A Machine Learning Based Approach of Performance Estimation for High-Pressure Compressor Airfoils

In recent years, the demands of Maintenance, Repair and Overhaul (MRO) customers to provide resource-efficient after market services have grown increasingly. One way to meet these requirements is by making use of predictive maintenance methods. These are ideas that involve the derivation of workscoping guidance by assessing and processing previously unused or undocumented service data. In this context a novel approach on predictive maintenance is presented in form of a performance-based classification method for high pressure compressor (HPC) airfoils. The procedure features machine learning algorithms that establish a relation between the airfoil geometry and the associated aerodynamic behavior and is hereby able to divide individual operating characteristics into a finite number of distinct aero-classes. By this means the introduced method not only provides a fast and simple way to assess piece part performance through geometrical data, but also facilitates the consideration of stage matching (axial as well as circumferential) in a simplified manner. It thus serves as prerequisite for an improved customary HPC performance workscope as well as for an automated optimization process for compressor buildup with used or repaired material that would be applicable in an MRO environment. The methods of machine learning that are used in the present work enable the formation of distinct groups of similar aero-performance by unsupervised (step 1) and supervised learning (step 2). The application of the overall classification procedure is shown exemplary on an artificially generated dataset based on real characteristics of a front and a rear rotor of a 10-stage axial compressor that contains both geometry as well as aerodynamic information. In step 1 of the investigation only the aerodynamic quantities in terms of multivariate functional data are used in order to benchmark different clustering algorithms and generate a foundation for a geometry-based aero-classification. Corresponding classifiers are created in step 2 by means of both, the k Nearest Neighbor and the linear Support Vector Machine algorithms. The methods’ fidelities are brought to the test with the attempt to recover the aero-based similarity classes solely by using normalized and reduced geometry data. This results in high classification probabilities of up to 96 % which is proven by using stratified k-fold cross-validation.

A Numerical Model for Simulating Liquid Particles Deposition on Surface

The deposition of liquid particles, which may be converted from solid particles due to high temperature gas heating, makes much more harm on turbine vane blades compared to solid particles, since it may block film-cooling holes, worsen the cooling efficiency and aerodynamic performance of the turbine vane blades. Due to the similarity between the deposition of liquid particles on a surface and the icing on a surface, a numerical model for simulating particles deposition was developed based on the Myers icing model, an extension of the Messinger model, which has been applied in predicting aircraft icing or aero-engine icing. Compared to the conventional liquid particle deposition model, the numerical model in this paper considers the heat transfer and the flow of liquid particles during the particles phase transition from liquid state to solid state. In this model, the change of the surface profile due to the particles deposition was also considered, which was implemented with dynamic mesh technique. To test this model, deposition distribution and thickness obtained from the numerical simulations were compared to the experimental results. Additionally, a numerical simulation was conducted for liquid particle deposition on a flat plate. The result showed that the deposition thickness at the leading edge was much larger than that on the upper surface where the deposition appeared mainly at the middle and rear of the plate. The deposition mass and thickness increased with the increasing in the particle size. The effect of the particle size on the deposition thickness was more notable on the upper surface compared to that at the leading edge.

Aeroelastic Optimization of an Industrial Compressor Rotor Blade Geometry

This paper describes a multidisciplinary optimization procedure applied to a compressor blade-row. The numerical procedure takes into account both aerodynamic (efficiency) and aeromechanic (flutter-free design) goals nowadays required by turbo-machinery industries and is applied to a low pressure compressor rotor geometry provided by Ansaldo Energia S.p.A.. Some typical geometrical parameters have been selected and modified during the automatic optimization process in order to generate an optimum geometry with an improved efficiency and, at the same time, a safety flutter margin. This new automatic optimization procedure, which now includes a flutter stability assessment, is an extension of an existing aerodynamic optimization process, which randomly perturbs a starting 3D blade geometry inside a constrained range of values, build the fluid mesh and run the CFD steady analysis. The new implementation provides the self-building of the solid mesh, the FEM analysis and finally the unsteady uncoupled aeroelastic analysis to assess the flutter occurrence. After simulating a wide range of geometries, a database with all the constraint parameters and objective functions is obtained and then used to train a neural network algorithm. Once the ANN validation error is converged, an optimization strategy is used to build the Pareto front and to provide a set of optimum geometries redesigning the original compressor rotor. The aim of this paper is to show the opportunity to also take into account the aeroelastic issues in optimization processes.

Fouling on a Multistage Axial Compressor in the Presence of a Third Substance at the Particle/Surface Interface

Solid particle ingestion is one of the principal degradation mechanisms in the compressor and turbine sections of gas turbines. In particular, in industrial applications, the micro-particles not captured by the air filtration system can cause deposits on blades and, consequently, can result in a decrease in compressor performance. It is of great interest to the industry to determine which zones of the compressor blades are impacted by these small particles. However, this information often refers to single stage analysis. This paper presents three-dimensional numerical simulations of the micro-particle ingestion (0.15 μm – 1.50 μm) in a multistage (i.e. eight stage) subsonic axial compressor, carried out by means of a commercial CFD code. Particle trajectory simulations use a stochastic Lagrangian tracking method that solves the equations of motion separately from the continuous phase. The effects of humidity, or more generally, the effects of a third substance at the particle/surface interface (which is considered one of the major promoters of fouling) is then studied. The behavior of wet and oiled particles, in addition to the usual dry particles, is taken into consideration. In the dry case, the particle deposition is established only by using the sticking probability. This quantity links the kinematic characteristics of particle impact on the blade with the fouling phenomenon. In the other two cases, the effect of the presence of a third substance at the particle/surface interface is considered by means of an energy-based model. Moreover, the influence of the tangential impact velocity on particle deposition is analyzed. Introducing the effect of a third substance, such as humidity or oil, the phenomenon of fouling concerns the same areas of the multistage compressor. The most significant results are obtained by combining the effect of the third substance with the effect of the tangential component of the impact velocity of the particles. The deposition trends obtained with these conditions are comparable with those reported in literature, highlighting how the deposits are mainly concentrated in the early stages of a multistage compressor. Particular fluid dynamic phenomena, such as corner separations and clearance vortices, strongly influence the location of particle deposits.

Effect of Pore Blockage on Transpiration Cooling With Additive Manufacturable Perforate Holes

Transpiration Cooling is an effective cooling technology to protect hot section components such as gas turbine airfoils, rocket heads and space craft. This external cooling method has much higher efficiency than film cooling with holes when consuming the same amount of coolant, due to the uniformity of coolant distribution. However, pore blockage, which frequently occur during the operation of transpiration cooled components, prevented its application in turbine components which require long term stability. Dust deposition was one the main reasons causing blockage of pores for transpiration cooling. A lot of effort was devoted into dust deposition and erosion while optimization for the components themselves were generally difficult as the blockage caused by dusts was unpredictable for traditional sintered porous media. Additive manufacturing, with capability to precisely construct structures in small scales, is a considerable tool to enhance the controllability of porous media, and furthermore, to find a good solution to minimize the blockage disadvantage. Present study selected a cooling configurations containing perforate straight holes with an additive manufacturable diameter of 0.4 mm. Computational Fluid Dynamics (CFD) methods were utilized to model the pore blockage and its effect on heat transfer. A scripting code in addition to the ANSYS CFX solver was utilized to simulate the random blockage conditions of the holes. Two hundred numerical cases with four different blockage probabilities were calculated and statistically evaluated to quantify the disadvantage of pore blockage on the cooling effectiveness. Results obtained from the numerical analysis indicated that the overall blockage ratio was a dominating parameter for the cooling effectiveness. Upstream regions of the cooled surface were more sensitive to local blockage compared to downstream regions. Randomness of the cooling effectiveness increased with the increase of blockage probability. Present study provided a quantitative understanding of the random blockage disadvantage on the specific transpiration cooling configuration, and could benefit further optimization effort to reduce the blockage disadvantage of transpiration cooling using additive manufacturing.

Design Optimization of Splitter Blades for Rocket Engine Turbopump

In this paper we present and validate a shape optimization framework for the design of splitter blades that extends the operative range under cavitation while maintaining the wetted performance of rocket engine turbopumps. For a target turbopump application, the optimization framework allows for independent changes to the blade angle distributions across the span and to the pitchwise position of the splitter blades while preserving the thickness distributions. The optimization is conducted with a surrogate-based gradient method. The geometry is optimized at a fixed cavitation number corresponding to a 5% head coefficient dropoff, while constraints are imposed on the wet pump performance. It is found that this approach, coupled with the optimal design points distribution provided by the Design of Experiment method, reduces the computational cost of the optimization process by minimizing the number of multiphase calculations. The numerical results suggest that the optimized splitter blades successfully increase the pump operative range by 2.2% and increase the head coefficient by 5.3% compared to the baseline case with non-optimized splitters. These results are corroborated by experiments conducted in a closed-loop water test facility. Several pump geometries are tested through rapid prototyping using additive manufacturing. The experimental data validate the optimization framework, demonstrating a 4.7% increase of pump operative range and a 7.6% increase in head coefficient. The calculations are used to gain insight in the physical mechanisms for the performance improvement. The analysis of the results indicates that the improved performance is due to the optimized position and shape of the splitter blades which increase the pump slip factor.

The Use of Fluid-Solid Cell Transformation to Model Volcanic Ash Deposition Within a Gas Turbine Hot Component

Creating robust empirical and computational models of the process of deposition of salts, dust, sand and volcanic ash has gained increased importance over the last two decades as civil aircraft flights in regions with particulate laden atmospheres have increased. This is associated with increased costs of maintenance to engine suppliers in a market where there is pressure from carriers to continue to fly. Thus, knowledge of the build-up of particulates within the engine over long or multiple deposition events is required in addition to predicting its onset. In this paper deposition in idealised geometries typical of internal cooling passages is examined. The fluid phase is modelled using the commercial flow solver FLUENT and a simple RANS approach. The discrete phase was then solved using Lagrangian particle tracking and a continuous random walk model using one-way coupling. Following identification of deposition fluxes, the local surface of the solid domain was modified using a bespoke cell transformation process. Particular care was taken to distribute deposited mass appropriately to surface cells to avoid large discontinuities at the boundaries. The model was implemented using user-defined functions. The functionality of the technique, is demonstrated through application two impingement cooling geometries for which experimental validation data are available. Here the solution was highly sensitive to the changing target surface geometry as deposition advanced temporally. Fair agreement was found with the experimental data of Burwash et al.[1] though the level of accretion found was an order of magnitude too high, highlighting the need to combine this approach with accurate stick-bounce and shedding models. Significant changes in deposition locations were observed as the deposition site grew in size. Comparison to a second validation case, by Clum et al [2], was used to test further the effect of deposition on the local flow field. Again, good qualitative agreement was obtained. The procedure is shown to create believable deposits of volcanic ash for all cases tested, without many of the typical problems encountered with mesh morphing — overlapping volumes and indeterminate boundary layer resolution. For the commercial computational fluid dynamic (CFD) code used, the process of identifying cells which are to be modified and their neighbours is, disappointingly, an order of magnitude slower than using a mesh morphing strategy. The procedure does, however maintain high, known, resolution throughout the thermal boundary layer, will allow the redistribution of particles to take into account features such as the fusing of neighbouring accretions and the breakaway of deposits from the surface as they grow.

On Deposition and Build-Up Detachment in Compressor Fouling

Compressor fouling is a severe problem for both heavy-duty and aero-propulsion gas turbines. Particles can impinge on the blade and annulus surfaces, sticking there. The consequences of particle deposition are the increase of the roughness and an uncontrolled variation in the surface shape. These problems have a major effect on the performance of the compressor over time. Variations in the flow field can make the flow quantities close to the deposit to change, and it may happen that the conditions for the sticking do not hold any longer. If this is the case, the build-up detachment may happen. This occurrence can mitigate the fouling effects and may be exploited for keeping the performance of the compressor as high as possible over the operating period. In this work, an innovative model is proposed in order to evaluate the adhesion forces and the possible detachment. Particularly, the same forces that keep a gecko stuck to a surface are considered: the van der Waals forces (due to the proximity of the two bodies) and the Laplace force (due to the curvature of the liquid film related to the humidity). The so formulated model, named gecko-like for such a reason, is used for the numerical analyses of a deposition problem. Both the sticking and possible build-up detachment are considered. The outcome of this work can be regarded as an a-priori estimate of the forces to be kept into account when dealing with compressor fouling problems.

Robust Fluid Topology Optimization Using Polynomial Chaos Expansions: TOffee

The paper presents an innovative solution to robust topology optimization developed for components that can be manufactured by additive manufacturing. Topology optimization has been used in fluid dynamics to optimize geometries based on a target set of performances required for the flow paths. These target performances can be defined as pressure losses or heat exchanges for example, and multiple optimized geometries can be found in the literature. However, none of these cases considered the impact of stochastic variations and are based on a deterministic optimization. It means the optimization has been done for a single boundary condition value. Would this boundary be random, as it is the case in real life gas turbines, then the optimized geometry, optimized for a single set of boundary conditions, will underperform. Robust topology optimization obtains a geometry able to cope with these random variations. The robust optimization method has been implemented in an in-house solver TOffee and relies on a multi-objective function. 2D and 3D robust optimized geometries are obtained and their performance compared to deterministic cases over a range of boundary conditions. Superiority of robust geometries as compared to deterministic geometries is shown. Robust topology optimization presents a great interest in the gas turbine industry due to the greater performance obtained by the optimized geometries while taking into consideration random variations of boundary conditions, making the simulations closer to real life conditions. For the first time in this work it is shown a fluid topology optimization solution with sedimentation that are inherently able to cope with uncertainty.