Novel Pressure Vessel for Pressurizing Corrosive Liquids With Applications in Accelerated Life Testing of Composite Materials

Accelerated life testing (ALT) is a common technique used across industries to test the performance of a design throughout, or in some cases to determine, its service life. Common examples of this include accelerating corrosion behavior [1] and determining fatigue and creep parameters [2,3]. It is a simulated method, the first step of which is to determine factors which are thought to degrade performance. Then, specimens that represent the design are exposed to an environment that accelerates the rate at which these factors act; a common application is to raise the temperature of a humidity chamber or water bath to accelerate the rate of corrosion or moisture ingression. This method has been used recently by various groups to study the effects of marine environments on polymer and composite structures [4–10], which is the primary motivator for this novel facility. As Davies et al. submitted in a review of composite pressure vessels for deep sea applications [11], the demand for greater operating depths and thus higher working pressures is ever increasing, and so this new method was developed motivated primarily by the desire to perform ALT at these deep sea pressures and thus aid design efforts in this regard.

The laboratory environments required to perform this new generation of accelerated testing created a unique challenge in procuring or designing a compatible pressure vessel. The required aging times were as long as 1000 h, wherein a pressure hold of up to 41.3 MPa (6000 psi) would need to be maintained at temperatures of up to 70 °C. This combination of saline media, temperature, and time constituted a severe corrosion risk. Coatings were researched for the interior of the vessel, but these coatings could not reliably be used on moving parts such as needle or pressure relief valves, where the extreme sealing pressures required of these devices would surely wear the coatings away after a few uses. Expense was also a concern with this approach, as it was also with procuring exotic metals such as Monel for use in the valves and other pressure equipment. Ultimately this option was deemed too costly and still potentially unreliable, as corrosion may yet attack even these metals after the durations of exposure required by the tests.

One design that was considered was inspired by the pressure vessel facility in use at the University of Rhode Island commonly used to study instability and implosion phenomena in thin shells [12–14]. This design uses pressurized nitrogen to fill a small gap of air at the top of the pressure vessel. The pressure is controlled from a remote location by modulating the nitrogen pressure, which by equilibrium, induces the same pressure in the underlying fluid region. A benefit of this design is that the pressure generating, monitoring, and relief equipment as well as any valves do not make contact with the fluid, yet by equilibrium, all measurements of pressure made on the gas are guaranteed to coincide with the actual pressure of the fluid. However, this particular approach was ultimately deemed unsafe for use in the ALT facility, as at the pressures required of a test and for the volume of the vessel, the energies contained in the gas pocket rivaled that of 10 g of TNT. Additionally, vapors of the process media may yet migrate through whatever gas line exists and still corrode equipment. This combination of explosive potential energy combined with the risk of a seized valve or pressure relief ruled out inert gas as a pressurizing method. Nevertheless, the simplicity of the approach and the concept of isolating the process media from valves and monitoring equipment were desirable qualities that the present design leverages. What follows is a complete description and discussion of that design.

An overview of the accelerated high-pressure aging facility is given in Fig. 1. The primary component and focus of this work is a large stainless steel pressure vessel, Fig. 1(a), the inner workings of which will be further elaborated on in Secs. 3 and 4. This vessel is attended by a gantry crane, which serves to aid transport and also to hoist the vessel into and out of a temperature controlled, filtered water bath (c), used to regulate the temperature of the primary vessel and its contents. A hand-operated hydraulic pump (d) pressurizes the vessel through a hose and high-pressure piping stem which contains safety and monitoring equipment. In typical operation, the process media is kept in nearby storage containers (e) and dispensed through a faucet. Also present is a small pressure vessel (b), which is used for small specimens and short duration exposures, as it does not have the same corrosion and chemical tolerance of the large vessel. Together, these form a self-contained system for the performance of water ingression type ALT; the smaller vessel is used to study the temperature dependence of water ingression rates through short-term diffusion studies, while the large vessel is for extended aging requiring constant high pressures and elevated temperatures for durations of the order of weeks.

The primary vessel is in most ways of a conventional design except for the diaphragm and manner of pressurizing. These details, while not the focus of this work, are still necessary for the proper understanding of its function and serve to offer some insight as to the reliability and reusability of the system, so they will be presented briefly in this section.

The vessel follows a fairly conventional flanged-end design, comprising five structural components: an upper and lower flat profile end cap, two flanges, and a central body section. All of these parts were fabricated from AISI 304 L stainless steel, which was chosen for its good balance of strength, corrosion resistance, machinability, and creep behavior. The body section fully defines the usable volume of the pressure vessel and is a single section of centrifugally cast, seamless tube 53.3 cm in length with an internal diameter of 15.2 cm and wall thickness of 3.81 cm. The ends of the body are externally threaded to accept the flanges, and the top end face is trepanned for an O-ring gland accepting a No. 441 O-ring used as a face seal between the body and top end cap. The bottom end cap is 30.5 cm in diameter and 8.26 cm thick, with a concentric through hole tapped for a 3/8-18-NPT thread on both ends by which the plumbing from the pump connects. Offset a small distance from this center hole is a plugged 1/8-27-NPT through-hole acting as an air bleeder. On the top surface in the region in contact with the body section are two concentric O-ring grooves sized for No. 439 and 444 nitrile rings. Only the innermost ring is sealing; the second trepan serves to concentrate clamping pressure on the diaphragm by reducing the area in contact, thereby affecting a better seal.

The top end cap is of identical base dimensions as the bottom except does not have the through holes. On the top surface are two shallow tapped blind holes into which threaded hoisting rings are fastened, which allow the vessel to be lifted by the gantry crane seen in Fig. 1. The flanges are threaded onto the body, and the caps are fastened to the flanges by 12 1/2-20-UNF Grade 9 bolts and nuts at both top and bottom. However, in the bottom, every third bolt is replaced by a longer one, which terminates in a swiveling foot, which gives the vessel approximately 75 mm of ground clearance and allows the plumbing to attach directly to the bottom of the lower end cap.

All surfaces expected to be in contact with the process media were coated with Cerakote Elite Series E190 (Cerakote, Inc., White City, OR). This includes the entire inside cylindrical surface of the body section, the top and bottom end faces of the body including the O-ring gland, and the bottom surface of the top end cap, which lies in contact with the face seal and encloses the pressurized volume. Cerakote Elite Series is a resin-based ceramic thin film coating, which was chosen primarily for its excellent hardness and corrosion resistance, with survivability of the coating exceeding 4000 h in an ASTM B117 salt spray test [15]. The coating was also noted for its flexibility and adhesion, important concerns when taking thermal expansion and cycling into account, combined with large compressive stresses. Application of the coating was conducted according to manufacturer's instructions by DiFruscia Industries, Inc. (Johnston, RI), a Cerakote licensed applicator. This coating allows the interior of the vessel to handle many different corrosive liquids directly, including the salt water used in many ALT's.

All manner of plumbing, pressurizing, and pressure monitoring equipment is attached to the vessel at the 3/8-18-NPT tap at the bottom of the lower end cap, as shown in Fig. 2. The assembly consists of an elbow threaded into the bottom end cap, which then routes a pipe nipple between the legs of the vessel. Following this is a cross into which a pressure transducer is threaded on one side, and a safety head with a rupture disk is threaded on the other. This rupture disk is set to burst at a pressure slightly higher than the vessel's maximum allowable working pressure of 41.3 MPa, and well below the design failure pressure. Following the cross is a needle valve which serves to fully isolate the vessel, terminating in a hydraulic coupler that allows connection to the pump.

The pump used was an Enerpac P39 (Enerpac, Inc., Menomonee Falls, WI) hand operated hydraulic pump, although the standard connections on the vessel allow for any hydraulic pump to be used. The vessel does not have a separate valve for pressure release, although the design is flexible enough to include one, which motivated the use of a single-acting pump with integral release valve. An elastomeric hydraulic hose connects the pump to the plumbing assembly.

The corrosive nature of the proposed process media at the design specified temperatures necessitated a novel approach for a method of pressurization for the vessel. As discussed, several possible options were considered and ruled out for their added expense and/or safety concerns. Ultimately, a novel diaphragm approach was chosen as a media isolator between the process media and any viable pressurizing media, in practice a stable hydraulic oil. The principles of its operation are shown in the cross section rendering of Fig. 3. The water domain contained in the pressure vessel's usable volume is seen at (a). Hydraulic oil is pumped through the plumbing stem at the connector (d), as discussed in Sec. 3.3. The oil flows through the central hole in the lower end cap and pushes against the diaphragm (b), occupying a volume (c). This decreases the volume of the water domain, inducing a pressure in it nearly equal to that in (c), save for the diaphragm's resistance to deflection. The diaphragm is sealed on the oil side by O-rings at (e), as well as by compression in the band of material between the glands. Because some pressure is required to deform the diaphragm into its bell-like shape, the oil pressure is naturally slightly higher, and so a gasket style seal is all that is required between the diaphragm and the lower end face of the body section. Details on the material and construction of the diaphragm as well as design calculations for the diaphragm are given in Secs. 3 and 4.

Two used diaphragms are shown in Fig. 4. Constructed of a nylon fabric reinforced Buna-N (nitrile) sheet with part number C2500-0125-121 (BRP Manufacturing, Inc., Lima, OH), the material has characteristics given in Table 1. The rubber itself has an ASTM D2000 [16] standard callout of 2BG515A14B14C12E014E034F19, as reported by the manufacturer [17]. The fabric is a plain woven, high-tensile strength nylon manufactured by DuPont (Wilmington, DE). This material was chosen because of its good chemical resistance to all the process media anticipated to be used with the facility. Nitrile is a common seal material particularly around hydraulic oils, and nylon was chosen because of its tolerance to both salt water and oils in case the face sheet of the rubber was penetrated in any way exposing the cloth. A cloth reinforcement is highly desirable as it was shown to maintain dimensional stability and limit compression set when used in the high-pressure seals of the vessel. Whereas unreinforced nitrile of similar hardness was seen to extrude at the sealing locations, much of this was seen to be mitigated by the presence of the cloth. More discussion on the effects of extrusion and compression set will be presented in Secs. 6.1 and 6.2.

For a Poisson's ratio of 0.49, this parameter evaluates to 0.656.

By substituting Eq. (1) into Eq. (5) and solving for w₀ given the value of ΔV calculated from Eq. (4), the maximum excursion of the diaphragm was found to be 22.2 mm. The average strain along the radius can be obtained by comparing the arc-length of half the deformed shape to the initial value of simply a. Doing so results in an increase from 76.2 mm to 80.0 mm, or a 5% average radial strain. This value is low enough in comparison to the failure strains of either material, the lower of which is reported in many handbooks to be at least a factor of five times greater, to conclude that the diaphragm should not tear from membrane strains and is also of a magnitude where the stiffness assumptions hold. It should be noted that this is a conservative approximation for the purpose of qualifying safe operation. These calculations do not consider the volume of specimens and fixtures, which displace water and lower the effective compressibility of the system, thereby reducing excursion and thus membrane stresses. Additionally, by substituting the maximum excursion value into Eq. (3) and solving for q, which is a good estimator of the pressure drop between the process and pressurizing domains, a value of 10.3 kPa is obtained, which is a negligible difference within the margin of error of most gauges used in the vessel's pressure range. It should be noted that this pressure drop and membrane strains will in practice be significantly less due to the Poisson effect. As the diaphragm experiences large compressive stresses through the thickness, these will tend to relieve the in-plane tension of the radial and circumferential stress components, possibly even overcompensating and resulting in net in-plane compression, leading to wrinkling of the diaphragm.

This system was designed primarily to conduct water ingression-type accelerated life tests. Many polymers and polymer matrix composites (PMCs) have been known to respond negatively to long-term exposure to a marine environment, with significant degradation in mechanical properties being the norm. Diffusion of water into polymers under large hydrostatic pressures is a relatively well studied phenomenon, yet the effects on mechanical properties have not been well investigated. Additionally, due to the complications of pressurizing heated saline water that this system addresses, previous work has almost exclusively used pure, de-ionized water with the assumption that the added salt does not have an appreciable effect. To bridge this gap in the body of knowledge, a study was performed which investigated both the quasi-static and dynamic performance of carbon fiber/epoxy PMCs after being subjected to sea-floor depth pressures for simulated time scales of the order of years. A summary of some of the results from this study is provided below, with the unexpected results highlighting the necessity to study composite materials in such an environment.

First, a diffusion study was performed which characterized the temperature dependence of the diffusivity of water into the composite. This was accomplished utilizing the small pressure vessel shown in Fig. 1(b) and drawing absorption curves at high pressure and over a range of temperatures. Once the temperature dependence was quantified, a quantity known as the acceleration factor, was calculated, which related in-laboratory exposure time at elevated temperatures with real-world service time at lower temperatures. For the specific PMCs used in this study, the acceleration factor was determined to be 0.64 yr/day when aged at a temperature of 70 °C and pressure of 41.3 MPa. These conditions were routinely monitored using the vessel's built-in pressure gauge and drum heater's digital thermometer at intervals not exceeding 12 h throughout the weathering duration. Three weathering cases were chosen, an unweathered baseline, 14 days, and 24 days, equivalent to 0, 8.9, and 15.3 years of actual service, respectively. Aging was performed using the Accelerated High Pressure Aging Facility as documented in this article. To protect the interior of the vessel from any galvanic reaction to the carbon fibers, specimens were held within polycarbonate weathering fixtures. Fixture contact with the specimens was limited to the edges only so as not to interfere with moisture ingression, and specimens were isolated from contact with each other. Once coupons were secured in the fixtures, the entire assemblies were placed centrally into the aging vessel avoiding contact with the side walls. Additionally, all specimens were desiccated for at least 48 h to purge them of moisture before aging and diffusion studies.

Specimens were cut using a water-cooled diamond abrasive saw to dimensions set forth in ASTM D3039 [21] for tension coupons and ASTM D7264-15 [22] for three-point bend coupons. Tensile coupons measured 178 mm long, 25 mm wide, and 1.3 mm thick nominally, while flexural coupons measured 292 mm long, 20 mm wide, and 3.8 mm thick with a 200 mm span nominally. Tensile tests were carried out in an Instron 5585 universal test frame equipped with a 250 kN load cell and taken to failure, while three-point bend tests were carried to 15 mm center deflection of a 20 cm span. The engineering stress–strain curves are shown in Fig. 5. The tensile behavior of the composite was unchanged by aging, with a linear trend persisting until ultimate failure, as expected. However, flexural behavior was highly affected by moisture ingression, decreasing significantly as aging time increases. Further analysis of the stress–strain data allowed the calculation of the average tensile moduli, flexural moduli, and tensile stresses and strains at failure for the three weathering cases, given in Table 2. All testing techniques and data reduction were performed according to the pertinent testing specifications [21,22].

Most notably, the tensile modulus is unchanged by weathering, the observed differences being within a 95% margin of error for each test, while flexural modulus decreases significantly, by 12.9% and 20.8% after 14 and 24 days of exposure, respectively. Due to the variability in general of failure of polymer–matrix composites, the differences in tensile failure stresses and strains are not statistically significant, and thus the common trend of increased degradation with aging was not present in tension during this high pressure study, though flexural properties degraded at similar rates to what would be expected. Although work is ongoing to further study this phenomenon, it is speculated that the large hydrostatic pressures, while aiding diffusion on a macroscopic level, close voids at the matrix–fiber interface and forces the ingressed water to concentrate between plies, thus leaving the in-plane properties relatively unchanged. This result was unexpected and shows the importance of using this system for performing high-pressure ALTs.

The diaphragm and O-ring seals, being composed of elastomers, are naturally sensitive to changing temperatures and thermocycling. Also, because the diaphragm is effectively a composite of nitrile and nylon, there are additional potential complications of delamination due to through-thickness stress and dissimilar rates of thermal expansion. In this section, discussion of these factors will be presented based on various observations made during the prototyping and testing stage.

The effects of temperature on the diaphragm can plainly be seen by comparing the two sample diaphragms shown in Fig. 4, both of which saw similar pressures but at different temperatures. Raising the temperature softened the diaphragm material and increased the amount of extrusion into the oil-side O-ring glands. During a 65 °C test where leakage occurred at approximately 35 MPa, inspection found catastrophic failure of the inner O-ring to be the immediate cause. Heavy extrusion and partial delamination of the outermost layer of rubber from the cloth can be seen at this location in Fig. 4(d), suggesting that the problem was multifaceted. While the delamination was likely the result of the violent tearing nature of the failure, the extrusion of both diaphragm and O-ring material, which were both similar grades of nitrile, suggests thermal softening of both parts was contributory. O-ring extrusion occurs in a face seal when the gap size between the faces reaches a critical size for the O-ring and pressure. A soft O-ring at higher pressures is much less tolerant of a gap than a hard O-ring at lower pressures. It is thought that because the flange was tightened at ambient temperature, as the temperature increased and both diaphragm and O-ring softened, the pressure was enough to open a gap between the now compliant diaphragm and lower end cap, a gap the O-ring could not tolerate.

To counteract this problem, two modifications were implemented with success. First, the nitrile O-rings were replaced with FKM of a higher durometer rating. FKM flouroelastomer rings are known to retain their stiffness better at higher temperatures, and the higher hardness was chosen to increase gap tolerance and lower the risk of extrusion. Additionally, the assembly procedure for the bottom seal was modified to incorporate preheating. The diaphragm and O-rings were installed as before at ambient temperatures and the bolts tightened, but then the entire vessel was heated at the maximum operating temperature of 70 °C and allowed to come to thermal equilibrium over the course of several hours. The vessel was empty during this time and any pressure caused by heating could release. Then the bolts were tightened to their torque specifications once again to take up the increased compliance of the diaphragm due to heating. A slight increase in the compliance was in fact noticed by the apparent loosening of the bolts before their retightening. Together, these two modifications proved successful in maintaining integrity of the seal, as the vessel was cycled several times and allowed to hold pressure for several days without issue before it was once again disassembled to check for any degradation of the diaphragm.

A detail of the diaphragm shown in Fig. 4(d) is shown in Fig. 6. The modes of degradation of this example serve as good examples of the types of deterioration that can occur in this system if proper countermeasures are not put into effect. In Fig. 6(a), the primary extrusion lobe can be seen, where the diaphragm material begins to be forced into the inner O-ring gland. Moving outward radially, some delamination of the outer rubber face from the lower cloth ply is seen. The severity of this delamination is most likely the result of the sudden failure of the O-ring and the large release of pressure that followed, and not from normal service. It should be noted that even with this delamination, only the outer face is affected; the core layer of rubber, which is sandwiched between two plies of cloth, is undamaged and so the process and pressurizing media remained isolated even after failure. More delamination is seen in Fig. 6(b), this time occurring over the secondary O-ring gland. Once again, only the outermost layer of rubber is affected, and the middle and top layers of rubber are not penetrated. The cause of this is thought to be the discontinuity in through-thickness compression at the edge of the O-ring gland combined with thermal softening of the nitrile. The compression on either side of the discontinuity combined with the Poisson effect most likely formed the softened rubber into a lobe or bubble, which was torn by the sudden friction of the extruding O-ring and the subsequent release of pressure.

Additionally, small amounts of delamination were observed on the water side, due to the clamping pressure of the body section against the lower end cap. This can be seen in Fig. 7, a cross section of a diaphragm that leaked slightly after several hours of test pressure hold but did not fail catastrophically. A primary delamination lobe is seen at (a), where in-plane compression due to bending of the diaphragm into the water section combined with the through thickness pressure to cause some separation of the rubber face. Once again, the reinforcing nature of the cloth restricted the inner rubber layer from doing the same, and thus there was no through-thickness penetration. A similar phenomenon was observed on the outer edge of the seal at (b), although it is not as pronounced as at (a) due to the lack of pressure on the outlying portions of the diaphragm. The cloth is exposed in a similar manner as can be seen at Fig. 6(b).

Additionally, extrusion at the O-ring glands can be seen at Fig. 7 location (c), but there does not appear to be delamination of the rubber at this location. On the interior portions of the diaphragm where the through-thickness stress is constant without discontinuities at the boundaries, seen at (d), the material is unaffected; the strength of the nylon cloth is enough to restrict the nitrile from extruding without delaminating.

In summary, most if not all damage to the diaphragm occurs at stress discontinuities at the boundaries of the water side seal and at the edges of the O-ring glands. The severity of the damage is heavily dependent on temperature and the softening of the nitrile at elevated temperatures. It is thought that hardening the nitrile at these locations, perhaps by a chemical treatment or the bonding of a hard polymer to the face layers, would significantly reduce the probability of this damage progressing and increase the reliability of the system as a whole.

Although the vessel was only tested and utilized with a 3.5% saline solution, the system may easily be adapted for other aging media. The ceramic thin film presented in this article is known to be inert to many corrosive agents; interested readers are referred to the manufacturer for an up to date listing. After a suitable coating is chosen, chemical compatibility is only a matter of choosing the correct sealing material. Particular attention should be made to hardness for the O-ring seals. Reliability of the diaphragm may drastically benefit from a tailored design. As described above, noticeable points of deterioration are around the edges of the clamping region and O-ring glands. A different choice of reinforcing material, denser reinforcement in this area, or a rigid insert or chemical hardening may yet bolster performance. Although the authors used a commercially available reinforced rubber for prototyping efforts, a purposefully designed seal is recommended for any future new construction.

A new system was developed for the pressurization of corrosive media at high pressures for extended periods of time and at elevated temperatures. The interior of this system was coated in an inert ceramic/epoxy coating, which provided adequate corrosion protection for all manner of static sealing. Using a fabric reinforced nitrile diaphragm combined with flouroelastomer supplementary seals, the corrosive process media, in testing a saline water solution, was isolated from all manner of pressure generating, monitoring, and safety equipment, while reliably transferring pressure from the hydraulic actuating media to the saline water. The design was shown to be simple and modular while still affording the same level of safety as traditional pressure vessels. The system was shown to have applications in accelerated life testing, particularly for deep ocean environments, where the effects of long-term storage on the sea floor on the flexural and in-plane tensile properties of carbon-fiber/epoxy laminates were investigated experimentally.

The authors would like to kindly thank Dr. Maria Medeiros from the Office of Naval Research for financial support under Grant No. N00014-18-1-2641, as well as colleagues at the Dynamic Photomechanics Laboratory, for their experimental advice. Additionally, the authors would like to extend appreciation to Dr. Carlos Javier and Dr. James LeBlanc from the Naval Undersea Warfare Center—Newport for their advice and support.

• Office of Naval Research (Grant No. N00014-18-1-2641; Funder ID: 10.13039/100000006).