Health Monitoring and Degradation Prognostics in Gas Turbine Engines Using Dynamic Neural Networks

In this paper two artificially intelligent methodologies are proposed and developed for degradation prognosis and health monitoring of gas turbine engines. Our objective is to predict the degradation trends by studying their effects on the engine measurable parameters, such as the temperature, at critical points of the gas turbine engine. The first prognostic scheme is based on a recurrent neural network (RNN) architecture. This architecture enables ONE to learn the engine degradations from the available measurable data. The second prognostic scheme is based on a nonlinear auto-regressive with exogenous input (NARX) neural network architecture. It is shown that this network can be trained with fewer data points and the prediction errors are lower as compared to the RNN architecture. To manage prognostic and prediction uncertainties upper and lower threshold bounds are defined and obtained. Various scenarios and case studies are presented to illustrate and demonstrate the effectiveness of our proposed neural network-based prognostic approaches. To evaluate and compare the prediction results between our two proposed neural network schemes, a metric known as the normalized Akaike information criterion (NAIC) is utilized. A smaller NAIC shows a better, a more accurate and a more effective prediction outcome. The NAIC values are obtained for each case and the networks are compared relatively with one another.

Multi-Angular Flame Measurements and Analysis in a Supersonic Wind Tunnel Using Fiber Based Endoscopes

This paper reports new measurements and analysis made in the Research Cell 19 supersonic wind-tunnel facility housed at the Air Force Research Laboratory. The measurements include planar chemiluminescence from multiple angular positions obtained using fiber based endoscopes (FBEs) and the accompanying velocity fields obtained using particle image velocimetry (PIV). The measurements capture the flame dynamics from different angles (e.g., the top and both sides) simultaneously. The analysis of such data by proper orthogonal decomposition (POD) will also be reported. Non-intrusive and full-field imaging measurements provide a wealth of information for model validation and design optimization of propulsion systems. However, it is challenging to obtain such measurements due to various implementation difficulties such as optical access, thermal management, and equipment cost. This work therefore explores the application of FBEs for non-intrusive imaging measurements in supersonic propulsion systems. The FBEs used in this work are demonstrated to overcome many of the practical difficulties and significantly facilitate the measurements. The FBEs are bendable and have relatively small footprints (compared to high-speed cameras), which facilitates line-of-sight optical access. Also, the FBEs can tolerate higher temperatures than high-speed cameras, ameliorating the thermal management issues. Lastly, the FBEs, after customization, can enable the capture of multiple images (e.g., images of the flowfields at multi-angles) onto the same camera chip, greatly reducing the equipment cost of the measurements. The multi-angle data sets, enabled by the FBEs as discussed above, were analyzed by POD to extract the dominating flame modes when examined from various angular positions. Similar analysis was performed on the accompanying PIV data to examine the corresponding modes of the flowfields. The POD analysis provides a quantitative measure of the dominating spatial modes of the flame and flow structures and is an effective mathematical tool to extract key physics from large data sets such as the high-speed measurements collected in this study. However, past POD analysis has been limited to data obtained from one orientation only. The availability of data at multiple angles in this study is expected to provide further insights into the flame and flow structures in high-speed propulsion systems.

Failure Analysis and Life Assessment of Thermal Fatigue Crack Growth in a Nickel-Base Superalloy Based on EBSD Method

In the case of failure incidents involving important components, it is necessary to clarify the fracture mechanism by failure analysis. In the case of conventional steel materials, according to the individual fracture mode the fracture surfaces have unique fracture morphology corresponding to tensile, impact, creep and fatigue conditions. We can identify the mechanism of a fracture by observing its fracture surface, and this is known as the fractography. However regarding nickel-base superalloys, any differences in fracture morphology are unfortunately barely distinguishable, which makes it difficult to conduct fractography. In this paper, in order to characterize the damage behavior of IN738LC, the misorientation analysis within grains by using the electron backscattered diffraction (EBSD) method across almost all the whole range of specimens has been carried out. As a result, it was found that the cross section of fracture samples have unique distinguishable morphology corresponding to the individual fracture mode. Furthermore, the striations corresponding to the fatigue crack growth rate was found in the crack cross-sectional sample. It was considered that the EBSD striation observed on the cross section reveals the fatigue crack growth rate, as with striations found in the fatigue fracture surface such as conventional steel materials. On the case study of the actual (service and damaged) gas turbine blade, the EBSD analysis as the fractography revealed the mechanism of cracking and the fatigue crack growth rate. Thus, it is concluded that the misorientation analysis of damage materials allows the qualitative estimation of the fracture mode and the quantitative life assessment of the fatigue crack growth.

Multi-Coordination of Actuators in Advanced Power Systems

Multi-coordination of actuators for a highly integrated, tightly coupled advanced power system was evaluated using the Hybrid Performance (Hyper) project facility at the U.S. Department of Energy’s National Energy Technology Laboratory (NETL). A two-by-two scenario in a fuel cell, turbine hybrid power system was utilized as a representative problem in terms of system component coupling during transients and setpoint changes. In this system, the gas turbine electric load is used to control the turbine speed, and the cold air bypass valve regulated fuel cell cathode mass flow. Perturbations in the turbine speed caused by variations in the waste heat from the fuel cell affect the cathode airflow, and the cold-air bypass control action required for constant cathode airflow strongly affects the turbine speed. Previous implementation of two single-input, single-output (SISO) controllers failed to provide acceptable disturbance rejection and setpoint tracking under these highly coupled conditions. A multiple-input, multiple-output (MIMO) controller based on the classic internal model control (IMC) concept was implemented and experimentally tested for the first time using the Hyper project facility. The state-space design of the MIMO configuration, the control law integration into the digital control platform, and the experimental comparison with the SISO case are presented.

Ion Current Measurements as a Method for the Detection of the Reaction Rate in Combustion With Swirl Stabilized Airblast Injection Systems

Due to strict emission legislation, the trend in the development of aero-engine gas turbine combustion is heading towards lean burning approaches. Lean combustion reduces the combustion temperatures and therefore also the nitrogen oxides emissions. Unfortunately, lean combustion suffers from instabilities and the operation close to the point of lean blowout increases the risk of imminent blowoff. Active stability control is therefore inevitable. The objective of this work is to evaluate the signal obtained from an ion current measurement technique to enable combustion control for aircraft propulsion applications in the near future. In the past ion current measurements have been used in several studies as flame turbulence analyzer and to detect the reaction rate. However, investigations in lean burning and swirl stabilized airblast injection combustors for future propulsion concepts are rare. The signal obtained from an ion current detector inside a combustor depends strongly on the measurement position. In this experimental investigation field measurements at atmospheric conditions of the ion concentration in a tubular combustor with a sampling rate of 8 kHz are compared with 4 kHz time resolved temperature and OH* chemiluminescence measurements in order to determine the position of the reaction inside the combustor. Variations were performed of the air to fuel ratio (AFR), the air preheating temperature and the pressure drop across the injection system to clarify the interpretation of the ion current signal. The results indicate a strong dependence of the ion current signal on the AFR and that the technique has distinct advantages compared to OH* chemiluminescence measurements: The measurement equipment is comparable non-expensive and the results reveal that the reaction rate is measured directly and are not interpreted from a 3D image. A transition in flame shape from a compact to a tornado flame can be clearly identified with the applied probe. Furthermore, regions with high temperature fluctuations do not necessarily reveal the reaction zone in a recirculating flow field.

3D Flame Measurements at 5 kHz on a Jet Fueled Aviation Combustor

This paper describes the measurements of three-dimensional (3D) flame measurements at 5 kHz from an optically accessible, atmospheric combustor burning Jet A. The measurements were performed by 3D tomography of flame chemiluminescence emissions. Such measurements resolve the instantaneous flame topography generated by the combustor in all three spatial directions and at a temporal resolution of 5 kHz (0.2 ms), essentially providing 4D information about the combustor. This paper reports the measurement technique, the experimental implementation, the results and their expected use for the continued development of future combustor concepts.

Assessment of Unsteady Pressure Measurement Uncertainty: Part 2 — Virtual Three Hole Probe

With the advancements in miniaturization and temperature capabilities of piezo-resistive pressure sensors, pneumatic probes — which are the long established standard for flow-path pressure measurements in gas turbine environments — are being replaced with unsteady pressure probes. On the other hand, any measured quantity is by definition inherently different from the ‘true’ value, requiring the estimation of the associated errors for determining the validity of the results and establishing respective confidence intervals. In the context of pressure measurements, the calibration uncertainty values, which differ from measurement uncertainties, are typically provided. Even then, the lack of a standard methodology is evident as uncertainties are often reported without appropriate confidence intervals. Moreover, no time-resolved measurement uncertainty analysis has come to the attention of the authors. The objective of this paper is to present a standard method for the estimation of the uncertainties related to measurements performed using single sensor unsteady pressure probes, with the help of measurements obtained in a one and a half stage low pressure high speed axial compressor test rig as an example. The methodology presented is also valid for similar applications involving the use of steady or unsteady sensors and instruments. The static calibration uncertainty, steady measurement uncertainties and unsteady measurement uncertainties based on phase-locked and ensemble averages are presented by the authors in [1]. Depending on the number of points used for the averaging, different values for uncertainty have been observed, underlining the importance of having greater number of samples. For unsteady flows, higher uncertainties have been observed at regions of higher unsteadiness such as tip leakage vortices, hub corner vortices and blade wakes. Unfortunately, the state of the art in single-sensor miniature unsteady pressure probes is comparable to multi-hole pneumatic probes in size, preventing the use of multi-hole unsteady probes in turbomachinery environments. However, the angular calibration properties of a single sensor probe obtained via an aerodynamic calibration may further be exploited as if a three-hole directional probe is employed, yielding corrected total pressure, unsteady yaw angle, static pressure and Mach number distributions based on the phase-locked averages with the expense of losing the time-correlation between the virtual ports. The aerodynamic calibration and derivation process are presented together with the assessment of the uncertainties associated to these derived quantities in this contribution. In the virtual three-hole mode, similar to that of a single-sensor probe, higher uncertainty values are observed at regions of higher unsteadiness.

Study and Analysis of a Power Turbine Preliminary Design for a Small Turboshaft

This work describes the continuous study that is being done in a small gas turbine that can be used for power generation purposes. Previous studies were conducted aiming to develop a gas generator able to be used in both applications, as a turbojet or as a turboshaft. The gas generator was designed, manufactured and is still under test. The thermodynamic cycle calculation was evaluated as a project-based class, hence, a power turbine was specified and its requirements were determined. The outlet conditions from the gas generator were used to perform the preliminary size of the power turbine. At this phase, the students must use 1D design models considering loss modeling to improve the machine design prediction. The meanline technique was used and the calculations at leading and trailing edges were extrapolated from hub-to-tip, using vortex design methods. With the airfoil stacking for each blade row was possible to determine the 3D geometry of the single stage axial flow turbine. This geometry was assembled in a CAD software to start the mesh generation procedure. After this step, a commercial CFD software was used to calculate the continuity, momentum and energy equations from fluid mechanics. The flow was considered fully turbulent and the two-equation SST turbulence model was set to determine the flow eddy viscosity. The results from preliminary design and 3D techniques were compared and evaluated to complete the first round of the design phase. In this work, experiences from the project-based class on turbomachinery design are described together with the challenges and difficulties that appeared during the project.

Drive Train Over-Speed Prediction for Gas Fuel Based Gas Turbine Power Generation Units

Under all circumstances, an engine and its driven equipment(s) must be prevented from operating at a speed above the maximum allowed speed — to ensure the safety of the equipment, plant and its personnel. However, meeting this requirement is particularly challenging for power generation units where the drive train is composed of electric generators driven by free power turbines (i.e. aerodynamically-coupled power turbines), since during load-shed events or circuit breaker failure, full loss of load happens almost instantly. During these events, usually the Fuel Metering Valve is fully closed by the Engine Control System and the Fuel Isolation Valve is closed by the safety system. But, fuel gas continues to flow to the system during the closing of the valves, and furthermore, the fuel gas trapped in the piping between the valve and the fuel injectors still has enough pressure to flow to the combustion chamber and add energy to the system, which at that point has almost no external load, thus likely to cause an over-speeding of the drive-train. This paper is to report a dynamic model created for drive-train over-speed predictions. In this model, fuel flow rate to the engine is calculated based on the principle of conservation of mass together with the fuel gas equation of state. The calculated fuel flow rate is then used to find the amount of power supply to the drive train, which in the next step is converted to the torque applied on the shaft. Finally, Newton’s second law is used to determine the angular acceleration and the angular speed. This approach is applied to two different variations of the Industrial RB211 Engine — the DLE (Dry Low Emission) RB211 and Non-DLE RB211 — which have different designs of the fuel gas system and the burners. For both cases, the results using the modeling approach presented in this paper demonstrate around 99% agreement with the actual measured over-speed values recorded during trip events. The model allows studying the drive train speed for different operating conditions and failure cases, and also makes it easy to understand and quantify the effect of fuel gas system parameter variation on drive-train over-speed.

Engineering Approach for LCF Assessment of Porous Alloys

Most components used in gas and steam turbines are metallic parts produced by either casting or forging processes. Although process control works to eliminate defects, there can be variation in micro-porosity from component to component. Previously this micro-porosity was only able to be detected destructively using metallography. Using Computer Tomography (CT), one can find voids in the range of a few tenths of a millimeter and know the location of the voids with high precision. This allows one to map the defects present in each component onto the stress and temperature fields for that component. However, there is not yet universal agreement upon a consistent method to evaluate the effect of these small porosities on a component’s lifetime. Having a robust analysis tool to understand the impact of micro-porosity would decrease development costs, decrease the time to bring a product to market, and increase the likelihood of failure-free operation. This paper presents an approach using equivalent LCF material properties which avoids the need to explicitly model the morphology of the microstructure in the region of the micro-porosity. The homogenization methodology calculates new LCF curves depending on porosity ratios in material. This approach uses Morrow’s correlation factor of LCF cycles to crack initiation regarding energy amount dissipated in stable cycling (shake-down) and ultimate strain energy under monotonic loading. The paper generalizes Morrow’s postulate and formulates the hypothesis that energy stored and dissipated in the material under shake-down conditions corresponds directly to the number of LCF cycles to crack initiation. The paper demonstrates that the reduction of LCF life based on the porosity ratio agrees well with the experimental results. These results also show that the methodology is very sensitive to the void orientation and loading direction.