Industrial Gas Turbines Engineering Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Centrifugal compressors form the major component used in the industry for various purposes ranging from automobile industry, aeronautics and industrial gas turbines.

Types of compressor are

Piston compressors or reciprocating compressors,

Rotary Screw Compressors

Rotary Sliding Vane

Compressors Centrifugal Compressors

As in view of its quiet run and because centrifugal compressor has not any part that oppose, except of bearings, centrifugal compressor is very reliable. Another advantage is the fact that specific mass of centrifugal compressor decreases with its increasing capacity. Due to all of these advantages they are widely used in the industry, but still centrifugal compressors account for major power consumption in the industries.

The varying operating range of the compressors is a major concern as the compressors are designed to have high efficiency in a given range and as in operation it is shifted there is a marked reduction in efficiency. In such environment it is desirable to have both high efficiency as well as quite large operational range. This is why optimization of the compressor geometry is carried out on a large scale in industries for shifting the high efficiency zone in the operating range in which the compressor will operate.

Under these challenging requirements, the improvement of aerodynamic performance is considered of prime importance. Major studies have been carried out to improve the efficiency of centrifugal compressors and industries thrive even for 0.5 % efficiency increase which can reduce the compressor work required to a large extent and thus reducing the energy consumption.

Computational fluid dynamics (CFD) has been used extensively over the past three decades to predict the performance of compressors for both single and multiple blade row configurations. Today, with the power of modern computers, steady-state solutions are carried out on a routine basis, and can be considered as part of the design process hence it is widely used to reduce the prototyping and hence reduce the design cycle time.

Simulation of high speed transonic air-compressors poses a challenge in CFD simulation as the compressibility effect becomes prominent and cannot be ignored.

Changes of energy in particular parts of compressor stage are shown on Figure 1 i.e. plotted against Total Enthalpy (H) v/s Entropy (S). When we carefully watch figure 2 we can see that highest increase of energy occurs in the impeller. From this point of view the impeller is main part of the compressor and its quality influences quality of the whole compressor. Nevertheless other parts of the compressor have also great importance on function and quality of designed compressors.

Figure 1: Changes of energy in particular parts of centrifugal compressor

Re-designing turbo machinery blades is a complex task involving many different objectives with lot of parameters and constraints coming from various disciplines viz. structural, manufacturability etc.

Various parameters associated with the impeller designing are as shown with help of velocity triangles below:

Figure 2: Velocity triangles (c2 -absolute speed, u2 - tip speed, w2 - relative speed

α2 - Outlet flow angle, β2 - Outlet blade angle)

Along with the parameters shown above there are other parameters available for optimization being the wrap angle, blade thickness, hub and shroud side contours etc.

Increasing performance levels and operating conditions requirements makes the task of designing a pump impeller very challenging. The role of internal flows and the viscous effects in centrifugal impeller blades are fundamental and must be taken properly into account in the design process in order to obtain optimum performance.

The major losses observed in the compressors is due to adverse pressure gradients on the blades of the compressors and the shock observed in the transonic compressors at the trailing edge of the compressor also account for the major losses as there is a reduction in total pressure across the shock which is a loss in the total energy i.e. the potential work done.

So a conscious attempt is made to reduce the losses by reducing the separation and dampening the shock intensity. Interestingly both the losses are due to improper aerodynamic design for the given operating range and reduction in losses can be achieved by improving aerodynamic design of the blades.

Due to involvement large number of geometrical parameter in a compressor blade design experimental studies are costly and time consuming to obtain the optimum blade design. Hence numerical studies are probably a more viable option for conducting parametric study and investigating the effects of various geometrical and numerical parameters.

This practical is carried out to understand the use of CFD tool to solve compressible flow problems. To get familiar with techniques of CFD tool used from CAD geometry import to optimized CFD solution.

To carry out this work, single blade of compressor is considered similar to blades used in experimental setup as shown in figure. Ansys 12.1 academic CFD software is used to perform the CFD analysis. A periodic model of blade passage between two blades is used, which helped in significantly reducing the efforts and computational time for simulation than considering complete model. Effect of the compressible is considered by assuming ideal gas law for density variation and Sutherland model for viscosity.

1.1Aim of the study:

Learn how to use the CFD tool to solve compressible periodic flow problems.

Become familiar with techniques used to import CAD geometry in the CFD software and creating a CFD model around it.

Improve the understanding of grid independence and convergence of solutions.

Being able to interpret CFD results compare them to experimental data and draw your own conclusions.

Appreciate the concept of periodicity in flows.

Become familiar with fundamental CFD concepts

Literature Review

Centrifugal compressors are widely used in small aeronautic and industrial gas turbines and in turbochargers for internal combustion engines consume a large amount of power. It is therefore important to minimize the energy consumption and widen the operation range and hence the main demand is to achieve the highest possible efficiency, especially for highly loaded machines, i.e. at high compression ratios.

For these reasons, the flows inside the impellers should be studied in detail in order to improve the design. The flows inside centrifugal impellers are very complex for computation and measurement because of the rotation, viscous effects and complex geometry. Some special phenomena, such as jet-wake structure, secondary flows and separation, have great effect on the losses inside the impeller which greatly affects the range of operation and limits efficiency.

Numerous theoretical and experimental investigations are done for optimization of the design of the centrifugal compressors. In recent years, a number of researchers have done considerable work on the computation and measurement of flows inside centrifugal impellers, including Eckardt [1], Johnson and Moore [2], Kjork and Lofdahl[3], Denton [4], Dawes [5], Casey et al. [6], Bansod and Rhie [7], Krain and Hoffman [8], Farge and Johnson [9] and Zhang et al. [l0]. Their achievements contribute significantly in revealing the flow mechanisms inside centrifugal impellers and are most valuable for designing centrifugal impellers with good performance.

One of the most widely-used validation cases for compressors is the study carried out by Eckardt [1] for a centrifugal compressor impeller [1], also known simply as the Eckardt rotor. Eckardt published a series of papers in the 1970s discussing experiments conducted on two 20 blade centrifugal compressor rotors. The resulting impeller performance maps as well as the investigation of local effects have been used extensively over the years to test the accuracy and robustness of CFD codes for turbomachinery.

Specifically, this study is for optimization of Eckardt impeller for overall efficiency and Pressure ratio thus it was desired that calculation be carried out for multiple speed lines from choke to near stall. The Eckardt rotor was best suited as the subject of this study given the availability of data for a wide range of conditions with which the numerical predictions could be compared.

Recently, the performance of centrifugal compressors has been substantially improved by the use of backswept impellers rather than the conventional types with radially tipped blades. Several researchers have studied the flows within backswept centrifugal impellers, including Farge and Johnson [9] and Krain and Hoffman [8], and some useful conclusions for improving the performance have been obtained.

In the paper of Farge and Johnson [9], the flow field inside a backswept shrouded centrifugal impeller was measured in detail. The measurement results show a strong jet-wake structure occurring near the outlet of the impeller even though the rotating speed is slow (500 rpm) and the impeller is operating at the design point.

As mentioned earlier the flow separation accounts for a major losses in a centrifugal compressor and can be reduced by optimizing the aerofoil shape of the compressor blades. Hence Turbomachinery designers are accustomed to work with two-dimensional sections that are then stacked to the three dimensional blade geometry. Lot of work has been carried out by many researchers to optimize the two-dimensional sections.

Christian Wauquiez [11] has carried out investigation on the shape optimization of the Low Speed Aerofoils ,for maximization of lift by varying the max-camber, max-camber location, max thickness and angle of attack. Limiting factors used were pitching moment and drag as both of them are limited from material properties and power requirement for compressor.

Some of the conclusions made by Christian Wauquiez are:

The large lift is obtained by strong camber, located approximately at the middle of the chord. The thickness, which has a greater negative effect on the drag than positive effect on the lift, is equal to its minimum value at this condition.

The angle of attack is equal to its maximum value, which gives a lot of lift, and the consequent increase in drag can be compensated by a smaller amount of camber.

For low drag the thickness is equal to its minimum value, lift is obtained by camber, located completely rearward.

Dr. S. Pierret & Prof. Ch. Hirsch [12] have performed a study on Integrated Optimization System for Turbomachinery Blade Shape Design by coupling a user interface, a blade geometry generator to an automatic CFD solution process and an optimization algorithm. They have converted the general optimization problem of minimization of an objective function (loss coefficient) in function of several variables (the geometrical parameters) subject to several constraints (mechanical, manufacturing and aerodynamic constraints), the objective function, and the constraints being non-linear by transforming the original constrained minimization problem into an unconstrained one by converting the constraints into penalty terms that are increasing when violating the constraints. A pseudo objective function is then created by summing up all the penalty terms and the original objective. By the above technique and by use of genetic algorithm for optimization of efficiency they have made some conclusions that blade shape with large lean near the end walls helps increase efficiency, which is in agreement with the secondary flow reduction guidelines found in the literature.

3. Basics of Turbomachinery

Basics of Turbomachinery:

A gas compressor is a mechanical device that increases the pressure of a gas by reducing its volume. The main types of gas compressors are illustrated below:


Figure 4: Compressors family tree

A positive displacement pump causes a fluid to move by trapping a fixed amount of it then forcing (displacing) that trapped volume into the discharge pipe. Whereas dynamic pumps are those in which kinetic energy is added to the fluid by increasing the flow velocity. This increase in energy is converted to a gain in potential energy (pressure) when the velocity is reduced prior to or as the flow exits the pump into the discharge pipe. This conversion of kinetic energy to pressure can be explained by the First law of thermodynamics or more specifically by Bernoulli's principle.

Dynamic type centrifugal pumps, centrifugal compressors are used throughout industry because they have fewer rubbing parts, are relatively energy efficient, and give higher airflow than a similarly sized reciprocating compressor (i.e. positive-displacement). Their primary drawback is that they cannot achieve the high compression ratio of reciprocating compressors without multiple stages.

Centrifugal compressors use a rotating disk or impeller in a shaped housing to force the gas to the rim of the impeller, increasing the velocity of the gas. A diffuser (divergent duct) section converts the velocity energy to pressure energy.

Figure 5 shows a typical compressor map which is a plot of pressure ratio versus mass flow rate. As shown in the figure a compressor map is confined by surge and choke line, also it can be seen that the highest efficiency is obtained in the centre region of the compressor map which is called as heart region (Blue region) and to operate at high efficiency the compressor needs to operate in this range which is not possible at all operating conditions. Thus the need for optimization of the compressor arises and is widely carried out in the industry by either increasing the map width or shifting the efficiency curve upwards at the operating conditions.

Figure 5: Compressor map


Mesh is flexible for faster mesh generation as we can take different sizing coarse ,medium,and fine element ,no of nodes

Accurate results can be obtained as compared to standard conditions

Analysis can be done on complex geometries

Both series and parallel processing are available


User should have good knowledge of CFD

For complex geometries error are difficult to solve.

Takes longer time if the geometry 3D

Software is prone to human error.

4.Modelling and Validation

4.1 Experimental setup:

A compressor cascade is a linear array of compressor blades that can be put in a wind tunnel for testing flow and pressure distribution under different flow conditions to optimise the design. Although compressors produce pressure rise by mechanical work imparted on the flow by rotating blades, wind tunnel cascades are stationary blades. An example of a wind tunnel compressor cascade is shown in Figure 6.

Figure 6: Example of a compressor cascade in a wind tunnel

Experimental data was collected using cascade geometry shown in Figure 4:

Figure 7: Schematic of the cascade

Sufficient blades were used to ensure that the flow was periodic between the blades (apart from the blades at the ends). The boundary layer was bled off on the top and bottom walls to attempt to make the 2D (i.e. with no variation in the z direction). Hence the flow through one blade should be modeled as a 2D blade with a periodic boundary condition in the y direction.

The flow conditions were:

Total temperature: 295 K

Total Pressure: 8x104 Pa

Inlet Mach Number: 0.7

Inlet Flow Angle: 510

Figure 8: Blade geometry and terminology

5.Computational setup and procedure:

5.1Geometry preparation:

3D Geometry of the blade is available in the form of .igs file. This represents one blade from the experimental setup shown in figure 7.3D geometry of the blade is change to 2 d by applying Boolean operation. 2D computational domain is generated in workbench as shown in figure below. To generate 3D computational domain 2D domain is sweep in z direction. All steps explained in lecture and lab is followed to generate 2D and 3D computational domain (geometry and mesh).

Figure 9: Geometry

Figure 10: Solution domain and boundaries








5.2 GRID Generation:


Grid generation is the most important and most time consuming part of CFD simulation.Quality of grid will decide the solution and analysis of the given geometry ,there are different method of making a grid for eg All tri method , sweep method, sphere of influence there are three different grid generations

Structured Grid

Unstructured Grid

Hybrid Grid

Structured Grid ;

In Structured grid method the grid is laid out in a regular repeating pattern called a block. These types of grids utilize quadrilateral elements in 2D and hexahedral elements in 3D .

Advantages of Structured Grid Generation: 1)high degree of control 2) freedom when positioning the mesh

Unstructured Grid : the arrangement of elements has no discernible pattern, the mesh is called unstructured mesh These types of grids utilize triangles in 2D and tetrahedral in 3D.It can be automated up to large extent.Advantages of Unstructured

grid: 1)require little user time or effort due to automation.2) unstructured grid methods

Hybrid Grid :This method is used when positive aspects of both structured and unstructured mesh is required. It contain hexahedral, tetrahedral, prismatic, and pyramid elements in 3D and triangles and quadrilaterals in 2D

Advantages of Hybrid Grid:1)Positive properties of both structured and unstructured grid can be used.


In the 2D grid generation 3 different sizing are taken coarse ,fine ,medium .All tri method is used for mesh generation.

In 2D geometry the edge sizing is given to the boundaries ,inflation is done by selecting the faces and inflation given to the edges. Edge sizing ,bias factor ,no of layers are adjusted according to the mach no.Proper inflation is given to the blade otherwise the solution will have convergence.In 2D there is no thickness .

The grid independency is checked using three different mesh size coarse, medium and fine. It is found that the coarse mesh result has more variations and different trends of pressure along blade surfaces. Medium and fine mesh size results shows approximately equal results. Due to complex physics involved medium and fine mesh is preferred for conducting 2D analysis which, resulted mesh size approximately 15000 cells. The average skewness for this mesh is 0.3 with maximum skewness of 0.6. Figure shows the typical mesh for 2D model.



Coarse mesh

Medium mesh

Figure 11: Mesh


Fine mesh



In the 3D grid generation medium sizing is taken .And sweep method is used,

In 3D geometry the edge sizing is given to the boundaries ,there is no inflation is done in 3D .Edge sizing ,bias factor ,no of layers are adjusted according to the mach no..In 3D there is 100MM thickness .


Fig12 3D medium mesh of domain

6.Boundary conditions:

Certain boundary condition required to run the calculation are:


Gauge pressure at inlet:80000pasc

Direction vector of X and Y are cos51=0.629 and sin51=0.777 where 510 is blade angle

Turbulent intensity and hydraulic dia =5% and 26mm respectively

Thermal conditions for inlet 295 k

Operating pressure at inlet is 00pasc

ρ=P/RT=80000/( 287 x 295)=0.9449kg/m3


Gauge pressure at outlet:70000pasc which is to be adjusted so that we can get accurate mach no which 0 .7 from the iteration we see as the outlet pressure increases mach no decreases .

Direction vector of X and Y are cos51=0.629 and sin51=0.777 where 510 is blade angle

Turbulent intensity and hydraulic dia =5% and 26mm respectively

Thermal conditions for inlet 295 k

Operating pressure at inlet is 00pasc

ρ=P/RT=80000/( 287 x 295)=0.9449kg/m3

Why K-E model?

The nature of the flow in compressor cascade is turbulent flow. Turbulent flow is mainly due to fluctuations in time. These velocity fluctuations give rise to additional stresses on the fluid, the so called Reynolds stresses. There are several approaches to the modeling of these extra stress terms. For the most engineering purposes it is unnecessary to resolve the details of the turbulent fluctuations. Only the effects of the turbulence on the mean flow are usually sought.

In particular, we always need expression for the Reynolds stresses in equations and the turbulent scalar transport terms. For turbulence model to be useful in a general purpose CFD code has wide applicability, accurate, simple and economical to run. Ansys 12.1 has many turbulence model and there applicability based on the certain complexity of the physics, solution accuracy, computation time and stability.

Two equations K-E model is one the available turbulence model available. The K-E model is the most widely used and validated turbulence model. This model required less computation time and gives accurate results for many industrial flow problems.

6.3 1ST Order of accuracy and second order accuracy:

In order to get faster convergence and stable results, we have to follow certain steps to solve numerical equations. The flow is highly compressible in nature hence chances of divergence and cause of incorrect solution. First run the solution using constant density and first order equation to get well initialization of domain. Solving constant density and first order accurate equations gives faster convergence and good initialization of solution for coupled density and second order. After getting first order converged solution changed density to ideal gas and viscosity to Sutherland model. After getting convergence changed first order to second and run solution further to get final convergence. In this way we can achieve well converged solution for compressible flow.

The periodic faces should be placed in the minimum interference location to reduce the errors induced due to periodic assumption as periodic boundaries in Fluent are first order accurate.

Due to periodicity of the flow domain, analysis is performed for the single blade of the compressor. The periodic boundary conditions are imposed on the upper and lower surface of the domain. Periodic boundary conditions are used when the flow across two opposite planes are identical. Pressure inlet and pressure outlet boundary conditions are imposed at inlet and outlet respectively. On blade surface no slip boundary is assumed to exist.

7.Solution and convergence:

Ansys 12.1 CFD software is used for the analysis which solves Navier-stokes and enegy equations. Ansys 12.1 CFD software uses a control volume based technique to convert the governing equations to algebraic equations that can be solved numerically. This involves subdividing the region in which the flow is to be solved into individual cells or control volume so that the equations can be integrated numerically on a cell-by-cell basis to produce discrete algebraic (finite volume) equations. All variables, including velocity components, pressure and temperature, are averages applied to a control volume.

A second order spatial interpolation method is employed to obtain velocity components, pressure and temperature on the control volume faces from those at the control volume centers. Second order interpolation gives more accurate results than linear interpolation. The control volume face of the dependent variables is used to evaluate the convective fluxes. Fluent's segregated steady-state solver is used for the numerical simulations. The SIMPLE algorithm is used to couple the pressure and velocity equations.

A second order upwind differencing scheme is used for the space discretization of the momentum, turbulence and energy equations in all simulations. The under relaxation factors for the update of computed variables at each iteration are for pressure = 0.3, momentum = 0.7, TKE = 0.8, TDR = 0.8 and energy = 1. The residuals are of the continuity, components of velocities, turbulent kinetic energy and turbulent dissipation rate are dropped below 10-5, while for energy it's dropped below 10-7 for converged solutions.


Figure 13: Residual plot of all parameters

8.Results and Discussion :

Analysis is performed for 2D coarse, medium and fine mesh to check the grid independency and then 3D with medium mesh. Air as a compressible (ideal gas) flow with Sutherland model for viscosity is considered for analysis. To achieve convergence and stable solution, first solution started with constant density and first order discretization. After some iteration density changed to ideal gas and Sutherland model for viscosity is used. After some iteration discretization method is changed to second order. Solution started with total pressure at inlet is 80000 Pa and outlet pressure is adjusted to achieve required Mach number and it comes around 700100 Pa. Similar settings are used in all solutions in 2D and 3D domain. Mach number monitor is plotted against iteration number to check the convergence and stable solution as shown below.

C:\Users\DHRUV\Desktop\mach num iter.jpg

Figure 14: 2 D FINE MESH Mach number monitor


Static pressure contours for 2D coarse, medium and fine mesh are shown below. From static pressure contour plots it is seen that increasing mesh size from coarse to fine, low pressure region at suction surface is increased. Low pressure is observed at suction side and higher pressure at pressure side surfaces. For compressible flow coarse mesh is not good to predict the flow around blade near stagnation zone and separation zone. It is observed that to achieve accurate results appropriate number of elements as well as fine mesh near wall is required. Comparing medium and fine mesh results shows insignificant changes in results and hence medium mesh size results can be assumed to be mesh independent results which is used for further 3D analysis.

COARSEFigure 15: Contours of static pressure near leading edge of bladeG:\my pendrive\2d coarse\2d_coarse\2d_coarse_in_static_press.tif

MEDIUMG:\my pendrive\2d fine\2d_15387_fine\2d_fine_in_pressure.tif


FineG:\my pendrive\2d fine\2d_15387_fine\2d_fine_in_pressure.tif

10 . Discussion of Contours of Mach number in domain

Higher Mach number is observed at suction side and lower at pressure side. Highest Mach number in coarse, medium and fine mesh observed is 0.838, 0.854 and 0.855 respectively. Contour plot shows the significant Mach number is changed from coarse mesh to medium mesh, but from medium mesh to fine it is not changed significantly. Thus medium and fine mesh size solution can be considered as grid independent solution. Mach number is plotted at inlet location, it can be seen that Mach number at inlet is approximately 0.7.

MEDIUMFigure 16: Contours of Mach number in domain

G:\my pendrive\2d medium\2d_15267_medium\2d_medium_in_machnumber.tif

FINEG:\my pendrive\2d fine\2d_15387_fine\2d_fine_all_machnumber.tif


G:\my pendrive\2d coarse\2d_coarse\2d_coarse_in_mach.tif


From velocity magnitude, it is seen that due to coarse mesh near wall velocities are spread out in large region. This region is improved in medium and fine mesh. Also higher velocity magnitude is observed with fine mesh. Thus it can be conclude that coarse mesh in not appropriate mesh to resolve near wall and stagnation zone. So the medium size or fine size mesh is the right choice for this kind of analysis where flow is highly compressible.

Figure 17:Contours of velocity magnitude





Coarse Figure 18: Graph of static pressure along blade on suction and pressure surface



Static pressure is plotted along blade length on suction and pressure side. It is observed that coarse mesh results shows the decreasing trend of pressure along the flow direction whereas in medium and fine mesh size results trend is increasing along the flow direction. To compare experimental data to non dimensional parameter pressure ratio (P/Poi) is used. Where, P is static pressure and Poi is total pressure at inlet.

The pressure ration results are compared with experimental data as shown in plot on suction surface and pressure surface. A simulation result for coarse mesh has higher deviation than medium and fine size mesh also trends looks reverse compared to experimental data. Similar trends are observed in experimental and medium and fine mesh results with close agreement between them. Coarse mesh results looks inaccurate due to insufficient cells to resolve boundary layer and wall functions.


13.Comparison of 2D MEDIUM AND 3D MEDIUM results:

From 2D analysis grid independent solution is achieved for medium and fine mesh size. Hence, 3 D analysis is performed using medium mesh size with similar operating and boundary conditions used in 2D analysis. Computed pressure, velocity and Mach number plots are shown in figure. From static pressure contour it is observed that, 3D analysis results are very similar to 2D analysis results because the flow is primarily along blade length. From pathlines it can be seen that flow is streamlined over blade and separation is observed near trailing edge of the blade.



Figure 20: Static pressure contour

Figure 21: Pathlines colored by velocity magnitudecontour_all_pathlinescontour_all_pathlines1

Figure 22: Contours of velocity magnitude


Pressure ratio is calculated on midspan of blade on pressure surface and suction surface. 3D analysis pressure ration results are compared with experimental data as shown in figure below. Analysis results are in close agreement with measured experimental data within 1-3 % accuracy.


14 .Conclusions:

Pressure ratio over compressor cascade blade is investigated on pressure and suction surface using 2D and 3D analysis. 2D and 3D steady state CFD analysis is performed and results are compared with experimental data. From 2D analysis it is found that, coarse mesh is not appropriate choice to predict the flow over blade when flow is compressible. To predict accurate flow near wall, stagnation and separation region appropriate number of elements are required. Grid independent solution is observed for medium and fine mesh size. So it is very important to conduct grid independency while performing CFD analysis. Medium size mesh is used for further 3D analysis and results are compared with experimental data. It is found that 3D results are very similar to 2D results due to streamlined nature of flow. Very good agreement is observed in CFD and experimental data. Simulation shows the feasibility of periodic boundaries as the simulation results has very less deviation from experimental results.