In This Issue
Winter Issue of The Bridge on Complex Unifiable Systems
December 15, 2020 Volume 50 Issue 4
The articles in this issue are a first step toward exploring the notion of unifiability, not merely as an engineering ethos but also as a broader cultural responsibility.

Cyberphysical Integration

Friday, December 18, 2020

Author: Chandrakant D. Patel

In the 19th and early 20th centuries engineering was about the industrialization of physical and electromechanical systems like the steam engine and utility grid. The latter half of the 20th century was about information management, cybersystems, and the internet. The 21st century is about the integration of the two and the proliferation of cyberphysical systems that address challenges stemming from global social, economic, and ecological trends such as resource constraints, demographic shifts, and human capital constraints.

Cyberphysical Solutions

The burden of negative externalities, from pandemics to environmental problems, is accelerating the need for cyberphysical systems. Solutions in the cyberphysical era will involve operating technologies (OT) and information technologies (IT) that function at the intersection of domain theories, data organization, and data science.

Digital manufacturing is an example of a cyberphysical solution in which the OT are fabrication devices, power delivery systems, and other support systems, and the IT are integrated with the OT to collect data, perform analysis, and drive automated management of manufacturing. In 3D digital manufacturing, the OT is made up of 3D printers and the “art-to-part” cyberphysical pipeline starts with part design anywhere in the world, transfer of digital data to print parts in a given locality, and delivery to the customer. The need-based provisioning capacity of 3D digital manufacturing has shown that it can enable resilient supply chains.

Indeed, 3D printing has created new realities for on-demand creation of customized and individualized parts. This was recently realized in the production of personalized protective equipment, medical devices, and isolation rooms in response to the covid-19 pandemic. Workflow for a personal protection mask can start by scanning an individual’s face, creating the “art” for custom-fitted design, and sending the design file to a local digital manufacturing site to produce the piece. Moreover, it is conceivable that 3D digital manufacturing units are distributed, resulting in microgrids of digital manufacturing supply near the sources of demand, such as in a hospital.

Design, Device, and Digital Factory

The success of 3D digital manufacturing necessitates a holistic perspective that encompasses design, device (3D printer), and digital factory (assembly of systems).

Design

Designers have immense possibilities in creating the art. With the range of additive 3D printing technologies they can create contours that were hitherto not possible in a cost-effective manner if at all. However, they must take into account new sources of variation in the final part outcome given the variable attributes of a given 3D printing technology and even a given printer.

Device

The 3D drawing digital “slices” enter the device (the 3D printing system), which uses a collection of sensors and actuators to additively build parts.

In the case of HP’s 3D Multi Jet Fusion printers, the slices of drawing are converted by a controller to signals that drive tens of thousands of thermal inkjet technology–based nozzles in a writing system that traverses the length of the raw material powder bed. The nozzles dispense picoliter-scale drops of fusing and detailing agents at thousands of hertz, layer by layer, on to the powder bed. Fusion heaters apply heat energy that is absorbed in the regions where fusing agent has been dispensed to create disparate solid bodies.

Precise control of the raw material positioning, writing system, and heaters in synchronization is critical to ensure that the right amount of fusing agent is provisioned and heat energy is applied proportionately in exactly the correct place. This makes for a complex multiple-input, multiple-output (MIMO) system that cannot be formulaically represented by a domain-based model alone. The system design uses domain theories, machine-generated data, and artificial intelligence (AI) algorithms to create a “digital twin” that is tuned to predict the successful operation of the machine.

The 3D printer cannot be treated as a black box with the assumption that large amounts of data and AI alone will create the digital twin. Deep domain understanding of engineering systems is fundamental to a successful model. As an example, computer vision algorithms applied with domain theories associated with the -powder bed can enable a machine vision system to discern the attributes of the powder bed and thereby control the deposition of the agents and heat flux from heaters in real time during production.

Digital Factory

The digital factory is a system of systems built with OT and IT. The heterogeneous mix of 3D printers is simulated with digital twins during the layout and design of the factory to allow for holistic factory optimization, minimizing operating costs while maximizing production output. For example, a digital factory with 100 3D printers draws more than 1 megawatt of power and thus requires appropriate power distribution system design.

The factory produces parts on demand based on customer service level agreements (SLAs), which spell out customization and engineering requirements given the application and the turnaround time. The SLAs are turned into service level objectives and applied to the pool of machines, managed via sophisticated enterprise resource management and manufacturing execution systems.

A communication layer collects machine-generated data from rich sensing subsystems (e.g., video sensors, actuators). In the future the communication layer for the 3D printers will be a 5G mm scale wavelength network, and a local microdata center (Cloud 2.0) may be used for storing and analyzing terabyte-scale data. Onsite computing will enable real-time analysis and action. Metadata and insights (e.g., tweets) will go to the current Cloud 1.0 for global coordination, which will facilitate other management such as software updates.

Digital factories can be further expanded for net zero operations with local power grids built using multiple sources of energy, creating a distributed network of such factories to meet growth in demand for need-based provisioning.

Cyberphysical Workforce Needs

Secure 3D printing suggests an exciting future in complex, unifiable cyberphysical systems. Multidisciplinary systemic instantiations show the impending need for cyberphysical systems professionals who operate at the intersection of domain, data, and AI. This is crucial given a prevalent, and often cavalier, view—particularly in light of AI success in ecommerce and social media—that any complex physical problem can be solved with data and AI alone.

AI for the 3D “art-to-part” pipeline is about detecting anomalies, prognostics, diagnostics, and closed-loop action to drive efficient operations of physical systems. It is not the same as cyber age examples of recognizing a cat or dog from pictures, recommending restaurants, or determining ad placement in social media based on an individual’s profile. The “black box” around a cyberphysical system is transparent.

Cyberphysical contributors must have depth in engineering fundamentals of the machine age and breadth in information sciences of the cyber age. This can be achieved through a variety of learning paths such as dual degrees and continuing education. Above all, these contributors must have a “learn by doing” aptitude with T-shaped knowledge to build and integrate systems.

About the Author:Chandrakant Patel (NAE) is chief engineer and senior fellow at HP, Inc.