The Next Industrial Revolution
Think for a minute about how the brain approaches a new task and learns how to master it. Let’s say you’ve never baked a cake, but you want to make one for a special occasion. What do you do?
Today, the first step might be to go online and look for how-to videos. Each time you watch one, your brain registers the steps in the process and begins to understand how they fit together. Watch enough videos and you’ll internalize that data and be able to proceed on your own.
The underlying premise of Industry 4.0 is quite similar, only on an exponentially larger, and unimaginably faster, scale. Imagine, every hour, thousands of bakers making millions of cakes. Consider each step in each baker’s recipe as a data point. How would you manage such a huge accumulation of information, and more importantly, how would you use it to improve the process?
These questions are especially relevant in the world of additive manufacturing (AM). This extremely data-intensive process builds three-dimensional objects by joining successive layers of material, one atop the other, through direct conversion of three-dimensional computer-aided design, or CAD, files that contain all the necessary design parameters.
Following is a look at the work of ASTM International’s committee on additive manufacturing technologies (F42), which is at the forefront of developing globally relevant AM standards that are key to the advancement of Industry 4.0. The committee has developed more than 25 critical AM standards since its formation in 2009, and the recent formation of a subcommittee (F42.08) specifically focused on AM data will move the I4.0 needle even further.
What Is Industry 4.0?
Industry 4.0 (I4.0 for short) is the term used to define the process of integrating digital technologies into physical manufacturing. Boiled down to its essence, I4.0 involves collecting the vast amounts of data generated by modern automated manufacturing systems; evaluating the data for patterns that may reveal insights into problems or better ways of doing things; and incorporating these solutions into the manufacturing process as part of both real-time decision-making and long-term quality-control analysis.
“Industry 4.0 is digital transformation,” says Alex Liu, Ph.D., head of ASTM’s additive manufacturing programs in the Asia-Pacific region. “It’s where cyber infrastructures such as artificial intelligence [AI] and machine learning [ML] collaborate and integrate with physical infrastructures such as additive manufacturing and robotics automation. It is the digitization of manufacturing.”
“Fundamentally, Industry 4.0 is about connecting machines, work cells, and factory floors to the information infrastructure within which decisions are made,” says David Rosen, Ph.D., a mechanical engineering professor at the Georgia Institute of Technology and chair of the AM subcommittee on design (F42.04).
The underlying digital ecosystem that enables the collection of this data is the so-called Industrial Internet of Things, or IIoT. Similar to the Internet of Things — smart appliances, for example, which can be controlled remotely by utilities via the internet — but more complex in scope, the IIoT refers to machines and systems that are connected to and communicate with one another via the Internet.
“AM machines are just another type of machine tool that will be connected,” Rosen says.
“In the case of AM, this technology is closely connected to other components of Industry 4.0 such as robotics and automation, big data, cybersecurity, and AI. The fusion of all these technologies offers the biggest opportunity, which is the extraction of new insights and uncovering patterns,” says Mahdi Jamshidinia, Ph.D., who is the additive manufacturing R&D project manager at ASTM.
It’s this connectivity, plus artificial intelligence algorithms that discern patterns (and anomalies) in the data flow, that could bring transformative benefits to the AM industry. “For example, machine learning is used to enable the rapid selection of optimal AM feedstock materials and can help to predict functionalities of AM parts based on multiple design parameters,” explains Mohsen Seifi, Ph.D., director of ASTM’s global additive manufacturing programs. “AI can also perform real-time AM manufacturing process monitoring and analyze root causes of production issues.”
Seifi also points out that artificial intelligence can be incorporated into AM systems via the IIoT. “For example, you could put a bar code or tag on each AM part and use the IIoT to track its performance and quality. Then, the data gathered from those sensors could undergo analysis to uncover trends or sift out other useful information.”
Additive Manufacturing and I4.0
The basic principles of I4.0 are applicable in most modern manufacturing environments, which employ digital sensors and controls to varying degrees. However, AM is a particularly instructive “laboratory” for testing and refining these principles because the amount of data collected, even in a small timeframe, is staggering.
“AM generates huge amounts of data — from AM feedstock, design, modeling, and manufacturing processes to the post-processing, inspection, testing, and ultimately performance of manufactured parts,” notes Jamshidinia. Just how huge? According to Georgia Tech’s Rosen, each minute of machine operation can translate into gigabytes of sensor data.
One source of all these data is the painstaking, layer-by-layer process that defines additive manufacturing (other data sources include product lifecycle and value chain activities). Each individual object produced via AM can comprise thousands of layers. Sensors capture data for each layer, conducting analyses and evaluating if it has been processed correctly.
These sensors can also be connected to the IIoT, providing access to the computing power needed to put all the collected information to good use. “There is a great opportunity to employ these data to accelerate the AM field through the application of Industry 4.0 disciplines such as artificial intelligence and big data analytics,” Seifi says.
The promise of I4.0 in the context of additive manufacturing is the ever-faster, ever-more-accurate automation of the analyses and evaluation processes for the accumulated data. “Full inspection of each single part is laborious, time consuming, and expensive. But with enough critical datasets, machine learning can significantly streamline the required inspection process by identifying suspicious regions of each part while ensuring quality components are produced,” Liu states.
Managing the Data
In view of the importance of data in AM, ASTM International’s Additive Manufacturing Center of Excellence (AM CoE), in collaboration with America Makes, organized the AM Data Management and Schema Workshop last December. This event brought together more than 90 experts from government, industry, and academia to cover recent AM data-management and data-enabled applications, and to discuss gaps, challenges, and potential solutions for AM data. One of the workshop outcomes was the formation of a stand-alone subcommittee for AM data (F42.08). “Moving forward, standards covering the compilation, organization, and dissemination of AM datasets will fall under the auspices of this new subcommittee,” says Jamshidinia.
Managing this gusher of information presents a number of challenges. First is the very nature of the data itself. Data subcommittee member and U.S. National Institute of Standards and Technology (NIST) scientist Yan Lu, Ph.D., explains. “Multi-modal data are generated in the AM product development lifecycle, including 3D models, in-situ measurement, and post-inspection data. Each of the data types represents spatial information of the part in different reference frames.”
Aligning these data types properly within a common reference frame is called data registration, an approach that allows relationships between the AM process, the properties of the materials being used, and the object itself to be correlated and interpreted. Delving a little deeper into AM technology shows the complexity of the alignment process.
For example, there are currently no fewer than seven distinct AM categories established by the document, standard terminology for additive manufacturing – general principles – terminology (ISO/ASTM52900): binder jetting, directed energy deposition, material extrusion, material jetting, powder bed fusion, sheet lamination, and vat photopolymerization.
Then there are the four basic types of materials used in additive manufacturing — ceramics, metals, polymers, and composites — and multiple options within each material type. Polymers can include ABS (acrylonitrile butadiene styrene), polycarbonate, polyamide, and epoxy resin. In the metals category, different alloys of steel, titanium, and aluminum are used.
READ MORE: The Manufacturing Supply Chain
Finally, an almost limitless range of products can be manufactured using AM, with all the variations in size and dimensions that such a blank canvas implies. Whether it’s rocket engine components, orthopedic implant devices, detailed architectural models, or after-market parts for vintage cars, AM is enabling the production of objects that simply cannot be made any other way.
Given all these factors, it becomes easier to appreciate the difficulty of aligning all the data related to AM processes, materials, and specific parts. “Currently, most data generated from AM activities are collected and stored in an ad hoc fashion — for example, in PDFs or house-made spreadsheets — which prevents their use for decision-making,” says NIST’s Yan Lu, who is also vice chair of the data subcommittee.
Alex Kitt, Ph.D., data subcommittee chair and product manager with engineering consultant EWI, echoes Lu’s comments regarding inefficient and inconsistent data acquisition. “Often, the person making the measurement will bring a USB drive to the data scientist, who then enters the data into the system,” Kitt says. “This discourages the acquisition of any data that are not critical.”
Lu and her colleagues believe standards are needed to address key questions such as how to collect, store, and handle data securely, and what formats should be used to share data among various industry stakeholders related to product lifecycle and supply chain management. “Standard AM information models and data change formats are necessary to share and make full use of the data generated from various stakeholders,” she says.
As manufacturing goes digital and processes become more modular, additive manufacturing requires sharing information among multiple stakeholders in the production chain, from design to end-users. “This requires robust cybersecurity so that data can be transferred securely in the cloud-computing environment,” says Seifi.
The first order of business for the data subcommittee is the creation of a common data dictionary (CDD) for additive manufacturing. Lu says, “NIST has been leading the development of the CDD since the fall of 2018 through an ad hoc working group with about 50 experts from industry, government, and academia.” Stakeholders represented in this group include NIST, EWI, Penn State University, Granta MI (a developer of materials information management systems), and various government groups.
According to Lu, the draft CDD version 1.0 is under review now, and parts of it will be incorporated into ASTM International standards. The next steps are to work on a common additive manufacturing data model, data exchange formats, and an AM data registration standard.
Another subcommittee focus area will be AM technologies that are driven by artificial intelligence, including machine learning, deep learning, and data mining. Seifi notes that, though rare, AI is already being used in additive manufacturing.
“For example, the Alchemite ML algorithm was used to design a nickel-based alloy for direct laser deposition,” he says. “Also, 3D analysis and reconstruction of additively manufactured materials have been done in a cloud-based infrastructure. Without standards to establish consensus-based practices for the use of artificial intelligence, it is difficult for organizations to repeat, build on, or expand these efforts at meaningful scales.”
Meanwhile, the AM CoE is developing a strategic guide for data in AM based on the input of various sectors. “This guide will be provided to the F42.08 subcommittee and AM community as a reference. It will allow stakeholders to identify existing gaps and challenges, and offer potential future solutions to improve data management and use through the timely development of related standards,” Jamshidinia says.
Cross-committee collaboration is a hallmark of ASTM’s standards development model. In the case of the additive manufacturing technologies committee (F42) and its new data subcommittee, the plan is to work with other committees that are addressing artificial intelligence and IoT/IIoT, including the committees on driverless automatic guided industrial vehicles (F45), unmanned aircraft systems (F38), consumer products (F15), and medical and surgical materials and devices (F04).
While autonomous forklifts, drones, and smart baby monitors provide opportunities to explore the intersection of data, connectivity, and performance, additive manufacturing takes these relationships to a new level. The work of the subcommittee on data will undoubtedly contribute to this upward trajectory.
“AM is the first new manufacturing process developed in the age of data,” concludes Alex Kitt. “However, ensuring that data are understandable, of high quality, and usable is hard. The new subcommittee will be developing the standards so that the community can build a data ecosystem in which data can easily be acquired, managed, and used.”
Jack Maxwell is a freelance writer based in Westmont, New Jersey.