1 Introduction

When the first additive manufacturing (AM) methods were introduced in the 1980s, they were mainly used to produce prototypes, leading to the widespread use of the term rapid prototy**. Other applications of additive manufacturing include the production of functional parts, also referred to as rapid manufacturing, and tools (rapid tooling) [1]. These last two applications became increasingly important.

Today, there are various AM methods on the market that operate with a variety of different materials, e.g., thermoplastics, photopolymers, and weldable metals. Since there are many products that are made of more than one material, the challenge is to combine AM methods to simultaneously process two different materials. In the present project, this was performed for thermoplastic polyurethane (TPU) and acrylonitrile butadiene rubber (NBR).

Hydraulic rod seals were chosen as an example of two-component parts made of these materials. The challenge is that the parts are not supposed to be just prototypes but functional parts, so functional testing on a test rig is a major part of the work. There are many papers dealing with the tribological properties of 3D-printed parts (e.g., [2,3,4,5,6,7,8,9]). Most of them work with the thermoplastics commonly used in AM [10]., but not with TPU. However, they show commonly used methodologies to investigate tribological properties and the influence of printing parameters, such as orientation and filling grade. Sood et al. [4], Roy et al. Mohamed et al., Maries et al. and Gurrala et al. investigated the tribological properties of acrylonitrile butadiene styrene (ABS). Roy et al. and Maries et al. included polylactic acid (PLA) in their investigations. Maries et al. also included polyethelene terephthalate glycol modified (PETG). Tey et al. [11] investigated the properties of additively manufactured parts made of TPU using the Additive Manufacturing method of Multi Jet Fusion, where the raw material in powder form is melted by using a fusing agent activated by an infrared heating element. Hossain et al. [12] studied the same approach for specimens made with a fused filament fabrication (FFF) 3D printer. However, no paper has been found on additively manufactured dynamic seals tested under realistic operational conditions.

A major motivation for choosing rod seals as an example was the fact that many seal types have long delivery times. If produced on order only, the delivery time can reach 3 months.Footnote 1 Therefore, producing spare parts in a short amount of time could significantly reduce costly machine shutdown times if industrially produced spare parts are unavailable.

The pivotal research question addressed in this study is as follows: Is it possible to produce fully functioning dynamic seals using additive manufacturing?

2 Experimental information and process

2.1 Reference part

To have a reference part to compare the specimen with, an existing rod seal type made of two components was chosen: a base body made of TPU and an energizing ring made of NBR. This rod seal type is known as a compact ring. The reference is an SKF PTB-50 × 60× 11-J1S [13] (Fig. 1). The reference was chosen because of its potentially good manufacturability, because of its rather simple shape, and easy availability. For the 3D-printed test parts, the shape of the energizing ring was changed. The reference part has an X-shape (Fig. 2), while the 3D-printed version is rectangular with fillets, as shown in Fig. 3. This was done due to better manufacturability.

Fig. 1
figure 1

Compact ring reference SKF PTB-50 × 60×11-J1S

Full size image
Fig. 2
figure 2

Polished cut image showing the cross section of the SKF reference shown in Fig. 1

Full size image
Fig. 3
figure 3

Left: original groove shape; right: groove shape of the test parts

Full size image

The focus is on such a two-component type because it reflects the objective of the research project behind this paper: the simultaneous additive manufacturing of a two-component part made from a thermoplastic material and NBR.

2.2 Materials

While the recipe for the NBR component was the same for all tests, different filaments were used to produce the TPU component for the test programs. In a preliminary investigation [14], that led to the research described in this paper, an AM method and a variety of materials were chosen to produce the first specimen. After using a systematic method screening, where specimens made with FFF, SLS (Selective Laser Sintering) and MJF (Multi Jet Fusion) have been examined. In the end FFF was chosen for further examination, primarily because the surface quality, achieved with the other methods, was too poor. In the subsequent investigation, 14 different filaments, mainly made of TPU, were characterized. The goal was to achieve properties comparable to those of conventionally produced rod seals. The aspects for the ranking of materials were: processability, temperature range, dry friction coefficient, radial force, and shore hardness.All of the examined filaments were commercial off-the-shelf products. In the end, the following filaments were chosen for the first functional tests:

  1. 1.

    Extrudr TPU Hard D58 [15]

  2. 2.

    BASF Ultrafuse TPU 95A [16]

  3. 3.

    Extrudr TPU Semisoft A85 [17]

  4. 4.

    Extrudr TPU Medium A98 [18]

  5. 5.

    Polymaker Polyflex TPU 95A [19]

2.3 Additive manufacturing of the TPU component

The manufacturing of the specimens for functional testing was performed separately for the two components. Moreover, simultaneous manufacturing was still not established at that time. The vulcanization of the additively manufactured NBR component was performed after mounting the ring in the additively manufactured TPU part. Figure 4 shows such a printed two-component seal.

Fig. 4
figure 4

3D-printed two-component seal

Full size image

The production of the TPU component of the test parts was performed on a commercially available 3D printer (PRUSA i3 Mk3 S +). A standard nozzle with a size of 0.4 mm was used, and the printing speed was set to 20 mm/s, because such a low speed worked with every filament. The layer height was set to 0.15 mm. Attempts to achieve smaller layer heights led to severe quality problems such as poor surface quality or insufficient bonding to the layer below. The temperature parameters recommended by the filament producers were applied as initial values.

A commercial adhesive spray [20] was used before printing to improve the adhesion of the first layer because TPU filaments tend to connect poorly to the building plate.

2.4 Additive manufacturing of the NBR component

The rubber compound used for 3D printing was an acrylonitrile rubber (NBR)-based mixture, which required optimized cure kinetics [21] and a further adjustment of the viscosity using a plasticizer content of 10 phr Mesamoll, which was described in detail by Sundermann et al. [22]. The recipe of the rubber mixture is demonstrated in Table 1.

Table 1 Rubber mixture recipe
Full size table

The printer used for additive manufacturing was a modified CNC milling machine with a twin-screw extruder as the material processing unit [23]. The nozzle diameter was 0.4 mm, the distance between adjacent stripes was 0.53 mm, and the layer height was set to 0.4 mm; however, the first layer was printed with a reduced distance of 0.26 mm to achieve better adhesion. The temperature of the extruder segments was set to Off/70 °C/100 °C. An additional water-cooling reflux system was installed in the feeding zone to reduce the local screw temperature to avoid disruption, as depicted in Fig. 5.

Fig. 5
figure 5

Water cooling reflux system close to the feeding zone of the extruder to prevent rupture of the feeding material

Full size image

To achieve a print speed of 21.6 mm/s, the screw speed was set to 12 rpm. A commercial adhesive spray (“Adhesive Spray detachable” by Weicon) was used before printing to improve the adhesion of the first layer, which was suitable for use in previous studies as a compromise between adhesion and detachability. After fabrication, the specimens were removed from the bed, placed in the TPU seal, and cured at 140 °C for 29 min in a convection oven.

2.5 Test rig

Figure 6 and Fig. 7 show the test rig for the functional tests. It was originally designed for optical investigations in the seal lip area [24]. The maximum speed achievable is 0.25 m/s, the maximum pressure is 15 MPa and the total traveling distance is 150 mm.

Fig. 6
figure 6

Rod seal test rig

Full size image
Fig. 7
figure 7

Rod seal test rig—schematic view of Fig. 6

Full size image

The terms instroke and outstroke refer to the test seal, which was mounted on the adapter plate facing toward the linear actuator. On the other side of the test chamber, a reference seal was always mounted (0–2; see also Table 2). This was done to be able to allocate every occurring effect to a specific test part and because the hydraulic pump is situated on that side. In the case of a severe defect, the necessary intervention by the operator could lead to a severe health threat. If this happens on the actuator side, intervention on the pump side occur place without danger.

Table 2 Results of the first test phase
Full size table

The axial force is measured with a load cell. This axial force is a result of the radial forces of the seals and their friction coefficients.

2.6 Test procedures

2.6.1 Examinations prior to the functional tests

Before mounting a test seal and after the test, the following values were measured and documented: the mass of the part, the inner diameter, and the radial force, measured on a test rig as described in detail in [14] and [25] and as depicted in Fig. 8. In [25] Debler described the original design of the test rig, while in [14] Graf et al. first described the application of such a test rig for measurements on rod seals.

Fig. 8
figure 8

Radial force test rig for rod seals with 50 mm diameter

Full size image

The test rig has a static and a movable jaw. The movable one is an angular segment of 120°. First the position of the movable jaw is calibrated using a steel ring with an inner diameter of 50 mm. After mounting a test part the displacement with regard to the calibrated position is measured and a control system compensates that displacement with a stepper motor. The exerted force Fsensor has then to be converted to the radial force FR of the entire circumference using the formula (1).

$${F}_{R}=\frac{2 \pi {F}_{Sensor}}{\sqrt{3}}$$
(1)

As described in these publications, the measured radial force is not the same as that of a mounted seal because of the different methods of mounting on the test rig. In the seal groove of a hydraulic cylinder the outer diameter is also defined leading to additional compression of the seal and thus to a higher radial force.

A tribometer (Anton Paar TRB3) test was also conducted to identify filaments that showed significant wear and disqualified them from further investigation. An example of a part made from such a filament is depicted in Fig. 9.

Fig. 9
figure 9

Example of a specimen made from a filament showing heavy wear

Full size image

2.6.2 Test program—phase 1

On the test rig, the axial force was recorded while the rod was moved back and forth 10 times per pressure stage with a maximum speed of 0.05 m/s, an acceleration of 0.8 m/s2, and a sliding distance of 100 mm. This was done with pressure stages 0, 1, 2, 3, 4, and 5 MPa with a 10 min delay after every pressure change and after the 10 in- and outstrokes.

The main purpose of the first test phase was to determine whether 3D-printed rod seals were likely to fulfill their main purpose: to avoid oil leakage. Therefore, in the early stage, only moderate pressure was applied.

The adapter plates of the test rig were not designed to allow a measurement of the amount of oil that eventually leaked during testing. Therefore, in the results, there is only a distinction between.

  • No noticeable leakage,

  • Slight leakage (i.e., an oil film is on the rod or single drops on the adapter plate, as in Fig. 10), or

  • Severe leakage that led to an abortion of the test cycle (Fig. 11). The definition was, that in case of a severe leakage, oil drips off the adapter plate.

Fig. 10
figure 10

Slight leakage

Full size image
Fig. 11
figure 11

Severe leakage

Full size image

2.6.3 Test program—phase 2

After optimizing the production parameters of the TPU component as described in chapter 3.2, the goal of the second test phase was to reach a maximum pressure of 15 MPa, which is the maximum pressure achievable with the test rig. After a test part was mounted, the part had to settle for at least 8 h before the test program started to achieve material relaxation.

The values for speed, acceleration, and distance remained the same as those in the first test phase, but the number of cycles per pressure stage was increased from 10 to 100. The pressure stages were 2.5, 5, 7.5, 10, 12.5, and 15 MPa.

Another difference from the first test program was the additional longer test with 1,000 cycles at 6.3 MPa. This procedure was done according to the standard testing procedure for rod seals in ISO 7986:1997 [26]. Other parameters could not be used according to ISO 7986:1997, because the test rig does not fulfil all the requirements such as a stroke length of 500 mm.

The test was again performed with three filaments and three test parts per filament. The filaments chosen for the second test program were as follows:

  • BASF Ultrafuse TPU 95A

  • Extrudr TPU Medium A98

  • Polymaker Polyflex TPU 95A

The material selection for the second test phase was performed based on the findings of the first phase and the results of the preliminary examinations. The Extrudr Medium A98 filament showed a more promising production quality than the Extrudr Semisoft filament. The Polymaker filament performed well in terms of surface quality and printability and replaced the Extrudr Hard filament, which turned out to be unfit for the compact ring shape used for the two-component parts. The reason was the high radial force these parts have shown. The parts were not mountable.

3 Results and discussion

3.1 First test phase

The general qualitative results are summarized in Table 2. If there is no comment in the result column, the test ended successfully.

The first figure in the test part numbers stands for the filament, according to the list in Chapter 2.6. The second is the sequential number of the test part made from that material. The first figure zero is used for reference parts. Figure 4 shows specimen 3–6 as an example before the test while Fig. 12 shows the same specimen after the test. There is no visible wear.

Fig. 12
figure 12

Specimen 3–6 after the test under the microscope. On the right side is the sealing edge (marked red)

Full size image

3.2 Second test phase

The parts for the second test phase were manufactured with optimized parameters in terms of filling degree. The main goal hereof was to achieve less porosity compared to the parts in the first test cycle. This was done by comparing the actual mass of the parts with their theoretical mass, based on the CAD data and the measured density of the filaments. Based on that, the extrusion factor was increased to achieve a nominal filling degree of approximately 100%. This was done because the first tests showed that the success was dependent mainly on the filling degree.

Due to the AM method, there was still some porosity, and the actual filling degree was less than 100%. Figure 13 shows a polished cut image of specimen 2–10 after the second test phase. The specimen clearly shows the porosity. Nevertheless, the entire test was successfully completed, demonstrating that a seal can work even if there is a certain porosity.

Fig. 13
figure 13

Polished cut image of test part 2–10 after the second test phase

Full size image

The results of the second phase are listed in Table 3. Because of the larger number of cycles in this test phase, it was possible to look at the drift over the cycles. In Table 4, the axial forces measured in Cycles 5 and 95 are compared. The values increase over the test cycles, and there is also a difference between the filaments used. The parts made of the BASF filament showed a lower drift than did those made of the other filaments and with the reference parts.

Table 3 Results of the second test phase
Full size table
Table 4 Deviation in the axial force between Cycles 5 and 95
Full size table

The axial forces are shown in Fig. 14 for the instroke only and in Fig. 15 for both the instroke and the outstroke. In this test phase, the values for the axial force are still significantly lower for the 3D-printed parts than for the reference parts, but the difference decreased slightly: for the parts made of the BASF filament, the factor between the reference and specimen for the 5 MPa value was 1.86 in the first test phase and 1.64 in the second one. The still-existing discrepancy between the values was presumably due to the higher compressibility of the 3D-printed parts, which is ascribed to the inevitable porosity of parts made with the FFF method. A significantly greater radial force of the printed specimens, combined with comparable friction coefficients, would otherwise lead to much greater axial forces.

Fig. 14
figure 14

Axial forces of all specimens in the second test program (instroke only)—(in the case that only one specimen was left, there is no error bar)

Full size image
Fig. 15
figure 15

Axial forces of all specimens in the second test program (instroke and outstroke))—(in the case that only one specimen was left, there is no error bar)

Full size image

The difference between the axial forces for the in- and outstrokes increases with pressure in the test rig, as shown in Fig. 15.

Since there was no literature on the application of FFF-produced parts as functional rod seals, the results could only be compared to the data achieved with the conventionally produced reference parts.

4 Conclusion

Before starting the functional tests, no data were available concerning the testing dynamic seals made by using additive manufacturing methods under realistic operational conditions. The main result of our investigation was that most of the specimens could be successfully tested up to 15 MPa. Additionally, the production parameters could be optimized to achieve such working parts.

The results showed that the performance of the FFF-produced parts is different from that of the conventionally produced reference parts. The additively manufactured parts are not entirely dense, the surface quality is not as good, and the anisotropy can also have an impact on the results. However, leakage of most of the test parts was avoided up to the maximum applied pressure of 15 MPa.

The possibility of producing spare parts in a short amount of time is a promising result. 3D-printed seals have the potential to bridge the time gap between a part failure and the delivery of a conventionally made seal. Further investigations must be performed to determine whether they can last as long as conventionally made seals.

Subsequently, the test rig will be modified to allow higher pressure and to quantify leakage. In addition, more filaments will be examined. It is also desirable to work with materials that are already used for the production of rod seals. The measured axial force is a very important parameter to define the main function of a rod seal (preventing leakage) since it is a result of the radial force that increases with the pressure. For a more detailed characterization the measurement of the film thickness in the contact area of the rod seal is planned. Additionally, the simultaneous production of TPU and the NBR component is planned for the future.