Assuring the Safety of Natural Gas Vehicles

Success for gas in this new role depends partially on defining effective procedures for fuel tank certification

By Stephen J. Hudak Jr., Ph.D.

When will natural gas vehicles (NGV) become commonplace? This question is being asked by automakers, engine manufacturers, makers of fuel storage cylinders, environmentalists, and even some consumers. Not surprisingly, the answer depends on whom one asks. This is true even within the microcosm of Southwest Research Institute (SwRI). For example, engineers in the Engine Research Department would be likely to say it will happen when advanced engine technology has succeeded in making cleaner, more reliable, and more fuel-efficient NGV engines. Others, however, assert it will come to pass when it can be demonstrated to regulators, auto manufacturers, and consumers that driving around with a 3,000-psi pressure vessel in one's car does not appreciably add to the risk of an already risky endeavor.

ngv1.gif (33795 bytes)

Institute engineers are evaluating a variety of compressed natural gas (CNG) cylinder designs to devise safety inspection procedures. Future research will examine the design of a cylinder optimized for weight, cost, and durability. Candidate designs consist of plastic or metal liners overwrapped with either glass or carbon fibers contained in an epoxy matrix.


While it is true that the performance of NGVs around the world have had an impressive record of safe operation, this record is not unblemished. For example, a catastrophic failure of a compressed natural gas (CNG) fuel cylinder occurred during refueling of a car in New Zealand in 1989. Fortunately no loss of life resulted. The failure was linked to improper heat treatment of the cylinder during manufacturing, a mistake that not only decreased the steel tolerance to cracks, but introduced the crack that subsequently grew to failure in service. The lesson learned from this failure is not that a generic safety problem exists in CNG cylinders. The faulty manufacturing step that caused this particular failure is easily avoidable. However, other human errors are likely to occur in the future during either manufacturing or service. The important lesson is that a more effective procedure is needed for initial certification, as well as for periodic recertification, of CNG cylinders. This need is emphasized when it is learned that the New Zealand failure occurred only months after this new cylinder had been introduced into service, even though the standard recertification test, consisting of an overpressurization to 5/3 the service pressure, was performed by the manufacturer.

The Institute is engaged in a Gas Technology Institute (GTI) sponsored program designed to eliminate such mishaps. The objective is to answer the following safety and regulatory questions about CNG cylinders: How often do CNG cylinders have to be recertified? What is the best recertification test, and can it be done without the costly operation of removing cylinders from the vehicle? This article highlights the technology being used to answer these questions for existing cylinders and indicates how it can be applied to future cylinders.

ngv2.gif (34215 bytes)

Cracks can develop in CNG cylinders as a result of manufacturing defects, improper heat treatment, or localized corrosion that may occur in aggressive environments. These micrographs show fissures that developed on the inside surface of a steel cylinder as the neck region was being formed.


The Approach

The general approach is a systems analysis that encompasses several different disciplines. Structural reliability is achieved through the integration of nondestructive evaluation (NDE), materials evaluation, and fracture mechanics analysis. The role of NDE is to find defects and estimate their orientation and size. Materials evaluation serves to measure the rate at which cracks grow in materials, often under the influence of aggressive environments, as well as the critical combination of defect size and stress required to bring the material to catastrophic failure. Fracture mechanics analysis combines all this information, along with a stress analysis, to predict the remaining useful life of a structure as a function of defect size. This information can then be used to establish a rational inspection limit, provided that the minimum defect size that can be reliably detected for the NDE technique is known. In some cases, it can be shown that in-service inspections are not needed if the computed remaining life for a detectable crack exceeds the anticipated service life.

This approach assumes that the useful life of the structure is controlled by the growth of cracks, and does not consider the time necessary for cracks to initiate. This assumption is valid for many structures that contain pre-existing cracks or micro-structural defects that form cracks relatively quickly. This assumption is applicable to CNG cylinders that can contain manufacturing defects, or to aggressive environments that can easily initiate cracks where localized corrosion occurs.

The most critical form of damage in CNG cylinders is crack growth due to corrosion-fatigue. The latter process occurs due to the combined action of the corrosive impurities in CNG -- hydrogen sulfide, carbon dioxide, water, or water vapor -- and the pressure cycling associated with periodically expending and replenishing the fuel.

Because of the interdisciplinary nature of this approach a team with diverse backgrounds and interests is required.

ngv3.gif (28359 bytes)

Research on crack detection in CNG cylinders showed that acoustic emissions (AE) technology was able to detect significant emissions from crack growth during both pressurization and depressurization. John Hanley, research engineer in the SwRI Nondestructive Evaluation Science and Technology Department, participated in the AE research program.


The Sounds of Failure

Since NGV fuel cylinders are relatively inexpensive in contrast, for example, with nuclear pressure vessels -- the NDE technique used to recertify cylinders needs to be quick and inexpensive. For this reason, the acoustic emission (AE) technique is the method of choice as, unlike ultrasonic inspection, it does not require the entire surface of the cylinder to be scanned with transducers. Instead AE transducers are attached to the cylinders in a few selected locations and passively listen for the sounds of damage as the cylinders are pressurized.

Acoustic emissions are elastic stress or pressure waves generated by local transient instabilities in materials. A variety of these processes are active during initiation and propagation of cracking, and range from dislocation motion on the atomic scale to crack advance on the macroscopic scale. In rare cases, cracking is audible to the unaided ear, as in the fracture of wood. However, in most cases sensitive instrumentation is required -- perhaps the most familiar example being the use of a seismograph to detect the shock waves produced by the earthquakes.

When using AE as a NDE technique, detection of cracking is of primary importance. However, it is preferable to detect a crack when it is relatively small so that it is still growing in a slow, stable manner. Although unstable cracking, which occurs at final failure, is relatively easy to detect with AE, this phenomenon is of little use for periodic in-service inspection. To be effective, the corresponding inspection intervals would need to be impractically short.

We began the GTI-sponsored program with a recognition that the primary weakness in previous studies was their inability to interpret the AE data properly. In other words, what did a given emission rate mean and how did it relate to the state of damage in the materials or structure? This situation often resulted from a lack of understanding of how the events of interest actually produced acoustic emissions. Thus, we set out to understand more clearly the mechanisms by which AEs were produced during corrosion-fatigue -- the most critical form of damage in CNG cylinders. This approach led to the discovery that the abrasion of the fracture surfaces, and associated corrosion products, was a more important source of acoustic activity than those arising from the relatively small advance of a crack on each cycle of loading. Furthermore, AE from abrasion occurred at low pressure as the crack surfaces were in contact, while those forming crack growth occurred near maximum pressure. As a result, we learned that data should be gathered during the entire pressurization cycle instead of focusing on the region of maximum pressure as is commonly done in conventional AE testing.

We also believed that it was crucial to locate the AE sources accurately, instead of merely counting the total number of emissions as is conventional. In this way, a map could be generated that clearly showed the distribution of emissions over the entire cylinder. Mapping the location of sources is particularly important for localized forms of damage such as cracking that can be identified by their high density of detected emission sources.

An accompanying figure shows a map of AE activity from a steel cylinder containing suspected defects. The rectangular region of the map represents the unrolled cylindrical region, while the two polar plots represent the hemispherical ends of the cylinder. The eight numbered circles show the locations of the transducers used to detect the emissions. Knowing the velocity of elastic wave propagation, the AE sources were located by triangulation. The high density of emissions in a spot in the cylindrical region is believed to be a crack, while the cluster of emissions in one of the hemispherical ends is believed to be due to manufacturing folds. Both locations are currently being sectioned to verify the presence of these defects.

ngv4.gif (21218 bytes)

A map of acoustic emissions (AE) activity from a steel CNG cylinder shows suspected defects. AE transducers are attached to the cylinder at the eight numbered circles to listen passively for the sounds of damage as the cylinder is pressurized. The rectangular map represents the unrolled cylindrical region, while the two polar maps represent the hemispherical ends of the cylinder.


In addition to these defects, AE characteristics have also been identified for weld arc strikes and general corrosion. Each of these forms of damage has been shown to have a unique AE signature. For example, in addition to emissions from cracks being highly localized, they also persist from one pressurization cycle to the next and grow more intense with time, or with each cycle. In contrast, emissions from corrosion tend to be widely dispersed and diminish significantly after the first pressurization cycle -- since most of the brittle corrosion products break away from the metal surface on the initial cycle. Thus, AE can be used not only to locate damage in cylinders, but to identify the specific form of damage. Remaining work is focusing on identifying the minimum damage that AE can reliably detect, and more importantly, the maximum extent of damage, particularly cracking, that AE is likely to miss.

How Fast Do Cracks Grow?

The presence of contaminants in CNG enhances the rate at which cracks grow in cylinder materials, compared to crack growth rates in air. Information on the degree to which growth rates in cylinder materials is essential in making a fracture mechanics based life prediction. Such measurements are being made at the Institute on both steel and aluminum alloys used to manufacture cylinders using a specially designed test chamber to contain this aggressive environment. This is combined with a computer-automated testing system to monitor the growth of cracks. The latter enables tests to continue around the clock for periods of several months, since tests must be performed at loading rates equivalent to those experienced during cylinder refueling in service.

Typical data, for the case of an alloy steel cylinder material, are also shown. The crack growth rate, expressed in inches per cycle of loading is shown as a function of the variation in the crack-tip stress intensity factor, DK. The latter factor uniquely characterizes the magnitude of the elastic stresses at the crack tip and thereby serves as a measure of the mechanical "driving force" for crack growth. (Values of DK can be computed for a given specimen or component geometry, provided one knows crack orientation.) Notice that the presence of the simulated CNG environment increases the rate at which cracks grow under fatigue loading by a factor of 20, compared to rates in a relatively inert environment of laboratory air. Fortunately, however, the detrimental effects of the simulated CNG environment on the crack growth rates diminish as DK decreases. This means that the environment becomes less effective at enhancing the crack growth rate for lower applied stresses, or smaller crack sizes. Detailed measurements suggest that this behavior is due to the build-up of corrosion products within the crack that serve to wedge the crack shut. This process is consistent with the enhanced emissions that occur at low pressure during AE testing of cylinders.

Measurements were also made of the fracture toughness of the material (Klc) for cylinder steels. This property defines the crack-tip stress intensity factor at which the crack becomes unstable and leads to catastrophic failure in a given material.

ngv5.gif (10362 bytes)

Since the minimum detectable crack size for AE technology has not yet been determined, it is convenient to express predicted remaining cylinder life as a function of the initial crack size. The calculations shown here are for a cylinder subjected to two repressurizations (refuelings) a day from 300 to 3,000 psi, a frequent refueling cycle typical of a taxi. Results indicate that the remaining life is relatively sensitive to the initial crack size. Specifically, an initial crack that is 25 percent of the cylinder wall thickness corresponds to a remaining life of about 95 years, while a deeper crack of 40 percent of the wall thickness corresponds to a cyclic life of only five years.


Defining Useful Cylinder Life

The primary advantage of the fracture mechanics approach is that these properties, when expressed in terms of crack-tip stress intensity factor, are independent of geometry. This feature enables laboratory measurements, made on conveniently sized specimens, to be transferred to full-scale structures, through appropriate analysis. In essence, this lets one compute the cyclic life for any structure by integrating the crack growth kinetics from some initial stress intensity factor, defined by the minimum detectable crack size, to some final stress intensity factor, defined by the materials fracture toughness.

This procedure has been implemented for the case of a steel CNG cylinder. Let us conservatively assume that this cylinder is subjected to two repressurizations (refuelings) a day from 300 to 3,000 psi. Such a frequent refueling cycle might, for example, be typical of a taxi. Since the minimum detectable crack size for AE has not yet been determined, it is convenient to express the remaining cyclic life as a function of the initial crack size. These results indicate that the remaining life is relatively sensitive to the initial crack size. Specifically, an initial crack that is 25 percent of the cylinder wall thickness corresponds to a remaining life of about 95 years, while a deeper crack of 40 percent of the wall thickness corresponds to a cyclic life of only five years. Consequently, recertification of the cylinder using AE testing at five year intervals would require that cracks of less than 40 percent of the wall thickness be reliably detected.

Future Challenges

Safety is not the only issue of current interest with regard to CNG cylinders. Future cylinders need to be lighter in weight and less expensive than today's cylinders. In fact, cylinder economics are more critical to the evolution of the NGV market than are engine economics. This is because the assembly line manufacturing costs of advanced, high efficiency NGV engines will be equivalent to competing gasoline powered engines. In contrast, even the most economically designed CNG storage cylinders are expected to add at least $l000 to the initial cost of an automobile. Thus, consumers will be forced to balance this higher initial cost against projected fuel savings and indirect benefits of a cleaner environment. The Gas Technology Institute, recognizing the overall impact that improved fuel storage cylinders can make on the growth of the NGV market, is embarking on a new research initiative to design lightweight, cost-effective cylinder, as well as verify their durability and safety. Candidate designs consist of plastic or metal liners overwrapped with either glass or carbon fibers contained in an expoxy matrix. The need to optimize for weight, cost, and durability represents a challenge to the designer. Successfully meeting this challenge will require additional NDE research to prove the usefulness of AE on composite cylinders, as well as the development of a rational life prediction methodology for fiber-reinforced composites. Because of the distributed nature of damage in these composites, fracture mechanics concepts are not directly applicable since failure is not controlled by the formation and growth of a dominant crack. The Institute is pursuing the development of a novel approach to predict the remaining useful life of such composite cylinders. This approach may soon be used to ensure that clean and efficient natural gas vehicles continue to operate safely and reliably.

Published in the Fall 1991 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Joe Fohn.

SwRI Publications separate.gif (834 bytes) SwRI Home