ML, AI, and the Smart Manufacturing of Smart Products
The IoT opens the door for smart manufacturing, but it has also created the need for vigorous product testing and security measures.
The IoT (Internet of Things) is dramatically changing how industries work. In the past decade alone, one of the defining trends in the global industrial sector has been the push for digital transformation and Industry 4.0. Manufacturing, in particular, is undergoing mass change as IoT technologies become more affordable, more accessible, and more accepted as the way of the future.
By 2020, BI Intelligence projects the installed base of manufacturing IoT devices to reach 923 million, and this investment in the IoT by manufacturers will translate to billions of dollars in spending globally. Spending will also ramp up as manufacturers graduate from more simplistic IoT solutions to more complex ones.
One example of this comes from AT&T, which announced earlier this year its goal to achieve “Zero Waste” at 100 facilities by the end of 2020. Some of the strategies include reducing waste and increasing recycling and composting.
Across the company, it is also working to make its network, fleet, and operations more efficient as well. Further, it is helping its customers reduce carbon emissions and use the IoT for Good, by connecting everything from trucks, to farm equipment, to city infrastructure, to manufacturing, and more.
Industrial automation through the use of technologies such as ML (machine learning) and AI (artificial intelligence) will be an important driver of digital transformation in manufacturing. Factories that were once defined by a complex system of buildings, machinery, and human workers are now being defined by automation and intelligence. In smart factories, industrial robots are replacing “boots on the ground” as AI-enabled systems become capable of managing an increasing number of core operations.
As they do, manufacturers will benefit from increased efficiencies resulting from streamlined processes and fewer errors. The growing demand for visibility throughout the manufacturing process, as well as safe, well-tested products will continue to push the space forward in the transition to Industry 4.0, but barriers to IoT implementation—including cybersecurity risks, difficulty determining ROI (return on investment), technical integration, and the threat of job loss—remain.
The Peggy Smedley Show:
AT&T Business Summit
AT&T Business Summit
Peggy Smedley is joined by Chris Penrose, president, Internet of Things, AT&T Business, and James Brehm, founder and chief technology evangelist, James Brehm & Associates, who discuss the verticals seeing the most IoT (Internet of Things) penetration. They also discuss the AT&T Foundry and why it is important to brainstorm and change how companies operate businesses. Finally, they talk about the trends going forward, including the fact that every business will change the way they operate by introducing IoT technologies.
Machine Learning and Artificial Learning
Thorsten Wuest, assistant professor of smart manufacturing at West Virginia University, says data analytics, ML, and AI are key to realizing smart manufacturing and the concept of Industry 4.0. “The ability not just to collect (large) amounts of manufacturing data but also to effectively and efficiently analyze it to learn previously unknown insights in the processes can not be valued high enough,” Wuest says. “And this is just one application where AI and ML are deemed valuable. In the future, we will see more automated design adaptation, again stressing the personalized product that essentially requires ‘batch size 1’ and ‘mass personalization’.”
Jagannath Rao, head of data-driven services at Siemens, says the biggest game-changing aspects of the IIoT (industrial IoT) include the ability to integrate disparate data sources in a manufacturing plant with ease and, using the latest sensor technologies, the ability to measure almost anything in an unobtrusive way.
“The combination enables the generation of large datasets with diverse features, and to these we can even add external data sources like weather, logistics, dynamic electricity pricing, etc.,” Rao says. “As a result, we can now apply cutting-edge modern technologies like machine learning, deep learning, AR (augmented reality), visual analytics, etc., on these datasets and build far more real models of any predictive aspect of manufacturing and process optimization.
This not only enables accurate decisionmaking but also reduces the response time to minutes and hours compared to traditional methods.”
What’s more, these technologies create the opportunity to gather insights into the entire manufacturing value chain.
Peggy Smedley sits down for a conversation with Michele Perchonok, PhD, about the science of food, family, and keeping astronauts eating healthy. Perchonok details her journey from Brown University, to NASA, and now IFT (Institute of Food Technologists). Coming from a family of foodies, she shares the greatest advice she ever received from her father, what she is most proud of today, and the importance of mentoring.
Running algorithms to find clusters in a dataset can indicate trends or process deficiencies, says Rao, thereby enabling changes in operations that can lead to productivity gains and/or reduced operations costs.
“The ability to monitor and analyze in near-realtime also enables businesses to embark on new business models which are more output based and customer centric,” he adds.
Becoming more customer centric and leveraging data to increase profitability are two areas of gain with the successful implementation of smart manufacturing technologies. And what makes smart manufacturing “smart”?
Machine learning and AI are in many ways responsible; for instance, these technologies can help increase the efficiency and, therefore, the speed of manufacturing, which can create ripple effects like lower labor costs and reduced equipment downtime.
AI and ML Bust Open Manufacturing
Artificial intelligence and machine learning are having a considerable impact on the changing tide in manufacturing. The change is so impressive it is actually playing a role in many industries. Its impact is so impressive that for this column I am going to dig into the difference between AI (artificial intelligence) and machine learning.
The Tech Battle for First Responders
This week has us all taking a moment to pause and remember the deadly attacks that brought America to its knees. Al-Qaeda terrorists hijacked four planes carrying innocent passengers, and two of them were flown into the two towers of the World Trade Center in New York, which toppled to the ground, leaving devastation in their wake.
Cyber Still a Threat to Manufacturers
Security is such a huge consideration for any business undergoing a digital transformation, and this is true for manufacturers that are automating their factories. The threat of hacks and breaches is real, but it’s not a good enough reason to not invest in the IoT (Internet of Things).
Manufacturing and Servitization
For manufacturers, a digital transformation often means looking for new ways to leverage technology and IoT (Internet of Things)-derived data to generate revenue.
Consumers, too, benefit from smart manufacturing thanks to improved product quality and consistency that comes from closely monitored and tightly controlled manufacturing processes. With ML and AI, for example, sensor data from various stages of a product’s manufacture can be monitored and analyzed to identify product quality issues and quickly correct them.
“In the future, we will see more automated design adaptation, again stressing the personalized product that essentially requires ‘batch size 1’ and ‘mass personalization’.” –Thorsten Wuest, West Virginia University
What Makes Up Smart Manufacturing
IT Based Production Management
3-D Printing/Additive Manufacturing
Data Sharing Systems
Energy Saving/Energy Efficiency
Law and Regulations
Innovative Education and Training
Source: Institute of Mechanical Engineers
Product Testing and Cybersecurity
Cheryl Ajluni, IoT solutions lead at Keysight Technologies, a manufacturer of electronics test and measurement equipment and software, says thanks to the IoT, manufacturing is evolving from large-scale production to a MaaS (manufacturing-as-a-service) model, with technologies like 3D printing enabling mass customization. Amidst the myriad benefits of IoT adoption, however, is an intensified need to ensure connected device security.
“As (the) IoT is rapidly making its way into our everyday lives, the importance of security testing becomes even more critical, especially during manufacturing,” explains Ajluni. “In many ways, it’s the last line of defense for the device. If a security vulnerability is not identified during this stage, a product may go into operation only to be promptly attacked by cybercriminals. When that occurs, it can devastate a company and its brand.”
Connected devices are especially prone to attacks because they often rely on end users to change a password before putting them into operation as their prime security measure. Ajluni says end users tend to connect these devices to the Internet before changing their passwords and, as a result, they instantly become vulnerable to an attack. “Security in IoT devices can no longer be an afterthought or an add-on feature,” Ajluni urges. “It must be built into the design, coding, and architecture of every IoT device. Testing throughout the manufacturing process is the only way to make sure every component of the IoT device is safe and secure for consumer use.”
The IoT has also created the need for new types of product testing, since each technology built in to a connected device introduces potential complications. These complications are only avoidable through appropriate precision tests. For example, Ajluni says a radio must be added to a connected device so it can send and receive information, and the device must work within proximity of other devices that are simultaneously sending and receiving information. “Wireless standard-specific testing is therefore essential to ensure IoT devices can operate unimpeded in the real world and without impacting other devices,” she explains. “As new standards emerge with their own unique modulation and interference-avoidance techniques and different technical requirements, product testing will have to be adjusted to ensure adherence.”
The way IoT devices are tested, in some cases, may also change, especially since many IoT devices are small, and connecting to them for the purposes of test is not always possible. “For this reason, testing that takes place over the air, as opposed to via direct physical contact, has emerged as a way to ensure IoT devices are sufficiently tested for defects like missing or wrong components, solder issues, and more,” Ajluni says.
Product security and safety is for the public good, and according to Ken Modeste, director of connected technologies and cybersecurity lead at UL, an independent safety science company, the concepts of safety are being expanded with connectivity. “This means with (the) IoT, safety concepts can now include privacy, securing data, interoperability, takeover of products to perform unintended functions, and possibly economic loss,” Modeste says. “Therefore, product testing needs to consider these additional elements when focusing on the public good. Being able to provide measurable criteria to assess and evaluate interoperability and security, along with safety, becomes an important aspect of product testing.”
The product-testing ecosystem has needed to adjust to accommodate the growing need for connected-device safety. “Being able to have standards that focus on testing and performance based on certain agreed-upon cyber criteria that are measurable, repeatable, and reproducible is a main factor that now needs to be considered,” adds Modeste. “Having the criteria aligned with risk-based methodologies is on the forefront of moving the ecosystem to accept these new innovative products.”
Companies need to secure not only the products they manufacture, but also their factories, which could contain hundreds or even thousands of Internet-connected devices. Manufacturers must consider the possibility of third parties, such as competitors, accessing a network and, with it, trade secrets, like the “recipe” for certain products’ design and manufacture. West Virginia University’s Wuest brings up another concern: the possibility of a cybercriminal changing a part or product design so that it’s manufactured to spec but does not perform as intended.
“Imagine a critical part of an airplane that all of a sudden only has half the strength as originally designed due to a criminal manipulation of the CAD (computer-aided design) data,” Wuest says. “If that part then fails during operation—the plane flying—(it) can cause serious damage and cost lives.”
While the promise of smart factories is compelling, producing safe connected products and maintaining a secure factory environment are two of several challenges manufacturers may face. Many manufacturers may also grapple with integration issues as they adopt IoT systems, since they’re likely to have a number of different systems in place that are of different ages, makes, and connectivity type. Manufacturers may also be hesitate to commit to a system for which they can’t fully estimate the true cost and potential ROI.
“Another factor is the understanding of what data should be acquired through IoT sensors,” says Wuest. “This is a critical point, as on the one hand, what is not measured cannot be analyzed, but also too much sensors and data exponentially increases the effort to implement and operate. And, what we see now is the question: When can we finally ‘delete’ data again? For some processes, such as SLM (service lifecycle management), the data generated per minute is significant.”
From a social perspective, there is also the challenge of dealing with the notion that robots will replace the human workforce—a perception that can be overcome with time and effort on manufacturers’ part. “Manufacturers and employees will need to adapt to a changing workforce, for sure, but it doesn’t have to mean a loss of all human jobs,” explains Keysight Technologies’ Ajluni. “People will still be needed to do the jobs on the factory floor that robots and sensors can’t. Additionally, new types of jobs will be created as manufacturers try and figure out how best to take action from all of the data collected on the factory floor. It will fall to manufacturers to determine how best to repurpose their existing workforce and in what positions.”
Increase in Labor Productivity in an AI World
Artificial intelligence promises to significantly boost the productivity of labor in developed economies.
Percentage difference between baseline in 2035 and AI steady state in 2035
Source: Accenture and Frontier Economics
Among the challenges Siemens’ Rao recognizes for manufacturers pursuing the IoT is a lack of clear vision from leadership, and he suggests manufacturers may struggle with their digital transformations unless they have top-down buy-in. Rao recommends manufacturers embark on their IoT journey by taking baby steps. “Take a unit part of a plant, connect some assets, start becoming familiar with data transparency and the operations, have wins with low-hanging fruits, pick one high-impact problem which these technologies can solve, and implement it successfully,” he says.
And while security is a hurdle that cannot be ignored, manufacturers should not allow security concerns to stifle innovation. “The damage done by hacking into a manufacturing environment and losses it can cause can be tremendous,” Rao says. “However, that cannot be a reason for not introducing the technologies. It is therefore necessary to specify and choose every device with security in mind, it is necessary to test them for vulnerability and keep them updated all the time, it is important to follow all the required standards in setting up networks, servers, and Wi-Fi to make sure that security standards are followed.”
As the IoT transforms every industry it touches, technologies like ML and AI will help lead manufacturing into the future. Manufacturers that accept the need to not only manufacture smart things but also to be smart in their own right will make the transition more easily. While the IoT isn’t all opportunity and no risk, manufacturers must not ignore its potential or write off the IoT and all the promise that goes with it as hype. In today’s connected world, there is no “opt out” option; manufacturers must embark on a digital transformation journey, but they must do so with product safety and system security in mind.
Want to tweets about this article? Use hashtags #M2M #IoT #industrial #IIoT #manufacturing #cybersecurity #security #factory #automation #robots #robotics #ML #machinelearning #AI #artificialintelligence #analytics #data #bigdata
Predicting Elevator Maintenance
August 21, 2018
Peggy and Chris Smith, vice president, service innovation, Otis Elevator Co., talk about how big-data analytics are changing how it provides services. He says it is a transition and it is learning as it goes, but the first thing it did was collect a lot of data. When he looks at what customers want, they want elevators up and running, so it looked at data and how to solve customers’ problems. The company also identified how to create a predictive model and turn it into a proactive scheduled maintenance visit, and still do the maintenance, but during a time that the customer wants to do it. Otis has built an algorithm around predictive door maintenance, and Smith explains the IoT helps because it can provide data to the customer about how the elevator is declining in health and how it can move the needle back to a healthy state.