Manufacturing and production have been big contributors to improved quality and sustainability of human life. Current market trends, such as consumer demand for variety, short product life cycles, high product quality and low cost, have resulted in the need for efficient, responsive, robust and sustainable manufacturing and production paradigm. 3D printing technologies hold the merit of affordability and customizability, while the key challenge in applying 3D printing for mass customization in real life is how to reduce the lead time per unit. The lead time of 3D printing a product unit comes from two sources, i.e., the pre-fabrication computation and manufacturing process. The pre-fabrication computation is increasingly significant and becomes the bottleneck in the manufacturing flow of mass customization in 3D Printing. This EArly-concept Grant for Exploratory Research (EAGER) project looks to address this problem through new computational methods with potential for two orders of magnitude reduction in time for pre-facbrication computation. This project aims to develop a transformative computational paradigm of 3D printing in mass customization. The project will pursue two novel and complementary objectives: 1) design a suite of quality-guaranteed geometric algorithms for the scalable and time-efficient pre-fabrication computation framework.; and 2) develop a low-complexity and efficient computing system to facilitate and accelerate the use of these methods and algorithms in Objective1. This new computer system focuses on domain-specific computing platforms as the next disruptive technology for power-performance-runtime efficiency improvement. Specifically, the team will develop accelerator-based architectures for computing primitives of geometric algorithms. This new hardware architecture will exploit the parallelism and customization to improve the efficiency of the new computational paradigm in 3D printing with less delay, lower complexity and higher computing power.

The evolution of manufacturing systems from loose collections of cyber and physical components into true cyber-physical systems has expanded the opportunities for cyber-attacks against manufacturing. To ensure the continued production of high-quality parts in this new environment requires the development of novel security tools that transcend both the cyber and physical worlds. Potential cyber-attacks can cause undetectable changes in a manufacturing system that can adversely affect the product's design intent, performance, quality, or perceived quality. The result of this could be financially devastating by delaying a product's launch, ruining equipment, increasing warranty costs, or losing customer trust. More importantly, these attacks pose a risk to human safety, as operators and consumers could be using faulty equipment/products. New methods for detecting and diagnosing cyber-physical attacks will be studied and evaluated through our established industrial partners. The expected results of this project will contribute significantly in further securing our nation's manufacturing infrastructure. This project establishes a new vision for manufacturing cyber-security based upon modeling and understanding the correlation between cyber events that occur in a product/process development-cycle and the physical data generated during manufacturing. Specifically, the proposed research will take advantage of this correlation to characterize the relationships between cyber-attacks, process data, product quality observations, and side-channel impacts for the purpose of attack detection and diagnosis. These process characterizations will be coupled with new manufacturing specific cyber-attack taxonomies to provide a comprehensive understanding of attack surfaces for advanced manufacturing systems and their cyber-physical manifestations in manufacturing processes. This is a fundamental missing element in the manufacturing cyber-security body of knowledge. Finally, new forensic techniques, based on constraint optimization and machine learning, will be researched to differentiate process changes indicative of cyber-attacks from common variations in manufacturing due to inherent system variability.

Additive Manufacturing holds the promise of revolutionizing manufacturing. One important trend is the emergence of cyber additive manufacturing communities for innovative design and fabrication. However, due to variations in materials and processes, design and computational algorithms currently have limited adaptability and scalability across different additive manufacturing systems. This award will establish the scientific foundation and engineering principles needed to achieve adaptability, extensibility, and system scalability in cyber-physical additive manufacturing systems, resulting in high efficiency and accuracy fabrication. The research will facilitate the evolution of existing isolated and loosely-connected additive manufacturing facilities into fully functioning cyber-physical additive manufacturing systems with increased capabilities. The application-based, smart interfacing infrastructure will complement existing cyber additive communities and enhance partnerships between academia, industry, and the general public. The research will contribute to the technology and engineering of Cyber-physical Systems and the economic competitiveness of US manufacturing. This interdisciplinary research will generate new curricular materials and help educate a new generation of cybermanufacturing workforce. The research will establish smart and dynamic system calibration methods and algorithms through deep learning that will enable high-confidence and interoperable cyber-physical additive manufacturing systems. The dynamic calibration and re-calibration algorithms will provide a smart interfacing layer of infrastructure between design models and physical additive manufacturing systems. Specific research tasks include: (1) Establishing smart and fast calibration algorithms to make physical additive manufacturing machines adaptable to design models; (2) Deriving prescriptive compensation algorithms to achieve extensible design models; (3) Dynamic recalibration through deep learning for improved predictive modeling and compensation; and (4) Developing a smart calibration server and APP prototype test bed for scalable additive cyberinfrastructures.

Cyber-physical systems of the near future will collaborate with humans. Such cognitive systems will need to understand what the humans are doing. They will need to interpret human action in real-time and predict the humans' immediate intention in complex, noisy and cluttered environments. This proposal puts forward a new architecture for cognitive cyber-physical systems that can understand complex human activities, and focuses specifically on manipulation activities. The proposed architecture, motivated by biological perception and control, consists of three layers. At the bottom layer are vision processes that detect, recognize and track humans, their body parts, objects, tools, and object geometry. The middle layer contains symbolic models of the human activity, and it assembles through a grammatical description the recognized signal components of the previous layer into a representation of the ongoing activity. Finally, at the top layer is the cognitive control, which decides which parts of the scene will be processed next and which algorithms will be applied where. It modulates the vision processes by fetching additional knowledge when needed, and directs the attention by controlling the active vision system to direct its sensors to specific places. Thus, the bottom layer is the perception, the middle layer is the cognition, and the top layer is the control. All layers have access to a knowledge base, built in offline processes, which contains the semantics about the actions. The feasibility of the approach will be demonstrated through the development of a smart manufacturing system, called MONA LISA, which assists humans in assembly tasks. This system will monitor humans as they perform assembly task. It will recognize the assembly action and determine whether it is correct and will communicate to the human possible errors and suggest ways to proceed. The system will have advanced visual sensing and perception; action understanding grounded in robotics and human studies; semantic and procedural-like memory and reasoning, and a control module linking high-level reasoning and low-level perception for real time, reactive and proactive engagement with the human assembler. The proposed work will bring new tools and methodology to the areas of sensor networks and robotics and is applicable, besides smart manufacturing, to a large variety of sectors and applications. Being able to analyze human behavior using vision sensors will have impact on many sectors, ranging from healthcare and advanced driver assistance to human robot collaboration. The project will also catalyze K-12 outreach, new courseware (undergraduate and graduate), publication and open-source software.

Cyber-physical additive layer manufacturing, e.g., 3D printing, has become a promising technology for providing cost, time, and space effective solution by reducing the gap between designers and manufacturers. However, the concern for the protection of intellectual property is arising in conjunction with the capabilities of supporting massive innovative designs and rapid prototyping. Intellectual property in the additive layer manufacturing system consists of: i) geometric design of an object; ii) attributes of an object; iii) process information; and iv) machine information. This Early-concept Grant for Exploratory Research (EAGER) project seeks to develop defense mechanisms for detecting malware and counterfeit articles using a variety of signals that are observed during the manufacturing process including acoustic, temperature, power, and others. The project is an EAGER because both the uniqueness of the observed signal signatures, and their utilization in securing the manufacturing process are high risk with potential for high reward in thwarting attacks. This project will demonstrate that during the life-cycle of the additive layer manufacturing system, the intellectual property information contained in the cyber domain can be recovered/reconstructed through attacks occurring during the manufacturing process in the physical domain through various non-intrusive techniques. It will then focus on creating both machine-dependent and machine-independent defense mechanisms for avoiding such an attack. This project will significantly impact US competitiveness over technology-oriented manufacturing. The attack model will provide feedback to 3D printer manufacturers and CAD tool designers to build defenses against these new types of attack. Moreover, it will have a significant societal impact to the explosively growing maker and crowd-sourcing community in protecting their intellectual property. In addition, the project's approach can be used in other manufacturing systems, e.g., CNC machines, manufacturing robots, etc. This is possibly the very first approach to create defense for additive layer manufacturing mechanisms against such attacks occurring in the physical domain to get access to information of the cyber domain. This project has three specific objectives: 1) It will demonstrate a proof of concept by presenting a novel attack model constructed using a combination of machine learning, signal processing, and pattern recognition techniques that utilize the side-channel information (power, temperature, acoustic, electromagnetic emission) obtained during the manufacturing process. 2) It will develop a machine-specific defense mechanism against the attack model for the 3D printer. New techniques to add additional physical process encryption, e.g. adding extra information to the G-code to obfuscate the printing process from the attack model between the G-code and the physical manufacturing process, will be demonstrated. 3) It will create a new security-aware 3D-printing algorithm for the machine-independent CAD tools that can protect against such side channel attacks. The 3D-printing algorithm will slice the STL and generate layer description language (e.g. G-code) randomly so that for the same 3D object, different instructions will be sent to the 3D printer and eventually different physical features will be extracted by the attackers.