Abstract Tip leakage flow (TLF) is a critical flow phenomenon governing the efficiency and stability of axial compressors. While its steady-state flow structure is broadly understood, its inherent unsteady nature and spatio-temporal evolution from stable to unstable conditions remain active areas of research. This research characterizes the dynamic behavior of the TLF in a low-speed axial compressor rotor using single-passage and full-annulus delayed detached eddy simulation. By performing mean flow analysis, transient flow analysis, and modal analysis, three dominant modes were identified: (i) vortex shedding after the main tip leakage vortex (TLV) breakdown, (ii) main TLV swing that triggers leading edge spillage, and (iii) post-stall compression system resonance. With decreasing flow coefficient, the spatio-temporal scales of TLF increase, and mode (ii) plays the dominant role. The full-annulus results show trends similar to those in single passage before the formation of stall cells. The findings elucidate the critical dynamic mechanisms by which the evolving TLF triggers instability, offering valuable insights for accurate prediction and flow control of compressor stall inception.
Abstract Accurate loss prediction is essential throughout the turbomachinery design process, from preliminary meanline calculations to detailed optimization. This article compares loss breakdowns across compressor topologies and supplements traditional design charts with a loss assessment based on flow features such as the shock criterion, vorticity orientation, and velocity gradients. This study therefore extends a phenomenological loss decomposition approach to axial, mixed-flow, and centrifugal compressors. Loss breakdowns are evaluated across a broad design space, which is defined by the flow coefficient, the loading coefficient, the hub-to-tip ratio (HTR), the inlet Mach number, and the Reynolds number. These breakdowns are then evaluated across the entire design space to reveal topology-dependent trends. The results show that loss distributions vary significantly within the design space and between compressor types. The proportions of secondary flow losses increase toward lower flow coefficients in axial and mixed-flow compressors, but toward lower loading coefficients in centrifugal machines. As radiality increases, secondary flow losses rise while profile losses diminish. Endwall losses mainly increase toward large flow and small loading coefficients in axial and mixed-flow compressors, but toward high loading in centrifugal compressors. In axial and mixed-flow compressors, variations in the HTR redistribute losses in opposite directions between the hub and the tip. The effect of an increase in the Reynolds number varies significantly between centrifugal and axial compressors. High inlet Mach numbers amplify secondary flow losses in all compressors. While profile losses are becoming more pronounced in axial compressors, no systematic trend has been identified in centrifugal compressors.
Abstract Based on our prior theoretical model for elliptical mode detection, compressive sensing is introduced in this work to overcome the limitations of Nyquist-Shannon sampling. The key contribution is showing how to apply compressive sensing to an elliptic duct, which holds potential for future blended-wing-body aircraft applications. The results show that the proposed method requires only about one-third of sensors under the traditional sampling theorem, while achieving comparable identification accuracy. Monte Carlo simulations further confirm that assigning a value of 2 to the coherence parameter between the sampling matrix and the sparse basis leads to a reconstruction error of 0.01. Furthermore, our work examines how various array configurations impact spinning elliptical mode detection, thereby informing the optimal design of sensor arrays for aeroengine applications of interest. This study proposes a compressive sensing method that minimizes sensor deployment for detecting acoustic modes in elliptical ducts, thereby enhancing the efficiency and scalability of aeroacoustic testing systems.
Abstract Airfoil flow field prediction is central to turbomachinery design; yet, computational fluid dynamics (CFD) imposes a computational bottleneck on iterative design and optimization. Recent studies have explored deep learning-based surrogate models to reduce the computational cost associated with aerodynamic analyses. Among these are generative models, a class of models that include diffusion models, generative adversarial networks (GANs) and variational autoencoders (VAEs). With a growing number of generative models in airfoil flow field prediction, there is limited systematic assessment of their performance and suitability. The purpose of this article is to review the state-of-the-art of generative models for airfoil flow field prediction, providing an overview of recent contributions, methodological advances, and outstanding challenges. We show that despite relatively higher computational costs when compared with alternative deep learning methods, generative models offer a promising approach for obtaining higher-fidelity flow fields and more accurate uncertainty modeling. Our analysis reveals that research in this area remains in its early stages, but applications point toward advancement beyond simple two-dimensional airfoil flow fields toward three-dimensional and multirow configurations reflective of real-world turbomachinery.
Abstract Aerothermal interaction between the combustor and the high-pressure turbine is critical for engine efficiency. Standard steady-state methods fail to capture the highly unsteady flow features at the combustor-turbine interface, leading to uncertainties in thermal load prediction. This study investigates the influence of combustor unsteadiness and scale-resolving simulation methods on the aerodynamic performance and thermal load of an engine-representative high-pressure turbine rotor. Numerical investigations utilize RANS, URANS, and Stress-Blended Eddy Simulations. To prescribe realistic unsteadiness, inlet boundary conditions derived from a combustor Large Eddy Simulation are applied via a one-way unsteady coupling method. Results demonstrate that unsteady inlet conditions substantially impact predicted mixing in the stator. This unsteadiness homogenizes the temperature distribution at the stator outlet, significantly increasing temperatures in the end wall regions. While combustor unsteadiness drives turbulent mixing in the stator, rotor passage flow physics are dominated by the choice of the turbulence model. The scale-resolving approach predicts deeper hot gas penetration into the rim seal cavity compared to URANS. Combined with the altered stator exit traverse, this results in higher thermal loads on the rotor blade suction side, particularly in the critical hub and tip regions. Furthermore, the study highlights the significance of inlet unsteadiness for predicting physically consistent flow fields in scale-resolving simulations. Consequently, high-fidelity, transient methods are shown to be essential for accurately predicting thermal loads, relevant for robust cooling design.
Abstract The aerodynamic design of axial compressor blades requires a careful trade-off between aerodynamic loading and efficiency. Fewer blades increase the aerodynamic load on each blade, while a reduced rotor blade count can lower profile losses due to the smaller number of wakes—potentially improving efficiency. It can also intensify secondary flows because of the higher local loading. To evaluate these opposing effects during preliminary design, three-dimensional RANS simulations remain common practice, although they are known to have weaknesses when predicting secondary flow phenomena. This paper presents experimental results from a four-stage, low-speed research compressor. Tests were carried out with two different blade counts and with both small and enlarged tip clearances to validate the numerical models. The rotor blade count was reduced by roughly 30%, and the rotor stagger anglewas increased to maintain the design-point pressure ratio and mass flow while compensating the stronger deviation. The study focuses on the influence of these changes on the tip-clearance vortex of the rotor blades; the measured data are used to validate simulations that support deeper analysis. The chosen reduction in blade count led to an overall efficiency gain and a markedly lower sensitivity to tip-clearance increases. Although the higher loading produces a steeper tip-leakage vortex, the operating range is not substantially narrowed. The larger blade pitch allows the vortex to traverse the passage with reduced interaction with adjacent blades. These results improve understanding of tip-clearance vortex behavior and can inform future compressor design.
Abstract The unsteady flow structure associated with acoustic resonance in a 4½-stage high-speed axial compressor is investigated through high-fidelity numerical analysis. Acoustic resonance in a multistage compressor is known to be one of the most severe phenomena potentially leading to violent blade vibrations and even structural damage. In this study, the driving mechanisms of acoustic resonance in the compressor tested, including tip leakage fluctuations, are examined using high-fidelity computational fluid dynamics. In addition, the flow structure during acoustic resonance and the relationship between acoustic resonance and stall inception are analyzed. The numerical results obtained using delayed detached eddy simulation show that acoustic resonance is initiated at an off-design operating point and that the resulting pressure wave exhibits a helical pressure mode. The fluctuation in the adverse pressure-gradient induced by this helical mode generates a non-axisymmetric velocity distribution within the rotor blade passages. Three-mode disturbances first appear near the tip region of the third-stage rotor and mutually interact with oscillations of the tip-leakage vortex at the acoustic-resonance frequency. Thus, the driving force of acoustic resonance is the circumferential oscillation of the tip-leakage vortex. During stall inception, flow spillage at the 1st and 2nd rotors, induced by acoustic-resonance effects, occurs earlier than in the rear stages. The resulting non-axisymmetric flow distribution is directly linked to the onset of stall.
Abstract The present study investigates the generalized k−ω (GEKO) turbulence model trained by a feed forward neural network for the improvement of gas turbine film cooling simulations. The key flow features from the GEKO simulation are used as the inputs for the neural network, and the turbulence kinetic energy source coefficient of the GEKO model serves as the output parameter. Both the inputs and the output are being updated by the adjoint optimization during each training cycle, known as a “design iteration.” The training of the GEKO model is successful on a very coarse mesh of the single-hole, jet-in-crossflow MIT Liner film cooling case. The trained GEKO solutions better match the large eddy simulation (LES) and the experiments than the untrained GEKO on the wall heat fluxes. In addition, the trained GEKO model can be extended to another flow blowing ratio and cooling material of the MIT Liner, without any retraining. Furthermore, the same trained GEKO can even be applied to a multi-hole, film-cooled perforated plate case with complex geometry, where the trained GEKO considerably improves the plate’s wall temperature predictions, better matching the LES and the measurements than the untrained GEKO model. The present automated GEKO training workflow, and the good generalizability of the trained GEKO model across different film cooling conditions and geometries, may help the gas turbine industry speed-up the film cooling design process for components such as combustor liners and turbine blades, saving a significant amount of computational time and resources.
Abstract Engine heat-transfer surfaces in ultra-high bypass ratio (UHBR) turbofan engines are exposed to particle-laden high-speed flows in a humid environment that causes particulate deposition and progressive degradation of thermal performance. Therefore, an experimental test facility is developed to study the influences of surface topology, flow velocity, and humidity on particle transport and deposition mechanisms in humidified bypass flow conditions. The presented computational fluid dynamics results are used for the design of the deposition test facility. The experimental test is performed with Arizona road dust and the flow velocities can be set between 90 m/s and 230 m/s under controlled thermodynamic conditions and relative humidity levels ranging from approximately 30% to 100%. The particle deposition and surface degradation is quantified using full-field optical profilometry, enabling spatially resolved measurements of the deposit thickness and roughness evolution. The test specimens consisted of smooth flat plates and engineered dimpled surfaces representative of heat-transfer-enhancing structures. The experiments show that no measurable particle deposition is observed at low humidity [30% relative humidity (RH)], indicating that dry adhesion forces are insufficient to overcome wall shear stresses at these velocities. In contrast, saturated conditions (100% RH) resulted in stable deposit formation, with particles preferentially accumulating in low-shear recirculation regions on the upstream side of the dimple cavities. Increasing the jet velocity from 96.5 to 116.8 m/s reduced the deposited mass by nearly one order of magnitude, demonstrating the dominant role of shear-driven removal. The experimentally observed deposition patterns suggest a correlation with the CFD-predicted wall shear stress distribution. Furthermore, it is observed that the surface roughness of the test specimen is altered throughout the deposition experiments.
Abstract In advanced gas turbines, thermal barrier coatings (TBCs) are commonly employed to improve the cooling performance of the holes. The application of a TBC often leads to the embedding of cooling holes within a two-dimensional trench cavity, which alters both the aerodynamic and thermal characteristics of the coolant flow. In this study, large eddy simulations (LES) were performed to investigate the influences of compound angle (CA) on film-cooling effectiveness and flow structures for a laidback fan-shaped hole embedded within a two-dimensional trench on a flat plate. The trench configuration was adopted from the authors' previous optimization work. Five different cooling hole orientations (CA0, CA15, CA30, CA45, and CA60) were analyzed using an LES framework. The results reveal that the cooling performance and turbulent flow dynamics are strongly dependent on the cooling hole orientation. Interestingly, the effect of compound angle on the coolant jet trajectory in the trenched holes exhibits an opposite trend compared with that reported for conventional non-trenched geometries in previous studies. Although the presence of a compound angle enhanced the cooling effectiveness on the cavity bottom surface in all investigated cases, the area-averaged cooling effectiveness on the downstream flat plate decreased with increasing compound angle. The CA60 case exhibited the lowest cooling performance, showing an approximately 7% reduction in cooling performance on the flat plate relative to the baseline configuration. The time-resolved analysis of the velocity field revealed that increasing the CA intensifies flow fluctuations near the trench exit.
Abstract This article deals with the computational fluid dynamics (CFD)-based modeling of surface roughness, a very challenging topic, which has been increasingly gaining popularity in turbomachinery applications. One approach to model the impact of roughness on the flow over a surface is to derive a roughness wall function that is responsible for shifting the log-law of the boundary layer downward, which is indeed a result of the increased wall shear stress caused by roughness. Another approach is to modify the boundary conditions of the turbulence model at the wall, in such a way that the produced turbulent viscosity mimics the roughness-generated turbulence. In this work, a new roughness model belonging to the latter category has been implemented in an in-house Reynolds-averaged Navier–Stokes (RANS) solver, based on the k–ω shear stress transport (SST) turbulence model. Using the standard equivalent sand-grain roughness parameter as input, the model modifies accordingly the specific dissipation rate (ω) at the wall, which ultimately leads to an increased turbulent viscosity. The model has been trained/calibrated to match pressure losses generated in roughness-resolved channel flow simulations, the roughness of which is representative of deposition-generated roughness and was obtained from previous investigations on in-service high-pressure compressor blades. Verification has been performed on various zero and nonzero pressure gradient channel flow configurations, and experimental validation on a high-pressure turbine cascade. Finally, as simply being a modification to the commonly employed k–ω SST turbulence model, this roughness modeling capability can also be extended to high-fidelity scale-resolving simulations, as demonstrated here via coupling with a hybrid unsteady Reynolds-averaged Navier–Stokes (URANS)–large eddy simulation (LES) methodology.
Abstract Wall-resolved large eddy simulations (LES) are performed to investigate turbulent heat transfer in a rough pipe over a range of Prandtl (Pr) numbers. The rough surface is generated using a controlled numerical procedure that allows independent specification of roughness correlation length, amplitude, and skewness; a statistically Gaussian height distribution is considered as a baseline case. The resulting geometry is used directly in an LES framework implemented in OpenFOAM. Simulations are conducted at a bulk Reynolds number (Re) of 11,700 for Pr = 0.5, 1, 2, and 5, and are extended to additional Re= (2000−15,000) to assess the robustness of global trends. Time-averaged velocity and temperature profiles, turbulent heat fluxes, and Reynolds stresses are analyzed to examine the interaction between roughness-induced mixing and Pr number-dependent thermal diffusion. Relative to a smooth pipe, the rough surface produces a systematic downward shift of the mean temperature profile, indicating enhanced heat transfer across all Pr numbers. The wall-normal turbulent heat flux closely follows the Reynolds shear stress, demonstrating a strong dynamical coupling between scalar and momentum transport within the roughness layer. Probability density functions of the local Nusselt number reveal pronounced spatial variability linked to surface geometry. A sheltering analysis based solely on surface visibility to the incoming flow distinguishes weakly ventilated regions with suppressed heat transfer from exposed windward faces that host intense thermal events. A complementary zonal analysis based on roughness height shows that local heat-transfer enhancement becomes increasingly sensitive to surface elevation as the Pr number increases. For the Gaussian-type roughness considered here, however, the overall influence of Pr number on global and local heat-transfer statistics remains comparatively modest, suggesting a dominant role of geometric sheltering effects. Re number sweeps of global metrics further demonstrate that roughness-induced heat-transfer enhancement and thermal performance increase monotonically with Re number, confirming that the identified mechanisms persist across a broader range of operating conditions.
Abstract Aviation propulsion is trending toward higher power density and more compact engine architectures to meet fuel-burn and emissions targets, thereby reducing available thermal margins across architectures. As a result, thermal management is a key driver of performance, reliability, and service life in modern aviation propulsion systems. An example of an innovative propulsion system is the free-double piston composite-cycle engine, which faces these challenges: it combines piston-engine efficiency with a turbofan architecture; however, its crankshaft-free mechanism and air-lubrication system require advanced thermal management to address elevated temperatures and heat loads in critical components. This article introduces a hybrid cooling strategy that integrates bypass-air convection with liquid-coolant passages and develops two predictive tools to assess and optimize that strategy: (i) a reduced-order analytical heat-transfer model and (ii) a corresponding computational thermal model. Using common boundary conditions and geometry, the models are cross-compared via parametric investigations of cooling configurations, material thermal properties, and the resulting temperature and heat-flux distributions across free-double piston-engine layers. To ensure a fair comparison, cases are evaluated under different operating conditions. Both approaches indicate that hybrid cooling lowers peak wall temperatures and mitigates the thermal loads. The findings offer a verified in-house tool for integrating effective hybrid cooling into lightweight thermal-management solutions for next-generation composite-cycle engines.
Abstract In this work, we investigate the mechanisms of entropy generation in the SPLEEN C1 high-speed low-pressure turbine cascade using high-fidelity large-eddy simulations. The analysis covers four exit Mach numbers representative of off-design operation. We compute the full entropy transport equation and separate all contributions into mean and turbulent terms. The method is applied to the pressure side, suction side, passage, and wake. Mean viscous dissipation is the dominant source of loss in the boundary layers, while turbulent dissipation governs the wake. Although the simulations are adiabatic, irreversible heat-transfer terms remain non-negligible near the blade surfaces, which highlights the need for a complete entropy-based framework in compressible cascades. The wake loss depends strongly on the vortex shedding regime. At low Mach number, coherent roll-ups generate high turbulent dissipation. At high Mach numbers, detached shedding reduces wake loss and shifts it downstream. At the intermediate Mach number, both mechanisms coexist and increase the wake contribution. Two turbulent terms usually neglected in turbine loss studies (the turbulent transport and the turbulent pressure diffusion) play a key role in shaping the turbulent entropy budget and must be retained for compressible flows. Off-design operation reduces total loss by weakening suction-side separation and lowering turbulent activity in the wake.
Abstract Reliable prediction of particle transport in turbomachinery is essential for evaluating deposition, erosion, and performance degradation. Most existing particle-laden flow simulations rely on Reynolds-averaged Navier–Stokes (RANS) solutions, yet the inherent structural uncertainty (i.e., uncertainty in the shape and orientation of the modeled stress tensor) of RANS models is often overlooked, raising concerns regarding the credibility of particle-behavior predictions. To address the gap between flow-field uncertainty and particle response, this work applies the improved eigenspace perturbation framework to NASA Rotor 67. Using six physically realizable perturbation modes, this study quantifies how RANS structural uncertainty propagates through particle-laden flows. The results demonstrate that structural bias in turbulence modeling introduces non-negligible modulation effects on particle deposition and exit-plane distribution. Specifically, the most influential 1C mode significantly alters the turbulent kinetic energy distribution within the passage. For 0.25 μm particles, whose near-wall deposition is dominated by turbulent diffusion, this leads to a relative uncertainty in capture efficiency of 54.5%. Concurrently, the 1C mode enhances hub-to-casing radial velocity, affecting exit-plane particle migration; the relative deviation in distribution uniformity (coefficient of variation) reaches 27.6%. These findings confirm that RANS structural uncertainty influences particle behavior across scales through distinct physical pathways, and neglecting this effect introduces systematic bias into deposition-risk assessment and performance prediction. By establishing a quantitative framework linking carrier-phase uncertainty to particle response, this work provides the foundation for high-reliability design and life prediction of turbomachinery in particle-laden environments.
Abstract Reliable, highly accurate prediction of off-design operating conditions using numerical methods is essential for overcoming the challenges of achieving greener aviation. To assess their capabilities, the high-speed TU Darmstadt axial compressor stage is considered at near stall conditions. Steady Reynolds-averaged Navier-Stokes (RANS) methods fail to capture the unsteady flow physics. Hence, the results show large deviations from experimental reference data. This was investigated by a comprehensive pre-study involving the variation of the mesh density, turbulence model extensions and operating point control mechanisms. Improvements were seen when using unsteady methods such as unsteady RANS (U-RANS) or the Delayed Detached-Eddy Simulation (DDES). Both methods show much closer agreement with the experiment while DDES outperforms U-RANS. These improvements were found to be due to the fact that DDES resolved both the deterministic and stochastic scales of the unsteady flow field, especially in separated flow regions. Further, a thorough assessment of two mechanisms to control the unsteady operating point near stall, namely the mass flow controller and a converging-diverging nozzle, was done. The mass flow controller performed best in terms of operability, predictive accuracy and reduced requirements on a priori knowledge from pre-cursor RANS simulations. The DDES method showed its potential and represents a powerful approach to investigating unsteady off-design operating conditions with high accuracy (below 1% average relative error in radial profiles of the total pressure and temperature ratio, as well as the circumferential flow angle).
Abstract To improve the understanding and enhance the predictability of nonsynchronous vibration (NSV) and rotating stall, an extensive experimental campaign was conducted with the open-test-case ECL5, which is representative of modern ultra-high-bypass-ratio (UHBR) fan architectures. These experiments indicate an influence of test conditions (absence of turbulence control screen (TCS), shortened intake configuration, and mechanical rotation speed) on the measured aerodynamic and aeroelastic behavior of the fan stage. This influence occurs superimposed on the actual parameter under investigation (for example, structural mistuning) and must be clearly characterized to avoid misleading data interpretation. Based on a detailed investigation of several test conditions, this article provides a characterization of different influential parameters. One main outcome is the impact of mechanical rotation speed on NSV amplitudes at subsonic operating conditions. This influence challenges the comparability of different rotor and facility configurations since experiments are usually conducted at corrected mechanical speed (to account for changing atmospheric conditions) to achieve Mach similarity. The findings of the study clearly emphasize that standard experimental procedures have to be adapted to account for the identified effects.
Abstract Aerodynamic instabilities such as stall and surge may occur in fans and compressors, potentially leading to severe flow disruption, blade structural damage, and ultimately engine shutdown. To elucidate the transient inter-stage aerodynamics and underlying flow mechanisms during deep surge, this paper employs full-annulus unsteady Reynolds-averaged Navier–Stokes (URANS) simulations to model the complete surge cycles in a high-speed two-stage fan, thereby predicting unsteady blade loadings. Results from simulation show that the time history of casing static pressure above the first rotor (R1) tip during deep surge cycle is in substantial agreement with experimental measurements. Both the single-passage and the full-annulus models yield similar predictions regarding the shape of the surge cycle. A notable distinction, however, lies in the capability of the full-annulus model to capture the formation, development and clearing of rotating stall, which leads to a longer surge inception period than that predicted by the single-passage model. At 100% rotational speed, spike-type stall inceptions emerge initially in the second stage, whereas at 80% rotational speed, the first stage exhibits stall inceptions prior to the second stage. This is primarily attributed to differences in the range of incidence angle variations across different blade rows under varying rotational speeds. Analysis of surge loading indicates that the S2 blade row experiences the most significant variation in axial force across all rotational speeds, and should therefore be prioritized in the structural design process.
A careful introduction of coolant airflows, which create the film protecting the turbine blade, is critical for a more fuel-efficient and environmentally friendly jet engine. Accurate prediction of streams mixing with the cooling airflows is of great interest for achieving a better design of a jet engine. The main objective of this article is to numerically investigate the rate of heat transfer in a high-pressure turbine rotor passage with purge flow at the hub using a large eddy simulation (LES). An in-house computational fluid dynamics solver, Glenn-HT from the NASA Glenn Research Center, was utilized. The three-dimensional blade and the conditions are those of the Penn State University START rotating rig. A high-quality 125 million cell structured grid, which adequately resolves the high Reynolds number flow (Re similar to 350,000) and the complex secondary flow structures, was constructed. To evaluate the adiabatic wall temperature, two LES simulations with different isothermal wall temperatures were carried out. This method is more robust especially when cooling air injections at the blade surface need to be considered. Our numerical simulations were able to capture a very accurate representation of three-dimensional unsteady flow structures near the tip as well as the secondary flow originating from the purge. Several distinct high heat transfer areas were identified. In addition, temporal "energy separation" in the high vorticity level regions was observed, which has not been reported before.
This study presents and evaluates three independent stall warning approaches based on unsteady casing pressure measurements acquired above the rotor blade tips. The methods are applied to two different axial compressors, each exhibiting distinct stall characteristics and inception mechanisms. The first approach utilizes a cross-correlation technique to detect small perturbations propagating circumferentially just prior to stall onset. The second method employs a convolutional neural network, trained on labeled time-resolved pressure data, to accurately classify operating ranges as stable and unstable. The third approach is based on an autoencoder trained on pressure data exclusively recorded at design flow conditions; deviations from the learned patterns are flagged as anomalies. This unsupervised technique enables the detection of any pre-stall anomalies-including those, not identifiable through visual inspection-that precede conventional stall indicators such as spikes. Importantly, none of the three approaches rely on manually tuned threshold values to trigger stall warnings. As a result, especially the autoencoder remains valid even with changing tip clearances, making this approach highly robust. All methods are validated using independent datasets, and their accuracy is quantified using stall margin-based metrics. Robustness under varying signal-to-noise conditions further demonstrates the applicability of these techniques for reliable stall warning. While the deep learning approaches consistently provide precise warnings at about 5% stall margin, the cross-correlation method only triggers at a similar level if the compressor is free of pre-stall instabilities such as rotating instability.
