Investigation of Failure in Gas Turbines: Part 2 — Engineering and Metallographic Aspects of Failure Investigation

This paper opens with a discussion of the various mechanisms of cracking and fracture encountered in gas turbine failures, and discusses the use of metallographic examination of crack and fracture surfaces. The various types of materials used in the major components of heavy-duty industrial and aeroderivative gas turbines are tabulated. A collection of macroscopic and microscopic fractographs of the various mechanisms of failure in gas turbine components is then presented for reference in failure investigation. A discussion of compressor damage due to surge, as well as some overall observations on component failures, follows. Finally, a listing of the most likely types of failure of the various major components is given.

Investigation of Failures in Gas Turbines: Part 1 — Techniques and Principles of Failure Investigation

This paper is Part 1 of a two-part paper on the principles and methods of failure investigation in gas turbines. The qualities of a successful failure investigator are presented, and the most efficacious approaches to an investigation are discussed. An example of an aircraft accident that might have been avoided is used to support the necessity for thorough and conclusive investigations into failures. Two case histories involving heavy-duty industrial gas turbines are described to demonstrate different aspects of the logical approach to construction of hypotheses and the determination of the essential cause of a failure — the one event without which the failure would not have occurred.

An Optimum Balance Weight Search Algorithm

A mathematical description for an optimum balance weight search algorithm for single plane multipoint balance is presented. The algorithm uses influence coefficients, either measured or known beforehand, and measured complex vibration data to determine an optimum balance correction weight. The solution minimizes the maximum residual vibration. The algorithm allows user defined balance weights to be analyzed and evaluated. A test case is presented showing actual results and comparison with a least square solution algorithm. An efficient multiplane influence coefficient calculation scheme is also presented.

An Application of Octant Analysis to Turbulent and Transitional Flow Data

A technique called “octant analysis” was used to examine the eddy structure of turbulent and transitional heated boundary layers on flat and curved surfaces. The intent was to identify important physical processes that play a role in boundary layer transition on flat and concave surfaces. Octant processing involves the partitioning of flow signals into octants based on the instantaneous signs of the fluctuating temperature, t′; streamwise velocity, u′; and cross-stream velocity, v′. Each octant is associated with a particular eddy motion. For example, u′<0, v′>0, t′>0 is associated with an ejection or “burst” of warm fluid away from a heated wall. Within each octant, the contribution to various quantities of interest (such as the turbulent shear stress, −u′v′, or the turbulent heat flux, v′t′) can be computed. By comparing and contrasting the relative contributions from each octant, the importance of particular types of motion can be determined. If the data within each octant is further segregated based on the magnitudes of the fluctuating components so that minor events are eliminated, the relative importance of particular types of motion to the events that are important can also be discussed. In fully-developed, turbulent boundary layers along flat plates, trends previously reported in the literature were confirmed. A fundamental difference was observed in the octant distribution between the transitional and fully-turbulent boundary layers, however, showing incomplete mixing and a lesser importance of small scales in the transitional boundary layer. Such observations were true on both flat and concave walls. The differences are attributed to incomplete development of the turbulent kinetic energy cascade in transitional flows. The findings have potential application to modelling, suggesting the utility of incorporating multiple length scales in transition models.

Effects of Free-Stream Turbulence Intensity on a Boundary Layer Recovering From Concave Curvature Effects

Experiments are conducted on a flat recovery wall downstream of sustained concave curvature in the presence of high free-stream turbulence (TI∼8%). This flow simulates some of the features of the flow on the latter parts of the pressure surface of a gas turbine airfoil. The combined effects of concave curvature and TI, both present in the flow over a turbine airfoil, have so far little been studied. Computation of such flows with standard turbulence closure models has not been particularly successful. This experiment attempts to characterize the turbulence characteristics of this flow. In the present study, a turbulent boundary layer grows from the leading edge of a concave wall then passes onto a downstream flat wall. Results show that turbulence intensities increase profoundly in the outer region of the boundary layer over the recovery wall. Near-wall turbulent eddies appear to lift off the recovery wall and a “stabilized” region forms near the wall. In contrast to a low-free-stream turbulence intensity flow, turbulent eddies penetrate the outer parts of the “stabilized” region where sharp velocity and temperature gradients exist. These eddies can more readily transfer momentum and heat. As a result, skin friction coefficients and Stanton numbers on the recovery wall are 20% and 10%, respectively, above their values in the low-free-stream turbulence intensity case. Stanton numbers do not undershoot flat-wall expectations at the same ReΔ2 values as seen in the low-TI case. Remarkably, the velocity distribution in the core of the flow over the recovery wall exhibits a negative gradient normal to the wall under high free-stream turbulence intensity conditions. This velocity distribution appears to be the result of two effects: 1) cross transport of kinetic energy by boundary work in the upstream curved flow and 2) readjustment of static pressure profiles in response to the removal of concave curvature.

Nitrogen Oxide Emissions Characteristics of Augmented Turbofan Engines

The exhaust plumes of modern military engines can be rendered visible at low augmentor power operation by the presence of nitrogen dioxide (NO2). Visible plumes have also been observed from some industrial gas turbines that have duct burners downstream of the power turbines. In 1986, gaseous emissions measurements were taken behind two F101 turbofan engines to determine the effect of reheat level on the degree of conversion of nitric oxide (NO) to nitrogen dioxide and to relate the plume visibility to nitrogen dioxide concentration.

Thermal Diffusivity Imaging of Ceramic Composites

Strong, tough, high temperature ceramic matrix composites are currently being developed for application in advanced heat engines. One of the most promising of these new materials is a SiC fiber-reinforced silicon nitride ceramic matrix composite (SiCf/Si3N4). The interfacial shear strength in such composites is dependant on the integrity of the fiber’s carbon coating at the fiber-matrix interface. The integrity of the carbon rich interface can be significantly reduced if the carbon is oxidized. Since the thermal diffusivity of the fiber is greater than that of the matrix material, the removal of carbon increases the contact resistance at the interface reducing the thermal diffusivity of the composite. Therefore thermal diffusivity images can be used to characterize the progression of carbon depletion and degradation of the composite. A new thermal imaging technique has been developed to provide rapid large area measurements of the thermal diffusivity perpendicular to the fiber direction in these composites. Results of diffusivity measurements will be presented for a series of SiCf/Si3N4 (reaction bonded silicon nitride) composite samples heat-treated under various conditions. Additionally, the ability of this technique to characterize damage in both ceramic and other high temperature composites will be shown.

Effects of Oxygen and Additives on the Thermal Stability of Jet Fuels

A previously developed flowing single-pass heat-exchanger test rig (Phoenix rig) has been used to evaluate the effectiveness of various additives and the kinetic mechanism of both deposit formation and oxygen consumption. The Phoenix rig has been modified to include not just a heated single tube, but also a cooling test section and both hot and cold filters. The effects of flow conditions, antioxidants, and metal deactivator additives on the location and amount of the deposit are discussed. In general, antioxidants were effective at reducing the deposits on the hot test section, but almost invariably caused increased plugging of cool downstream filters. Downstream plugging of cool filters also increased with decreased temperatures in the heated section or with increased flow. Tests with both oxygen-saturated and oxygen-depleted fuels have shown that the solubility of oxygen is linearly related to the fraction of oxygen in a sparge gas, and that the amount of deposit is linearly related to the total quantity of dissolved oxygen passed. Finally, in contrast to initial modelling efforts, the consumption of oxygen is shown to be significantly more complex than a simple bimolecular, pseudo-first-order in oxygen, process. It is found to be much closer to pseudo-zero-order in the early stages, decaying to pseudo-first-order when the oxygen nears depletion.

Nondestructive Eddy Current Evaluation of Anisotropic Conductivities of Silicon Carbide Reinforced Aluminum Metal-Matrix Composite Extrusions

Nondestructive eddy current methods were used to evaluate the electrical conductivity behavior of silicon-carbide particulate (SiCp) reinforced aluminum (Al) metal-matrix composite extrusions. The composites investigated included 2124, 6061 and 7091 Al base alloys reinforced by SiCp. The composite extrusions exhibited anisotropic conductivities with the maximum conductivity occurring along the extrusion plane. Microstructural characterization showed that the observed anisotropic conductivities could result from the preferred orientation distribution of SiCp. A theoretical model was formulated to quantify the influence of composite constituents (SiCp, intermetallics and Al base alloy) on the anisotropic conductivities of the composites. The theoretical predictions of conductivities were found to be in good agreement with the experimental results.

Fault Diagnosis in Gas Turbines Using a Model-Based Technique

Reliable methods for diagnosing faults and detecting degraded performance in gas turbine engines are continually being sought. In this paper, a model-based technique is applied to the problem of detecting degraded performance in a military turbofan engine from take-off acceleration type transients. In the past, difficulty has been experienced in isolating effects of some of the physical processes involved. One such effect is the influence of the bulk metal temperature on the measured engine parameters during large power excursions. It will be shown that the model-based technique provides a simple and convenient way of separating this effect from the faster dynamic components. The important conclusion from this work is that good fault coverage can be gleaned from the resultant pseudo steady-state gain estimates derived in this way.