River Rapids Conference Proceedings of the Society for Experimental Mechanics Series Joining Technologies for Composites and Dissimilar Materials, Volume 10 Gary L. Cloud Eann Patterson David Backman Proceedings of the 2016 Annual Conference on Experimental and Applied Mechanics River Publishers
Conference Proceedings of the Society for Experimental Mechanics Series Series Editor Kristin B. Zimmerman, Ph.D. Society for Experimental Mechanics, Inc Bethel, CT, USA
River Publishers Gary L. Cloud • Eann Patterson • David Backman Editors Joining Technologies for Composites and Dissimilar Materials, Volume 10 Proceedings of the 2016 Annual Conference on Experimental and Applied Mechanics
Published, sold and distributed by: River Publishers Broagervej 10 9260 Gistrup Denmark www.riverpublishers.com ISBN 978-87-7004-944-3 (eBook) Conference Proceedings of the Society for Experimental Mechanics An imprint of River Publishers © The Society for Experimental Mechanics, Inc. 2017 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, or reproduction in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
v Joining Technologies for Composites and Dissimilar Materials represents one of ten volumes of technical papers presented at the 2016 SEM Annual Conference & Exposition on Experimental and Applied Mechanics organized by the Society for Experimental Mechanics and held in Orlando, FL, on June 6–9, 2016. The complete Proceedings also includes volumes on Dynamic Behavior of Materials; Challenges in Mechanics of Time-Dependent Materials; Advancement of Optical Methods in Experimental Mechanics; Experimental and Applied Mechanics; Micro and Nanomechanics; Mechanics of Biological Systems and Materials; Fracture, Fatigue, Failure and Damage Evolution; and Residual Stress, Thermomechanics & Infrared Imaging, Hybrid Techniques and Inverse Problems. Composite materials are being increasingly utilized at multiple levels in application areas including automotive, aerospace, marine, biomechanical, and civil infrastructure, so the need for improved joining of these materials has become critical. While the design of the composite laminate is important, it is the ability to join sections of composite to one another or to components made of dissimilar materials that is the enabling technology for creating structures that approach optimum in function, weight, durability, and cost. Composite joining technologies have been routinely classified in the past as either mechanical or adhesive. Increasingly, joint optimization requires combinations of the two types as well as the introduction of innovative new methods, such as composite welding, that provide high strength and light weight. Hybrid composite joints that allow the joining of composites to monolithic or other classes of material comprise another important technology that will facilitate the use of composites in many new application areas. Today, developments in composite joining technologies are progressing at a rapid rate, driven by both technology and user requirements. This symposium addresses pertinent issues relating to design, analysis, fabrication, testing, optimization, reliability, and applications of composite joints, especially as these issues relate to experimental mechanics of macroscale and microscale structures. East Lansing, MI, USA Gary L. Cloud Liverpool, UK Eann A. Patterson Ottawa, ON, Canada David Backman Preface
vii Contents 1 How to Join Fiber-Reinforced Composite Parts: An Experimental Investigation ............................................ 1 Yaomin Dong, Arnaldo Mazzei, Javad Baqersad, and Azadeh Sheidaei 2 Analysis of a Composite Pi/T-Joint Using an FE Model and DIC ...................................................................... 11 Chris Sebastian, Mahmoodul Haq, and Eann Patterson 3 5xxx Aluminum Sensitization and Application of Laminated Composite Patch Repairs ................................ 21 Daniel C. Hart 4 Investigation and Improvement of Composite T-Joints with Metallic Arrow-Pin Reinforcement.................. 33 Sebastian Heimbs, Michael Jürgens, Christoph Breu, Georg Ganzenmüller, and Johannes Wolfrum 5 Review of Natural Joints and Bio-Inspired CFRP to Steel joints ....................................................................... 41 Evangelos I. Avgoulas and Michael P.F. Sutcliffe 6 Fabrication of 3D Thermoplastic Sandwich Structures Utilizing Ultrasonic Spot Welding ............................ 49 Cassandra M. Degen and Navaraj Gurung 7 Impact and Lap Shear Properties of Ultrasonically Spot Welded Composite Lap Joints ............................... 59 Cassandra M. Degen, Lidvin Kjerengtroen, Eirik Valseth, and Joseph R. Newkirk 8 Numerical and Experimental Characterization of Hybrid Fastening System in Composite Joints ............... 71 Ermias G. Koricho, Mahmoodul Haq, and Gary L. Cloud 9 Application of Digital Image Correlation to the Thick Adherend Shear Test .................................................. 81 Jared Van Blitterswyk, David Backman, Jeremy Laliberté, and Richard Cole 10 Interfacial Strength of Thin Film Measurement by Laser-Spallation ............................................................... 91 Leila Seyed Faraji, Dale Teeters, and Michel W. Keller 11 Joining of UHTC Composites Using Metallic Interlayer .................................................................................... 99 Noritaka Saito, Laura Esposito, Toshio Yoneima, Koichi Hayashi, and Kunihiko Nakashima 12 Metal-to-Composite Structural Joining for Drivetrain Applications ................................................................. 107 Peter J. Fritz, Kelly A. Williams, and Javed A. Mapkar 13 Short-term Preload Relaxation in Composite Bolted Joints Monitored with Reusable Optical Sensors ....... 115 Anton Khomenko, Ermias G. Koricho, Mahmoodul Haq, and Gary L. Cloud
1 © The Society for Experimental Mechanics, Inc. 2017 G.L. Cloud et al. (eds.), Joining Technologies for Composites and Dissimilar Materials, Volume 10 Conference Proceedings of the Society for Experimental Mechanics Series, DOI 10.1007/978-3-319-42426-2_1 Chapter 1 How to Join Fiber-Reinforced Composite Parts: An Experimental Investigation Yaomin Dong, Arnaldo Mazzei, Javad Baqersad, and Azadeh Sheidaei Abstract A coupler has been developed to prevent windshield wiper systems from being damaged by excessive loads that can occur when the normal wiping pattern is restricted. Unlike the traditional steel coupler used in wiper systems, the composite coupler will buckle at a prescribed compressive load threshold and become extremely compliant. As a result, the peak loading of the coupler and the entire wiper system can be greatly reduced. The coupler is composed of a pultruded composite rod with injection-molded plastic spherical sockets attached at either end. The sockets are used to attach the coupler to the crank and rocker of the windshield wiper linkage. Because the loads exerted on a coupler vary in magnitude and direction during a wiping cycle, the joint between the sockets and the pultruded composite rod must be robust. The paradigm for attaching sockets to steel couplers (i.e. over-molding the sockets around holes stamped into the ends of traditional steel couplers) was tested and found to produce inadequate joint strength. This paper details the methodology that was employed to produce and optimize an acceptable means to join the injection-molded sockets to the fiber glass pultruded rods. Specifically, a designed experiment based on the Robust Design Strategy of Taguchi was used to identify the process, processing parameters, and materials that yield a sufficiently strong joint at a reasonable manufacturing cost without damaging the integrity of the underlying composite structure. Keywords Composite materials • Buckling • Wiper systems • Durability • Joint strength 1.1 Introduction Automotive windshield wiper systems, in conjunction with washer systems, are used in vehicles to remove contaminants such as rain, sleet, snow, and dirt from the windshield. As shown in Fig. 1.1, a typical wiper system consists of an electric motor, a linkage to transform the rotational motion from the motor to oscillatory motion, and a pair of wiper arms and blades. The areas of the windshield that must be wiped by the wiper system are mandated by the federal motor vehicle safety standards FMVSS 104 [1]. Figure 1.2 shows that snow has accumulated above the cowl screen and caused the normal wiping pattern of the system to be restricted. Under such conditions, the loads in the wiper system have been observed to be approximately four times greater than those encountered under normal wiping conditions for a particular application. The elevated loads are due to two factors: (1) As the arm(s) and/or blade(s) come into contact with the restriction, the speed of the wiper system decreases and the output torque of the motor increases. (2) This factor is aggravated by the very large mechanical advantage of the wiper linkage near the reversal positions. In extreme cases, the loads that result from restricting the wipe pattern can cause failures. For example, Fig. 1.3 depicts a 5-mm thick hardened steel rocker arm that has fractured as a result of such a loading. Clearly, such a damage is unacceptable from both safety and warranty standpoints. Penrod and Dong [2] reviewed various methods that have been developed to solve this problem. One such solution is to use a coupler fabricated from a thermoset polyester/glass fiber pultruded composite material [3–6]. This coupler—referred to as the composite link—has the advantage of being sufficiently stiff during normal operation providing very good pattern control. However, in the event that elevated loads occur, the coupler is designed to buckle at a prescribed load level (i.e. the critical load). Once buckled, the coupler becomes extremely compliant. As a result, the peak loading of the coupler and, moreover, the entire system, can be limited and greatly reduced when compared to the same system without the composite link. Y. Dong, Ph.D. (*) • A. Mazzei, Ph.D. • J. Baqersad, Ph.D., P.E. • A. Sheidaei, Ph.D. Department of Mechanical Engineering, Kettering University, 1700 University Avenue, Flint, MI 48504-4898, USA e-mail: ydong@kettering.edu
2 Figure 1.4 shows the operating principle of the composite link. In the illustration the wiper arms have encountered snow/ ice accumulation (depicted by the hatched pattern) above the cowl screen that blocks the normal wiping pattern of the system. As a result, the wiper system loads have increased sufficiently to buckle the composite link (depicted by the arcuate member Fig. 1.1 A typical wiper system consists of wipers, motor, and a linkage, and wiper arms Fig. 1.2 Restricted wiping pattern due to snow accumulation Fig. 1.3 Fractured rocker arm resulting from snow accumulation Y. Dong et al.
3 of the wiper linkage). Once buckled, the composite link behaves nearly perfectly-plastic. As a result, the chord length of the composite link decreases significantly without a corresponding appreciable increase in axial load. Thus the crank is able to continue rotating and the wiper system is enabled to wiper at a reduced area above the restriction. The composite link is an assembly of a 16.7 mm × 5.2 mm rectangular cross-section pultrusion and glass-filled plastic sockets attached at either end, as shown in Fig. 1.5. The design methodology, material selection, validation testing, and application guidelines for the composite link are discussed in detail by Penrod, Dong, and Buchanan [3]. Because the loads exerted on the coupler vary in magnitude and direction during each wiping cycle, the joint between the sockets and the pultrusion is critical and must be robust. Initially, it was planned to attach the sockets to the pultrusion rod using the same approach that sockets have been traditionally joined to steel couplers (i.e. over-molding the sockets around holes stamped into the ends of the traditional steel coupler). In the case of the composite link, holes were drilled into the ends of the pultrusion and the sockets were overmolded onto the pultrusions. Unfortunately, this joining method performed dismally as it was characterized by bearing stress failures of the pultrusion at relatively low tensile loads. In retrospect, this failure mode was no surprise. The pultrusions consist of longitudinally oriented continuous glass fibers bound together by a thermosetting polyester matrix. Drilling holes on the pultruded rod cuts the continuous fibers and weakens the composite structure. As such, the material has little strength on short-transverse planes where the bearing failure was observed to occur. Since the paradigm for attaching sockets to steel links did not work, a Greenfield approach to attaching sockets to the pultrusions was embarked upon. This paper details the approach that was employed to generate, evaluate, and optimize an acceptable means of attaching the injection-molded plastic sockets to the pultrusions. Fig. 1.4 Operating principle of the composite link Fig. 1.5 Composite link assembly 1 How to Join Fiber-Reinforced Composite Parts: An Experimental Investigation
4 1.2 Designed Experiment for Joining Socket to Pultrusion How can the plastic sockets be attached to the pultruded composite rod? The traditional approach of over-molding the sockets to steel links proved inadequate as was discussed previously. In order to answer this question, a set of design specifications were generated to assess design concepts against. Qualitatively, the specifications of paramount importance were that the attachment method. • Provide adequate joint strength in tension and compression when subjected to monotonic and cyclic loads. • Avoid damaging the integrity of the glass fibers to the extent that the pultrusion would no longer be capable of providing a predictable critical load and adequate post-buckled axial compressive deflection. • Be cost-effective. Once the aforementioned specifications were established, three phases of the development were sequentially undertaken. Specifically, • System Design : the first phase of the development where a variety of state-of-the-art candidate processes were generated using input from a cross-functional team of material, manufacturing, and product specialists. Once identified, the candidate processes were then narrowed to the most promising approaches via conceptual evaluation by the cross-functional team. Additionally, some ad hoc testing was completed in this phase of the development that suggested adding grooves to the ends of the pultrusion so that the over-molded socket could form a positive lock was the most viable approach. • Parameter Design : the second phase of the development where a series of designed experiments were used to quantitatively ascertain the optimal combination of design and processing parameters. • Tolerance Design : the third and final phase of the development where the allowable tolerance range on the critical dimensions and processing parameters were determined. The focus of this paper is on the series of designed experiments that were used to identify the optimal combination of designs, materials, and processes during the parameter design phase of the development. 1.2.1 Parameter Selection Stage 1: Screening The Robust Design Strategy of G. Taguchi [7] has gained broad acceptance in Japan, the United States, and elsewhere for use in product design and development. Taguchi’s approach to parameter design provides a systematic and efficient method for determining near-optimum design parameters. The Taguchi method allows for the sensitivity of the design to a large number of parameters to be evaluated using a relatively small number of sample parts by utilizing orthogonal arrays from design of experiments theory. The conclusions drawn from such small scale experiments are valid over the entire experimental range spanned by the control factors and control factor levels. As mentioned previously, adding grooves to the pultrusion ends so that the over-molded socket would be positively locked to the pultrusion was deemed the most viable approach. However, the influence of other factors on the joint strength was still unknown. Specifically, what is the influence of the surface finish of the pultrusion on joint strength? What role does the location of the grooves play? Does the socket material significantly impact the joint strength? What benefit might be derived from the use of adhesives? What is the influence of glass content in the pultrusion on joint strength? To answer these questions, an experiment was designed to screen these factors for relative importance and identify where the various factors might interact—both in a positive and negative sense. Table 1.1 depicts the L18 Orthogonal Array that was used to screen the design candidates. Notice that the five aforementioned design parameters (A ~ F) are chosen as control factors, with each factor having three levels. Descriptions of the control levels indicated in the array are as follows. • Surface Finish: Smooth denotes the surface finish of the pultrusion without any additional processing. Abrasive-fine, and abrasive-coarse denote that the ends of the pultrusion have been grit blast with a fine and coarse medium, respectively. Chemical etched refers to the resin-rich surface of the pultrusion ends being removed using an acidic solution. Laser etches #1 and #2 denote two power levels used to remove the resin-rich surface of the pultrusion via the use of a laser. • Machined Grooves: Smooth denotes no grooves, top–bottom denotes grooves machined into the top and bottom surfaces of the pultrusions. Side-side denotes grooves machined into the edges of the pultrusions. Ultimately, axial grooves which Y. Dong et al.
5 are oriented longitudinally were also considered. (Please see Sect. 2.3 for details.) All grooves were full-filleted as indicated in Fig. 1.6 and were machined by grinding. • Socket Material: Nylon denotes a glass-filled nylon 6/6. Acetal #1 refers to 20 % glass-filled acetal, whereas acetal #2 refers to 30 % glass-filled acetal. • Attachment/adhesives: Three possibilities were considered. Insert molding the sockets over the ends of the pultrusion was done both without adhesives (i.e. insert mold/no adhesive) and with adhesives (i.e. insert mold/hot curing). Hot curing refers to the heat available from the injection molded material promoting rapid curing of the adhesive.) Mechanical attachment of the socket to the pultrusion was also considered and used an adhesive to bond the sockets to the pultrusions. In this case an accelerant was used to promote curing of the adhesive. Thus, this approach is denoted by mech attach/chem cure in Table 1.1. In all cases, the adhesive used was a commercially available cyanoacrylate. • Pultrusions: Low-, moderate-, and high-modulus refer to elastic modulus of the pultusions resulting from three different glass content levels being used. The higher the glass content, the higher the modulus. It should be noted that a full-factorial experiment would require evaluating the performance of (6 × 3 × 3 × 3 × 3 =) 486 combinations of control factors and control levels. Using the Taguchi approach, only 18 combinations need to be considered with the 18 specific combinations of control factors and control levels dictated by the method. Table 1.1 L18 Orthogonal array for stage 1 (screening) 1&2 3 4 5 6 A B D E F Surface finish Machined grooves Socket material Attachment/adhesives Pultrusions 1 Smooth Smooth Nylon Insert mold/hot curing Low modulus 2 Smooth Top-bottom Acetal #1 Mech attach/chem. cure Moderate modulus 3 Smooth Side-side Acetal #2 Insert mold/no adhesive High modulus 4 Abrasive-fine Smooth Nylon Mech attach/chem. cure Moderate modulus 5 Abrasive-fine Top-bottom Acetal #1 Insert mold/no adhesive High modulus 6 Abrasive-fine Side-side Acetal #2 Insert mold/hot curing Low modulus 7 Abrasive-coarse Smooth Acetal #1 Insert mold/hot curing High modulus 8 Abrasive-coarse Top-bottom Acetal #2 Mech attach/chem. cure Low modulus 9 Abrasive-coarse Side-side Nylon Insert mold/no adhesive Moderate modulus 10 Chemical etch Smooth Acetal #2 Insert mold/no adhesive Moderate modulus 11 Chemical etch Top-bottom Nylon Insert mold/hot curing High modulus 12 Chemical etch Side-side Acetal #1 Mech attach/chem. cure Low modulus 13 Laser etch #1 Smooth Acetal #1 Insert mold/no adhesive Low modulus 14 Laser etch #1 Top-bottom Acetal #2 Insert mold/hot curing Moderate modulus 15 laser etch #1 Side-side Nylon Mech attach/chem. cure High modulus 16 Laser etch #2 Smooth Acetal #2 Mech attach/chem. cure High modulus 17 Laser etch #2 Top-bottom Nylon Insert mold/no adhesive Low modulus 18 Laser etch #2 Side-side Acetal #1 Insert mold/hot curing Moderate modulus Fig. 1.6 Groove configurations 1 How to Join Fiber-Reinforced Composite Parts: An Experimental Investigation
6 Once the 18 combinations were prototyped, a strip test was run on each configuration to determine the ultimate strength of each joint in tension. Figure 1.7 depicts the test. The assembled composite link was placed in a specially designed fixture that captured the socket only. The tensile load was increased until the assembly of the socket and pultrusion separated. Analyzing the data collected from the testing using the Taguchi approached allowed for the contribution to the joint strength of each control factor and control level, and interactions of the control factors and control levels to be quantified. From this analysis, it was found that grooves added to the top and bottom faces of the pultrusion were the single greatest contributor to joint strength. The use of adhesives to produce bonding between the socket and pultrusion was found to be the second largest contributor to strength, but was ruled out due to processing difficulties. Additionally, altering the surface finish of the pultrusion provided little benefit and was quite costly. As such, it was also disqualified from further consideration. 1.2.2 Parameter Selection Stage 2: Groove Design Based on the results of the screening designed experiment, a second designed experiment was constructed to further optimize the groove design. Table 1.2 depicts the L9 Orthogonal Array that was constructed for this purpose. In this experiment three control factors (i.e. groove size, groove spacing, and number of grooves) were considered at three different control levels. Once again it is noteworthy that the number of configurations to be tested is only nine and that a full factorial experiment would require 27 (i.e. 3 × 3 × 3) configurations. The control levels indicated in Table 1.2 are described as follows. • Groove Size: Small denotes 0.5 mm depth, while medium corresponds to 1.0 mm depth, and large denotes a 1.5 mm depth. • Groove Spacing: The spacing between adjacent grooves with level #1 corresponding to 2 mm spacing, level #2 corresponding to 3 mm spacing, and level #3 corresponding to 4 mm spacing. • Number of Grooves: Few denotes two grooves per face, moderate denotes four grooves per face; and many corresponding to six grooves per face. Fig. 1.7 Strip test used to determine the ultimate strength of the joint Table 1.2 L9 Orthogonal array for stage 2 (groove design) 1 2 3 A B C Groove size Groove spacing Number of grooves 1 Small #1 Few 2 Small #2 Moderate 3 Small #3 Many 4 Medium #1 Moderate 5 Medium #2 Many 6 Medium #3 Few 7 Large #1 Many 8 Large #2 Few 9 Large #3 Moderate Y. Dong et al.
7 Prototype parts of each configuration were fabricated and evaluated against two criteria. The first criterion was simply evaluating the tensile strength of the joint as was performed in the screening experiment. The second criterion involved examining the performance of the composite links in compression after being buckled and axially compressed to a preset deflection. Doing so revealed that under certain circumstances, the pultruded material tends to delaminate on long-transverse planes with the delamination initiating at the root of the groove. Analyzing the data using the Taguchi approach once again allowed for the contribution to the joint strength of each control factor and control level, and interactions of the control factors and control levels to be determined. From this analysis and delamination considerations, it was found that numerous small, closely-spaced grooves provided the optimal combination of joint strength in tension and delamination resistance after buckling. 1.2.3 Parameter Selection Stage 3: Final Design Once the groove optimization designed experiment was completed, a third and final designed experiment was constructed to examine various processing methods to produce the grooves. Additionally, surface finish was once again considered and the use of adhesives was revisited. An additional six configurations were added to the experiment for checking purposes. Table 1.3 depicts the L18 Orthogonal Array that was constructed for this final investigation. Table 1.4 contains six additional configurations of interest that were included in the experiment. Prototype parts representing the 24 configurations specified in Tables 1.3 and 1.4 were fabricated and evaluated for tensile strength per the procedure used in the previous experiments. Notice that four different processes were considered to produce the grooves both across the grain (cross grooves) and axially—machining (i.e. cutting), grinding, grit blasting, and laser ablation, as well as, a control case of no grooves. Table 1.3 L18 Orthogonal array for stage 3 (final design) 1&2 3 4 A C D Grooves Surface finish Attachment method 1 Smooth Smooth Insert mold 2 Smooth Ground axial fiber exposure Adhesive assy W/o surface primer 3 Smooth Laser axial fiber exposure Adhesive assy W/surface primer 4 Ground axial grooves Smooth Insert mold 5 Ground axial grooves Ground axial fiber exposure Adhesive assy w/o surface primer 6 Ground axial grooves Laser axial fiber exposure Adhesive assy W/ surface primer 7 Ground top-bottom grooves Smooth Adhesive assy W/o surface primer 8 Ground top-bottom grooves Ground axial fiber exposure Adhesive assy W/ surface primer 9 Ground top-bottom grooves Laser axial fiber exposure Insert mold 10 Ground side-side grooves Smooth Adhesive assy W/surface primer 11 Ground side-side grooves Ground axial fiber exposure Insert mold 12 Ground side-side grooves Laser axial fiber exposure Adhesive assy W/o surface primer 13 Laser axial grooves Smooth Adhesive assy W/o surface primer 14 Laser axial grooves Ground axial fiber exposure Adhesive assy W/ surface primer 15 Laser axial grooves Laser axial fiber exposure Insert mold 16 Cut top-bottom grooves Smooth Adhesive assy W/ surface primer 17 Cut top-bottom grooves Ground axial fiber exposure Insert mold 18 Cut top-bottom grooves Laser axial fiber exposure Adhesive assy W/o surface primer Table 1.4 Six additional configurations of interest 19 Ground cross grooves Plasma axial fiber exposure Insert mold 20 Ground cross grooves Plasma axial fiber exposure Adhesive assy W/o surface primer 21 Grit blast cross grooves Smooth Adhesive assy W/o surface primer 22 Grit blast cross grooves Laser axial fiber exposure Insert mold 23 Laser cross grooves Smooth Adhesive assy W/o surface primer 24 Laser cross grooves Laser axial fiber exposure Insert mold 1 How to Join Fiber-Reinforced Composite Parts: An Experimental Investigation
8 Three surface treatments were considered—grinding, laser ablation, and plasma cutting, as well as, untreated (smooth) finish. Additionally, insert-molding the sockets was compared to attaching the sockets using adhesives with and without a surface primer. It is again worth noting that a full factorial experiment would require 54 (i.e. 6 × 3 × 3) configurations, whereas the designed experiment requires only 18 configurations. Figure 1.8 depicts the test results for the top five performing configurations. In all five cases, the joint strength requirement was exceeded by a factor of two. Once again the strip test method depicted in Fig. 1.7 was used to obtain these results. The non-linear portion of the curves near the origin is attributable to the small amount of “play” in the joint due to shrinkage of the injection-molded socket material. Then, the loading noted to increase linearly with respect to deflection until the peak load-carrying capability of the joint is reached. At this point, either the socket material filling the grooves sheared off or the pultruded sheared off on a transverse plane. 1.3 Conclusions Based on the aforementioned testing, and considering cost and quality control, the results of the investigation can be summarized as follows: • When machining grooves into the pultrusions, grinding is superior to cutting due to the nature of the glass-filled composite structure. • Avoid attaching parts to pultrusions via through holes as the transverse planes of the pultrusion which are sheared by tensile loading have little strength. JOINT STRENGTH TEST 0 500 1000 1500 2000 2500 3000 3500 4000 0.00 0.50 1.00 1.50 2.00 2.50 3.00 Extension (mm) Load (N) 8 14 7 6 9 Fig. 1.8 Joint strength of the top five performing configurations Y. Dong et al.
9 • Grooves should be on the faces of the pultrusion having the greatest surface area (e.g. the top and bottom in the present design). Offsetting the grooves has been observed to make the design less susceptible to delamination. • Numerous fine grooves are superior to a fewer coarse grooves. • Adhesives can be used to significantly improve joint strength, but are noted to be difficult to process since they are very time-sensitive and difficult to control during the molding process. • Surface roughness has minimal impact on joint strength. Due to the extra cost associated with this processing, it should only be considered when the load requirement cannot be satisfied otherwise. In lieu of these assertions, the final, optimized groove configuration and dimensions of the joint between the sockets and the pultrusion is depicted in Fig. 1.9. References 1. FMVSS 571-104, 49 CFR. Federal Motor Vehicle Safety Standards; Windshield Wiping and Washing Systems. National Highway Traffic Safety Administration, DOT, Washington, DC 2. Penrod, J., Dong, Y.: An Application Flexible Method to Limit the Loads in Windshield Wiper System. SAE Technical Paper Series, 2005-011835 (2005) 3. Penrod, J., Dong, Y., Buchanan, H. C.: A Novel Use of a Composite Material to Limit the Loads in Windshield Wiper System, SAE Technical Paper Series, SP-1575, 2001-01-0104 (2001) 4. Buchanan, H.C., Dong, Y.: Windshield wiping system, US Patent US 6,148,470, Valeo Electrical Systems (2000) 5. Buchanan, H.C., Dong, Y.: Windshield wiping system, US Patent US 6,381,800 B1, Valeo Electrical Systems (2002) 6. Bryson, B.A., Buchanan, H.C., Dong, Y., Penrod, J.P.: Windshield wiping system manufacturing method, US Patent US 068 81 373, Valeo Electrical Systems (2005) 7. Peace, G.: Taguchi Methods: A hands-On Approach to Quality Engineering. Addison, Reading, MA (1993) 298.3 BALL SOCKET 11.0 8.0 5.0 R 0.5 0.5 5.2 16.7 21.0 11.0 R 13.0 Ø 16.0 21.0 10.5 17.2 2.2 2.0 3.5 6.5 9.5 12.5 1.0 R 1.8 4 CORNERS 2.0 COMPOSITE LINK Fig. 1.9 Composite link socket attachment 1 How to Join Fiber-Reinforced Composite Parts: An Experimental Investigation
11 Chapter 2 Analysis of a Composite Pi/T-Joint Using an FE Model and DIC Chris Sebastian, Mahmoodul Haq, and Eann Patterson C. Sebastian (*) • E. Patterson School of Engineering, University of Liverpool, Liverpool L69 3GH, UK e-mail: c.sebastian@liverpool.ac.uk M. Haq Composite Vehicle Research Center, Department of Civil and Environmental Engineering, Michigan State University, Lansing, MI 48910, USA Abstract An analysis of composite Pi/T-joint was performed through the comparison of a finite element simulation and experimental results obtained using digital image correlation. The finite element simulation was performed in ABAQUS using a realistically modeled adhesive layer consisting of cohesive zone elements. The composite Pi/T-joints were manufactured using vacuum assisted resin transfer molding (VARTM). The resulting Pi/T-joints were loaded in out-of-plane (web pull-out) until failure, during which three-dimensional digital image correlation (3D-DIC) was used to measure the in-plane displacements and strains. From the experiments, it was found that the average peak pull-out force was 12.56 ± 0.82 kN, which was 10 % less than that from the FE simulation at 13.77 kN. However, this does not give much information about whether the stress and strain distribution has been accurately predicted by the simulation. To better understand this, the experimental data obtained using DIC was compared to that from the FE simulation using image decomposition. The image decomposition process reduces the large, full-field data maps to a feature vector of 130 or so elements which can be compared much more easily. In this case, there was good agreement between the experiment and simulation, when taking into account the experimental uncertainty. Keywords Composite joint • VARTM • Digital image correlation • Numerical simulation 2.1 Introduction Composite Pi-joints are being extensively used in the marine and aerospace industries and are called T-joints, or Tee joints in some sectors. ‘T’ refers to the parts being joined while ‘Pi’ describes the shape of the preform used to form the joint. These joints are of special importance due to their complex geometry and criticality to overall structural integrity. T-joints are found at composite bulkhead-to-skin, rib-to-skin and spar-to-skin interfaces [1]. T-shaped stiffeners are extensively used in aircraft wings [1]. Additionally, the requirement for reduced structural weight and the ever-increasing demand for efficient aerospace structures has driven the development of adhesively bonded T-joints. However, these joints have been found to be sensitive to peel and, or delamination and through-the-thickness stresses. It has been reported that composite T-joints often fail near the web/skin interface [2]. Various configurations of composite T-joints have been studied extensively, mostly in the context of marine structures [3–18], including tee-joints made of sandwich panels [9–12]. The damage and failure analysis of such marine composite joints has also been studied [13–18]. A good overview on adhesively bonded composite joints is provided by Banea and Silva [19]. Moreover most of the aforementioned T-joints were manufactured by connecting the horizontal (flange) and vertical (web) laminates with a hand-layup laminate/skin (overlaminate). Similarly, literature on T-joints using a pi-preform (hence a Pi-joint) is relatively limited [20]. The use of Pi-preforms compared to hand-layup skins offers ease of manufacturing and speedy construction. Researchers from the Wright Patterson Air Force Research Laboratory (AFRL), Ohio, US, have studied adhesively bonded Pi-preform T-joints and also incorporated them in an aircraft wing and reported enhanced performance relative to a similar wing manufactured by conventional methods [2, 21, 22]. In addition, they tested the wings with T-joints © The Society for Experimental Mechanics, Inc. 2017 G.L. Cloud et al. (eds.), Joining Technologies for Composites and Dissimilar Materials, Volume 10 Conference Proceedings of the Society for Experimental Mechanics Series, DOI 10.1007/978-3-319-42426-2_2
12 in fatigue and found no failure in the joints [21, 22]. They reported considerable savings in manufacturing time (days instead of weeks) and up to 75 % reduction in costs. Apart from the AFRL study [2, 21, 22], the literature on pre-form Pi-joints is relatively limited. Hence, this work focuses on evaluating the pull-out performance using both structural tests to destruction and integrated experimental and simulation approach to analyzing the strain fields in structural Pi-preform composite T—joints. In this work, the out-of-plane behavior or pull-out performance behavior of the composite Pi-joints was experimentally studied by using three-dimensional Digital Image Correlation (3D DIC). To the best of author’s knowledge, the use of DIC to study the composite Pi-joints is unique and no other work has been reported, except the preliminary results reported earlier by the authors [23, 24]. The following sections provide brief details on the manufacturing process, materials used, and the experimental and numerical results. 2.2 Experimental Method 2.2.1 Manufacture of the Pi-joints The Pi-joints of dimensions shown in Fig. 2.1 were manufactured by connecting flange and web plates with a Pi-preform and infusing the resin (adhesive) using vacuum assisted resin transfer molding (VARTM). The VARTM technique was also used to manufacture the glass fiber composite adherends (base and web plates) for the Pi-joints. The reinforcement used for the adherends was Owens Corning ShieldStrand S, S2-glass plain weave fabric with areal weight of 818 g/m 2 . The base and web plates had 16 and eight layers of plain weave glass fabric respectively resulting in cured thicknesses of 9.53 mm (3/8 in.) and 4.76 mm (3/16 in.), respectively. The resin used was a two part toughened epoxy (SC-15, Applied Poleramic). The base and web plates were then connected with the Pi-preform and infused with SC-15 adhesive. The Pi-preform used was a carbon-fiber 3D woven preform (Albany Engineered Composites, Inc., Rochester, NH, USA.). A constant adhesive bond-line thickness in Pi-joints was maintained by placing steel wires of 0.127 mm (0.005 in.) diameter in the bond-line. These spacers were placed strategically such that they did not influence the resulting performance. The tensile properties of the adhesive were obtained from experimental tests following the ASTM D638 standard. The fracture properties of the resin were obtained experimentally from Mode-I and Mode-II tests. The properties of the flange and web plates were also obtained experimentally. A summary of the material properties is provided in Table 2.1. All bonded surfaces were grit-blasted and then cleaned with acetone. Figure 2.2 provides a step-by-step overview of the Pi-joint manufacturing process. Figure 2.2b shows the aluminum mold used to obtain the desired pi-preform dimensions. The aluminum mold has inlet and outlet pipe-fittings that attach to the vacuum pump and the resin-bath, respectively. The inlet 4.57 2.54 14.48 3.30 17.27 60.00 139.20 69.60 29.98 14.99 12.70 Fig. 2.1 Pi-joint dimensions (in mm) with the adhesive thickness omitted. Not to scale C. Sebastian et al.
13 Fig. 2.2 Manufacturing of the Pi-joint using VARTM: ( a ) aluminum mold, ( b ) assembled pi-joint during VARTM, ( c ) completed pi-joints, and ( d ) a 50 mm wide section of pi-joint during a tensile pull-out test (a) Adhesive tensile properties (experimental) Modulus (GPa) 2.69 ± 0.19 Tensile strength (MPa) 58.10 ± 3.94 Tensile failure elongations (%) 3.45 ± 0.37 (b) Adhesive fracture toughness [29] GI c (N/m) 0.132 GII c (N/m) 0.146 (c) Plates—GFRP (experimental) Web, tensile modulus (GPa) 17.25 Base, flexural modulus (GPa) 7.00 Poisson’s ratio [5] 0.17 (d) 3D-Preform—CFRP [30] Elastic, isotroPic modulus (GPa) 32.8 Poisson’s ratio 0.17 Table 2.1 Material properties used in the numerical simulation 2 Analysis of a Composite Pi/T-Joint Using an FE Model and DIC
14 and outlet are located just above the web-location of the Pi-preform. The permeability of the preform was found to be sufficient for successful manufacture of the Pi-joints, and hence, a distribution media was not required. The aluminum mold was fitted with end-plates and enclosed with vacuum tape prior to infusion as shown in Fig. 2.2c. The resin-infused joint was cured in a convection oven at 60 °C for 2 h and post cured at 94 °C for 4 h. The completed joints, which are shown in Fig. 2.2c, were then cut into sections of 50 mm wide using a water jet prior to being subjected to experimental testing as shown in Fig. 2.2d. 2.2.2 Pull-Out and Damage Resistance (DIC) The out-of-plane behavior of the Pi-joints was evaluated by performing pull-out tests on the manufactured Pi-joints using a specially designed rig in a servo-hydraulic test machine, as shown in Fig. 2.3a. The specimens were tested at a loading rate of 1 mm/min until failure. The load and displacement from the test machine cross-head was recorded. Additionally, an external LVDT (linear voltage displacement transducer) was used to measure the relative displacement of the web with respect to the support. In addition to the load and displacement values measured during the pull-out tests, a 3-D Digital Image Correlation system was employed to measure the full-field displacement and strains on the cut surface of the pi-joint. This was a commercially available DIC system (Dantec Dynamics Q-400) which consisted of a pair of cameras (AVT F-125B) along with a matched pair of lenses with a 12 mm focal length. Illumination was provided using a small panel of green LEDs. The cut surface of the Pi-joint was prepared by painting it first with a coat of matte white spray paint (Krylon matte white) and then misting with matte black (Krylon matte black) to provide the contrasting speckle pattern. Reference images for the correlation process were captured with a small pre-load applied to the joint, in order to remove any slack from the experimental setup. Images were then captured in 500 N increments up until failure of the joint occurred. 2.3 Numerical Simulations An integrated approach to numerical simulations and experiments was taken with the intention of using simulations to facilitate the design of the experiments, sensor placement and also to support the interpretation of the experimental data, as well as creating a validated numerical model that could be used in future design work with a high-level of confidence. In this work, simulations of Pi-joint pull-out behavior were performed using a commercially-available finite element package, ABAQUS ® [ 25 ]. Fig. 2.3 Picture of a typical pull-out test setup ( a ); schematic of the experimental test setup ( b ) C. Sebastian et al.
15 A two-dimensional plane-stress model of the Pi-joint was created in ABAQUS ® . The web, flange and the pre-form were created as separate parts for ease of defining the properties and also the meshing. The adhesive was also created as a separate part but with a much finer mesh using cohesive elements. The adhesive surfaces were appropriately connected with respective surfaces of the web, flange and pre-form to assemble the model. Models that reproduced the three cases studied in the experiments were analyzed. 2.3.1 Mesh, Gap Creation and Material Models Four-node quadrilateral plane-stress elements were used to mesh all the parts of the model in ABAQUS ® . All parts were given an out-of-plane thickness of 50 mm corresponding to the width of the Pi-joint in the experiment. The flange, web and Pi-preform were modeled as linear elastic materials. An element size of 0.5 mm was used for all parts except the adhesive layers. A meshed finite element model for a Pi-joint is shown in Fig. 2.4. For adhesive layers, cohesive elements, the mesh was more refined as illustrated in the inset of Fig. 2.4. The element size of cohesive elements was equal to bond-line thickness (0.005 in., 0.127 mm), and an aspect ratio of 1 (unity) prior to load application. The adhesive was modeled in greater detail using cohesive elements and the delamination was simulated using a traction-separation material model [25]. The damage evolution was modeled in ABAQUS by defining the fracture energy as a function of the mixed modes using the analytical power law fracture criterion. The material properties used in the simulations are provided in Table 2.1. 2.3.2 Boundary Conditions and Loading The loading pattern and boundary conditions used in the experiments were replicated in the finite element models. A schematic of the FE model is provided in Fig. 2.4. It has been reported that the number of degrees of freedom at the constraints and the angle of loading play a vital role in the accuracy of the results [6]. Kesavan et al. [6] simulated marine composite T-joints and applied the loading at an angle of 0.55° to the positive Y-axis in Fig. 2.4, to take into account manufacturing inconsistencies, and found good agreement with their experimental results. In this work, the web was assumed to be perfectly perpendicular and the angle of loading was not considered. The applied load was distributed along the top nodes in the web. The model was constrained symmetrically at two points on the flange simulating a simply-supported condition with a total span of 100 mm. The left-hand side support constraint does not permit translation in either the X- or Y-direction while the right-hand support constraint does not permit translation in the Y-direction. All simulations were performed on an eight-core 3.16 GHz machine with 64 gigabytes of RAM. Fig. 2.4 Details of the FE model showing the boundary conditions and applied loading. The inset shows a more detailed view of the mesh with the quadrilateral elements and the adhesive elements 2 Analysis of a Composite Pi/T-Joint Using an FE Model and DIC
16 2.4 Results and Discussion 2.4.1 Experimental Pull-Out Results Figure 2.5 provides the pull-out performance of the Pi-joints from the experiments. The general failure mode in the experiments for the Pi-joints consisted of an increasing load until failure, which led to the separation of the Pi-preform from the base. The failure was instantaneous and the direction of failure propagation could not be seen with the naked eye. A highframe rate camera would be required to confirm the propagation path of failure. Since symmetric failure initiation was observed, the delamination propagation was also expected to be symmetric. Instead an asymmetric failure was observed. Such a discrepancy may be caused by a combination of eccentricity in load application and manufacturing flaws. The use of a brittle adhesive, like the one used in this work, makes the joint highly susceptible to manufacturing flaws and less tolerant to damage. 2.4.2 Digital Image Correlation Results Digital Image Correlation was used to capture the displacement and strain on the cut surface of the Pi-joint as it was loaded in the tensile test. The images were processed using commercial software (Dantec Dynamics Istra 4D) to calculate the displacements and strains in the Pi-joint. A facet (or subset) size of 25 pixels was used, with an offset (or shift) of 15 pixels. The measured surface strain in the x-direction is shown in Fig. 2.7 at an applied load of 3 kN. The two red boxes indicate the areas for comparison with the FE results. 2.4.3 Numerical Simulation Results The finite element (FE) simulations successfully modeled the complete pull-out response of the damaged and undamaged Pi-joints and the results are shown in Fig. 2.5. Figure 2.6 shows the failure mechanisms observed in the FE simulation. The initial onset of failure occurs in the web/Pi region (see inset in Fig. 2.6) and continues asymmetrically on one side of the web. This onset of failure does not reduce the load carrying capacity as the load transfer continues mostly across the other bonded 0 2000 4000 6000 8000 10000 12000 14000 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 Peak Pull-out Force (N) Top-Web Displacement (mm.) Experimental Response Simula on Predic ons Fig. 2.5 Comparison of the complete pull-out response of the experiment and the FE simulation C. Sebastian et al.
17 side of the web/Pi. As the applied load increases, the subsequent onset of failure occurs in the adhesive/bond between the preform and the base. This failure initiated from the termination of the preform with the base and continues until separation of the Pi-preform from the base. The peak pull-out loads from the simulations agreed very well with the average experimental data as shown in Fig. 2.5. Similar load eccentricity and mis-alignment issues have been reported by Kesavan et al. [6] who applied the pull-out load at an angle of 0.55° counterclockwise to the vertical axis (web) in order to obtain better agreement between experimentation and computational results. Such modifications were not considered made in this study and all loads were applied to the web. Additionally, the experimental data provided in Fig. 2.5 shows the standard deviation (as error bars) for each Pi-joint studied. The variation in experimental data may be due to many factors including manufacturing inconsistencies, variation in material properties of constituents, errors induced in experimental test-setup, instruments etc., Similarly, the numerical simulations assume average material properties and ideal test configurations. 2.4.4 Comparison of DIC and FE Using Image Decomposition Although the pull-out force data matched well between the experiments and the simulation, it was desired to perform a more thorough validation of the simulation. Figure 2.7 shows the measured strain in the x-direction from the DIC experiment (left) compared to the x-strain found from the FE simulation (right). A recently published CEN document [26] recommends the use of image decomposition using Tchebichef polynomials to make quantitative comparisons of measured and predicted strain fields as part of a model validation process, This approach, which has been used previously to validate models of a Fig. 2.7 Comparison of the principal strain from DIC ( left ) and FE ( right ). The highlighted boxes indicates where there the image decomposition comparison was performed Fig. 2.6 Failure mechanism for the pi-joint in the simulation 2 Analysis of a Composite Pi/T-Joint Using an FE Model and DIC
18 vibrating plate [27] and a composite protective panel [28], was adopted here. The image decomposition using the discrete Tchebichef moments can only be performed over a rectangular domain, so the pi-joint was divided into two section for analysis. One rectangular patch was taken from the vertical section, and another from the horizontal section. The Tchebichef polynomials were used to represent the strain component in the x-direction from both the experiment and the simulation following the procedure described in [28]. A reconstruction of the original strain image was performed from the Tchebichef coefficients to ensure that they were an accurate representation of the original data. In this fashion, 136 coefficients were used to represent the strain data from the vertical section and 176 coefficients from the horizontal section. The larger number of coefficients required to describe the horizontal section would indicate that the strain distribution was more complicated in this region. The decomposition was performed on both the experimental and simulation data sets, producing a set of coefficients representing each strain field. These coefficients were then plotted against each other for the horizontal and vertical sections, as shown in Fig. 2.8. The experimental coefficients were plotted along the x-axis, and the simulation coefficients along the y-axis. If there were perfect agreement between the two sets of coefficients, the points would all lie along the line with gradient of unity (solid line in the figure). Deviation from this line indicates differences between the experimental and simulation data sets. In reality, uncertainty in both the experiment and the simulation will cause deviations, and so it is unrealistic to expect all the points to fall on this line. For that reason, Fig. 2.8 has been plotted along with confidence bounds (dashed lines) which relate to the experimental uncertainty. The bands are defined by combining of the uncertainty from the experiment and from the image decomposition reconstruction process. For the case shown here, all of the moments fell within the confidence bounds. The points are primarily concentrated along the solid line, but there are a few moments close to the dashed lines, particularly in the comparison of the vertical section. 2.5 Conclusions The performance of adhesively bonded structural composite Pi/T-joints was evaluated by experimental pull-out tests and numerical simulations. The values of the peak pull-out force recorded experimentally matched well with those from the numerical simulation (~10 % difference). The vacuum assisted resin transfer molding (VARTM) technique was found to be a successful technique to produce Pi/T-joints. 3-D digital image correlation was used to capture the displacements and strains in the pi-joint during the pull-out tests. The CEN procedure for making quantitative comparisons of measured and predicted Fig. 2.8 Comparison plots of the most significant moments from the experiment and the simulation for the horizontal section ( left ) and the vertical section ( right ). If the data were in perfect agreement, all points would fall on a line with a gradient of unity ( solid line ) the dashed lines represent confidence bands equal to twice the uncertainty in the data from the experiments C. Sebastian et al.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==