Open Manufacturing for Advanced Material Systems

Transcription

1 Open Manufacturing for Advanced Material Systems Michael Maher Program Manager, Defense Advanced Research Projects Agency ISABE Prior to joining DARPA, Michael Maher was Chief of the Composite and Hybrid Materials Branch and Materials Applications Branch at the U.S. Army Research Laboratory (ARL). At ARL, his focus was expediting transition of new technologies to the field by guiding them through the acquisition process and successfully placing them into the hands of U.S. warfighters. Before joining the staff at ARL, Mr. Maher held various positions in the manufacturing industry over a span of 20 years, ranging from production support engineer to factory design and product planning. Throughout his career, Mr. Maher has observed, first-hand, the increasing difficulty of implementing new technologies in materials and manufacturing processes, and he brings that experience to his current position as Program Manager for DARPA s Open Manufacturing program. 1. Introducing New Manufacturing Technologies is Hard Over the last three decades, novel manufacturing techniques and processes have developed an unfortunate legacy of high risk. Introduction of a new manufacturing technology now inevitably leads decision-makers chief engineers, plant managers, operations managers, and even certification officials to expect delayed production, defective parts, and/or cost overruns. To make matters worse, even when new manufacturing technologies are successfully implemented, engineers often expect unaccounted defects and variability in manufactured products to arise during the production program that will lead to failures in the field and/or unanticipated maintenance requirements. These concerns are not unwarranted. Problems related to the introduction of new manufacturing methods have, in fact, surfaced repeatedly over the last three decades and are responsible for today s tepid response. The net impact is well illustrated by the Joint Strike Fighter, which is largely using the same materials and detailed part manufacturing processes that were used over 25 years ago on the legacy F-22 and F-18 aircraft systems. The Air Force Research Laboratory Material and Manufacturing Directorate performed failure analyses from 2006 to 2008 and found that manufacturing variability was the root cause of approximately 45 percent of all failures 1. Three examples illustrate the recent course of events: Ballistic window system for the Mine-Resistant Ambush Protected ATV: When the Mine- Resistant Ambush Protected (MRAP) ATV was fielded, it included a new ballistic window system. The window was designed to protect U.S. warfighters from new threats emerging overseas, and its armor solution was a novel design. The design and manufacturing process was qualified and the system certified for use. However, problems soon appeared after the ballistic windows were delivered overseas: soldiers checking their vehicles in the morning frequently discovered that the MRAP windows had cracked and crazed overnight, resulting in the ATV being taken out of service for repair. An investigation revealed that the windows had been 1 Results of approximately 145 failure analyses conducted by the Systems Support Division of the Materials & Manufacturing Directorate at the Air Force Research Laboratory Wright-Patterson Air Force Base ( )

2 fabricated with an unexpected level of residual stress. As a result, the extreme overnight temperature excursions in-theater caused the windows to craze or crack. The decision was made to replace the windows on 3400 ATVs. Advanced Composite Cargo Aircraft: A second conspicuous example is the Advanced Composite Cargo Aircraft (ACCA). In the ACCA program, an aluminum-framed aircraft fuselage was converted to a unitized, bonded composite structure, which reduced the parts count by 90 percent and obviated the need for 15,000 fasteners and the labor to drill and install them during assembly of the aerodynamic surface. The structural unitization allowed the ACCA to be developed, built, and flown around the Palmdale Airport in only 19 months. However, the ACCA program was not without its challenges. First, building the large bonded skin panels required multiple attempts. Second, while trying to build test articles to support flight envelope expansion, the contractor was unable to successfully fabricate the test articles using representative production processes. As a result, the ACCA aircraft itself was grounded and has, to-date, failed to fly outside the Palmdale airspace. Therefore, although the ACCA program demonstrated the potential cost and schedule advantages of unitized bonded structure, it also highlighted that numerous producibility, process understanding, and repeatability issues remain before this type of structure can be considered for actual production applications. DDG 1000 bonded-steel-to-composite joint: A third remarkable case is a DDG 1000 bondedsteel-to-composite joint design that won a 2008 OSD Manufacturing Technology (ManTech) Achievement award. The developer met all of the requirements for a joint that would attach the upper composite deck housing to the steel hull. However, when preparations were made to implement this technology, the shipbuilder was directed to install tie-down bolts through the bonded joint because of an inherent lack of confidence in the new technology resulting in a 60 percent weight and cost impact. Again, this serves as an example of the embedded lack of trust in adopting new technology, even when it has been qualified. Such case histories illustrate why there is a lack of confidence in bringing new technology into the manufacturing environment to the point where, even if a truly beneficial capability were to appear to be in-hand, the natural instinct is to let someone else take the risk by being the first to field it. This paper describes a Department of Defense program that is developing a new approach to building confidence in new manufacturing technologies by treating process data statistically, and tying this data to material properties and defects. This is a very difficult and ambitious goal, but successfully reaching it would establish the groundwork for Rapid Qualification of manufactured parts and components. 2. Limitations in the Current Certification Building-Block Test Structure Approach Why are equipment, supplies, and finished parts that have been produced using novel manufacturing processes still being fielded with defects or problems not detected during certification testing? And why are new manufacturing technologies synonymous with cost overruns and schedule delays in the development cycle? Addressing these questions requires going back and reassessing the fundamentals the building-block test structure required for certification (Figure 1)

3 Figure 1: Current DoD qualification/certification approach does not capture impact of manufacturing variability across all size scales. Initial manufacturing steps typically involve fabricating a large quantity of small-scale samples (e.g., ,000 coupons and elements), followed by scale-up in size and geometric complexity to subcomponents, components, and, ultimately, full-scale articles. At the coupon- and element-levels, statistically valid populations of measured material property data are collected, which enables generation of A-basis and B-basis material property allowables used for the design of larger-scale structural articles. These design allowables are minima that are not measured but are statistically derived. This practice has emerged as the developmental standard for producing design allowables for new processes or materials because of the prohibitively high cost of producing a statistically valid set of data for larger-scale specimens. The process parameters used to build the test coupons are thereafter frozen, without either a full understanding of how variation in processing affects material properties or a definition of the truly acceptable process window. Validation of design allowables is subsequently performed on only a handful of subcomponents, components, and full-scale items, with each level of validation becoming more expensive and based on a smaller number of test objects. The net result is a very costly and time-consuming process, which even if successful at each step along the way can still result in failure in the field. These experiences give evidence that material properties measured on small-scale coupons do not always scale to larger, full-sized components, which calls into question the reliance upon coupon measurements as the basis for designing and manufacturing subcomponents, components, and full-scale articles. Experience shows that the building-block material and product development system can fail when the manufacturing process effects are different at various size scales and geometric complexities, and/or the manufacturing variability across all size scales is not captured or understood. When this approach fails, the consequences are significant: the material or manufacturing process under development may have to be restarted, and such unplanned iteration can require extensive retesting, with corresponding schedule delays and cost impacts. In addition, the - 3 -

4 system or application may be compromised by unexpected changes in capability, performance, or cost. In these cases, the plan to insert new manufacturing technology will likely be abandoned due to cost, schedule, or performance risks. Consider the material and processing methodology used to fabricate bonded, composite structures. The manufacturing process used to produce a composite coupon actually changes when adapted for a more complex configuration, such as a pi joint. Tooling, radii, and other manufacturing features irrelevant at the coupon level must now be taken into account, i.e., variation in thermal histories and pressure gradients. Moreover, the range of material aging characteristics is inadequately captured at the coupon level to allow robust and successful design and manufacture of larger-scale parts. This illustrates the inherent, overall limitation of traditional building-block test procedures: a failure at the subcomponent, component, or full-scale article level forces designers to work back down through the building-blocks to, possibly, a fundamental redesign of constitutive materials, processes, and/or parts. In current manufacturing canon, this is considered to be an unanticipated iteration through the building blocks that leads to unacceptable schedule delays and significant, adverse cost impact. 3. Open Manufacturing Program Approach Based on the problems associated with inserting new technologies into manufacturing and the limitations of the building-block test protocol, DARPA s Open Manufacturing program makes three assertions: 1. The critical flaw in the building-block test structure is that the effects of scale-up are not captured until subcomponent, component, and/or full-scale testing of the article is completed. Of even greater concern is that these effects may not be observed until the component or system actually enters service. Generally speaking, this is caused by stochastic process variation and inability to transfer processes from well-controlled developmental environments to the production environment. Consequently, properties of larger-scale items may not be adequately predicted based on coupon- and element-level measurements. 2. While variations in measured coupon and element properties are routinely captured, the impact of manufacturing parameters and manufacturing variability on material properties is almost never captured, analyzed, or controlled. That is, in traditional manufacturing the process windows are not constructed with an understanding of the combined impact of all processing parameters. Open Manufacturing seeks to rigorously understand these combinations of effects, identify the part s location with respect to the process window, and characterize the implications of that location not just determine whether a part is in-or-out of a particular specification. 3. To overcome these challenges and help determine and control critical manufacturing parameters, a comprehensive understanding is needed of the effect of the processing parameters and their variability at each size-scale, from coupon-level through the full-scale article

5 The Open Manufacturing program aims to fundamentally change how manufacturing variability is captured, analyzed, and understood to enable predicting and characterizing the manufactured outcome as never before resulting in a dramatically better assessment of risk associated with the manufacturing process. The overall goal of the Open Manufacturing program is extraordinarily difficult: predict location-specific probabilistic performance including the tails of the distribution functions with predetermined confidence levels. At an idealized, practical level, this is equivalent to understanding an as-manufactured part s microstructure and defects at voxel-level resolution, thereby enabling a prediction and understanding of the part s performance in-service. Broadly speaking, the program has two components, notionally illustrated in Figure 2: Figure 2: Open Manufacturing fundamentally changes how manufacturing variability is captured, analyzed and controlled Informatics-based methodology to fully parameterize and monitor the factory floor Informatics methodology is the process of capturing and understanding the manufacturing environment by fully parameterizing the factory floor. The manufacturing environment for structures especially for the manufacture of defense vehicles is often very data-rich. However, the data is often not captured, and, when it is, it is very rarely analyzed or leveraged. The Open Manufacturing informatics methodology is based on identifying, capturing, and analyzing the set of all relevant manufacturing process and material information. Open Manufacturing, however, goes one critical step further: rather than merely capturing the variability of resulting material properties, the program requires that the processing parameters be probabilistically captured in order to understand which manufacturing parameters are critical and to quantify the impact on performance. Identifying the relevant parameters for a given manufacturing process and then capturing that data is no small task. Not only must myriad parameters be disregarded, but there is usually a great deal of variability in the parameters that remain. Traditionally, those parameters are captured as discrete numbers, when there is often an inherent variability about their mean, which influences the outcome. The informatics-based methodology in Open Manufacturing requires that parameters be captured probabilistically, because the performance of an as-manufactured part can only be truly - 5 -

6 characterized in terms of probabilities, not with discrete numbers. As discussed below, understanding the statistical distributions of manufacturing parameter values is foundational to developing simulations that can be used with confidence. By contrast, propagating discrete numeric point-values of specific manufacturing parameters through a numeric manufacturing simulation typically results in a single-digit output, which does not reflect reality. The informatics-based methodology also requires new sensing approaches be developed and implemented to collect the required data to fully monitor the manufacturing process itself. In some cases, entirely new classes of manufacturing sensors and appropriate sensor technologies will need to be developed. This will require the involvement of research communities currently developing sensor technologies for other purposes, but which could be adapted to manufacturing needs. Moreover, populating the manufacturing process with sensors may necessitate fundamental changes in the basic operations on the manufacturing shop floor Probabilistic computational process models and design tools In isolation, accumulating all of the data described above in the informatics methodology would certainly inform the manufacturing environment. But application of this data to quantitatively and accurately predict the as-manufactured state of a part requires developing computational tools to simulate the process numerically. The actual data can be used as input for the simulation tools and will be useful in model validation. The objective of the manufacturing simulation tools is to capture the real variability of the manufacturing parameters as inputs to process models that probabilistically predict process variability and, using property-based models, the resulting part performance in a locationspecific, point-by-point manner. To link the manufacturing process parameters to the asmanufactured state, simulations are comprised of three basic modules: the manufacturing process model; the material microstructure model; and the material properties and performance model. The inputs for each module are probabilistic in nature; in turn, probabilistic outputs are exported to the next module in the process. Said another way, parameters characterized by probabilistic distribution functions not discrete quantities are propagated through the simulation models and characterize the final performance output. In practical terms, the measured process parameter values would be fed to process models to simulate relevant, unmeasured process parameters, such as the temperature distribution throughout a part over time. Process model outputs would feed material property models to simulate such features as micro/meso-structure, including defects. Material data would be fed to property models that simulate such characteristics as strength, ductility and fatigue performance, and their distribution throughout the manufactured part to ensure that properties in critical locations are above minimum design threshold values. As simulation calculations iteratively loop through this sequence (upper right-hand side of Figure 1), errors are compounded, and cumulative errors or uncertainties for a calculated result could become unacceptably large. The simulations can identify the primary sources of variability among the process steps and assist the material and process engineers in understanding which process parameter distribution functions require additional data and, hence, refinement

7 In the Open Manufacturing program, three broad approaches have been taken to the manufacturing process simulation models, and each has its strengths and weaknesses: (i) (ii) Ab initio, or first principles physics-based models have the advantage of producing results with essentially unlimited spatial resolution in calculating the details of micro/mesostructures and with accuracy limited only by the quality of the input data available for verification and validation. However, this approach is computationally intensive, e.g., numeric simulation of a material volume 500 µm x 500 µm x 100 µm using today s computational technology could require several days, which eliminates the possibility of real-time solutions. Accordingly, this approach is not practically scalable to the size of real world, manufactured parts. The current way around this issue is to select critical locations in a real part and use actual location-based variation data to define the minimum volume required to capture the worst case or lower bound values with 95/99 confidence. While the initial simulation will be very computationally intensive, lessons learned can be used to create subsequent, high-fidelity reduced models that are much more computationally efficient. Data-centric models lie at the other end of the spectrum, abandoning physics-based modeling altogether and relying completely on collected data. Empirical in basis, the datacentric approach typically begins with an initial model for each parameter based on expert elicitation. The initial model is then improved and refined using a plethora of data and, for example, Bayesian updating. Given enough time and enough data, this approach can work quite well. The drawbacks are the time and expense required to obtain adequate additional data, and the companion concern that flawed data may have an equal vote with legitimate data, leading to incorrect predictions. (iii) Physics-based phenomenological models comprise a middle ground approach that offers near-real-time simulation of microstructures and material properties. The weakness of this approach is that it is largely based on empirical models, and an accurate empirical model may not exist for a specific manufacturing process of interest. 4. Open Manufacturing Rapid Qualification Framework Ultimately, the Open Manufacturing program seeks to develop a Rapid Qualification framework that facilitates accelerated maturation, reduced risk, and improved implementation of new materials processing and manufacturing technologies. To be successful, a Rapid Qualification framework must address and improve several key issues that have historically affected the legacy technology development process and have ultimately inhibited insertion, in many cases. These include: early identification of technology risks for targeted mitigation; an accelerated development cycle with reduced iterations; reduced dependence on extensive, empirical validation processes; and more rapid demonstration of process understanding and control limits. These issues have been especially important to implementation of new manufacturing technologies, since the impact to a program or to a major defense system can be enormous. Consequently, program or systems management may be very conservative regarding commitment to new manufacturing technologies

8 The approach taken under Open Manufacturing to address these issues and achieve a successful Rapid Qualification framework includes several elements: Increased use of computational modeling for materials and processes (Integrated Computational Materials Engineering (ICME)): The Open Manufacturing program is the first known program to systematically incorporate ICME 2,3 and related support activities to accelerate manufacturing process qualification and reduce risk of implementation of new materials and manufacturing technology. Development and implementation of comprehensive, validated ICME methods and models will reduce dependence on lengthy, expensive test programs, reduce the number of development iterations, accelerate process optimization, enable property tailoring to design and structural requirements, and guide informed process control and qualification. Verification and Validation (V&V): For an ICME application to be successful and credible for decision-making particularly decisions made by program management, chief engineers, and regulatory authorities rigorous V&V processes are essential. Rigorous V&V techniques have already been established for other, more computationally focused engineering disciplines 4, and these V&V methods have been extended specifically for ICME applications under AFRL sponsorship. 5,6 Open Manufacturing is promoting the adoption of these methodologies to ICME practitioners. Uncertainty Quantification (UQ): Uncertainty Quantification (UQ) is a critical element of ICME model development, V&V, and process qualification. UQ represents assessment of model input parameters and their known or expected variation, followed by quantitative exercise of the computational model (or models) in a manner that permits assessment of the sensitivity and response of the model to the input parameters over the range of intended application. UQ results in a quantitative assessment of the uncertainty of the model and its sensitivity to parameters of interest, which provides a means for guiding: (i) experimental efforts to support model development and validation; and (ii) model refinements that may be needed to achieve desired or acceptable fidelity. Feedback and Updating Process: Experimental data is continuously generated as a normal part of manufacturing technology and in production processes. In the Open Manufacturing Rapid Qualification framework, this data also is important for enabling a rigorous and systematic 2 National Research Council (NRC), Integrated Computational Materials Engineering: A Transformational Discipline for Improved Competitiveness and National Security, Washington, D.C., The National Academies Press, Office of Science and Technology Policy (OSTP), Materials Genome Initiative for Global Competitiveness, June Available at: 4 The American Society of Mechanical Engineers, Guide for Verification and Validation in Computational Solid Mechanics, ASME V&V Cowles, Backman, and Dutton, Verification and validation of ICME methods and models for aerospace applications. Integrating Materials and Manufacturing Innovation :2. See also: 6 Cowles BA, Backman DG, Dutton RE: Update to recommended best practice for verification and validation of ICME methods and models for aerospace applications. Integrating Materials and Manufacturing Innovation, 2015, Volume 4, Issue 1. See also:

9 updating process to refine and improve ICME models and parameters, including their variation, which are deemed important to represent the materials and manufacturing processes of interest. Ideally, this would continue beyond process qualification into the production phase, differing substantially from legacy approaches that define a process with empirical substantiation and then freeze that process. In-process Monitoring and Advanced Inspection Techniques: Open Manufacturing has invested in advanced inspection methods and/or integrated in-process monitoring to validate process output, ensure that processes are operating within defined limits, and provide potential for closed-loop control of key process parameters. These technologies are being coupled with experimental data, ICME, V&V, and UQ. These elements, when employed together and enabled by a robust informatics approach, can effectively guide verification and validation of computational tools, provide confidence levels for model outputs, and guide efficient testing. Ultimately, these technologies will be combined to identify process windows or limits for multiple processes, optimize manufacturing processes based on improved understanding, and build confidence in new technologies. Collectively, they represent the common themes in the DARPA Open Manufacturing framework for Rapid Qualification. They are intended to work synergistically to accelerate maturation of technology and to increase confidence in new materials and processes. This Rapid Qualification framework will also increase confidence and enable rapid response when system or program iterations are inevitably required whether the changes involve system design, product configuration, or the performance environment. Open Manufacturing programs are developing and exercising the various elements of an overall Rapid Qualification framework. 7 The Open Manufacturing Rapid Qualification Framework offers potential to greatly accelerate maturation and implementation of new manufacturing technologies. While it is fully consistent with requirements of the established, gated technology processes described by Manufacturing Readiness Levels (MRLs), the elements and activities of the Open Manufacturing Rapid Qualification Framework are intended to be employed, integrated, and completed at earlier MRL gates than legacy processes. The result is increased confidence in new manufacturing technologies, better quantification of uncertainty and risk, and earlier achievement of technology maturation. This acceleration in maturation is essential to successful implementation of new technologies in any system application. 5. Open Manufacturing Program Goals and Metrics The vision for the Open Manufacturing program is the use of probabilistic techniques based on manufacturing input parameters to predict the properties of a specific part at any given location, and to understand the confidence level of this prediction for a given set of manufacturing parameters. The program goals are, therefore, also stated in terms of distribution functions: demonstrate a capability to predict the spatially dependent and volumetric properties of a manufactured part with as high a resolution as possible and with a calculated level of probabilistic accuracy. That is, perform a reliable and repeatable prediction of both the mean 7 See Additional Information at

10 and the distribution for the properties of a manufactured item, and demonstrate that this has been achieved in practice within a metric for accuracy that is specific to each process being pursued. Open Manufacturing seeks to verify that: (i) predictions have converged on measured performance; (ii) variances are quantified and the sources are thoroughly understood; and (iii) measured envelope performance information is rigorously derivable from the body of informatics measurement data. This emphasis is motivated by the paucity of measurement data in the tails of performance distribution functions in conventional material design and testing methods yet they influence the A-Basis and B-Basis design allowables. A key success metric for each of the manufacturing research processes being explored in the program is the degree to which the specific process research has successfully reduced the time and expense of qualifying an extension of that same process in the future. The informatics and process simulation models would be exercised extensively to probabilistically predict performance in a virtual testing paradigm, and would combine that with a limited set of real-life data to reduce manufacturing cost and improve efficiency. If, through application of its methodologies, the Open Manufacturing program achieves higher fidelity definition of the tails of these distribution functions thereby enabling improved confidence in the specified allowables the program will have succeeded in meeting its goals. Tying this understanding of probabilistic character back to allowables would achieve a common language that communicates with the designers and the certification officials to facilitate acceptance of the technology and the approach. 6. Applications of the Open Manufacturing Framework and Methodologies From the universe of manufacturing process technologies, the Open Manufacturing program selected Bonded Composite Structures and Metals Additive Manufacturing to bound the application space to which the program s framework and methodologies would be applied. Bonded composite processes are sensitive to many environmental factors and involve a great deal of human touch labor. On the other hand, Metals Additive Manufacturing is a very machine-driven process involving little direct human interaction, but which is very sensitive to input material and process equipment parameters. If Open Manufacturing could solve these issues, the program will have resolved fundamental problems inherent in most of the manufacturing processes either under development or currently in-practice Bonded Composite Structures Bonded Composite Structures have been the Holy Grail for composites for the last 30 years, hence the motivation and emphasis in this program. Success in a program such as Advanced Composite Cargo Aircraft means full advantage could be taken of the inherent attributes of composite materials, i.e., reduction of part count and lightweight performance. As noted above, however, bonded composite manufacturing processes are sensitive to myriad environmental factors that drive variability, including humidity and temperature of the lay-up room, curing temperature/time, length of time the materials have been out of the refrigerator, bonding surface cleanliness, and pressure

11 One effort identified over 500 separate composite manufacturing parameters that could potentially have a measurable impact on bonded structure performance. Use is made of a series of Design of Experiments to determine which parameters are important, which relationships between the parameters are important, and whether any dependencies exist between them with the goal of beginning to quantify what matters most for performance. As a result of these efforts, the Open Manufacturing program has identified several technologies that can be brought to bear to remove variability from the process. Prior to the Open Manufacturing program, only a fraction of these parameters had been monitored, let alone probabilistic measurements taken of process sensitivity to parameters in each manufacturing step. Until these processes and their contributions to overall performance are understood scientifically, quantitatively, and predictably, wholehearted embrace of the advantages of bonded composite structures in manufacturing is unlikely. The data gathered through these carefully designed experiments are stored in a data warehouse, which is coupled with a probabilistic toolset to construct a predictive model of the composite bonding process. Bayesian inference is used to quantify the uncertainty in the predictive capability of this model, guide additional data collection via sensitivity analysis, and provide a framework through which the process model can be updated as additional data are obtained. This system is intended to monitor the factory floor to compute the reliability of produced parts in real time and to gather operational data to update the underlying models and ensure they always reflect the best state of knowledge of the bonding process. The Open Manufacturing informatics methodology and computational process and design tools will be adapted to the increasing body of data as more composite bonds and more bonded parts are fabricated in the program. The objective is to achieve a predictive process model and a better understanding of what drives bond reliability. Ultimately, these results will be linked to design models to predict the bond strength anywhere along the bonded area with quantified confidence in that prediction. These data will then be coordinated with designers and certification officials to more rapidly qualify the processes and certify the parts. The Open Manufacturing program has begun to quantify the bond surface preparation process and produce the data to be fed back into the probabilistic process model for predictions of actual bond strength, remove variability wherever possible, and identify measures of bond reliability. Seemingly mundane physical characteristics, such as surface preparation, have a dramatic impact on bond strength and reliability. For example, hand-sanding, grit-blasting, and atmospheric plasma surface preparation methods are being subjected to analyses that provide a quantitative understanding of the relationship to bond strength. Sensing technologies to assess surface preparation, including handheld water contact angle measurement and handheld Fourier Transform Infrared (FTIR) spectroscopy, are being developed and introduced for practical use in the manufacturing environment. Another key issue that strongly influences bonded composite certification is the inability to locate kissing bonds in which, despite intimate contact between the different structures, an actual bond is not achieved because of a contaminant or perhaps a problem in the resin flow system. Nondestructive test methods and technologies are being explored to locate these kinds of defects and quantify bond strength based on measurable diagnostic signals

12 6.2. Metals Additive Manufacturing The second technology focus, Metals Additive Manufacturing, is an emerging production technology that offers potential for enhanced design flexibility, reduced feedstock material usage, and elimination of tooling. The two types of additive manufacturing technologies in the Open Manufacturing program involve melting either a powder (Direct Metal Laser Sintering (DMLS)) or a wire (Electron Beam Direct Manufacturing (EBDM)) and building up a structure ab initio, rather than actually machining material away from a beam, ingot, or plate to obtain the part. (This latter method can be called subtractive manufacturing, in contrast to additive manufacturing. ) Despite the potential advantages of metals additive manufacturing, implementing this technology for production is currently challenged by inadequate understanding and control of materials, machines and processes. Further, a conventional building-block approach may prove to be too costly and lengthy. Typically, certification officials and designers view metals additive manufacturing as a welding process, which has been an impediment to progress for this technology. Unlike a casting process, in which the thermal history throughout a given part is fairly homogenous, residual stress and a variation in microstructure in an additive-type process results from the very different thermal histories in different sections of the part. The result is a lack of trust that the thermal distribution throughout a part being built up can be quantified reliably and accurately, leading to uncertainty in the residual stress in the part and the microstructure-associated properties that result from this type of manufacturing process. This has been carried to the point where, today, use of additive manufacturing processes must usually be accompanied by the traditional rigorous test regimen to ensure that the part being built actually matches the original design intent which obviates some advantages of additive manufacturing and slows its adoption in practice. To further complicate the matter, additive manufacturing equipment manufacturers have traditionally tied use of their equipment to exclusive, proprietary control processes and materials relegating the equipment to a black box functionality and limiting widespread application in the marketplace. In fact, most pre-programmed machines take a component s geometry as input and autonomously determine the process parameters and build-sequence used to manufacture that component. This business model has restricted the ability of the users and operators of the equipment to understand and control the technical details of the manufacturing process itself. The Open Manufacturing program is actively addressing these issues. Researchers in Direct Metal Laser Sintering have designed experiments to identify which parameters are important and to quantify the impact of these parameters on the performance of the part at the end of the additive fabrication cycles. Material defects are being evaluated: how do manufacturing parameters affect the defects that do occur and where they are likely to occur? Understanding these uncertainties and relative sensitivities enables determination of which parameters are primarily responsible for affecting the performance mean, and which are primarily responsible for affecting the standard deviation. And in Electron Beam Direct Manufacturing, neural networks are being used to examine which parameters are important and the interdependencies that exist among them and how to tie those to known phenomenological models to optimize yield strength

13 Process monitoring is central to the application of informatics methodology to additive manufacturing. To validate that a manufactured part has been fabricated under specified, required, or prescribed conditions, the Open Manufacturing program is pursuing in-process sensing technologies to quantitatively capture and assess the key physical processes underway. For example, sensors are being developed to monitor the beams whether laser or electron to understand how much energy is imparted to the powder bed, the directed powder stream, or the directed wire. For Direct Metal Laser Sintering, pyrometric technology is being developed to image the powder and the melt pool to begin to understand the thermal dynamics and history of the process. To this end, physics-based process models are being developed to model the heat source of the laser and the melt pool itself, allowing greater understanding of how the laser beam affects the melt pool and the thermal environment surrounding it. 7. Manufacturing Demonstration Facilities The Open Manufacturing program has established two Manufacturing Demonstration Facilities (MDFs) one at Penn State focused on Additive Manufacturing and the other at the Army Research Laboratory focused on Bonded Composites. The collective mission of the MDFs is to provide central, non-commercial facilities where Open Manufacturing program models and supporting data can be archived in a pedigreed, protected, and retrievable manner including Integrated Computational Materials Engineering (ICME) models, manufacturing process tools and methods, analytical models, and all the supporting data and relevant metadata. The mission of these facilities is to: (i) curate and independently assess, and validate all data models and qualification frameworks that are being developed in the program; (ii) accelerate qualification through parameterized material and process data schema, data archiving, Metallic Materials Properties Development and Standardization (MMPDS) protocol, and model interoperability standards; (iii) generate advanced design, analysis, and simulation tools; (iv) create and host a controlled, cyber-based system enabling industry use of internal and external tools; and (v) work with defense and commercial industry as a trusted agent to independently demonstrate designs, manufacturing processes, and manufactured products. Defense and industry community outreach and interaction are a key feature of the MDFs. They are chartered to promote, disseminate, and sustain the application of manufacturing technologies through data exchange, workshops, conferences, internships, education, and direct industry collaboration. The institutional principle behind the MDFs is to establish permanent reference repositories that endure long after the Open Manufacturing program concludes, where individuals can access various contributed approaches and processes models. The MDFs serve as trusted agents and transition points for the models and data that are being developed in the program. They also serve as agents to demonstrate applications of the technology being developed for the Department of Defense and its industrial base, and other agencies. To date, the Additive Manufacturing MDF at Penn State has demonstrated applications associated with aerospace and ground based systems. Their integrated capabilities,

14 ranging from modeling to processing to testing and inspection, have served as a catalyst to accelerate the adoption of the technology Data Collection and Curation Historically, government-sponsored materials and manufacturing research programs have not retained adequate documentation that couples manufacturing/processing data to the resulting materials properties. Rigorous record keeping of data pedigree is essential to ensuring that the data can be accessed and used with confidence and reliability in the future. Incomplete or inadequate processing documentation is costly, since it forces the government or industry to go back and reinvest to regenerate the data as has happened all too frequently in the past. To support Open Manufacturing s development of a new generation of processing models (Section 3), one of the program s core objectives is to rigorously archive, and ensure future access to the processing, post-processing, microstructure, and resultant materials properties data generated during the program. The program s MDFs were charged with developing schemas for data retention: the Additive Manufacturing MDF at Penn State prepared schemas for Direct Metal Laser Sintering (DMLS), Laser Engineered Net Shape (LENS) processing, and the Sciaky Electron Beam Additive Manufacturing (EBAM) process. These schemas have been vetted by numerous government agencies and activities, private industry partners, standards bodies, and professional groups. To assure that the data is indeed fully and robustly archived and accessible in the future, the MDF at the U.S. Army Research Laboratory is integrating the schemas, and ARL s Materials Selection and Analysis Tool (MSAT) 8 serves as the repository for all the data being generated. MSAT has been integrated into NASA s Materials and Processes Technical Information System (MAPTIS) 9 to ensure the longevity of the data and to utilize MSAT/MAPTIS capabilities for maintaining/tracking data pedigree and data integrity with version control, as well as offering data analysis tools. The MAPTIS/MSAT database archival system is accessible by government personnel, both while the respective government-sponsored program is contributing data and after the program has ended. With proper permissions, data is accessible by industry at a variety of levels on a case-by-case basis, depending on need. Data control features include data reliability (experimental; prototypical; validated) and national security partitions, as well as data mining and validation capabilities. As Open Manufacturing moves forward with data and code curation, collaboration has been initiated with the standardization bodies, particularly Material Properties Development and Standardization (MMPDS). For example, with regard to qualification and development of design allowables for additively produced metals, no mature standardized analysis practice currently exists because of the unique processing aspects of different additive processes. Through its novel research efforts, the Open Manufacturing program has the unique opportunity

15 to participate in standardizing the processing parameters to be monitored and the methodology to capture this information during the manufacturing process, along with the microstructure and properties for as-manufactured material and parts Code Curation A fundamental objective of the Open Manufacturing program is reduction of the amount of data that must be generated for process qualification. The program is accomplishing this by developing and validating a series of models that can virtually generate some of the necessary data, thereby accelerating process qualification. It is well known that models are continually being generated by researchers in government, industry and academia. Frequently, these models are not properly validated with referee-quality data, and updating as part of support for the software is infrequent. Consequently these models often are quickly relegated to obscurity. Therefore, Open Manufacturing is developing a process by which models are evaluated, maintained, and retained at the Additive Manufacturing Demonstration Facility at Penn State. A framework is being established for rapid process qualification (Section 4) that will be driven by a combination of data and models, both empirical and physics-based. The framework will be powered by more than one dataset because manufacturing processing drives material microstructure, which, in turn, drives material properties. Also of interest are models associated with post-process heat treatments and models that will predict the likelihood of discontinuities. Models developed by any source can be validated at the MDF using a refereed dataset. If warranted, the models can be exercised within the Rapid Qualification framework to determine whether the framework performance is improved. If desired, the model can then become part of the curation activity at the MDF. This should help overcome the loss of knowledge regarding the codes intentions, assumptions, approaches, range of applicability, and nuances when the specific project or thesis draws to a close. 8. Summary DARPA s Open Manufacturing program is developing, verifying, and validating a framework that allows new manufacturing methodologies to be qualified rapidly with reduced testing. The program is working to promote early adoption by the broader DoD communities, including critical constituents such as the OSD Manufacturing Technology program, service labs, industry leaders, and certification authorities. Open Manufacturing is building the foundation for a framework to be applied to a broad range of manufacturing processes, while allowing further refinement of the ones being initially addressed. Specific platform transition planning is underway; these will be described in future publications