LAYERED-MANUFACTURING OF FIBER-REINFORCED COMPOSITES G. Zak
M. N. Sela
V. Yevko
C. B. Park
B. Benhabib
Computer Integrated Manufacturing Laboratory Department of Mechanical and Industrial Engineering University of Toronto 5 King's College Road, Toronto, Ontario, Canada M5S 3G8 e-mail:
[email protected]
ABSTRACT In this paper, we present a rapid manufacturing process for the layered fabrication of polymer-based composite parts using short discontinuous fibers as reinforcements. This process uses a UV-laser-based system for the selective solidification of the composite liquid. The primary components of the prototype are: (1) fiber-resin mixing subsystem, (2) composite-liquid deposition subsystem, (3) liquid leveling subsystem, and (4) laser-light delivery subsystem. Axiomatic Design Theory was used to validate the design selected for the experimental embodiment of the process. Extensive microscopic examination of the layered composite parts verified that the prototype system can yield comparable layer quality, in terms of accuracy and uniformity, to that of pure-resin parts made by a photopolymer-based commercial system. Furthermore, mechanical testing of these composite specimens showed up to 60% improvement in modulus over the unreinforced layered specimens.
July 1998
1. INTRODUCTION Reinforcement of plastics by fibers has been employed successfully for over fifty years as means of improving the mechanical properties of the manufactured products [1]. Combining high-modulus, high-strength fibers with a polymeric matrix produces a composite material with higher stiffness and strength, and lower thermal-expansion coefficient. The reinforcing fibers can be introduced either in a continuous (long) or discontinuous (short) form. While continuous fibers provide greater relative improvement of the mechanical properties, they also significantly complicate composite-material processing. Short-fiber composites, on the other hand, can be easily manufactured by automated, and hence more economical, methods. Since the late 1980’s, several Rapid Layered Manufacturing (RLM) techniques have been investigated, and some commercially developed. These techniques allow free-form fabrication of complex-geometry parts directly from their CAD models [2]. The most commonly used RLM technique for the production of plastic parts is Stereolithography (SL). As a building material, it employs a liquid photosensitive resin, which is selectively solidified by an ultraviolet (UV) laser beam. Recent research work has targeted improvement of the mechanical properties of polymerbased parts produced by SL methods. For example, in [3], long fibers were added to the polymer matrix by stacking rings with arrays of parallel horizontal fibers stretched across, and then curing the polymer via a standard SL procedure. In another approach [4], continuous fibers were laid out by a dedicated apparatus before curing each layer of the part. Other approaches include using solid inserts within the polymer [5], or building fiber-reinforced shells around solidified resin part [6]. Feasibility studies were also reported regarding the use of discontinuous reinforcements in the form of either 10-15 mm chopped glass fiber bundles or 55 m diameter glass microspheres [7, 8]. Composite samples several-layers thick were produced by manually spreading the glass fibers over the liquid resin on each layer. Fibers were not premixed due to the very high viscosity of the resulting liquid. Improvements of material mechanical properties were reported for fiberbased reinforcements, while no improvement was attained by using microspheres. The above methods of reinforcement (especially those using medium-length or continuous fibers, as in [3, 4, 7, and 8]) are suitable for the rapid layered manufacturing of objects with relatively simple geometric shapes. However, difficulties may arise when applying these methods to the production of objects with small-scale features, thin walls or intricate shapes.
-2-
Thus, herein we propose reinforcement of RLM-fabricated complex-geometry objects by short fibers introduced into the photopolymer matrix. The fact that the photopolymer remains liquid at room temperature simplifies the process of adding the fibers, storing and handling the mixture, as well as controlling the amount of fibers added. Glass fibers were selected for reinforcement due to material’s transparency (for photo-curing) and relatively low cost [1]. Development of our proposed process for RLM of glass fiber-reinforced plastic parts commenced with the investigation of the constituent materials’ properties, which was followed by several iterations of the experimental system design and process analysis. The constituent materials studies concentrated on the identification of the rheological properties of the fiber and resin mixtures and on the surface thermodynamics of the photopolymeric resins. In designing the process, the Axiomatic Design Theory was extensively employed to validate the proposed design embodiments [9]. Process analysis tasks were in two areas: layer formation studies and identification of mechanical properties. The paper will first briefly discuss the improvements achievable by the addition of shortfiber reinforcements. Subsequently, the process design and the current experimental system will be described, followed by a report on our investigations into the part layer quality and mechanical properties. 2. REINFORCEMENT BENEFITS This section will examine the relationship between the reinforcing fiber content and the part’s mechanical properties. A theoretical model based on a “rule-of-mixtures” principle is described first for the case of short glass fiber reinforcements. Next, we present the results of mechanical tests conducted on molded specimens made from photopolymers containing varying concentrations of short glass fibers. 2.1 Theoretical Predictions Beneficial effects of the reinforcements on composite’s mechanical properties derive from the transfer of stresses from the matrix to the reinforcing fibers. By assuming elastic stress transfer, the shear lag theory [10] describes a relationship between the composite’s mechanical properties and the fiber volume fraction, aspect ratio, orientation, and material moduli. The tensile modulus is given by Ec =
1 2 f
Ef +
(1)
m Em
where E f and E m are, respectively, the fiber and matrix tensile moduli, is the matrix volume fractions. The orientation-efficiency factor -3-
1
f
is the fiber and
m
ranges from 0.2 for the three-
dimensionally randomly oriented fibers to 1 for unidirectionally aligned fibers. The fiber-length correction factor 2 depends on the fiber aspect ratio; its value approaches 1 for aspect ratios over 100. The most common formulation for 2 is 2
=1
tanh(na ) na
(2)
where n=
2G m E f ln(2 R d )
,
Gm =
Em , 2(1 + )
2R = d
4
,
and
a=
f
l d
(3)
In the above equation, a is the aspect ratio for fibers of length l and diameter d. Prediction of the tensile strength for randomly oriented short-fiber composites is a difficult task, and no universally accepted theory exists on this subject. The difficulty arises because the material’s ultimate strength in the case of composites is determined by the onset of fracture, and not via a yielding mechanism. Predictions are particularly difficult in the case of randomly oriented fibers, as cracks tend to propagate by fiber avoidance process as opposed to fiber pullout or fracture [11]. When predictions are attempted, they most frequently take the form: c
where
fu
is the fiber tensile strength,
= * m
3
4
f
fu
+
m
* m
(4)
is the tensile stress in the matrix at composite failure
strain, and 3 and 4 are the fiber-length and orientation-correction factors, respectively. The difficulty lies in the estimation of the correction factors. However, even the general form of this expression may be only applicable over a small range of fiber volume fractions, as a non-linear relationship has been observed for higher volume fractions [12]. The range of applicability of the above models is limited by the maximum volume fraction that can be accommodated by the matrix. This limit is in turn governed by uniformity of the fiber orientation and the fiber length. For uniformly oriented fibers, the concentration can reach 50-60% by volume; for randomly oriented fibers, the limit can be 10-40%, depending on the fiber length. A tensile modulus estimate for an exemplary composite (Ciba Geigy SL5170 resin and 1.6 mm glass fibers, Table 1) is plotted as a function of fiber volume fraction in Figure 1 for 3-D randomly oriented fibers. The plot demonstrates the linear nature of the relationship and predicts that the tensile modulus may double, from 1.5 GPa for the unreinforced resin to 2.9 GPa for the composite, with the addition of 15% (by volume) of short fiber strands. For comparison
-4-
purposes, the matrix modulus value, Em, in Equation (1) has been assumed to be equal to the experimentally observed value for pure-resin mold-fabricated specimens. Table 1. Material data for reinforcing fibers (Owens Corning 737BD milled glass fiber) Tensile Modulus, E f (GPa)
72.4
Diameter, d ( m)
15.8
Average Length, l (mm)
1.6
4.0
Tensile Modulus (GPa)
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0
5
10 15 Fiber Volume Fraction (%)
20
25
Figure 1. Tensile modulus estimate for short-fiber reinforced composites
2.2 Experimental Verification with Mold-Fabricated Composites While there exists an extensive body of knowledge on short-fiber reinforced engineering plastics (a common raw material for injection molding operations), we were not aware of any information regarding short-fiber reinforced photopolymers. This has motivated us in the early phase of this research to conduct mechanical tests on mold-fabricated fiber-reinforced photopolymers. The tests enabled us to verify the improvement in the mechanical properties, and to observe curability and processability of the resin-fiber mixtures [13]. Tests were conducted on dogbone-shaped specimens, 66 mm long with 3x3 mm crosssections (ASTM Standard Test Method D638-91a, Type M-III), which were produced by pouring the liquid (either a pure resin or a resin-fiber mixture) into an open Teflon mold and UV-curing. The constituent materials were CibaTool SL5170 resin and short (1-1.5 mm) milled glass fibers (either 737BD from Owens Corning or MFX from Phoenix Fiberglass). Figure 2 shows the tensile modulus and Figure 3 the tensile strength observations plotted against the fiber volume fraction. Addition of only 15% by volume of short glass fibers can be seen to increase the modulus by 64%, from 1.4 GPa to 2.3 GPa, and the strength by 20%, from 53 MPa to 63 MPa. The experimental measurements follow the expected linear relationship for both strength and modulus; however, the slope of the line fitted to the modulus data is less than theoretically expected. The discrepancy may exist either because the theoretical model is not -5-
accurate or because the measured fiber-length distribution does not equal that within the specimens. The fiber-length difference may be due to the fiber breakage during the composite preparation. 4.0
Tensile Modulus (GPa)
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0
5
10 15 Fiber Volume Fraction (%)
20
25
Figure 2. Tensile modulus of mold-fabricated short-fiber composites
75
Tensile Strength (MPa)
70
65
60
55
50
45
40 0
5
10 15 Fiber Volum e Fraction (%)
20
25
Figure 3. Tensile strength of mold-fabricated short-fiber composites
3. THE PROPOSED RAPID LAYERED MANUFACTURING PROCESS The high viscosity of a composite liquid, fiber settling, lack of interlayer fiber penetration, and other engineering issues necessitate development of a new process for the layered fabrication of reinforced polymeric parts. In regard to the first issue mentioned above, rheological studies have demonstrated that the addition of fibers to a photopolymer resin produces a highly viscous liquid [14]. At low shear rates, such as those encountered in layer formation by liquid spreading, the viscosity of the -6-
composite liquid can be up to five times greater than that of the pure resin (Appendix). In the commercial SL process, a wiper blade is used to spread the liquid photopolymer onto the newly solidified layer. With a significant rise in viscosity, as in our case, the composite liquid would have difficulty spreading over the entire solidified surface, especially for larger surface areas. Secondly, if the composite liquid were left undisturbed in a vat during part building, the fibers would settle continuously due to having a density more than twice that of a typical resin (2.54 g/cm3 for glass vs. 1.13-1.15 g/cm3 for resin). Thirdly, if the composite layered parts were produced via complete layer-by-layer solidification (as in the regular SL process), there would be no fibers extending across the layer boundaries; the fibers would be prevented from protruding above the surface by the liquid surface tension. This section first presents a brief overview of the proposed RLM process, followed by a detailed discussion addressing the use of Axiomatic Design theory in arriving at the final process design solution. 3.1 Process Overview The primary fabrication steps of the current process are as follows, Figure 4: (1) A precise volume of the composite liquid is withdrawn from an external source and deposited from above for each layer (Figure 4a); (2) A wiper levels the liquid at the required height (Figure 4b); (3) The layer is selectively cured by a UV laser in a pattern to be discussed in subsection 3.3 (Figure 4c); and, (4) A platform supporting the part is lowered into the vat (Figure 4d). The process steps are repeated until the entire part is built. Subsequently, the platform is raised, and the part is removed, cleaned, and post-cured. As the fabrication process continues, the composite liquid is continuously mixed in a separate container, i.e., an external raw-material source. This process solves all the primary problems addressed in the introduction of this section.
-7-
(a)
(b)
(c)
(d)
Figure 4: Steps of the proposed RLM process
3.2 Axiomatic Design of the Proposed RLM Process The design objective for the RLM process has been formulated as: Design a process which produces fiber-reinforced plastic parts by selectively solidifying thin layers of liquid photopolymer containing short glass fibers. Axiomatic Design Theory defines a design solution as a mapping between the Functional Requirements (FRs) of the functional domain and the Design Parameters (DPs) of the physical domain [9]. FRs comprise a minimum set of independent requirements that completely characterize the design objective. Examining the above design objective for our process, three Functional Requirements can be identified, Table 2. The Independence Axiom provides guidance for finding the appropriate DPs by stating that a good design must maintain independence of the Functional Requirements. In other words, DPs must be chosen in such a way that each FR can be satisfied by the adjustment of the corresponding DP without affecting other FRs. The design parameters for the current RLM embodiment have been selected based on the above guidelines (Table 2). Table 2. Top level FRs and DPs
FR1 = Build part layers of accurate height
DP1 = Liquid-Layer Formation subsystem
FR2 = Build part layers with accurate horizontal dimensions
DP2 = Laser-Light Delivery subsystem
FR3 = Build part layers with a specified fiber content
DP3 = Fiber-Resin Mixing subsystem
Each of the subsystems listed in Table 2 only affects the corresponding functional requirement, as expressed by the following design matrix
-8-
FR 1 X 0 0 FR 2 = 0 X 0 FR 3 0 0 X
DP1 DP2 DP3
(5)
where X at (i,j) position indicates that DPj affects FRi, and 0 indicates that it does not. Such a diagonal square design matrix is considered ideal, as it represents an uncoupled design, and thus satisfies the Independence Axiom. Equation (5) defines our design at the highest level of the design hierarchy. To be completed, the design must be “decomposed” down to the lowest level. For each DP on one level, a corresponding set of FRs must be identified on the next lower level. The lower level design then continues with the selection of DPs to satisfy the lower-level FRs. In our design, DP1, the Liquid-Layer-Formation Subsystem, is decomposed as shown in Table 3. Table 3. Lower level FRs and DPs
FR11 = Provide sufficient liquid for layer creation
DP11 = Liquid-delivery subsystem
FR12 = Create layers with minimum height variation
DP12 = Liquid-leveling subsystem
FR13 = Create layers with specified average thickness
DP13 = Z-platform subsystem
The relationships between the FRs and DPs in Table 3 are described by: FR 11 X 0 0 FR 12 = X X 0 FR 13 X X X
DP11 DP12 DP13
(6)
The matrix in (6) indicates that liquid delivery (DP11) must occur before the layers can be formed, and, therefore, DP11 affects both FR12 and FR13 in addition to FR11. Liquid leveling (DP12) not only smoothes the layer (FR12) but also levels it to the specified thickness (FR13). Lowering of the Z-platform (DP13) only affects the layer thickness (FR13). A triangular design matrix, such as that in (6), means that the design is decoupled. A decoupled design matrix requires that the DPs must be adjusted in a particular sequence in order to assure independence of FRs. In this case, the sequence is DP11, DP12, DP13. DP11 affects all FRs, and so must be adjusted first. Once DP11 is set, DP12 is adjusted. Note that adjustment of DP12 does not affect FR11. Since DP11 is already set, FR12 can be achieved by adjusting only DP12. Finally, with DP11 and DP12 set, FR13 can be achieved by adjusting DP13. For the case at -9-
hand, this establishes the correct sequence for process parameter adjustments: (1) liquid delivery, (2) liquid leveling, and (3) Z-platform. This completes the process design specification at the subsystem level. Further decomposition would be specific to the particular hardware selected to achieve the desired functionality of the subsystem. The following sections will now describe in detail the key elements of the proposed RLM process. 3.3 Process Details Composite-Liquid Delivery Process
The method of composite-liquid delivery has significant impact on achieving the desired fiber content in the finished parts. The composite liquid has to be delivered directly onto the solidified areas of the previous layer in order to achieve the desired fiber content. The fibers within the liquid surrounding the solidified part settle out relatively quickly; after reaching the depth of several millimeters, a layer of mostly pure resin forms near the surface. Thus, if the newly delivered composite liquid were not deposited directly on top of the previously solidified areas, being denser than the pure resin, the fibers would sink quickly, leading to a wiping operation that covers the part with a liquid of lower fiber concentration. To quantify the performance of the delivery system, we can use the ratio between the fiber content of the liquid delivered into the vat ( f1) and the fiber content of the finished parts ( f2); ideally, this ratio should be 1.0. In our experiments, by implementing the proposed liquid delivery process, the ratio ( f2 / f1) was improved from about 0.78 for non-direct deposition to within 2-3% of 1.0 for the current process. Ideally, the composite liquid must be also spread as evenly as possible. This is achieved in our process by translating the nozzle over the vat while delivering the liquid through a narrow nozzle in a tightly controlled manner. Liquid-Leveling Process
The liquid-leveling operation must achieve two objectives: create a liquid layer of specified thickness and of minimum height variability. Formation of a thin liquid layer by a straight-edge blade is a complex fluid mechanics problem affected by a number of parameters. These include the shape of the blade edge, the wiping speed, the amount of material accumulated at the leading edge, and the liquid viscosity, Figure 5. The height of the resulting liquid layer, h, is normally not equal to the gap between the underlying solid surface and the blade bottom, h0 [15]. - 10 -
h0
h
Figure 5. Layer formation with a straight-edge blade
The interrelationship of these parameters can be approximated by a simplified fluid flow model based on a combination of Couette and Poisseille flows, Figure 6 [16]. The model is derived by assuming (a) laminar viscous flow, (b) constant viscosity, and (c) channel length L is much greater than the gap h0. The last assumption is only a rough approximation in our case as the gap h0 is 0.3-0.6 mm, while the flat bottom surface of the wiper is only 1 mm wide (L). L g H
h
h0
V
Figure 6. Simplified flow model of the blade wiping
In the model, the wiper blade is stationary, and the solid surface underneath is moving with a velocity V. The liquid reservoir with liquid elevation, H, represents the liquid accumulated at the wiper’s leading edge during translation. The model assumptions allow application of standard flow-rate expressions, which gives the final layer thickness as h=
h0 h03 gH + 2 12 VL
(7)
where is the liquid density, is its viscosity, and g is the gravitational acceleration. By selecting representative values for the parameters in Equation (7) (h0 = 0.6 mm, H = 2.0 mm, L = 1.0 mm, = 1.15 kg/m3, and V = 1.5 cm/s), the ratio h/h0 can be shown to decrease from 0.73 to 0.55 as viscosity increases from 200 to 1000 cP. The model points to the significance of the liquid bulge formed at the wiper’s leading edge. As the wiper moves, the bulge height changes. This variation can affect the layer height locally, - 11 -
leading to variation in layer thickness. The proposed wiping process aims to minimize this effect through performing two wiping stroke sequences to form each layer: the first stroke sequence is carried out after the platform has been lowered by a depth of several layers, and the second stroke sequence when the platform has been raised back to the height one-layer lower than the last layer built. Pre-wiping prior to the final leveling helps to spread the liquid more evenly by removing most of the excess liquid and roughly spreading the liquid during the first stroke. Pre-wiping also reduces the leading edge liquid accumulation during the final wiping stroke. Part cross-section studies have confirmed the effectiveness of this approach. X-Y Scanning Process and Rivets
As laser scanning converts the liquid resin into a solid part, it determines the part’s structural integrity, correctness of its final shape, and presence of residual stresses within the part. In the scanning method adopted for our process, a border scan is followed by parallel hatch lines along the X and Y axes. The hatch lines are offset by half of line separation from layer to layer to improve the part integrity. The scanning sequence is also varied from layer to layer, so that the lines are drawn either in ascending or descending order (in terms of their position along the coordinate axis), which equalizes build up of residual stresses within the part due to resin shrinkage. A novel aspect incorporated into our X-Y scanning method is the creation of interlayer rivets. The motivation was to improve the layer-to-layer bonding. Since the liquid surface tension keeps the fibers from protruding above the surface, with the normal complete-surface scanning pattern, there would be no fibers crossing the layer boundaries. The rivets, in our case, are small volumes of resin intentionally left uncured within each layer. Fibers protrude into these volumes from the surrounding solidified resin. When the subsequent layer is spread and cured, it will extend into the uncured pockets left in the preceding layer, forming interlayer rivets, with the fibers extending into the rivets. Figure 7 attempts to graphically illustrate this concept. It shows the sequential formation of three consecutive layers. (Darker shading indicates solidified resin, while lighter semi-transparent volumes denote uncured liquid.) For the first layer, there are three small volumes left with the uncured liquid. The uncured volumes are shifted by half of the inter-rivet distance in the next layer. When this second layer is cured, the uncured volumes of the previous layer are partially solidified; the solidification is completed during post-curing. 4. DESIGN OF EXPERIMENTAL PROTOTYPE An experimental prototype system was built to assist us with the process development. Since our emphasis was on the process analysis, and not on designing a commercial product, the - 12 -
choice of some components was dictated not by the greatest suitability to the task but by the equipment availability and simplicity of implementation. For example, to achieve fast building speeds, majority of commercial systems utilize galvanometrically actuated mirrors for X-Y laser beam scanning. In our case, on the other hand, since building time was not a significant issue, the system uses a precision X-Y translation system for the same purpose due to the availability of the equipment.
Figure 7. Schematic illustration of the interlayer rivets: half of a cross-section is shown
Current design is a result of several iterations. For each of these, Axiomatic Design principles were used to help with the design analysis and to supply ideas for the new designs, leading to significantly improved system performance. The subsystems of the current prototype RLM machine (Figure 9) are individually discussed below: The fiber-resin mixing subsystem keeps the fibers in suspension throughout the building process. It consists of an open-top container and a helical-screw stirring device inserted into the container from above. The helical-screw arrangement achieves adequate mixing, when turning slowly, without causing agitation or foaming of the liquid. While significant air entrapment occurs during the initial preparation of the mixture, the liquid is not used until sufficient time has elapsed for degassing (typically 24 hours). No air bubbles are visible when the mixture is used. The composite-liquid delivery subsystem transfers desired amounts of resin into the vat. This function is performed via a peristaltic pump, which uses rollers to squeeze the liquid - 13 -
through a flexible plastic tube. One end of the tubing is inserted into the container of the mixing subsystem; the other end is attached to the platform of the X-Y translator, Figure 8. The pumping action is combined with simultaneous translation of the dispensing nozzle to achieve even spreading of the composite liquid. The advantages of delivery via a peristaltic pump are: (i) accurate metering of the liquid volume delivered, since the pump is of a positive displacement type, and (ii) simplified system maintenance, since there is no contact between the pump mechanism and the liquid, and since the tubing is easily replaceable. In our machine, the liquid was delivered into a vat with inside dimensions of 93 93 mm. Deposition from above did not cause any observable air bubble entrapment.
Nozzle
Mixer
Pump
Vat
Figure 8: Composite-Liquid Delivery and Fiber-Resin Mixing subsystems
The liquid-leveling subsystem assures uniform spreading of the liquid over the vat’s top surface to create a layer of consistent thickness. This is achieved by translating a wiper with a triangular edge profile. The wiper movement is actuated by a pneumatic cylinder. The laser-light delivery subsystem delivers a focused beam of UV light to the surface of the composite liquid and translates this beam in the X-Y plane. The light is delivered via a fiberoptic cable attached to a focusing lens. The lens is translated over the vat by an X-Y table with a 1 micrometer resolution. The Z-platform subsystem moves the supporting platform vertically. Since the height of the platform directly affects the layer thickness, its vertical displacement must be accurately controlled. To achieve the required accuracy, the Z platform is actuated by a stepper motor driving a micrometer attached to a vertical translation stage with a 70 mm range of motion.
- 14 -
Figure 9. Prototype RLM system for fabrication of fiber-reinforced composites
5. PROCESS ANALYSIS In analyzing the proposed process, we have selected two process output characteristics to serve as metrics of quality. The first metric is the quality of the layers (and hence the parts) in terms of their geometry, and the second one is the quality of the parts in terms of their mechanical properties. Test specimens were fabricated using the building parameters listed in Table 4. To arrive at an appropriate scan speed value, lines were drawn in pure resin and composite liquids at varying scan speeds to establish the working curve – cure depth vs. the scan speed relationship. No significant differences were observed between the cure depth of glass-fiber composite and pure resins, concurring with the observations in [17]. Scan speed was chosen to achieve cure depth approximately equal to the layer thickness. Table 4. Building parameters for the process analysis experiments Layer Thickness
0.3 mm
Scan speed
15-25 mm/s
Hatch-line separation
0.3 mm
Laser Power at the surface
12-18 mW
5.1 Layer Quality Since parts produced by an RLM process comprise individual layers, good surface quality and dimensional accuracy can be achieved only by building high-quality individual layers. The experimental system was therefore used to build a set of pure and composite rectangular parts (25 30 4.8 mm) whose cross-sectional profiles were examined microscopically. The composite liquid comprised Allied Signal 2202SF photopolymer and Owens Corning 737BD 1.6 mm milled - 15 -
glass fibers (15% by volume). An additional pure-resin test part was built on a commercial SL machine (Sony JSC-2000 Solid Creator) using the DeSolite SCR310 photopolymer. Measurements obtained from this part served as a benchmark for the layer quality of our own parts. Several vertical sections of each part were made. Figure 10 shows the layer profiles for one cross-section of a pure-resin part built on the Sony machine (“SL_PURE”) and for an equivalent pure-resin part built on our prototype system (“RLM_PURE”). Figure 11, on the other hand, shows layer profiles for a composite part of the same dimensions built on our system (“RLM_COMP”). Ideally, the plots should consist of straight horizontal lines, representing layer boundaries, separated by the nominal layer thickness of 0.3 mm.
5.4
5.4
4.8
4.8
4.2
4.2
3.6
3.6 Z (mm)
Z (mm)
Statistical analysis of the measurements from several sections of the above three parts was carried out after discarding the data for the first 8-10 layers in order to allow for process stabilization. Three parameters are shown in Table 5 for each part: the average layer thickness, which is the mean of layer boundary separation values measured across all the part sections; the standard deviation within the layers, which is a measure of the unevenness of the layers; and the standard deviation between the layers, which is a measure of the layer-to-layer variability of the average layer thickness. The results verify that our prototype system is able to build layers with a mean thickness very close to the nominal value. Also, the variabilities in the building process of the layers are quite close to those obtained on the Sony machine, especially when one considers that our system is a University prototype. The composite part layer variability is only slightly higher than that for the pure-resin part. Equipment refinement should reduce the difference even further.
3.0
3.0
2.4
2.4
1.8
1.8
1.2
1.2
0.6
0.6 0.0
0.0 0
5
10
15 X (mm)
20
0
25
(a)
5
10
15 X (mm)
20
25
(b)
Figure 10. Layer profiles for a pure-resin part made (a) on the Sony machine and (b) on the prototype of the proposed RLM process
- 16 -
30
5.4 4.8 4.2
Z (mm)
3.6 3.0 2.4 1.8 1.2 0.6 0.0 0
5
10
15 X (mm)
20
25
30
Figure 11. Layer profiles for a composite test part made on the prototype of the proposed RLM process Table 5. Statistical data for layer profiles of multiple sections Average Layer
St. Dev. Within
St. Dev. Between
Thickness (mm)
Layers (mm)
Layers (mm)
SL_PURE
0.313
0.008
0.002
RLM_PURE
0.306
0.016
0.007
RLM_COMP
0.315
0.021
0.010
Part cross-sectioning was also employed to microscopically examine the interlayer rivets within the dogbone tensile test specimens. Figure 12 shows a vertical section through the central plane of the specimen. All the layer boundaries are visible on the right-half of the photograph. The interruptions of the boundary seen on every second layer in the middle correspond to interlayer rivets. The rivet boundaries can be more clearly discerned in a photo taken at a higher magnification (Figure 13). Layer boundaries are indicated by triangular markers on both photographs.
Figure 12. Part cross-section showing interlayer rivets - 17 -
Figure 13. Close-up view of an interlayer rivet
5.2 Mechanical Properties A number of dogbone-shaped specimens fabricated on the prototype system were subjected to tensile tests. The tests aimed first to evaluate the tensile strength and moduli of the composite parts, and second to compare the properties of composite parts to those of pure-resin parts fabricated via the same method. An additional objective was to quantify the effect of the interlayer rivets on the part properties. The test specimens (ASTM Standard Test Method D638-91a, specimen type M-III) were fabricated utilizing the CibaTool SL5170 photopolymer and Owens Corning 737BD 1.6 mm fibers. The specimens measured 60 mm in length, with a 4.0 3.4 mm cross-section at the waist. After layered fabrication, all specimens were post-cured for two hours using a UV lamp. The change in resin type from that used in the layer quality studies was made for logistical reasons. However, both resins have similar rheological properties and the change did not have a negative effect on the part quality. The fiber contents of the finished parts were measured via ASTM Standard Test Method D792-91 (Density and Specific Gravity (Relative Density) of Plastics by Displacement). The composite specimens contained 20 2% fibers (by volume). As an additional check, the fiber content within the narrow waist section of the dogbones was measured for selected specimens and compared with the value observed for the dogbone as a whole. Measurements matched typically within 0.5%. The specimens were tested in tension until failure at an extension rate of 1 mm/min. Table 6 and Figure 14 show the experimental results for the four sets of specimens prepared: (a) Pure resin - with rivets, (b) Pure resin - no rivets, (c) Composite - with rivets, and - 18 -
(d) Composite - no rivets. The results show a 60% increase in the modulus for the (riveted) composite specimens over the pure-resin specimens. Rivets had no effect on the modulus of pure-resin parts, while increasing the modulus of composite parts by about 26%. It was noted that the tensile strength was not affected by the addition of fibers, nor by the introduction of interlayer rivets. Although tensile strength is generally expected to increase with the addition of reinforcements (Section 2.2), a possible explanation for lack of such improvement could be that the specimens fail by fracture initiated at the layer boundaries. Since there are relatively few fibers extending across the boundaries (even with the addition of the interlayer rivets), the presence of fibers may not affect this fracture process significantly. Table 6. Mechanical test results Set
Modulus (GPa)
Strength (MPa)
Ave.
St. Dev.
Ave.
St. Dev.
Pure Resin - Rivets
1.20
0.09
45
0.9
Pure Resin - No Rivets
1.26
0.10
43
0.9
Composite - Rivets
1.98
0.17
44
1.3
Composite - No Rivets
1.57
0.11
44
1.7
2.5
Pure Composite
Modulus (GPa)
2.0
1.5
1.0
0.5
0.0 Rivets
No Rivets
Figure 14. Tensile modulus – pure-resin and composite specimens built on the prototype system
6. CONCLUSIONS The paper presented a novel process for the layered fabrication of fiber-reinforced plastic composites. The process development made extensive use of the Axiomatic Design Theory for - 19 -
guidance and validation. A distinguishing feature of the process is the deposition of composite liquid from above for each layer, with specific recommendations for the deposition and liquidleveling methods made in Section 3.3. Another novel feature of the proposed process is the enhancement of layer-to-layer bonding through the introduction of interlayer rivets. Process analysis studies have demonstrated the process’ ability to produce composite parts comparable in quality to pure-resin layered parts manufactured on a commercial system. The mechanical testing of composite and pure-resin parts also verified a significant increase in modulus through the addition of short-fiber reinforcements to the layered parts.
7. REFERENCES
[1]
M.R. Piggott, Load-Bearing Fibre Composites, Pergamon Press, New York, 1981.
[2]
L.D. Schmidt, “How Chrysler is using stereolithography rapid prototyping survey results,” Proc., Fifth Int. Conf. on Rapid Prototyping, June 1994, Dayton, Ohio, pp. 359-370.
[3]
W.B. Barlage, C.C. Jara-Almonte, A. Bagchi, A.A. Ogale, R.L. Dodey, “Fiber/resin composite manufacturing using solid freeform fabrication,” Proc., Third Int. Conf. on Rapid Prototyping, June 1992, Dayton, Ohio, pp. 15-24.
[4]
R. Charan, T. Renault, A.A. Ogale, and A. Bagchi, “Automated fiber-reinforced composite prototypes,” Proc. Fifth Int. Conf. Rapid Prototyping, June 1994, Dayton, Ohio, pp. 91-97.
[5]
D.E. Reiff, D.W. Dorinski, and S.D. Hunt, “Method of manufacturing a three-dimensional plastic article,” Motorola Inc., U.S. Patent 5173220, December 1992.
[6]
S.W. Thomas, R.D. Key, K.G. Fluegel, W.K. Jackson, K.D. Elwell, and G.L. White, “Method for manufacturing fiber-reinforced parts utilizing stereolithography tooling,” E-Systems Inc., U.S. Patent 5296335, March 1994.
[7]
A.A. Ogale, T. Renault, R.L. Dooley, A. Bagchi, C.C. Jara-Almonte, “3-D photolithography for composite development: discontinuous reinforcements,” SAMPE Quarterly, October 1991, pp. 2838.
[8]
T. Renault, and A.A. Ogale, “3-D photolithography: mechanical properties of glass and quartz fiber composites,” Proc., ANTEC 1992 - Conf. SPE, Vol. 1, May 1992, pp. 745-747.
[9]
N. P. Suh, The Principles of Design, Oxford University Press, 1990.
[10] H.L. Cox, “The elasticity and strength of paper and other fibrous materials,” Br. J. Appl. Phys., Vol. 3, 1952, pp. 72-79. [11] M.R. Piggott, “Short fiber polymer composites: a fracture-based theory of fiber reinforcement,” J. of Composite Materials, Vol. 28, No. 7, 1994, pp. 588-606. [12] A.R. Sanadi and M.R. Piggott, “Interfacial effects in carbon-epoxies: Part 2, Strength and modulus with short random fibers,” J. of Materials Science, Vol. 20, 1985, pp. 431-437.
- 20 -
[13] G. Zak, A. Y. F. Chan, C. B. Park, and B. Benhabib, "Rapid layered manufacturing of short-fiberreinforced parts: investigation of the material properties," CSME Forum 1996: 13th Symposium on Engineering Applications of Mechanics, Hamilton, Ontario, May 1996, pp. 458-464. [14] G. Zak, A. Y. F. Chan, C. B. Park, and B. Benhabib, "Viscosity analysis of photopolymer and glassfibre composites for rapid layered manufacturing," Rapid Prototyping Journal, Vol. 2, No. 3, 1996, pp. 16-23. [15] K. Renap, and J.P. Kruth, “Recoating issues in stereolithography,” Rapid Prototyping Journal, Vol. 1, No. 3, 1995, pp. 4-16. [16] J.A. Fay, Introduction to Fluid Mechanics, MIT Press, 1994, p. 285. [17] T. Renault, A. A. Ogale, M. J. Drews, "Influence of reinforcements on photocuring: photo dynamic mechanical analysis," ANTEC 93, Society of Plastics Engineers, Conference Proceedings, New Orleans, May 9-13, 1993, pp. 2352-2355.
- 21 -