Smart factories that use breakthrough technologies to drive efficiencies within production processes and across value chains have captured the attention of manufacturing executives.
Digitalization, so the story goes, offers a wide array of advantages. They include predictive maintenance that will reduce downtime through the creation of “digital twins,” enhanced quality control, demand-driven production, inventory optimization, reduced energy and material costs, and improved safety and environmental performance.
Numerous estimates attempt to quantify the value proposition. Consulting firm McKinsey says the economic impact could be between $1.2 and $3.7 trillion by 2025. A recent U.S. Department of Commerce survey of U.S. manufacturers and smart manufacturing vendors suggests $57 billion dollars in annual cost reductions.
Of course, there is a catch. Several, in fact. Investment cycles in the manufacturing sector are extremely long. Robust processes and devices will not spring up overnight. Critically, the needed technologies—such as artificial intelligence, or AI—are not yet fully developed.
AI as the Catalyst
Smart factories leverage the Industrial Internet of Things (IIoT), Big Data, and advanced analytics. Information technology (IT) and operations technology (OT) converge. Devices that communicate with each other lead to real-time decision-making that optimizes value creation.
It occurs both within the factory and across the entire value chain, from raw material acquisition all the way through order fulfillment and customer service.
The underlying catalyst for this transformation is artificial intelligence (AI). Much of the current excitement over AI is about machine learning—a set of techniques that couple real-world data and experience with statistical analysis to draw conclusions and predict outcomes.
Machine learning is not a new AI field, but the development of the Internet, the proliferation of massive amounts of data, and the growing processing power of computers have greatly advanced the depth, breadth, and accuracy of its predictive power.
Although AI is clearly advancing, it has its limitations. The underlying algorithms are tricky to design, which can lead to both vulnerabilities and unintended bias. The training step often requires an extremely large amount of data and real-world experience that may be difficult to obtain. Neural networks often take a long time to train. And when an AI-enabled decision goes wrong, it is often difficult to determine the reason, which is a major concern in safety-critical systems.
Why is AI now being applied in a factory setting? For sure, technology is a driver: the availability of massive amounts of data, the evolution of machine learning, the advent of cloud computing (for network-wide monitoring and optimization) and edge computing (which provides machine learning for real-time decisions), and the marriage of information technology (IT) systems with operations technology (OT) systems.
But current social trends are also important, including the increasing complexity of global supply chains and ongoing challenges in attracting skilled production workers. In other words, the emergence of smart factories is a function of both technology push and market pull.
Should all of the AI wrinkles get ironed out—and eventually they will—smart factories still won’t proliferate absent the optimal information governance.
Three such governance issues include technical standards, cybersecurity/privacy, and spectrum allocation.
Technical Standards
Smart factories rely on information flow and system responsiveness that cannot occur without standards—basically norms or requirements in relation to technical systems.
Hundreds or perhaps thousands of standards are utilized in manufacturing processes, and many new standards are needed to enable smart factories. A February 2016 report from the National Institute of Standards and Technology (NIST) noted that the smart manufacturing ecosystem can be seen as a pyramid of four progressing levels: the device level, supervisory control and data acquisition (SCADA) level, manufacturing operations management (MOM) level, and enterprise level. Information must flow within and across each level, and dozens of standards have been developed or are under development to quicken this collaboration.
According to NIST, “within the manufacturing pyramid, communication standards are well established but interoperability among systems is somewhat limited, meaning that manufacturers typically are locked into a single vendor solution. Along the business cycle, several well-established standards exist; however, the level to which information is able to interconnect with the production systems is quite limited.”
Aside from the development of standards to fill these gaps, the report identified two other standard-related barriers to smart factories: (1) the lack of tracking of standards and standards adoption, and (2) overlap and redundancy between standards. Harmonization and collaboration among SDOs are necessary to address these barriers, and some of this is under way.
Standards are also being developed to promote the use of blockchain technology. Blockchain is a digital ledger that can record transactions in a verifiable and secure manner. The Department of Homeland Security (DHS) is conducting blockchain pilots with industry to see if the technology can thwart counterfeit products and intellectual property theft. Security and defined interoperability standards will be needed to promote use of the technology.
Cybersecurity/Privacy
Smart factories require interconnection among equipment and devices both within the factory and across the value chain. Such connectivity increases the risk of cyberattack, espionage, and data theft for manufacturers.
These are not hypothetical problems. For example, in 2014 hackers damaged a German steel mill after gaining access through phishing emails. A recent UK survey found that 50% of manufacturers acknowledge being hacked and half of those attacked suffered a loss as a result. According to the U.S. Department of Homeland Security, manufacturers are a top target of cyberattacks against critical infrastructure.
Given the fatter target presented by smart factories over traditional factories, security becomes more of an issue. Security goals include the maintenance of production (no downtime or delay), prevention of a system failure that results in physical damage to property or human injury/fatality, prevention of espionage, and protection of the privacy of customers and employees.
Meeting these goals is not at all straightforward nor easy. To protect smart factories requires a multiplicity of approaches and systems, including security architectures for cyber-physical systems, verification of software integrity through attestation (a process that enables the detection of malware or unintended code), and secure device management (especially challenging given heterogeneity of devices in a manufacturing setting).
Vendors offering smart manufacturing devices and services are clearly involved in these security developments. As are governments. The U.S., working in collaboration with industry, has developed a risk-based and voluntary framework for cybersecurity for critical infrastructure broadly applicable to a range of businesses, including manufacturers. NIST has also published a framework for smart cities that is relevant for smart factories.
Another growing issue relates to the privacy of personal information. The EU’s General Data Protection Regulation (GDPR) is a legal framework that sets guidelines for the collection and use of personal information of individuals. This new law has implications for smart factories. For example, technology which measures output on a production line might collect individual worker data. Manufacturers will need to ensure that they are transparent about the personal information that is collected using these technologies by updating privacy notices and ensuring those notices comply with the GDPR.
Finally, smart factories will drive changes in insurance, and the insurance industry will face a demand to build solutions to manage changes in risk.
Spectrum Allocation
The number of devices that will be needed to realize the promise of smart factories is an important consideration for information governance. These devices are expected to operate via wireless communication. Wireless devices currently number in the billions, and that number is expected to grow exponentially, due to IoT and IIoT.
All of this demand for wireless communications will require spectrum, a scarce public resource. For smart factories to succeed, governments will have to allocate spectrum sufficient to meet this increase in demand. In the U.S., the Federal Communications Commission (FCC) allocates spectrum for consumer and commercial use.
Last year, the U.S. Government Accountability Office (GAO) investigated this issue. According to the GAO report, FCC believes currently available spectrum is sufficient for the growth of the IoT in the near future, unless there is a surge in devices that use a disproportionately large amount of spectrum. GAO also noted that “managing interference is becoming more challenging as the number of wireless devices grows, particularly in bands that do not require a wireless license.” GAO recommended that FCC start tracking growth of the IoT to ensure sufficient spectrum remains available.
Should additional spectrum be needed to support smart factories, will it be licensed spectrum, unlicensed spectrum, or shared spectrum? The FCC decides how available spectrum is allocated across each type and within which frequency bands.
These governmental decisions will impact the availability and quality of spectrum for smart factories in the US. Other countries are also grappling with how to allocate spectrum for industrial uses. The GAO report noted that each country is taking a different approach, with at least one, South Korea, dedicating spectrum for industrial use.
Keith Belton is director of the Manufacturing Policy Initiative in the School of Public and Environmental Affairs at Indiana University.
