2012 William M. Murray Lecture: Some Curious Unresolved Problems, Speculations, and Advances in Mechanical Fastening

Cloud, G. L.

doi:10.1007/s11340-013-9736-3

2012 William M. Murray Lecture: Some Curious Unresolved Problems, Speculations, and Advances in Mechanical Fastening

Issues and Opportunities for Research in Joining

Published: 26 April 2013

Volume 53, pages 1073–1104, (2013)
Cite this article

Download PDF

Access provided by Autonomous University of Puebla

Experimental Mechanics Aims and scope Submit manuscript

2012 William M. Murray Lecture: Some Curious Unresolved Problems, Speculations, and Advances in Mechanical Fastening

Download PDF

G. L. Cloud¹

670 Accesses
8 Citations
8 Altmetric
1 Mention
Explore all metrics

Abstract

Mechanical joining is one of the oldest, most important, and most neglected aspects of engineering design of machines and structures of all types and sizes. Approximately 250 U.S. companies manufacture fasteners worth over $8 billion per year. There are 18,000 fasteners in a common fighter jet airframe, and fasteners account for roughly one-third the cost of a typical airplane. Yet, most failures of structures, including aircraft, originate at fasteners, suggesting that improved understanding of fastener mechanics, better design criteria, and informed applications of fundamental knowledge are required. This issue is exacerbated by increased demands on systems, particularly in the transportation and military sectors, and by the growing use of composites, for which current fastening practice seems to be underdeveloped owing to the complexities of material structure and response. This lecture first traces a brief history of mechanical joining, its importance, and the problems faced by engineers in designing for fastening. Research and development of fasteners through analysis and experiment are complicated by the large array of variables involved, and investigators must have at hand an extensive array of experimental and analytical techniques as well as an appreciation of the practicalities of fastening. Verification and validation of findings are crucial, and extrapolation is fraught with pitfalls. Described subsequently are some examples of experimental results that generate speculation and that might provide points of entry for investigators who are willing to take on difficult challenges where progress would be valuable but is not easily realized. These cases include, among others, odd and perhaps dangerous behaviors resulting from coldworking holes in engines and structures, impact stresses caused by joint slippage in composites, the use of inserts to control stress concentrations, difficulties in applying sufficient clamping force in composites, merits of hole shaping, and unusual configurations of conical washers. Finally, some ideas for hybrid joining and for systems that allow quick field assembly/disassembly are briefly described.

Introduction

An Overview of Durability and Damage Tolerance Methodology at NASA Langley Research Center

Materials considerations for aerospace applications

Article 27 November 2015

Use our pre-submission checklist

Avoid common mistakes on your manuscript.

Introduction and Orientation

Purpose, Origins, Scope, Limitations

We offer this written version of the SEM 2012 William M. Murray Lecture with abiding humility and devout appreciation upon being selected as the Lecturer. The journey has been long and enlightening. May this paper be accepted with interest as well as the granting of grace for its probable shortfalls.

This presentation is limited to mechanical fastening, already too large a topic for a single paper, but with the addition of some results and thoughts about hybrid fastening. We hope that a similar treatment of adhesive fastening for metals, plastics, and composites will soon appear.

A main purpose of this presentation is to point out several areas in which quality research on mechanical fasteners is needed in order to gain understanding of the basic mechanics of fastening as well as to improve fastening systems as we seek to design structures and vehicles that are more efficient in both fabrication and performance. Works performed by the author and his colleagues over a span of many years for manufacturing concerns, the U.S. Army, and the U.S. Air Force are briefly described to illustrate the origins of the problems still to be faced and the limited progress achieved thus far. The list of cases is neither exclusive nor is it all-inclusive. It is possible, indeed likely, that, in the interim, some of the issues mentioned have already been addressed.

A variety of suggestive experiments and results are offered minus complete descriptions of techniques and with less-than-thorough discussion of the full implications of what was learned. Much of this work has not been published before, partly because it is still not complete. Further, considerable speculation is indulged because it leads to insights as to what needs to be done and it creates an environment for the development of potential advances in fastening technology. A fundamental engineering point of view is adopted whenever the option is offered.

Finally, this exposition is not a proper research paper, as most experimental details are missing. Nor is it a review paper. Given the unique character of the Murray Lecture and space considerations for this print version, little attempt is made to provide a comprehensive list of citations to the works of the author, let alone of the many other investigators, particularly experimentalists, who have labored mightily to gain useful results. Rather, only a variety of the most relevant and most accessible publications are cited whenever possible, perhaps with an unintended bias that might be perceived as ego-centric; and these will lead to the greater body of literature. We hope that none feel slighted by the necessary choices made.

Does Mechanical Fastening Deserve our Attention?

To begin to answer this question, let us ponder just a few of the available relevant statistics:

Over 200 billion industrial fasteners are used in the USA every year, and 26 billion of these are absorbed by the automotive industry alone.
The current annual value of fasteners used in the USA alone is about $10 billion and is projected to reach $13 billion in 2013.
There are about 350 dedicated manufacturers of fasteners in the USA.
Approximately 18,000 fasteners are used in the F-18 jet fighter, which means, if the average weight of a fastener is only 0.5 oz., then this rather small airplane is carrying over 500 lbs. of bolts.
Fasteners typically comprise about 1/3 the cost of an airplane, the same as the engines.
Most machinery and infrastructure failures originate at fasteners. In the late 1970’s, an air force study showed that approximately 70 % of aircraft failures involved mechanical fasteners.
Field assembly, disassembly, repair, modification, and reassembly are important in many applications, particularly in military vehicles and heavy equipment.

Many of us take fasteners for granted both in our design work and in everyday life. After all, we can select from hundreds of types at any home center, the local hardware, or a machine shop supply depot; so it is natural that we limit our design options to what is easily accessible. This trend is unfortunate from both engineering and economic viewpoints. A senior engineer at a large maker of mechanical fasteners told the author many years ago that, “Fastener choices are almost always made too late in a project, resulting in necessary redesign and serious delays, if not outright failures and recalls.”

In fact, many unanswered questions and opportunities for research exist in this field, and these questions are of increasing importance with growing demands for structures that are lighter, safer, and stronger. Unfortunately, fastener research tends to be ignored outside a relatively small but dedicated cadre of investigators, and necessity often dictates that fastener design choices be empirically energized. Two factors seem to drive this trend. One is that fastening problems are very difficult to investigate experimentally or analytically because they are inherently nonlinear and because of the large number of parameters involved, so generalized publishable results are not easily acquired. Another factor is that, given value in the competitive marketplace, development is done in-house, is not broadly based, is not sponsored, and publication of results is not rewarded.

An Example—Airplane Fastener Failure

To illustrate the insidious nature of fastener failure and the huge cost penalties derived from such failures, consider the recently discovered and highly publicized problems with mechanical joinings that grounded the new super jet airliners that are manufactured by one of the world’s largest makers of aircraft. In one case, two types of cracks were found in the wing ribs of an airplane that had experienced only 399 flight cycles in an operating time of 2,454 h. These airframes incorporate around 2000 rib brackets with L-shaped feet that carry bolt holes for attaching the wing skins to the ribs. Two types of cracks were found. The ones of most interest were hairline cracks in the area where the rib foot is bolted to the wing skin. The cracks ran from the bolt hole to the edge of the rib foot. The manufacturer claimed that these failures are caused by high stress and the aluminum alloy chosen for the brackets. He further stated that the cracks are not the result of a design flaw!

Another Example—Composite Automobile Wheel

Ponder another unforeseen joining problem that contributed to the eventual termination of a promising and expensive engineering program. Figure 1 is a photograph showing a composite automobile wheel that was developed several years ago by a major manufacturer. This wheel demonstrated many favorable features including an excellent ride, reduced unsprung mass compared with a typical steel wheel, outstanding appearance, freedom from corrosion, and predicted economies in production. It was actually introduced with success on a famous line of “muscle cars.” Extensive road testing brought to light two related problems, namely: (1) during mounting of the wheel, it was difficult to obtain consistent and sufficient clamping forces when tightening the lug nuts, and (2) the lug nut clamping force was not dependably retained during use.

We were asked by the manufacturer to identify the causes of these problems and develop, if possible, an acceptable solution. Preliminary examination suggested that a key factor was the nature of the lug nut bearing chamfers, which were machined after fabrication of the wheel. Many sample sections were cut from wheels for optical and electron microscopy of the bearing chamfers in various combinationsof conditions including as-manufactured, after various treatments of the chamfers, and after mounting the wheels following standard practice. Figure 2 shows one sample electron microphotograph of the bearing chamfer in its virgin state. Note the fine, sharp stubble of fiber ends that protrude above the machined surface. These and many similar observations suggested that the fiber stubble was preventing full tightening of the lug nuts, as represented in the sketch of Fig. 3. The short stubble, which is sharp and stiff, evidently resists bending and compaction and so creates high friction, even digging into and scratching the lug nut surface, with the result that the lug nuts cannot be fully tightened onto the chamfer. We reasoned that this effect would occur even if sharp carbide tooling were used in the post-machining process so as to cut the fibers flush, or nearly so, with the matrix material. Most of the bearing forces would still be resisted by the fiber ends.

Several solutions to this problem were suggested. The best, which was discovered by the manufacturer even before we were asked to look at the problem, was to use a common antiseize product on the chamfers. The particles in the antiseize fill the spaces between the fiber ends so as to smooth the bearing surface, and the lubricant facilitated tightening the lug nut. A single application was found to be effective through numerous mountings of the wheel. Other solutions suggested and tried with various degrees of success included, among others: (1) case hardening the lug nuts to prevent the digging-in of the fiber ends; (2) dry lubricants containing Teflon ®, graphite, and similar materials; (3) simply removing the soft galvanize layer from the nuts; (4) using a thin metal interface that snaps into the chamfer and remains in the wheel; and (5) not post-machining the chamfers so that the lug nuts would bear directly on the resin-rich surface layer, a solution which raised the possibility of creep and wear with subsequent loss of clamping.

The difficulty with all solutions examined is that they are not acceptable in aftermarket sectors because of safety concerns. Tire emporiums have a standard practice for mounting steel and aluminum wheels. They cannot guarantee that, for example, adequate antiseize compound would be in place after, say, the fifth mounting of a given wheel during a tire rotation. As mentioned above, this promising wheel development program was dropped in near-market stage for a number of reasons, one of which was the lack of an appropriate solution to the lug tightening problem.

Parameters Involved in Fastening

A major issue in fastener investigations at any level is that a great number of parameters must be considered. Some of these are inherent in the nature of mechanical fastening (e.g. clamping force), and others might be introduced as a potential mechanism for improving fastener performance (e.g. changing hole shape). The following is an incomplete, non-exclusive list of variables that immediately come to mind as important:

material thicknesses,
fastener type (bolt, cap screw, rivet, etc.),
material properties and the mix of materials (e.g. composite to metal),
hybrid joinings (mechanical plus adhesive or insert),
preload clamping,
loss of preload (creep and relaxation),
fastener size and scaling effects,
bolt-to-hole clearance,
interference fits,
tapered threads,
washers (flat, spring, conical),
hole shape (non-round holes, tapered holes, etc.),
fastener shape (non-cylindrical, tapered, etc.),
hole treatment (coldworked, peened),
inserts and bushings,
fastener arrays (shape and spacing),
edge distance,
torque (shear) effect upon tightening (especially for laminates),
blind fasteners (one-side access),
impact, blast, stress and shock waves,
manufacturing methods.

We conclude that, when performing theoretical analysis, numerical modeling, or experimental investigation, the test matrix tends to become very large if any generally useful results are to be obtained. A related very important fact is that validation and verification of all models and experiments are necessary, given the complexity of fastener mechanics. Major dedication of resources is required. This prospect can be either daunting or stimulating, depending on viewpoint.

Concerning Methodologies

Owing to the complexities of fastening research, particularly as experimentation is necessary, critical attention must be given to measurement and test methods. At the fundamental level, the investigator should hold at the forefront of the mind and preach vigorously to all involved the principles that undergird good experimental practice, which may be summarized as follows:

Choose the experimental method to fit the problem, not modify the problem to fit a favored technique, as is too often the case.
Know the difference between accuracy and sensitivity.
Measure exactly what is needed to minimize data manipulation.
Analyze error propagation, or it might burn you.
Keep experiments as simple as possible.
Calibrate often.
Work in differential mode if you are able.
Extrapolation can be useful, but it cannot be trusted.
Validate and verify results by independent methods.
Avoid circular reasoning when interpreting results.
Design a double-blind protocol if possible, although this option is often not viable in the physical sciences.

The complexity of fastener research and the principles of good experimentation dictate that the investigator must have at hand and thoroughly understand a diverse array of measurement techniques. Chosen from the many approaches available in our laboratory, the following measuring devices and methods have been used in our fastening research.

photoelasticity
- transmission
- embedded polariscope 3-D
digital speckle pattern interferometry
- surface
- embedded 3-D
white-light speckle photography
three-axis, six-beam moiré interferometry
moiré grating photography with optical Fourier processing
- surface
- 3-D with multiple embedded gratings
resistance strain gages
- surface and embedded
fiber optic strain gages
- Fabry Perot and Bragg types
- surface and embedded
instrumented bolts and washers
optical measuring microscopy
electron microscopy
laser extensometry
assorted testing machines.

Issues Arising from Coldworking Fastener Holes

Following a brief introduction to the coldworking process and experimental methodology, this part describes in varying degrees of detail six experiments that address certain questions connected with the common practice of cold-expanding fastener holes to improve fatigue performance. No claim is made that these questions have been fully and generally answered. A summary of arising questions that remain to be answered through experimental and analytical research is offered at the end of this section.

The Cold-Expansion Process and Summary of Potential Problems

For many years, fastener holes have been cold-expanded in order to induce a residual stress regime around the hole boundary that resists the initiation and propagation of cracks, particularly in structures that are subject to cyclic loading [cf. 1–4]. The coldworking is usually implemented by one or another of several available proprietary procedures that involve forcing an oversize mandrel or ball through the hole. Often a sleeve is used, which may or may not be left in the hole depending on the process and application. Coldworking has been extensively used in rework or salvage situations to extend the fatigue life of structures, but it is also implemented in OEM applications. Typically applied to metals, coldworking and interference fits have been explored in recent years for use on certain types of composites and sandwich structures [5, 6].

The efficacy of coldworking fastener holes in joining practice has been well demonstrated. There remain, however, several areas of concern with this procedure that have safety implications, and it seems that these issues have not yet been adequately investigated. The following is a summary list of these issues, which are closely inter-related:

effects of in-plane compression on the residual stress/strain field,
interactions between holes,
hole shape and position after coldworking,
effect of coldworking order on the residual stress field,
stress induced in near plate edges,
fastener arrays: pattern and spacing,
stress concentration increased in certain locations,
three-dimensional nature of the stress field.

The following sections outline first the experimental methodology and then present sample results drawn from research on the problems mentioned that highlight behaviors that are unexpected and that could raise questions about the wisdom of prescribing coldworking in the absence of justification through adequate testing. The coldworking equipment and process marketed by J.O King, Inc., which uses a solid sleeve, were utilized as the standard in these studies. One experiment used the so-called “split sleeve” technique. Not all of the issues mentioned above are specifically addressed in this presentation.

Methodology

Characteristic of problems that involve elasto-plastic material deformations is that the process involves several discrete non-reversible steps and a very broad range of strain. Most of our studies of coldworked fastener holes were accomplished through the use of intermediate-sensitivity moiré, which requires high-resolution grating photography plus Fourier optical filtering. The procedure is described in detail in various reports and books [3, 4, 7–12], so it will not be repeated here. It does seem worth the while to remind ourselves of a few advantages of this technique that make it especially suitable for the problem at hand:

Data for all stages of the experiment are permanently stored.
Any two stages of the experiment can be compared in differential mode.
Sensitivity multiplication is easily attained after data storage.
Sign and degree of pitch mismatch can be chosen during data processing.
Signal/noise ratio ranges from acceptable to excellent even after some grating damage.
Fringes are closely packed so dependable displacement and strain values can be obtained.

Figure 4 shows a much-reduced clip from a moiré fringe pattern that was extracted from these experiments. In this instance, pitch mismatch is imposed, but there was no fringe multiplication. In other cases, fringe multiplications of two and three were utilized to create a fringe density that is too high to be adequately reproduced in this publication.

Effect of In-Plane Compression Stress on Residual Stress/Strain

The Question and its Origins

The question posed is this, “Do compressive in-plane loads modify the favorable residual stress and so reduce or even negate the benefits of coldworking?” The question apparently arose when coldworking was being adopted to extend the service life of military aircraft. The sketches of Fig. 5 illustrate the origins of the problem. Suppose that the fastener holes in the wing skins of an airplane were coldworked. The finished airframe is then subjected to several landing and take-off cycles, some of which are probably severe, so as to create large alternating in-plane compressive and tensile loads in the wing surfaces. This effect is easily seen if you look out the window of a airliner so as to observe wing deflection during even ordinary take-off or landing. If you happen to look along the wing surface under glancing light, you might notice bulging of the skin between the rows of fasteners, which indicates that localized elastic plate buckling is occurring. Such behavior is entirely acceptable, but imagine what would happen if one of the rows started to “unzip.”

Experiments

Laboratory tests were conducted to determine whether cyclic in-plane compressive loads modified the residual stress/strain state near a single coldworked hole in coupons of 7075-T6 aluminum that were 0.25 in. thick [13, 14]. Figure 6 shows the essentials of the experimental scheme as well as the results for one of the three experiments conducted.

After applying the moiré gratings, photographing them, and analyzing the results to establish baseline data, the holes were coldworked and the residual strain profiles along several longitudinal x-axes at various y-distances from the hole were determined by the technique named above. Of particular interest were the maximum longitudinal strains (ε_x) at points where the chosen x-axes intersected a y-axis passing through the center of the hole, and these values could be simply picked off the strain profiles. In other words, ε_x at the critical points is determined as a function both of applied load and of y, where y is the distance from the hole edge. The first set of points in the figure, marked “after coldwork” show these initial strain values, which agreed with those obtained in comprehensive static experiments [3, 4].

Now, after the coldworking and initial strain measurement, the specimen was placed in slotted sockets in a testing machine so at to apply an increment of in-plane load that created a far-field compressive stress of 8 ksi. Subsequently, the specimen was removed from the loading rig and grating photographs were recorded and processed in order to measure again the strains at the points mentioned above, and these values were plotted as shown in the figure as the second set of data points. This process was repeated for 16 and 24 ksi far-field stresses.

After measuring the strains following application of static 24 ksi stress, the specimen was put back into the testing machine and subjected to 50 cycles of 0–24 ksi far-field stress, after which the strains were again measured. Twenty-four ksi was chosen as the peak value of the cyclic stress because, at the time, this value was considered to be a target design stress level for aircraft wing skins.

Subsequent to the cyclic load, static far-field stresses were repeated at increments up to 60 ksi, at which level plastic buckling of the coupon became evident and the experiment was terminated.

Observations

As expected (or feared), the residual strain produced by coldworking decreased consistently with even small values of static in-plane load and continued decreasing until plate buckling occurred. A most startling observation was that the residual strains at points within .04 in. of the hole edge somehow recovered to approximately the starting (after coldwork) values during the load cycling at 24 ksi. After the first increment of static load following cyclic exposure, the residual strain at points nearest the hole was found to again decrease, but the strains at the two points furthest from the hole recovered to original values. Subsequently, increased static loading caused consistent decline of the residual strains at all points, reaching, in the end, values approximately half the starting values.

The recovery of residual strain during or immediately after cyclic loading caused considerable anxiety. It was contrary to thoughtful expectation. Is this the truth? If so, what is the cause and what is the meaning? Confidence in these findings was gained by re-analyzing the stored grating data several times with the same results. A control specimen that was not coldworked was tested with the identical procedure, and there was no effect of in-plane load. Later, a second coldworked specimen was created and tested by different people with somewhat different equipment and procedures in a different laboratory, and behavior that was similar but not quite identical to that of the first specimen was observed.

Summary

The following general remarks were derived from investigations of the effects of in-plane loads on coldworked fastener holes:

The moiré technique involving grating photography and optical spatial filtering proved to be well suited for research on coldworked holes.
Except when cyclic loads were applied, the residual strains decreased linearly with static load.
The residual strain and, likely, the favorable residual stress that is the goal of coldworking are compromised by application of in-plane compression.
The residual strain recovered to its original static value upon the application of cyclic loading at the standard design stress.
The recovery of residual strain with cyclic loading could not be rationalized except in a general way having to do with redistribution of residual stress. This behavior seems to be not widely known, nor, to our knowledge, has it been explained.

Effect on Residual Strains from Coldworking Adjacent Holes

The Question

Usually, mechanical fasteners are used in an array of one sort or another, and, in many cases, the fasteners are relatively close together. Careful measurements have shown that, in metals, cold expansion of a hole creates elastic stresses well outside the local plastic zone. The following question arises: How are residual strains between holes affected by coldworking adjacent holes in an array? There is no easy answer to this question given the large number of parameters involved, and space limitations prohibit exhaustive discussion. We present here only a few limited results that underline the complexity of the problem and the potential risks involved.

Example Experiment

Figure 7 illustrates a 0.25-in.-thick specimen of 7075-T6 aluminum with a 5-hole in-line array of fastener holes that were coldworked in order left to right, one of a number of array patterns and coldwork sequences that were examined [13–16]. The holes in this particular array are moderately close together (1.75 x hole diameter), so it might represent a worst-case example. This figure also presents a much-reduced moiré pattern for the region surrounding holes 2–4 of a 4-hole array in a different specimen after all holes were expanded in the order 1-2-4-3. Some fringes are lost in this reduction, but one need not be a moiré expert to perceive that the strain profiles between the holes differ widely from one another.

Observations

Figure 8 shows the variations of the residual strains created between the first few holes as coldworking progresses along the line of holes for the 5-hole specimen. Unfortunately, the hole sequence was swapped end-for-end in these graphs so the first hole is now on the right. Also, here, compression strain is positive, contrary to usual practice.

Consider first the left-hand graph of Fig. 8, which gives the radial strain profiles between holes 1 and 2. The dotted line shows that, when only the first hole is expanded, a compressive strain of about 4 % is created at that hole edge, but the residual strain falls off and changes sign to 4 % tension at the edge of the adjacent non-coldworked hole. The solid line shows that, when hole #2 is subsequently cold-expanded, the residual strain at the edge of hole #1 remains at about 4 % compression, but the radial strain at the edge of the second hole is now drastically amplified to a value of about 10 % compression.

Now, look at the right-hand graph of Fig. 8, which shows the strain histories between holes 2 and 3 as coldworking progresses. The dotted line is the result when only holes 1 and 2 have been coldworked. The profile is similar to that appearing in the left panel, but the maximum compressive strain near hole #2 is now twice as high at 8 %, and it oscillates and drops to about 4.5 % tension at the edge of non-coldworked hole #3. The solid line shows the strain profile after hole #3 has been coldworked. The residual strain reaches 6 % compression at about .005 in. from this hole, but, for some unknown reason, drops off to only about 1 % at the hole edge.

Summary

These results, and many others obtained from these experiments, demonstrate some characteristics that consistently arise when coldworking arrays of holes, namely:

When coldworking one hole in an array, unfavorable large (e.g. 4 %) tensile residual strain and, likely, a matching tensile residual stress is produced at the edge of the adjacent non-coldworked hole.
An implication is that if one hole in a “tight” array is coldworked, then all of them should be coldworked.
Large and unpredictable variations of residual strain will likely arise between adjacent coldworked holes.
The residual stress state near the holes is not consistent, nor is it as expected from tests on single holes.
The interactions of the residual strain fields between cold-expanded fastener holes is very complex and depends on many parameters and process variables including hole spacing, type of array, and, especially, the sequence of coldworking.

Hole Movement and Change of Shape

The Question

Fundamental mechanics knowledge suggests that coldworking a hole in an array will cause adjacent holes to permanently move, possibly causing partial interference fits, gaps between fastener and workpiece, and misalignment of holes in workpieces that are being fastened together. Hole motion and shape change have been observed in the field, and one is led to suspect that these effects modify the expected stress states in the workpieces and the fastener.

Experiment

Optical measuring microscopy was utilized to measure the dimensions and shapes of the coldworked holes in various arrays [15, 16]. A sample result for the 5-hole in-line pattern of Fig. 7 is presented in Fig. 9. The shaded circles represent the original holes, and the outlines of the final holes are shown as a solid line. The hole boundary movement is magnified 5X for purposes of visualization.

Observations

The findings from these experiments are largely as would be visualized based on experience with material behavior. After hole #1 is coldworked, its boundary is nearly concentric with the original hole; but the act of expanding that hole appears to push the adjacent inter-hole material and hole #2 to the right. This process continues on down the line, and all the holes get pushed over except for the last one, which, like hole #1, ends up being distorted but not moved laterally, probably because the material to the right of that hole is solid and blocks movement. Further, if a hole is between two others, it is both moved and squashed laterally into an oval shape. The maximum boundary displacements are on the order of 0.008 in., which might not seem excessive until one realizes that it exceeds the coldworking radial interference of 0.0073 in. used in this example.

Summary and implications

These experiments suggest that approximately 80 % of the expansion of any fastener hole occurs toward the adjacent unexpanded hole.
All but the first hole coldworked in a linear array are moved and undergo a change of shape, becoming oval.

A serious implication is that hole movement caused by coldworking might cause misalignment of the holes in the workpieces being fastened together, thereby creating additional stresses if the fastening process forces the holes into alignment. Another implication is that a partial interference fit of the fastener might result and/or gaps could exist between the fastener and the workpieces. If the fastener were a bucked rivet, its shank would be deformed to accommodate the non-round holes that are likely misaligned. All of these effects depend on the spacing of the fasteners and the order in which the holes are coldworked.

Stress/Strain State in Close-by Plate Edge

The Question

In many structures, including aircraft, boats, and civil infrastructure, fasteners are necessarily placed near the edges of plates and shear panels. The question arises as to whether coldworking these fastener holes creates a problematic stress/strain state in the close-by edges of the workpieces. Of course, the result depends on many factors including the obvious ones of edge distance, hole spacing, and order of coldworking. Again, we present here some suggestive results from only one set of our experiments.

Example Experiment

As in the studies outlined above, moiré grating photography with optical processing was used to explore the residual strain states between a row of holes and a plate edge as coldworking progressed [13, 17]. Various coldworking orders were used. Figure 10 shows one such specimen, the coldworking specifications, and the quantities that were to be measured. The coldworking order for this 4-hole specimen was 1-2-4-3.

Observations

Figure 11 is a summary plot of the compressive strain profiles between each hole and the plate edge after all the holes in the specimen of Fig. 10 were expanded. The most striking features of these data are that the strain profiles vary widely and, for three of the holes, the maximum strains are remote from the hole edges where it is expected. Further, the last hole expanded, #3, has very low levels of residual strain in the region, being almost zero at the hole edge where it is needed. Possibly the top edge of the sleeve slipped below the specimen surface during the coldworking of this hole or something else went wrong. But, the strain fields do seem to be interacting in this region between the holes and the plate edge. An encouraging finding is that, even though the edge distance is fairly small, the compressive strain perpendicular to the edge falls to zero at some distance from the edge for all the holes.

The residual tensile strain profiles along the plate edge for three cases are presented in Fig. 12. Concentrate attention on the upper profile, which shows the final residual strain after all the holes are coldworked. While it oscillates somewhat, the strain in the plate edge is seen to be about 1 % tensile, meaning that a large tensile stress exists all along the edge, likely with some bulging of the edge in the regions adjacent to the fasteners. This finding elicits the specters of stress corrosion and/or fatigue crack initiation somewhere in the edge.

Summary

These experiments on the strain states caused in near plate edges by coldworking holes suggest that:

Large variations in the residual compressive strain profiles exist between the hole and the plate edge.
The maximum compressive strain often appears somewhere between the hole and the plate edge rather than at the hole edge, contrary to expectations and desired results.
This experiment, the preceding one, and others conducted suggest that the edge becomes bulged slightly outward adjacent to the holes.
Large residual tensile strain (e.g. 1 %) is induced along the edge of the plate, which creates an environment conducive to stress corrosion or fatigue cracking.

Stress Concentrations Created by Coldworking

The Question

That cold-expanding fastener holes can actually create significant stress concentrations is best demonstrated by an example drawn from our consulting practice. Figure 13 is a photograph of a jet engine turbine disc that is manufactured of a super alloy. The disc carries near its hub a circumferential array of holes through which pass the bolts or cap screws that bind a stack of discs and spacers together. Standard practice called for coldworking of these holes using the so-called “split-sleeve” process to reduce the probability of fatigue cracks originating at a hole boundary. Given anxieties about stress concentrations in the vicinity of the split in the sleeves, the shop technicians are instructed to align the split so that it points toward the center of the hub. The engine maker remained concerned about the stress concentrations and, if they existed, whether they were aligned according to specifications and whether they were serious enough to consider modifications of procedure.

Experiment

To gain insight into the magnitude and location of the possible stress concentrations, the moiré technique was again used. Photoresist was applied to most of the web of the disc, appearing as the yellow stain in the photograph. Gratings were created in the photoresist in regions surrounding three of the fastener holes in a non-coldworked disc. The grating orientation was set so as to measure the displacements in the circumferential direction in the regions of each hole boundary nearest and furthest from the hub. These gratings were photographed using the usual high-resolution techniques in order to establish a baseline moiré record for each hole. The disc was returned to the maker for coldworking of the holes using his standard technique. Subsequently, the gratings were photographed again using the same apparatus as was used to record the original gratings. The grating records were then superimposed with master gratings in the Fourier optical filtering system to obtain moiré patterns before and after coldworking for each hole.

Observations

Figure 14 is a much-reduced typical moiré fringe pattern for one of the holes after cold-expansion. Given the difficulty of photographing the gratings on this reflective specimen that had suffered through two-way shipping and standard in-shop coldworking of the series of holes, the fringe pattern is somewhat noisy in spite of the spatial filtering that was done. Still, you need not be a moiré magus to conclude from merely qualitative observation that there exist potential problems arising from the coldworking procedure. The two areas where the fringes are concentrated at the hole boundary bracket the location of the split in the sleeve, and it was not aligned with the center of the hub. Now, the pitch mismatch used was chosen so that, compared with the fringe spacing in the far field, the fringes would come closer together in regions of compressive residual strain. The fringes along the hole boundary furthest from the hub suggest a nicely uniform compressive strain there. But, notice that the fringes are very tightly packed near the edges of the split in the sleeve, so the compressive strains there must be very high. We see also that the fringes in the region between the edges of the split are spaced further apart than they are in the far field, so this segment of the hole boundary must contain residual tensile strain and so possibly be susceptible to fatigue cracking. Finally, the change from large compression to significant tension occurs over a small segment of the hole boundary, so the strain gradient must be very large, a situation not conducive to long life. It is not known whether the manufacturer changed his coldworking process or his procedures.

Summary

Coldworking mounting holes in a turbine disc using the split sleeve process resulted in:

large residual tensile strain in the hole boundary at the gap of the slit in the sleeve,
high strain gradients ranging from large compression to severe tension in the regions where the edges of the slit contacted the hole boundary,
larger than nominal compressive strain concentration in the areas near the slit in the sleeve,
stress concentrations that are not necessarily aligned with the axis of the hub as specified.

Three-Dimensional Nature of Residual Strain Field

That the residual stresses/strains created by the usual process of drawing a mandrel through the hole vary through the thickness of the workpiece has long been suspected. We conducted a study to investigate this question using a novel multiple-embedded- grating moiré technique [11] on specimens of polycarbonate resin [18–21]. Space limitations prohibit discussion of these experiments, but we found that the strain state is indeed three-dimensional, with the interior radial strain being smaller than the surface values. Further, the coldworking created a tensile strain perpendicular to the workpiece near its surface where the mandrel enters. This transverse strain changed to compression in the interior, reached a maximum at about mid-plane, decreased toward the bottom, but then increased again near the mandrel exit side. The minimal radial strain in the interior and the presence of tensile transverse strain cause worry, so the three-dimensional aspects of the residual stress state created by coldworking needs considerable elucidation.

Research Questions Connected With Cold-Expanded Holes

Only a very limited assortment of potentially problematic issues involving coldworked fastener holes have been studied, and the inquiries have not been exhaustive. Many challenging inter-related questions remain for investigators to answer, including:

What is the change of residual stress, if any, in the zone surrounding the fastener hole after static and cyclic in-plane compression?
Where is the elastic–plastic boundary after in-plane static and cyclic loading? Poolsuk and Sharpe [22] skillfully measured this boundary for coldworked holes and provided a basis for further investigations involving remote loading and interactions.
What happens to the residual stress if in-plane static tension and/or fully reversed tension-compression cycles are imposed?
Is coldworking worthwhile or even safe for structural elements that are exposed to in-plane compressive or cyclic loads?
When hole and edge interactions are involved, what are the final residual stresses in the plastically deformed zone near the hole?
Where is the elastic–plastic boundary after an array of holes has been coldworked?
Is fastener performance compromised by hole movement and shape change caused by coldworking adjacent holes?
What is the optimum array of fasteners?
What is the optimum order of coldworking holes in an array?
Do fatigue and stress-corrosion effects arise from the tensile stresses created at plate edges and in the boundary of non-coldworked holes in an array?
Do the stress concentrations created by the split sleeve technique compromise performance?
Can the three-dimensional nature of the strain field created by coldworking be adequately defined, particularly as to the effects of tensile transverse strains.
Can these multi-step elasto-plastic cold-expansion processes be adequately modeled with verification/validation for cases involving in-plane loads, hole arrays, various coldworking sequences, near plate edges, and complex materials?
What are the answers to these and similar questions for materials other than those studied, particularly composites and sandwich plates.

Bolting and Washers for Thick Composites

This section offers first some orientation as to the general mission undertaken. Subsequently, the importance of bolted joining of so-called thick composite sections to one another and to sections of other materials is described briefly. The rationale for limiting studies to the single-lap joint is presented along with some discussion of the problems inherent in this type of joining. Then, we examine a limited array of experiments and numerical analyses that were designed to advance the fundamental understanding of mechanical joining of thick sections; to identify, refine and validate suitable measurement techniques; and to develop useful design protocols for these joints. Not all goals were achieved, and, as often happens, more questions arose than were answered. Some research gaps are summarized at the end of this segment.

The Mission and its Relevance

The general mission of this research has been to develop validated design protocols for vehicle structures made of thick composites for use in severe service environments. The objective is approached through: (1) expanded fundamental research that advances understanding of the basic mechanics of fastening; (2) function-specific studies for aircraft, ground vehicles, watercraft and subsets of these categories; and (3) the eventual development of design standards, especially for severe service environments.

Most composite structures, e.g. civilian transport and sport vehicles, are made of relatively thin sections. Heavy-duty vehicles, particularly those used in the military sector, must be made of much thicker laminates so as to be resistant to ordnance impact, blast, and fire as well as to tolerate heavy loads resulting from extreme usage in rough terrain or heavy seas. Composites are being introduced as the material of choice for some components of military vehicles, for example the workhorse HMMWV. A goal is to eventually fabricate entire load-carrying structural components and outer hulls of land, air, and water machines of combinations of composite materials. Armored vehicles and ships are included in this initiative, and some prototypes of armored combat platforms have been constructed and tested. Imagine going into armed conflict inside a plastic tank! Of course it is not that simple. Composites have been widely used in military aircraft, and the all-composite plane is supposed to be just around the corner. Beyond the military sector, composites are increasingly used in heavy-haul road and construction equipment. In all these applications, “lighter, safer, stronger” are the watch-words.

Some important hard-learned lessons and criteria must be remembered when introducing composite components to existing platforms or starting from scratch with a new all-composite design. These include:

Simple substitution of a composite part for a metal one is generally not satisfactory.
Using with composites a fastening scheme that is known to serve for metal components is not a good idea, and early failures will result if this lesson is ignored.
For combat and remote site applications, quick field assembly and disassembly are important.
Inspection, repair, and validation of repair are necessary for operation in severe environments, no matter the material used.
Composites can be fabricated into large complex shapes, e.g. an entire monocoque or monolithic chassis, which reduces the number of joints but which imposes a new set of fastening criteria.
The failure mechanisms of composites, particularly in blast and fatigue, are not yet well understood, and safety factors that include fastener design must be incorporated until we gain sufficient knowledge to take full advantage of the possibilities offered by these remarkable materials.

The Single-Lap Bolted Joint

The single-lap bolted joint, shown in the simple illustration of Fig. 15, is the first choice for perhaps the majority of joinings in structures and vehicles. On the one hand, this mechanical joining is the simplest possible. On the other hand, it is the most complex because, especially for joining thick workpieces, three-dimensional effects derived from secondary bending and tilting of the fastener create large stresses and amplify stress concentrations. In general, these three-dimensional problems are difficult of solution either analytically or experimentally. Inevitable joint slippage arising from bolt-to-hole clearance is a complicating factor. Further difficulties arise when complex materials, such as composites, are to be joined in this way. Curiously, relatively few published works about this seemingly simple joint are extant, possibly because of the forbidding difficulty of obtaining fundamental, validated, publishable results that would lead to design guidelines. We wryly observe, based on our own experience, that it is much easier to analyze or perform experiments upon symmetrical double-lap pin joints; and there exist many reputable papers on that subject.

Even though, from a mechanics viewpoint, the single-lap joint is complex, its mechanical simplicity makes it attractive for severe applications. It facilitates field assembly, disassembly, inspection, part substitution, and repair; and these characteristics alone might be sufficient to choose this type of fastening.

Given its practical importance and the relative dearth of useful relevant design information, the experiments described below focus solely on the single-lap bolted joint.

Validation/Verification Example

Before examining the single-lap joint, it seems worthwhile to emphasize through an example the importance of validation and verification of numerical and experimental analyses [23, 24]. Figure 16 shows a false-color whole-field surface strain map for a pin-loaded FGRP laminate as obtained via digital speckle pattern interferometry (DSPI). The strain concentration factor along the axis of symmetry was extracted from the measured strain values (not the fringe map), and these are plotted as the solid lines in Fig. 17. The values obtained from finite element analysis (FEA) are plotted as dotted lines in this figure. The agreement is excellent except right at the loaded hole boundary. The DSPI values near this boundary are likely contaminated by error resulting from filtering and/or taking the derivative of the displacement map. The important point is that, overall, the agreement is fine. Probably these independent studies are not both wrong, so confidence that both are correct seems appropriate.

Single-Lap Bolting of Thick Isotropic Materials

Purpose

Preliminary to the experimental and numerical analyses of single-lap bolted joints of composite and aluminum, we explored, developed, and refined suitable techniques on similar joinings of isotropic materials, in this case a birefringent plastic and aluminum [25–27]. The objective was to determine, with validation, the strain fields near the fastener, particularly in the load-bearing region, on the surface and in the interior of a single-lap bolted joint of “thick” plates of these materials for various clamping forces and applied loads. Although seen as a pilot study, the results were expected to be useful in themselves because of the previously mentioned shortfall of information about the behavior of such a joining.

Methodologies

The techniques chosen for stress analysis of the bolted joining of thick isotropic materials were:

embedded polariscope three-dimensional photoelasticity,
surface and embedded resistance strain gages,
finite element analysis using HyperMesh 5.0, ABAQUS 6.2 and ABAQUS/Viewer 6.3-1.

Embedded polariscope photoelasticity is one of the very few methods that can be used to examine the stress distribution inside an object, and it is quite easy to implement. The technique is well known [cf. 28], so it will not be described in any detail here even though it is, perhaps, on the verge of being forgotten. Its main disadvantages are that it yields the stresses in only one thin slice of the object; it gives directly only the maximum shear stress distribution; goniometric compensation is not possible because the polarizers are fixed; and, unless several specimens are made with linear polarizers, isoclinic data indicating principal directions are not to be obtained.

Figures 18 and 19 show how the specimen was made. It was machined from a plate of PSM-9 (CR-39) photoelastic resin having a thickness of 0.5 in.. The bolt hole was then rough-drilled. Next, a thin slice was carefully cut from the bearing section below the hole. Resistance strain gages were applied to the cut interior face of the slice. Then, the slice with its strain gages, along with properly oriented circular polarizing material (combined linear polarizer and quarter-wave plate), were glued into the slot from which it was originally cut using the matching cement. The saw kerf provides just enough space for the polarizers, glue, and strain gages. Finally, the hole was finish reamed to provide a smooth sliding fit for ground shoulder bolts having a shank diameter of 0.5 in. Strain gages were installed inside the bolts, converting them into transducers that were calibrated to provide accurate measurement of joint clamping force.

After appropriate curing and hooking-up of the strain gages, the specimen was attached to a similar aluminum plate using the shoulder bolt mentioned above. The bolted specimen was then placed into a dead-weight loading apparatus with a light source behind it and a lens and digital camera on the near side for viewing the isochromatic fringes for various levels of bolting torque and applied load. To maintain similitude to real-life applications, the load path was not arranged so that the load was artificially aligned with the plane of contact. Remember that, in this case, the observer is looking through the specimen from edge to edge.

Sample Results

Overall, good quantitative agreement between the results from the strain gages, photoelasticity, and finite element analysis was attained. Figure 20 shows a sample comparison of the photoelasticity fringe pattern and the maximum shear stress from FEA for one case of bolting torque and applied load. The small strain gages on the interior slice and their lead wires can be seen in the photoelasticity fringe photograph.

A serious problem arose during the testing, however. At higher loads, the cement binding the polarizers and the slice into the specimen failed, probably because of the high shear stresses at those surfaces in the bearing region.

Quantitative observations from this program are too extensive for inclusion here, but they may be found in the papers and the thesis cited above.

Summary

These studies suggested that embedded polariscope photoelasticity and embedded strain gages coupled with carefully implemented finite element analyses are dependable for investigating the three-dimensional stress distribution at critical areas in the region of a mechanical fastener. Subsequently, the FEA analysis was extended to explore the effects of joint modification, such as chamfering the corners of the holes in the inside faces of the workpieces [27]. This simple but uncommon procedure was found to improve the stress distributions by a significant factor, and it should be considered for critical applications pending experimental verification.

Bolted Joining of Thick Composite to Aluminum

Objectives

The major goals of this project were to determine by independent experimental and analytical techniques the stress/strain distributions in the interior and on the surface of thick composite plates that were bolted by means of a single-lap joint to aluminum plates with varying degrees of bolt-to-hole clearance and subjected to various clamping forces and applied loads. If the FEA and Boundary Element Analysis (BEA) results could be validated, then these numerical models could be used to analyze an extensive matrix of cases with only a limited number of experimental checks being necessary. The eventual objective was to establish dependable design protocols for this type of joining [29–35].

Methodologies

As mentioned above, the experimental techniques and numerical analysis methods developed and proven for joinings of isotropic materials were intended to be used to study bolted joints of composites to metals. Reliance was to be placed on the transparent birefringent composite materials that were developed many years ago by Dally and Prabhakaran [36] and by Daniel, et. al. [37]. It turned out that the resins for these materials are no longer available. Thus, considerable effort was expended in developing a new recipe for a composite that could be use in photoelasticity experiments [38]. The main problem is matching the indexes of refraction of the glass fibers and a transparent resin while at the same time achieving appropriate mechanical and stress-birefringence properties. This initiative was eventually quite successful. However, the results obtained from photoelastic calibration tests and experiments on a simple model were disappointing because of creep and relaxation effects in the matrix resin. Clearly, more research is needed in order to develop a successful recipe, so photoelasticity was abandoned for the duration of the study.

Given the necessary adjournment of three-dimensional photoelasticity as the method of choice, several surface resistance strain gages and embedded fiber optic strain gages were adopted in order to measure the strains at several critical locations. Adding only a few (e.g. 6) glass fibers to the FGRP composite should not modify the material properties enough to cause discernible error. To eliminate the scatter in the well known torque-clamping force relationship for bolted joints, an annular bolt transducer made specifically for the purpose was used to measure directly the joint clamping force.

Two specimens were fabricated by hand layup of 60 layers of plain weave E-glass plus epoxy resin to achieve a thickness of 0.5 in. with a fiber volume fraction of 0.35 [34]. For the second specimen, which is the only one that will be discussed here, five Bragg-grating-type fiber optic strain gages (FOSG’s) were embedded during layup by attaching them to the appropriate glass cloth layers with care taken to assure that they did not move while compacting the assembly for curing. Still, keeping track of the locations of the sensing elements was expected to be a significant issue. They were tagged using tiny droplets of a resin mixed with metal powder at specific locations from each of the sensing elements so that their locations could be determined by x-ray if necessary. As it turned out, the tag and gage locations could be observed visually by backlighting the translucent specimen. Figure 21 shows a schematic of the specimen and its final FOSG locations. Following machining of the specimen to dimensions and drilling and reaming the hole, resistance strain gages were applied to the surface of one of the specimens. Figure 22 is a photograph of a finished specimen with the clamping-load transducer in place as it is ready to be loaded.

Finite element analysis was accomplished using LS-DYNA, the method of choice for these problems that involve many contact surfaces, point contact, clearance, slippage, impact, and rigid body motion. The number of mesh nodes typically ran to about 5000 for analysis of these problems. Details of the modeling scheme are omitted from this presentation, which focuses on experimentation. But, understand that the details are important and care must be taken; this problem cannot be simply meshed and plugged into a solver. Some of the FEA results are in the form of “movies” that show the progression of selected stress or strain components in various views as the load is increased.

Sample Observations

Owing to the number of parameters involved, the results for even this one specimen are voluminous. We offer here only a nibble.

Figure 23 shows the color-coded FEA map of the longitudinal component of strain at a rather small maximum load for one level of clamping load, this being the last frame of a time-sequence. The composite plate is on the bottom in this view. Notice first the large (ca. 0.07 %) tensile strains that are created at the top surface of the composite remote from the fastener by the secondary bending inherent in this joint. Second, observe that, even though the bolt-to-hole clearance is small (smooth sliding fit), the tilting of the bolt causes very large (ca. -0.05 %) compressive strain near to but not coincident with the corner where the bolt contacts the composite. These effects are consistently found in analyses of the single-lap joint. The strains observed in the aluminum are, of course, much smaller than those found in the composite.

Figure 24 is a comparison of the measured strains and the FEA results from one specimen at one level of clamping torque and over a load range from 0 to 200 lb. The solid lines are from the FOSG’s and the dashed lines show the values at the FOSG locations as obtained from the FEA. Gage locations correlate with the diagram of Fig. 21. The results from these two independent means agree satisfactorily except for gage #5, which was damaged during specimen handling, as was gage #4. Improved protection of the fiber leads where they exit the specimen has been implemented for subsequent studies.

An interesting feature of these results is that the strains remained very small up to an applied load of about 100 lb., at which point the joint began to slip and take up the clearance. At this point, the strains started to increase rapidly owing to contact between the bolt and the workpieces. An unexpected result is that both the experimental and numerical analyses showed that the highest strain occurred at the location of gage #2 rather than at gage #1, which is closer to the contacting corner. Figure 23 shows this happening in the composite but not in the aluminum, where the contact stresses are overall much smaller than they are in the composite. Unless the hole diameter was not uniform, a possible reason for this behavior, which has been consistently seen, is that the composite might slightly bulge as the clamping load is applied.

Torsional Strains in Composite

Although the experiments are not described here, worth mentioning is that measurement of surface strain by moiré interferometry showed that high torsional strains are created when tightening a nut onto a laminated composite even when a washer is present [29, 39–41]. These strains are caused by the friction between the nut and the washer or laminate; and they would, evidently, cause very high inter-lamina shear stresses near the surface. The only way found to totally eliminate these strains was to use a thrust ball bearing between the nut and the washer, a practice certainly unacceptable in manufacturing.

Summary

The results of both the numerical and experimental investigations are far-reaching, quantitative, and provocative. We must be content here with limited qualitative observations.

A suitable recipe using currently available components for a birefringent transparent composite would be extremely useful for research on mechanical or adhesive fastening of composites using photoelasticity.
Fiber-optic strain gages serve well for measuring internal strains at specific points in composites, and installation is reasonably easy during layup of the material.
Instrumented bolts or a bolt load transducer must be used to accurately measure the clamping load; calculating the clamping force from the bolting torque is not acceptable for fundamental research.
Validation and verification are especially important for these problems that are characterized by stress states and materials that are complex.
Numerical analysis of problems of this type must be implemented with great care.
Secondary bending in the single lap joint, when loaded as it would be in practical applications, creates large strains adjacent to the lap.
Even with small clearance, tilting of the bolt causes large contact strains near the inside hole corner.
The contact strain is maximum at a point on the hole boundary that is removed from the contact surface by roughly 20 % of the plate thickness.
The strains in the region of the bolt remain small up to the point at which the joint slips to take up the clearance, after which the strains increase quickly at a rate that is roughly proportional to the additional load. This result underlines the importance of adequate clamping load to the performance of a bolted joint.
Tightening a nut, with or without a washer, can create high torsional strains near the surface of the composite, with concomitant large inter-lamina shear stresses. This effect is minimized with difficulty.
The fundamental mechanics of bolted joints involving composites, and particularly single-lap bolted joints, have not yet been sufficiently investigated to develop anything approaching adequate general design criteria.

Impact Strain Caused by Joint Slip

The Question

The studies summarized above suggested strongly that additional investigation should be undertaken in order to understand the limiting effects of clearance, clamping force, and slip.