Manufacturing Engineering: PARC, a Xerox company, is involved in developing new concepts for digital design and manufacturing. Describe how new representations and algorithms are helping to design, analyze and plan the manufacture of highly complex structures via new technologies.
Ersin Uzun: Advancements in manufacturing and materials technology are setting the stage to create a future of product design where non-intuitive combinations of shape and material layout can be designed and fabricated to produce high-performance functional parts at comparable cost. But design freedom and complexity are two sides of the same coin. To truly enable a future where such complexity can be harnessed by design tools, we need to rethink the foundations of CAD/CAM/CAE (CAx) systems. Today’s CAx systems are very good at representing shapes with homogeneous, isotropic material distribution.
This approach works well for parts made with traditional manufacturing processes but will not scale well to represent, for example, 3D printed parts with multiple materials, or organic structures such as human bones. We are working on a novel CAx system that represents manufacturable shapes as 3D images called fields at extremely high addressable resolution, so that microstructural details may be captured within a bulk shape without incurring a corresponding memory blowup.
Also, we believe that design, manufacturing planning and engineering analysis cannot be broken into separate software systems because a) they introduce serious interoperability problems and b) manufacturing strategies affect fabricated material properties and therefore the design.
These two critical issues are addressed with our novel representation of views in a computational design workflow. The as-designed, as-manufactured and as-analyzed views capture the information essential for a stakeholder operating in that view (e.g. a designer may care about representing the shape and material complexity while balancing performance requirements, but a manufacturer may think about how best to represent the shape in terms of their machine and tooling capabilities).
The view centric approach allows us to re-cast the interoperability problem from one of data transfer (which requires converting representations leading to numerical inaccuracies) to that of information transfer and exchange. We’ve had some very interesting results using this approach.
ME: Can this type of technology also be used in traditional manufacturing processes, such as subtractive machining?
Uzun: Yes. Indeed we started with automating the process planning of subtractive machining in our uFab technology. However, we moved later to representations and algorithms that can support new complex materials, designing and simulations at multiple size scales and both additive and subtractive processes.
We believe the future of manufacturing is neither just subtractive or additive, but a seamless hybrid of the two.
ME: The Cyber-Physical Systems group is focused on helping manufacturing clients develop complex new products that incorporate both digital and physical components. What is an example of this type of “hybrid” component and how does PARC assist in the manufacturing process?
Uzun: The term ‘hybrid’ may be overloaded, and can refer to heterogeneity in the manufacturing process or in the manufactured artifact. We are interested in both interpretations. Hybrid machines (additive and subtractive manufacturing in the same machine) such as the Mazak Integrex i-400 AM allow the freedom to 3D print metal on top of curved surfaces and then machine these printed features to impart CNC-like surface finish. One can then imagine the next generation of functional mechanical parts enabled by this process.
In other words, it may be too expensive or simply impossible to make these artifacts any other way. When the additive manufacturing component can fabricate multiple materials (e.g. when heterogeneous metal powder is used in a laser sintering process) the resulting artifact will have composite and even graded material layout and properties.
PARC has also developed a 3D printer that utilizes “inks” or slurries of metal and ceramic inks as its feedstock, where the relative composition of any voxel is controlled by a micro-mixer, and another printer where electronic interconnects can be printed along with plastic features possibly leading to objects with integrated electronics printed by a 3D printer.
ME: What are the software tools and algorithms the Cyber-Physical group has developed to create digital twins of their designs?
Uzun: Ever increasing complexity and new types of components make it expensive to build accurate models of cyber physical systems (CPS), resulting in design faults, delays and unanticipated interactions. Typically, system models are incomplete (models for subcomponents may not be known or available) and their behavior needs to be inferred to perform any realistic physical analysis or design. Training deep neural networks (DNNs) to infer behavior is predicated on implicit assumptions of large data sets, spanning multiple system modalities.
Such assumptions are rarely accurate for physical systems (i.e., systems are more reliable and do not fail often so very little failure data is typically available). Moreover, such models are not suitable for every application. For example, in a vehicle, the transmission acts as a bidirectional conduit for energy depending on the mode of operation. A trained DNN to represent the transmission when driving uphill (the motor provides input power) is not applicable when driving downhill (the wheels provide input power, and the engine dissipates energy together with the brakes). In other words, the DNN is not reversible.
Our research has uncovered approaches where a model of a physical system may now be driven both by physics-based models that describe the nominal expected behavior of the CPS, and data driven models extracted from the operating physical counterpart. This modeling approach results in a refined representation of the digital twin as an ongoing computation where the algorithms take advantage of both model based and data driven AI.
We have helped our partners incorporate incomplete physics-based models and limited data to generate functional digital-twins that can be effectively used for diagnosis, prognosis, simulation and optimization purposes.
ME: How can digital twins be used to analyze manufacturing outcomes and improve product performance?
Uzun: Digital models of both the manufacturing process and the manufactured artifact must be used to create an accurate representation of manufacturing outcomes. This involves the integration of fabrication planning along with engineering design and analysis so that performance of the fabricated part may be analyzed in the presence of manufacturing imperfections.
For example, our research in automated planning for hybrid manufacturing can generate valid process plans (if possible) that are within manufacturing limitations. To execute the process, we must then convert these plans into machine executable instructions (e.g. G-code) that will actually control the motors that drive the material addition/removal. During the manufacturing process, it is possible to imagine an online control system that continuously analyzes the structural details being fabricated and adaptively controls the fabrication ‘in-situ’ to minimize manufacturing discrepancy from designed intent.
All of these operations (planning, execution, control) can and should have hybrid models in a digital twin of the fabrication process that use both model-based AI (e.g for planning) and/or data driven AI (e.g. for the online control). This will then allow us to provide immediate design feedback on manufacturability, cost and product performance across all views in the design workflow, eventually resulting in a digital twin of the fabricated product.
Siemens PLM Software (Plano, TX) has announced, with Bentley Systems, an integrated solution for enterprises to deliver capital projects more efficiently, combining the Teamcenter portfolio with Bentley’s ProjectWise and its Connected Data Environment (CDE). Teamcenter is a product lifecycle management (PLM) system, and ProjectWise is a project delivery collaboration platform. The new offering continues Siemens’ and Bentley’s strategic alliance that was announced in 2016, and will extend enterprise visibility across program management of capital project engineering and construction.
The leading capabilities for systems engineering and requirements management within Teamcenter, and lifecycle simulation of engineered components, are now complemented by Bentley’s CDE to take advantage of a project digital twin. Project digital twins automate digital alignment and change synchronization across the project supply chain, enabling continuous and comprehensive status reviews.
Flexera (Itasca, IL), a software developer, and KPMG LLP (New York), the U.S. audit, tax and advisory firm, have announced an expanded strategic alliance to help development, legal and security teams with open source license and security management. FlexNet Code Insight, Flexera’s open source security and compliance platform, is now the technology behind KPMG’s software composition analysis offering, which helps clients detect open source license compliance risk and security vulnerabilities.
KPMG’s new software composition analysis, featuring Flexera’s FlexNet Code Insight, helps suppliers and buyers of software and Internet of Things solutions uncover OSS risks through regular monitoring that reviews soft-ware developed, used and distributed by an organization. The service results in a detailed Software Bill of Materials (BOM) that defines the organization’s OSS footprint, vulnerabilities that need to be patched and licensing risks that require action.
Fast, high-level scans can be performed to identify critical problems, and deeper scans and analysis are available on high-risk code. The analysis also helps companies navigating technical due diligence because of a pending merger or acquisition.
CADWorx & Analysis Solutions (Houston), a unit of Hexagon, has released GT STRUDL 2018 R1, a solution for structural analysis and design. GT STRUDL 2018 R1 offers integrated, database-driven software for beam and general finite element analysis and comprehensive structural engineering design. The solution includes 10 functional areas that operate seamlessly with one another.
GT STRUDL 2018 R1 features numerous enhancements, including new capabilities that allow the user to place, edit, split and copy physical members and display graphics and labels, apply loads and set parameters for these members while maintaining mapping between the CADWorx Structure/Intergraph Smart 3D and GT STRUDL models.