PWR Engineering - Threshold Journal: Designing for the Long Term

The science of predicting and establishing the lifespan of working mechanisms has existed at least as long as the design parameters of the mechanism themselves have been considered. Designers have always directed their attention not only to how well a device would work but for how long. As device sophistication has increased, concern has moved to how long a device would function well.

Those dual considerations are especially crucial in the design of high-performance propulsion devices, such as rocket engines, particularly when they are involved in the conveyance of human beings. The difference, in fact, between the reliability requirements for unmanned flight versus manned flight does constitute "a whole new ball game." And now, with multiple-use engines being implemented for manned flight-as opposed to the expendable propulsion systems of the 1960s and 1970s-the importance of an ever more accurate gauge of genuine reliability has grown to very high levels.

Life prediction methodologies are an evolving factor in accord with expanding hardware design requirements, improving analytical methods and an emphasis on hardware reliability and cost. The continual development of accurate means for predicting structural capability are essential for current mission success as well as in the support of future endeavors in high-quality, low-cost products.

Engines such as the Atlas, Delta, J-2 and F-1 that were designed in the 1950 to 1965 time frame were expendable engines with a single flight requirement. Those engines were designed primarily on the basis of static strength, with limited structural analysis, material failure modeling and definition of dynamic environments. Components were evaluated based upon their ability to sustain or survive internal or external pressure, the effects of body forces induced by high-speed rotation, and external flight loads. Thus, the procedural approach was to subsidize limited analysis with extensive component and engine testing until all known hardware issues were resolved. As analytical methods matured, analysis was used to supplant some of the dependence on testing to expose limitations in the design.

In contrast, the Space Shuttle Main Engine (SSME), designed in the 1970 time frame, required a fourfold increase in the energy density of its turbopumps, a threefold increase in dynamic flow loads, chamber heat flux and internal system pressures and a reuse criterion of 55 flights. It was the first rocket engine with a reusable, man-rated requirement, so it necessitated the additional consideration of detailed dynamic loads and the performance of structural response and life prediction through fatigue analysis on an overall engine basis.

Fatigue is a measure of resistance of a structure to the initiation of cracking when exposed to repeated fluctuations (cycles) of load. The effect of high frequency vibration, as well as repeated missions, must be assessed to define usable mission lifetime.

The implementation of a fatigue-based approach involves retiring hardware when a designated fraction (e.g., 0.25) of the predicted life is expended. This analytical approach considers upper bound loading and minimum properties for the fleet design. Individual pieces of hardware typically have significantly more capability. However, since fatigue damage prior to crack initiation cannot be seen or detected, all components, good and bad alike, would be retired when this calculated damage fraction was reached.

The need for fracture control was recognized at the start of the SSME design. An approach was needed that accepted the possibility that cracks or defects could exist in hardware as a result of material processing, fabrication, handling or usage. The subsequent development of the SSME clearly demonstrated that fracture control considerations must be utilized to have a cost-effective, yet reliable engine. As a result, improvements continue to be developed for fracture control in the next major engine design.

An integral part of fracture control is the use of nondestructive inspection procedures that can reliably detect cracks of a given size and fracture mechanics-a technical discipline that quantitatively predicts the capability of a structure to survive with the existence of cracking and progressive crack growth. The inherent advantage of a fracture control design is that damage is observable, with the potential for failure quantifiable; in addition, hardware limitations can be identified by the detection of detrimental-sized defects. Therefore, deficient hardware can be differentiated from those that are acceptable and retired, allowing remaining hardware to be utilized further, with significant leverage in cost to the program. The SSME injector plate incorporates components that are observed closely for the occurrence of defects.

The function of an injector is to introduce and meter the propellant flow to the combustion chamber. The main injector on the SSME contains 600 injection elements-or LOX posts-spaced uniformly across the face of the injector, both radially and circumferentially, into 13 annular rows. The posts are tubes that provide flow area for one propellant - liquid oxygen - injected at low velocity through the center of the tube. The second propellant, hydrogen-rich hot gas, is injected at high velocity through the surrounding annulus around the outside of the tubes in the array.

The loading experienced by the posts is both static and dynamic. Static loading occurs in each flight as a result of (1) thermal loads caused by a high-temperature gradient through the tube wall, (2) drag loads from hydrogen-rich steam impinging on the post and (3) direct loads from differential pressure across the faceplate and between the inside and outside of the post. Dynamic loads result from mechanical vibration of the powerhead and flow-induced vibration.

Early LOX post design incorporated the use of corrosion resistant steel, 316L CRES. Cracks occurred in the retainer thread and tip (end) regions at posts in high hot gas flow regions.

Simple material models were incorporated for assessing the cycle-dependent and time-dependent capabilities of the posts. Basic life prediction models were used to address high cycle fatigue (HCF), low cycle fatigue (LCF) and creep rupture (CR). Both high- and low-cycle fatigue describe time-independent damage caused by load cycling, while CR describes damage that is time-dependent. A simplified linear cumulative damage law was used to assess the effects of a combination of the above types of damage.

High cycle fatigue relates life, in terms of cycles, to the level of fluctuating applied stress (force per unit area). This is applicable when the material's cyclic deformation response to load is linear (elastic). It generally occurs in structural regions where the stress fluctuations are small and stress amplification is relatively low because of gradual geometric contour changes. Cycle life requirements are generally in the millions. Typical hardware that are designed for HCF include engine ducting and lines, and injector elements, as well as the internal hardware in pumps and turbines, such as impellers, turbine blades, stators and sheet metal.

SSME LOX post cracks in the thread and tip regions were attributed to HCF loading resulting from flow-induced vibration. A design change was instituted, after which no further cracking was observed. The 316L CRES was replaced with a material called Haynes 188 for the entire post. Analysis for the Haynes 188 post design indicated that the HCF capability was sufficiently raised so that LCF-type loads became more important.

Low cycle fatigue relates life to the level of applied total strain range. Strains indicate the magnitude of the relative distortion of a material under load. This approach is applied when the material's cyclic deformation response to load is not linear (plastic). Nonlinear material response occurs, for the most part, when stress fluctuations are large, thermal gradients are significant, or abrupt geometric contour changes exist. Typical hardware designed for LCF includes almost all components internal to turbines; i.e., blades, stators, etc., and combustion devices components, such as injector elements, liners and nozzle tubes.

Creep rupture relates life, in terms of time, to the level of applied sustained stress. Time-dependent effects occur, as a rule of thumb, above half the homologous (melting) temperature of the material. Creep, as referred to here, relates to a monotonic material deformation response, but it can also be related to a cyclic deformation effect and can occur both in the elastic and plastic regions.

The complex load histories experienced by space propulsion hardware, in general, highlight the need for accurate prediction of cumulative damage. The analyst must combine the effects of HCF, LCF and CR.

Existing inspection techniques have a limit as to the size of the defect that can be detected 90 percent of the time, with a 95-percent level of confidence-an accepted criteria. The analyst must assume that a worst-case defect of this size exists in the worst location on the component. But the analysis must demonstrate that this defect will not grow to failure condition under the operating loads.

Crack growth assessments are addressed only in the low-frequency pressure and thermal cycling of engine components. Environmental effects resulting from high-pressure hydrogen enhancement of the rate of crack growth are included for hydrogen-sensitive materials. Inspected components are accepted for service only if no crack-like defects were detected. Previously, if cracks were detected, the basic policy required that the defect be removed for flight-ready hardware. More recently, components with detected crack-like defects could be accepted for flight by engineering analysis using fracture mechanics techniques on a part-by-part basis.

Excessively stringent analysis requirements may have resulted from the compounding effect of conservative input. A review of the overall methodology has led to the Fracture Control Strategic Initiative to develop a more representative methodology that considers the integrated effect of the uncertainty of all the variables to the overall fracture mechanics analysis. This effort includes involvement of all areas of Pratt & Whitney Rocketdyne (now Rocketdyne Propulsion and Power, a part of Pratt & Whitney ) that affect life prediction: engineering, manufacturing and quality assurance. The current deterministic analysis may consider the worst load, thickness, environment, material and flaw condition to make a conservative assessment. The evolving approach, in contrast, considers an overall combination of the variables and their uncertainty through a probabilistic life prediction analysis. The initial assessment of critical welds considered in this fashion appears to more effectively reflect the actual cond itions and successful operation of engine components. Pratt & Whitney Rocketdyne intends to have a fully functional capability in this area by 1995.

Already, fracture mechanics is used for anomaly resolution to help explain how hardware structural capability degrades and when the part would cease to function, identifying viable scenarios as well as possible courses of corrective action.

Physical evidence on the fracture surface can provide invaluable information relating to the cause or severity of the problem. Arrest marks, striation spacings, crack shape advance patterns, initial and final defect sizes, etc., obtained from fractographic examination can be related to the duty cycle loading. Leakages that adversely affect engine performance can be quantified through crack opening displacement and flow considerations. If the applied loading is well known, observed cracking can be related to material capability. When actual load measurements or accurate calculations are not available, detailed knowledge of crack propagation can be used to estimate the magnitude and gradient of the applied loading. Fracture mechanics prediction has provided valuable insight into occurrences of cracking on the LOX post at both the base and inertia weld regions.

The challenge facing implementation of improved life prediction and damage assessment capabilities for future product design is fourfold. First, basic studies need to be performed using specimen test results to obtain a better quantitative understanding of the accumulation of damage for situations applicable to engine operation. Second, the extension of rather simplistic test specimen data to the much more complicated conditions that exist on engine hardware must be accomplished. Third, better understanding of actual hardware flaw characteristics are needed. Flaws are not necessarily crack-like and can include anomalies such as pores, lack of fusion, lack of penetration and inclusions. And, finally, one must demonstrate that the benefits associated with improvement in life prediction outweigh the costs associated with implementation of the new approach.

Significant advancements through research have been achieved that have yet to be incorporated into the basic design process for rocket engine components. Many of these advances have been born out of identified needs related to the severe operating conditions inherent to space propulsion operation. In addition, simplifications and approximations have been put forth to minimize cost in test requirements and limit analysis complexity for some of these approaches to make them more attractive for component design analysis.

A unified approach to constant temperature (isothermal) fatigue crack initiation can be realized through the use of full-range fatigue curves. Instead of treating HCF and LCF as separate entities, the two can be combined as the sum of an elastic strain contribution and a plastic strain contribution. Consistent material deformation (constitutive) response is automatically defined so that the use of inconsistent constitutive and life prediction material properties can be avoided.

An extension of this same concept has been adopted for describing the combined effects of creep and fatigue on cyclic crack initiation, called the Total Strain Version of Strainrange Partitioning. Four different types of damage-producing cycles are presumed to exist resulting from deforming the material nonlinearly. Systematic rules have been identified to break down, or partition, a complicated load cycle experienced by a component into a summed combination of these four basic elements.

A potential drawback of this approach is the amount of testing required to support development of the model. To temper this concern, methods to estimate model parameters from very basic material properties have been developed based upon normalization of a myriad of empirical test results. The engineering estimations-the Method of Universal Slopes and Ductility Normalized Strainrange Partitioning-are generally considered adequate to typically produce life estimates within a factor of three of actual test results.

Further extension of the above models exist to address variable temperature (nonisothermal) cyclic loading conditions and multiaxial states of stress which are much more representative of actual engine operating conditions. In numerous instances it has been observed that damage produced during nonisothermal loading proceeds more rapidly than that predicted using isothermal data obtained at either extreme of the variation in temperature. In addition, the inherent capability of the material (ductility and crack initiation resistance) may be impacted by the multiaxial state of stress.

The advanced models have been developed under conditions of oxidizing environments for simple laboratory experiments. Applicability of these models to aggressive environments is a critical concern relative to hydrogen-fueled propulsion systems and needs further investigation. Some elements relative to the applicability of these models to a high-pressure hydrogen environment have been evaluated at Pratt & Whitney Rocketdyne under NASA-sponsored technical contract work.

The advanced life prediction models have addressed the effect of different types of damage-producing cycles. However, the loading throughout the life of the component is assumed not to vary from cycle to cycle (termed constant amplitude loading). To account for variations in applied loading from one cycle to the next, cumulative damage laws have been proposed.

Nonlinear cumulative damage models have been developed to more accurately predict available life when variable amplitude loading is present. Examples of such models include the Damage Curve Approach and the Double Linear Damage Rule. These models have been developed from smooth specimen test results. However, limited testing of notched specimens has shown results contrary to smooth specimen testing. Therefore, caution should be exercised by identifying the limitations of the existing models prior to their indiscriminate usage.

Issues related to the practical application of these advanced models and quantification of risk through probabilistic analysis are being pursued at Pratt & Whitney Rocketdyne under complementary NASA-sponsored and IR&D efforts. Application of these models is progressing for several SSME components, including the LOX posts, the main combustion chamber liner and turbine blades. An end goal is to develop a relationship between engine performance variables and life for diagnostic and health-monitoring purposes

Relative to fatigue crack growth, traditional methods have been developed and implemented into design analysis, assuming that the structural response is predominantly linear elastic. Substantial effort has been expended in recent years in the area of nonlinear or elastic-plastic fracture mechanics (EPFM) analysis. Complementary efforts for the development of practical procedures for EPFM design analysis is currently under way at Pratt & Whitney Rocketdyne, sponsored by internal funding as well as major engine and NASA technical contract work.

Both time and cost do not allow empirical test results to be performed on all materials and conditions relative to advanced propulsion system applications. With probabilistic life prediction in design, interrelationships among various material properties, including tensile, crack initiation and crack propagation, are being sought. The primary goals of this investigation are to prioritize and focus testing, to project material capability from limited tests, including data scatter, and to ensure that consistent sets of material data are being utilized for component design analysis.

Significant advancements have been achieved in life-prediction material modeling applicable to component analysis for advanced propulsion system hardware. The combination of enhanced engine performance and duty cycle requirements, increasingly severe operating conditions and the far-reaching consequences of engine failure dictate that further advances should be pursued. To incorporate these improvements into component design, the new methods must be responsive to the specific needs of the application and extendible to more complicated situations than those tested. They must also project a tangible benefit that produces a net gain when implemented. This requires that the models be developed for efficient utilization in component design, and users are aptly educated as to their advantages. With the ever- increasing emphasis on product quality and safety, as well as cost, the continual evolution of improved life prediction capabilities is essential to provide the most cost-effect ive, highly reliable product possible.