects on a big data reference architecture, use cases and a framework for business management of big data analytics.

Finally, as the world of AI and data science is rapidly changing, SC 42 is looking at several study areas. An example is the implications and challenges of developing and integrating AI into different applications, such as smart manufacturing through an advisory group on AI Systems Engineering. Concepts such as integration, maintenance, adaptation of best practices to AI systems and redeployment are being discussed. Another area of study is looking at a management systems standard that would provide AI-specific process requirements which would in turn allow for conformity assessment.

Building out the industry ecosystem

The opportunity of AI enabling smart manufacturing and AI applications more generally, is not only large, it’s transformative. Consequently, no single organization or standards body can go it alone. While SC 42 is taking a broad look at the entire AI ecosystem, it works collaboratively with other IEC and ISO committees, which cover biometrics, blockchain, coding of audio, picture, multimedia and hypermedia information, digital twins, health informatics and risk management and IT, as well as with external organizations through liaisons.

For smart manufacturing, SC 42 is working with a number of IEC and ISO committees such as IEC TC 65 for industrial-process measurement, control and automation, IEC Systems Committee for Smart Manufacturing, ISO Smart Manufacturing Coordinating Committee (SMCC) and other JTC 1 committees which cover cloud computing, data management and interchange and IoT. 

Smart manufacturing AI standards help ensure AI systems in smart manufacturing are trustworthy