A Parametric Blade Design System - Anders

A Parametric Blade Design System (Part I + II)
Jrgen M. Anders and Jrg Haarmeyer

BMW Rolls-Royce GmbH, Dahlewitz, Germany
Hans Heukenkamp
atech GmbH engineering software technology, Germany
Abstract
A complete 3-D Parametric Blade geometry design system for industrial design project applications is presented. The system generates a 3-D free-form surface of an axial compressor or turbine blade by means of individual definition of quasi 2-D stream surface blade profiles. These profiles are then stacked relative to a reference point for each section along a so-called "stacking line". All design properties both for the quasi-2-D profiles and the 3-D stacking line are parametrised in a problem-adapted way to support the understanding of the aerodynamic engineer. These parameters are then transformed into a fully CAD-compatible B-spline representation. The geometry engine is completed with a CFD-code integration, a blade profile optimising package, a parametric database and a correlation utility to find good starting solutions for new design tasks based on existing proven technology.
Abbildung 1 BR715 Turbine Blades
The complete design systems offers a number of project-related advantages to the user: the system is fully integrated into the complex engineering design process for turbomachinery modules like a multistage compressor or turbine. It offers a bi-directional data exchange between different design disciplines such as stress & mechanical analysis, aerodynamics, checking of mechanical constraints, thermal analysis and CAD-design for platform, root, shrouds etc. The system is completed with optimisation strategies based on parametric free-form-solids generation. It is linked to an engineering database and into an EDM/PDM system.
VKI LECTURE SERIE1.doc
To support preliminary design activities in an early stage of a design project the system contains a correlation utility to interfere from existing designs and get a reasonable initial multistage design blading for a compressor or turbine. The presentation will be splitted in two parts: Part I 1. 2. 3. 4. 5. 6. Part II 1. Technical Details: geometry representation, optimisation, correlation rules, parametric stacking 2. Curvature distribution 3. Standard profile properties 4. Cloning 5. System integration 6. Profile editor b2d 7. Implementation details 8. Graphical user interface 9. Integration into CAD/CAE environment 10.General remarks Introduction Basic Principles of the design task Outline on the major system components Experiences with industrial applications Parametric 3-D blade manipulations Recommendations
Abbildung 2 The new Boing717 twinjet with BR715 engines
Introduction
Blade design & gas turbine business
The gas turbine business has changed over the last five to ten years. In the seventies and eighties a large number of physics-based technology programs were carried out, especially in the
aero/thermo and acoustics area. These programs all together have led to an essential reduction of SFC and hence improved the cost situation for the airlines. In the nineties engine maker have put more emphasis on cost and quality issues. The common core family concept of the BR700 engine family is one successful example of an innovative technical approach to tackle both engine development costs, time and quality for a number of similar derivative products with a significant amount of common components. On the other hand the cost competition has gradually changed into a time- and innovation competition (time based competition). Every player in the global aeroengines market is trying to find and strengthen its unique selling proposition to stay in the market and make profit with reduced time-to-market and innovative engine concepts. The main condition for that is for a single company to be able to design and certify the right engine at the right time for the right application. Technology is still very important and will be developed. But with limited resources, companies must re-think the cost benefit of certain purely physics-based research programs of the past because further return of investment will be quite low. High tech as the one and only unique selling proposition (USP) is not enough for todays customers, they assume it and demand reduced costs for the product as well. Thats why companies have to focus efforts and technology programs on those things that impact units cost most: reduction of hardware costs, design & re-design costs and reduction of expensive test series by numerical simulation and concurrent simultaneous engineering.
Abbildung 3 The dressed BR715 engine
One way to proceed is improvements of methods in the engineering development areas. The objective is to improve the engineering design process, which to a large extend now is computer based. Concurrent simultaneous engineering is the key word for these activities in the past. The complex blade design process is one major part of that.
CAD developments
The design process in aeroengine design is fundamentally driven by physical properties of the components, its weight, performance, stress behaviour and life. All these properties are estimated by various numerical simulation methods such as FEM, CFD and crack propagation calculations. On the design side engineers are using more and more advanced CAD system features. Originally the CAD was just a replacement of the drawing board to get 2-D sketches out rapidly and with the chance to correct without starting from scratch. Then the 3-D CAD systems came up where designers could create and combine standard geometric entities such as cylinders, cubes, tetrahedra, sphere and free-form surfaces. Then more and more emphasis was put onto processing the topological information of a part or assembly. People started generating watertight solid models in CAD. Geometric library standards were established such as ACIS, Parasolids and STEP to uniquely describe functions and properties of the geometric representation of a solid part. In order to process iterative changes and adjustments of a given part without changing topology parametric CAD systems were developed. The idea of introducing parametrics into CAD revolutionised the design work and a new level of efficiency in CAD work could be reached. The parametric idea was so important to the CAD community that one of the pioneers and market leaders in this business decided to name themselves Parametric technology Corporation. Now with parametric 3-D solids as a standard, the original work of step-by-step developing the CAD part and creating sketches, drawings and views is more and more loosing its importance in the whole development cycle. Parameters of a topological master model can be treated outside the CAD world and can be subject to optimising processes or application of artificial intelligence, neuronal networks, expert and decision support systems. Once the master model is generated, the job of finally generating the solid model and all required drawings and views is just a matter of pressing a single button. After a final optimised parametric configuration has been detected, the parameter set is passed to the CAD and replaces the generic parameters of the master model to create the final model.
Abbildung 4 Sketch of the BR715 engine
In the typical business of an aeroengine company the number of major different master model parts is fairly restricted and will not change so frequently. This is basically the turbine & compressor blades with their respective platforms and feet - only a limited number of different
designs is in common use today - , discs, shrouds, drums, sealings and the annular combustor just to name the most important parts of a core engine. Once a library of all relevant master model parts is created, future designs will largely consist of recombination, adjustment and small incremental improvements of these fundamental entities. This process is driven by engine specifications and numerical simulations of the aero/thermal performance and other objectives. So we end up with an engineering process were design and simulation are very much integrated. Bidirectional data exchange between the different simulation models and the CAD is essential. Also meshing of the models or their respective surrounding volumes (for CFD) is essential. It is vital for that process to use geometric standards like ACIS, Parasolids, STEP or IGES for the model descriptions throughout the whole system. Most commercial meshing, CFD and FEM tools have interfaces serving to these standards. We need to make sure that also the CAD and the other parts of the process are using the same standards.
Aerodynamic blade design process

The blade geometry design process generally plays an important rule in the development and verification of a new gas turbine or aeroengine. It is the inner loop of most design iterations and hence - taking into account the high number of different blades required for a multistage compressor or turbine - is a critical cost factor for the task. However, from its very nature it is a seasonss job and in many aspects not very attractive to the scientific community in this area. The limited number of recent publications in the field and the fact that very often students and young engineers are given the ugly job to create the topologically all-so-simple geometry speak a clear language. But is it really so simple? Aerodynamic design deals with boundary layers, shock losses, secondary flow patterns, surface curvature distributions, blade metal and air angles like wedge, inlet & exit angle, stagger & camber. In the early years of aerodynamics standard profile families were developed like NACA in order to describe the profile with a small number of parameters and be able to compare designs and test results of different aerodynamic shapes. Later on the aerodynamisits learned more about the physical effects of boundary layer separations, turbulence, reattachment, skin friction and shape factors and nonsteady loss generating mechanisms. As a consequence of that a new style of optimised compressor blades came up, the so called controlled diffusion aerofoils (CDA). The CDA profile philosophy is mainly based on the understanding of the suction side profile boundary layer physics. The CDA profiles have extended the design space for the blade geometry virtually to infinity. In the low pressure turbine area people realised the positive effect of calming for the stability of the boundary layer and increased blade loading, simply took off up to 20% of the blades out of a single bladerow.
Abbildung 5 Aerodynamic performance prediction for LP turbine blade section
Inverse design methods where developed in which engineers start from an optimal surface pressure distribution and try to generate the accompanying blade profile which creates the desired pressure distribution. However, especially the CDA developments stated new requirements to the geometry design process of a compressor blade. In general, standard CAD systems were not suited to meet the requirements of these new CDA design tasks. People in all major aeroengines companies started to develop highly specialised blade geometry design software in order to generate the best possible CDA profiles. Design space simply exploded to huge dimensions. Every single section out of 20 or more on different radial positions for a single blade in a 10 or more staged compressor with 20 or more bladerows was individually designed to the objectives of the CDA design rules. In all these individual Blade design software systems each individual blade section had usually more than 200 degrees of freedom. Companies started to view this technology as core competence and invested a lot of effort to develop and maintain these highly complex software systems. They were usually backed up with numerous strategies to stack the set of 2-D sections over different height to a complete 3-D blade in a smooth and aerodynamically efficient way. Most 2-D base systems were actually designing the profiles on so-called stream surfaces (surfaces of revolution with the assumption that circumferentially averaged normal flow vectors vanish). This puts an extra difficulty into the geometry design software.
Abbildung 6 Stator blade with stream sections and bowed hub and tip region
Also, systems were coupled with 2-D flow solvers to interactively simulate the flow and improve the design. With all that in hand the design task for a 10stage compressor became a nightmare. The systems offered all kinds of options and even the most odd design idea could be created. The systems were really at an endpoint were they could do everything for you but it was simply too much. Users very often got lost in the enormous design space dimensions and the effort to improve a given design did very often not compare well to the technical benefit of the improvement. Also people developed individual different design strategies to work with these systems leading to very similar designs on completely different ways. This has led to the question wether the approach of total design space was not suitable for an efficient process. Would it be better to limit the degrees of fredom for the blade design to a reasonable number, i.e. specify the design space as large as necessary to do what you want and as small as necessary to restrict complexity. A number of attempts have been made and published dealing with the question of how many parameters are actually needed to design a single section. [cf. /1/,/2/,/7/,/8/,/9/,/10/] The general outcome of all these research work was that for a compressor blade section 20-22 design parameters should be sufficient with a turbine blade potentially requiring 2-4 parameters more to meet all possible design cases. A standard way to proceed now was to use a B-spline geometry representation and translate the B-spline parameters 1-to-1 into the language of an aerodynamic engineer. Now, B-slines is the dialect that most CAD systems can understand. The only task in-between is to transfer the parameters without loss into the 3-D solid CAD world. Some extra work is required to do that, especially when the design works on real stream-surfaces with varying radial height along the axis. The local co-ordinate systems on these stream surfaces can be extremely dangerous, since they are not angle- or distance -preserving, i.e. Pythagoras is not valid, depending on how you define them. So some integration and approximation is necessary in the general case for the data exchange between CAD and aerodynamic design system. But this
process is numerically very stable and accurate and does not deteriorate the fundamental principle of bi-directional data exchange between both systems.
Aspects of commercialisation
When the technical problems of non-linear behaviour of physical behaviour and the intractability of the underlying partial differential equations and their relevance to the development and improvement of a technical solution have been detected first, a number of companies started to develop their own expertise in the respective areas. Later, when in the past decades a number of technologies have reached a high degree of maturity, more and more commercialisation took places and tasks were treated in a standardised way. A number of companies settled down offering technical solutions to the problems, very often a combination of a specialised software package + consulting and training service. The FEM area is a well known example for that development. From the very start of recognising the problem to the de-facto industrial standard of FEM software solutions and vendors like MSC/Nastran for stress FEM analysis was just 2 or 3 decades. This was back in the Eighties. Since then even competing companies have no problem to work with identical the same software package to simulate and predict the stress of their components. This technology is no longer a unique selling proposition for them. It has become a standard. Now we experience the same concentration and commercialisation of software solutions in the CFD area. Companies like Fluent and NREC (just to name two companies without any attempt to completeness) have reached a very high level of expertise. It is nearly impractical, even for a large aeroengines company with quite specific requirements with respect to the capabilities of the numerical code, to develop and maintain their own CFD software package. It is economical and for time and development risk reasons much more appropriate to buy a commercial package, integrate it into the companys engineering process environment and discuss software extensions with the vendor, if necessary. The question now is whether we will see the same story for the complex of integrated aerodynamic blade design In this area it is also getting very expensive for the aeroengines companies to maintain the required degree of expertise in their in-house developed systems. On the other hand, this technology is more and more not viewed as a core competence for the company. Aspects like the right general arrangement of the engine, family concepts, design for ease of maintenance, design for low cost, design for reliability are playing a much more important role for the chances to make profit. Thus the technical and economical relevance of a in-house core-competence in advanced blade design has to be reviewed potentially. Taking all that into account it is not unlikely in the future, that the worlds aeroengines industry will have a (presumably small number of 2-3) highly specialised engineering software companies supplying state-of-the-art blade design technology to this industry. The design software package presented here is viewed to have a degree of maturity to potentially establish a base technology for this task.
The design task

Constraints and design targets
The design of a turbomachinery blade is an iterative process, where a number of (sometimes competing) targets have to be met. The following is a list of the major design constraints and variables which are checked during the design phase:
Constraints
Mechanical Thermodynamic constraints
aerodynamic block - axial 2-D, quasi 3-D positioning of the blade annulus definition, curved throughflow values blade edges axial position
3-D stack (positioning of blade sections along in the radial direction)
torsion and bending modes air inlet & exit angles, mach end bends of the blade numbers, pressures, temperatures life (for turbine blades) FOD (foreign damage) turbulence ratios endwall treatment
object passage area, throat area
stagger camber position of maximum thickness
3-D blading tip leakage flow casing & hub treatment
curvature distribution on suction non-axisymmetric end-walls and pressure side leading edge geometry (circle, radial smoothing of the blade ellipse, spline) blade angles inlet metal angle, exit metal angle inlet & exit wedge angle high aspect ratio of surface mesh
The BRR parametric blade design system This chapter gives an overview of the BMW Rolls-Royce parametric blade design system.
The parametric blade design system is an engineering software package designed at BRR to make the complex compressor & turbine geometry design process better faster more reliable standardised. The introduction of the parametric blade design system will influence other related processes, such as FEM- and CFD-analysis and CAD-design. This has the potential to simplify and improve the whole 3-D blade design process. The system consist of several modules: AutoBlading Blade Profile Optimisation parametric Blade stacking radial Blade smoothing and interpolation parametric CAD interface parametric database
These modules together form an base prototype of the system. With the limited resource and time constraints of an industrial company in the aeroengines business it is clearly not possible to implement such a complex system up to the standard of a commercial software product in terms of reliability, easy of use and user interface integration. However, with the application of best software technology and highly specialised external software consultancy for implementation a maximum value software result could be achieved. More work is at hand to test the parametric blade design system and integrate it into the design process. This activity has to be driven by the user community within engineering.
Parametric blade representation

The blade profile parameters are the basis of the system.
10
Abbildung 7 Blade parameter definition
The blade representation consists of two independent patches of higher order Bzier curves plus leading & trailing edge geometry. Special attention is to be paid to the treatment of the leading edge region. Currently both a circular leading edge and an elliptical leading edge definition are used. Slope continuity is always maintained at the joint points.
Rules-based design
In the first specification phase a so-called AutoBlading (Automatic blading) project was set up to test the principal capabilities of the parametric approach and a rules based system for the design of a compressor or turbine blade. The objective of the second project phase was to convert the trials system of phase I into an complete test system. During the AutoBlading project phase II the scope was extended according to requirements from the compressor department and the availability of the new design program b2d. On the other hand with the given resources it is not possible to create a complete design system. Therefor the work focused on the implementation and testing of major parts of a design system to establish an effective alternative to the existing system.
11
The first idea of the AutoBlading project was to transform existing designs into the same unique parametric representation and detect similarity rules and correlations between the aerodynamic performance and the throughflow axisymmetric conditions. First of all, a tool had to be developed that derives these rules from the analysis of existing configurations. A simple SAS application has been developed to support this task. Based on this tool, rules and blades for various existing and new compressors were created, e.g. the BR715 Booster, HP9 research compressor, Trent500 and Trent800 HP compressor. Finally, the rules-generated AutoBlading blades were analysed and compared with its conventional design. In this sense, conventional design means design with the traditional blading geometry generators. Derivation and editing of rules - software auto_blader Within the project a rules generator was developed. The generation and modification of correlations between the throughflow- and profile parameters is implemented as an SAS tool. With this tool it is possible to calculate algebraic formulas for the x and y axis with the following set of parameters:
Throughflow-parameter AI Air inlet angle AO Air outlet angle MAI Air inlet Mach number MAO Air outlet Mach number REYNOLD Outlet Reynolds number S2 H2H1 Streamline contraction SCX Space chord ratio CL Lift coefficient PI Pressure ratio XCHORD x-Part of chord PHSEC Percentage of height NOB Number of blades ZWEIFEL Zweifel criterion RS Rotor or stator
profile parameter MUI MUO BIA BOA RADT RADL STAG CHORD CURVISS CURVIPS CURVOSS CURVOPS TANGISS TANGIPS TANGOSS TANGOPS STIFFISS STIFFIPS STIFFOSS STIFFOPS
Inlet wedge angle Outlet wedge angle Blade inlet angle Blade outlet angle Radius trailing edge Radius leading edge Stagger Chord length Inlet curvature suction side Inlet curvature pressure side Outlet curvature suction side Outlet curvature pressure side Inlet tangents suction side Inlet tangents pressure side Outlet tangents suction side Outlet tangents pressure side Inlet stiffness suction side Inlet stiffness pressure side Outlet stiffness suction side Outlet stiffness pressure side
The throughflow-parameters were extended for both compressor and turbine use (Zweifel criterion, lift coefficient etc.). The formulas for x and y are allowed to be mixed up by both parameter types. However, it has to be kept in mind that the rules must built up a linear system of equations which should be consecutively solvable. In addition, x and y are stochastically analysed and plotted in a diagram for existing compressor data by SAS. For the rules, fit functions in polynomial form with grad 0 to 3 were used. With
12
SAS it is also possible to select area restriction in the plot and in the analysis so that exceptions of the rule can be extracted.
Similarity rules Empirical rules Geometric equations Spline conversion
Physical Throughflow values
Abstract Throughflow parameters
Profile parameters
Spline parameters
Streamsection profile geometry
Abbildung 8 Rules-based AutoBlading process
Rules generation process Two ways of rules creation are found: Investigation of data of conventional designed compressors and turbines The data of existing gas turbine components are reviewed. Three ways of making rules were found: usage of physical dependencies and simulation of technical formulas trial-and-error methods automatic correlation of each single throughflow-parameter with each single profile parameter in order to find polynomial dependencies of degree 0 to 3 In many cases, these rules fit well for the existing compressors and turbines. Only Method three turned out to be not flexible enough to get acceptable correlations. However, with an aerodynamically determined domain decomposition of the complete set of blade sections the correlations could be highly improved. Separate rules were generated for rotors and stators, transonic and subsonic compressor sections etc. Adaptation of existing rules Existing rules of conventional designed compressors were taken and adapted for the target compressor in view of the resulting blade. The rules adaptation is tested for some sections and transferred to the whole compressor. In many cases, these rules fit only for the objected compressor. However, this can help detect scattered design sections. This is helpful for the quality check within large teams of aerodynamic experts working on a single component. Based on the SAS tool, different sets of rules and blades were analysed and generated:
1. BR715 Booster
Based on the first BR715 booster design, rules were made and applied to the booster aerodynamic. The resulting blades showed an improved efficiency and performance compared with the conventional blades. The reason of this improvement is not quite clear. Most likely a
13
smoothing of the section geometry parameters with respect to the throughflow data may have caused the effect. However, also the improved section curve smoothing can be reason for the better performance.
2. HP9 research compressor
Here, the rules of the PRIME HPC were applied to the planned aerodynamics of the HP9 high loaded research compressor. The resulting blades did not reach the desired performance due to the advanced loading. It was found, that the definition area of the rules was extended in Mach number and camber. The adaptation of existing rules showed reasonable results.
3. Three spool engine HP compressor
The blading of 2 different HP compressors of three spool aeroengines were analysed. Highly loaded HPC The highly loaded HPC had a new design style with an unconventional 3D design. The first simple approach of rules derivation for these blades failed, the data looks scattered. Thus, the fit function resulted in large Root-Mean-Squares and standard deviations. It was impossible to recreate the blades by using the existing rules and the desired aerodynamics. Lower loaded case Based on the existing rules BR715 HPC, BR715 Booster and three-spool-high load HPC, different bladings were created. The three-spool-high load HPC rules showed the best results, because of the aerodynamic similarity of both compressors, and because of the low loading of the compressor compared with the highly loaded case. Only incidence, deviation and curvature distribution on suction side had to be adjusted to come to a reasonable 2D blading. The adjustment was fulfilled manually with the b2d parametric section design program.
4. High lift turbine
BRR turbine blades consist of elliptically shaped leading edges. The quintic spline fit of elliptical leading edges combined with profile shapes is poor. Thus, a special leading edge parameter adjustment is necessary to improve the parametric fit for turbines and to find satisfactory rules. However, according to the test cases the rules-based profile generation is also feasible for turbines.
Parametric 3-D blade manipulation To complete the idea of automatic creation and simple manipulation of a complete parametric compressor, a set of additional activities was required. First of all, a simple tool b2d for the generation and modification of a single parametric profile section was developed. This b2d tool was gradually extended during the project and is now in a form to control and modify all major aerodynamic parameters of a 2-D section. The idea of using B-spline curves and surfaces for the generation of smooth blade geometry came up already 20 years ago. The problem with that is, that the terminology of B-splines has very little to do with the parameters the aerodynamic engineer uses for his design. The idea of b2d was to replace the B-spline parameters with aerodynamically meaningful parameters as far as possible.
14
However, some parameters are not transferable one-to-one into the B-spline description. Especially the standard procedures for the adjustment of profile thickness and suction side curvature distribution were requested. Both tasks have been solved with numerical procedures. These procedures were inserted as standard options into the AutoBlading system. Furthermore a simple low-parametric representation of the 3-D stack for a complete blade consisting of a number of sections was requested. This approach has been implemented with a parametric stacking program stack3d (ref. /20/,/21/). stack3d gives aerodynamic designers a powerful, fast and flexible tool for creating specific 3-D blade properties. A major disadvantage of most existing design systems is the principal inability to carry lessons learned automatically from one engine design project to the next. AutoBlading inherits the best assumptions to solve this problem. A parametric cloning option was developed to help carry experiences of previous designs to a new project. Finally, the pilot implementation of AutoBlading did reach a degree of complexity and functionality, that a systematic analysis and improvement of the heterogeneous implementation was necessary. After that a user-friendly desktop-integration was developed and an concept for the integration of that into the BRR engineering environment was set up.
Linear & standard suction side curvature

The suction side curvature distribution is one of the major design items for compressor profile design. Currently different philosophies are applied to get best aerodynamic profile properties. To support the suction side curvature design work a set of standard curvature distribution options have been developed. These options create a linear or convex curvature distribution for the suction side profile geometry. Changes to the geometry are normally minimal due to the properties of the second derivative equivalent curvature. Pressure side geometry as well as blade inlet and exit angles remain unchanged. Blade thickness may change slightly. It is recommended to apply the standard curvature options first and the correct the blade thickness. This will result in a profile with an optimal curvature + required thickness. More details of the mathematical procedures for standard & linear curvature design are given in the next section.
Mathematics of Standard Curvature Distribution A curvature distribution of a 2D curve is said to be standard if it is monotone and slightly convex (almost linear). Denoting the two components of a curve x(t), y(t), with t being a parameter living in the unit interval [0,1], the curvature is defined: c( t ) = x (t ) y (t ) x (t ) y (t ) x (t ) 2 + y (t ) 2
3
(1)
The definition of a standard curvature may be described as: (2)

c ( t ) = w( t )
15
with a non-negative but small right-hand-side. Since scaling and rotating does not change the curvature pattern, it is no loss of generality to consider only curves that start at the origin and end at (0,1). Now restricting to the curve which are closed to the x-axis one may replace the curve parameter t by the x- co-ordinate and thus considering only functions. With this approximation (1) simplifies to: c( x ) = y( x) 1 + y( x) 2
3
(3) Now denoting: (4)
y1 ( x ) = y ( x ), y 2 ( x ) = y ( x ), y 3 ( x ) = c( x ), y 4 ( x ) = c( x )
(1), (2) may be rewritten as: (5a) (5b) (5c) (5d) y1 ( x ) = y 2 ( x ) y2 ( x) = y3 ( x) 1 + y2 ( x ) 2 y3 ( x) = y4 ( x) y 4 ( x ) = w( x )

3
w(x) denoting a non-negative small prescribed weight function. In order to obtain a unique solution 4 initial or boundary values must be provided. It is already assumed that (6a) (6b) y1 (0) = 0 y1 (1) = 0
In blading users usually want to control the tangent of the shape. Thus the remaining two boundary conditions will be: (6c) (6d) with given angles 0 and 1. A solution of (5), (6) can be calculated by utilising a standard software package. In b2d the IMSL routine BVPFD is used. Experience shows that the choice w(x) = (0 - 1).x/50 or w(x) =(1 - 0).(1-x)/50 gives good results. The output of a boundary value problem solver BVPFD is a number pointwise given curve. Thus it remains to find curve parameters such that the resulting curve fits best to the points. This is done with the fit algorithm that is already introduced in the system. y1 (0) = tan( 0 ) y1 (1) = tan( 1 )
Suction side curvature distribution, form factor H and loss loop

16
For the geometry design of compressor blades companies normally use a set of standards and rules describing the principal shape properties of an acceptable blade geometry. Amongst those is a recommendation for the suction side curvature distribution. This recommendations frequently used basically say that the suction side curvature should have a decreasing distribution from the leading edge to a a minimum at the trailing edge of the blade. The analysis of various public available profiles from research projects reveals different design strategies for this particular feature. Some profiles have obviously been designed with a strictly linear suction side curvature distribution ending at zero curvature at the trailing edge. The principal setup of the 2 most common styles is sketched in figure 1:
Style 1 Style 2 Suction side Curvature
Suction side surface length x/c
The effects of these different approaches have been compared in a little design study. The AutoBlading geometry tool b2d has been used in a quasi-inverse design mode starting from two existing profiles. The objective was to implement the style 1-like linear decreasing curvature feature with a minimum of geometry change. Test-Profiles Case 1:
original design redesign original design redesign
BR710 HPC Rotor 5 mid section BR710 HPC Rotor 5 mid section HPC Rotor 1 near hub section HPC Rotor 1 near hub section
Case 2:
Figure 1 shows the original and the redesigned profiles with their respective curvature distributions.
17
Results and discussion The resulting loss loop for case 1 is shown in figure 2 in the appendix. This plot clearly shows lower losses in the design area an extended off-design range The blade tolerates a broader incidence interval. Moreover, looking at the shape factor H for the design point calculation, we get the following result (case 2, figure 3 - 5 in the appendix): original redesign linear increasing shape factor up to 2.7 (at the trailing edge) convex increasing shape factor up to 2.3
From these results it can be estimated that the boundary layer of the redesigned blades is less loaded and will sustain off-design situations more stable. Another (presumably partly numerical) effect is that the linear curvature profiles have a better convergence within the used blade-to-blade solver MISES thus reducing the computation times. This effect may be partly due to the influence of surface curvature terms in the boundary layer calculations. The results presented above are clearly only indications of possible advantages of the linear curvature design rule. More investigations are at hand to validate these results and find possible conclusions for future design work. The results of these investigations may have an profound impact on the AutoBlading integration project. A rule like the one from style 1 type blades will reduce the number of free geometric parameters. It is possible to implement this rule in the AutoBlading project which leads to a remarkably simplified design procedure.
Thickness correction With the approach of separately defining suction and pressure side profiles blade thickness is not a direct design parameter. To define blade thickness and thickness distribution (i.e. position of maximum thickness) a separate method had to be developed. The methods asks the user for the value of the section thickness. The required thickness will then be maintained by a minimum change of the pressure side geometry. The result will be a smooth airfoil with the required thickness. Blade inlet and exit angles and suction side geometry remain unchanged throughout the procedure. Details of the thickness correction procedure are outlined in appendix 6.
In the setup of the AutoBlading rules neither the blade thickness nor the position of maximum blade thickness appear as parameters. Both thickness and position of maximum thickness are fundamental values for the design, particularly of high pressure rotors. For each of them usually design constraints and intervals are given to meet the stress and vibration criteria. In order to insert these parameters we have different possibilities. Within the parametric blading design system we have selected one out of these different alternatives, which appeared to cause the most uncritical changes to the blade geometry. So the objective was to
18
meet the required blade thickness and position of maximum thickness with minimal variation of blade geometry and rules. Distance function A possible distance function is given by the following sum of squares of weighted differences diff=100*incidence+100*deviation+10*wedge_inlet+10*wedge_outlet+2*curvatu re_inlet+... For that we have to answer 2 questions: 1. Which parameter can be free in order to set thickness? 2. How do we have to weight these degrees of freedom? Variant 1: On the pressure side of the profile knot point 3 of the B-spline polygon (from the leading edge) is set free both axially and circumferentially. These extra 2 degrees of freedom are sufficient in most cases, to correct between 25%-50% maximum thickness position and correct maximum thickness. Variant 2: Release the leading edge wedge angle and knot point 3 of the B-spline-Polygon (counting from the leading edge) in both directions. With these extra 3 degrees of freedom most likely all practical design changes for thickness and thickness position can be carried out.
B-Spline-Polygon suction side
Modification of the 3rd knot point on the pressure side
Picture 9 Variant 1: Knot point 3 on the pressure side is free design parameter Both alternatives have the advantage, that they do not change the suction side profile, i.e. they do not affect the diffuser performance of the cascade. More other variants are possible.
19
Variant 2 has been implemented with a little IMSL-Routine uminf (unconstrained optimisation of a multivariate function with finite difference gradient). Based on that other variants can be implemented easily. Defining target functions - general approach To proceed further, a set of general questions has to be answered: 1. Which additional design data has to be considered in the rules for the AutoBlading? This could be loading distribution (front/rear loading, radial variation of incidence), choke margin, stall margin etc. 2. Which geometric parameters should be free for adjusting these additional design targets? 3. What is the weighting for these target parameters? To answer this questions, we have to take into account the different process approach of the parametric design compared to the traditional design style. If the designers previously directly manipulated geometry in order to achieve the desired aerodynamics, we are now in the position to manipulate the target function and its components and weights. The profiles are then generated automatically according to the target specifications. This is usually named inverse design. The variability and flexibility of the target function and the quality of the change methods is important for the work of the engineer. Therefor the above questions have to be answered carefully to get the right answers for an individual design task.
Parametric 3-D stacking Once the 2-D streamsurface profiles are defined, they have to be stacked to give the final 3-D blade shape.
20
Abbildung 10 TecPlot display of stacked blade
This is a tedious and complex task for real 3-D shaped blades. Therefor a simple procedure has been developed to simplify this and make it repeatable. This procedure is based on a parametrisation of the major stacking options. The specification of the stacking parameters is described in reference 20.
Abbildung 11 GUI for the parametric 3-D stacking program
21
The parametric 3-D stacking was implemented in a program stack3d. A detailed description of stack3d including user guide and implementation details in contained in Ref. 21.
Abbildung 12 Standard views for parametric stacking
Radial blade smoothing and interpolation Usually the complete blade is defined only on 3 to 7 intermediate sections along the total height of the blade. To convert the blade into the CAD system 21 sections are necessary to give the required accuracy. To get the 21 sections out of 3, 5 or 7 originally defined sections an interpolation process is carried out. With the parametric section the best approach is to interpolate these sections in the parameter space. This will result in smooth airfoils. The interpolation result can be checked in the parameter space. If further radial smoothing of the blade is necessary, this can be done by smoothing the parameter distribution along the blade height. A radial blade smoothing and interpolation tool rit has been developed to carry out these tasks. This tool is fully integrated into the design system. It can also be used to optimally smooth existing profiles very easy. rit has a direct interface to the parametric CAD systems and can directly create a CADDS part.
22
Abbildung 13 Radial Blade smooting & interpolation tool
Ref. 22 contains a description of the rit tool.
Profile optimisation A blade section that is result of the AutoBlading algorithm is expected to be approximately 90%95% as good as an optimum solution. The quality of these section is considered to be sufficient to work with in a 3D design strategy. With a set of autobladed sections one may do further 3D design by applying sophisticated stacking methods e.g. sweeping, bowing, and leaning of blade shapes and analysing these geometries with a 3D analysis tool. Another application which may be even more important is optimisation. An autobladed section is an excellent starting point for an automated optimisation algorithm. The current discussion both internally and with external contacts shows that the major activities to take place are as follows: Adjustment of the parametrisation to the requirements of usage in automated optimisation environment Development of an objective function that is capable to support the engineer's design intend Selection of an optimising algorithm applicable to the problem of blade section design The authors of this report are convinced that there will be a major benefit for our company in terms of lead time and development costs to have an optimisation tool in place. It is known that in
23
other companies (e.g. ABB, SIEMENS) there are existing running projects on blade section optimisation. Hence it was recommended to start according activities. As a rapid prototyping result a section geometry optimiser BladeOpt was developed. This tool optimises blade sections based on MISES calculations. The user can interactively define target functions and constraints. BladeOpt will the create an optimum blade shape within a couple of minutes on an average workstation.
Abbildung 14 BladeOpt profile optimisation tool - parameter and constraint management window
24
The optimisation routine in BladeOpt will find geometrically similar profiles. The user can interactively specify which blade properties should remain unchanged and which properties can be optimised. AutoBlading results in smooth parameter set Optimisation in sub-space (of the total parameter space, eg. normal to blade surface) 8 => 4 free parameters will result in smooth parameter set Target function components thickness f1 outlet slope f2 pressure ratio f3 loss (visous, wave) f4(total) spike ss-ps f7,f8 choke margin shape factor H, Hbar machnumber distribution loading characteristics
f5(visc), f6(wave)
A detailed description of BladeOpt is given in Ref. 23.
Design process
Fundamentally, it is possible to create blades supported by the parametric blade design system. This could be demonstrated with the AutoBlading testcases BR715 Booster and the three-spool engine HPC. However, the blading algorithm is a complex and sequential task. The original intention of fully automatic creation of a final blade design by means of rules applied to aerodynamic throughflow conditions turned out to be not a feasible objective. However, the AutoBlading system generates a reasonable design within seconds based on rules and experience knowledge. Keeping that in mind a new blading procedure was defined. This new process starts with an initial preliminary design either from AutoBlading or from an existing cloning blade. In the next steps the applied rules have to be optimised in view of the resulting blades. Fine tuning is to take place in the parametric profile modification program b2d where each section must be modified manually. The aerodynamic blade design process basically consists of the following steps: setting of incidence and deviation supported by 2D CFD calculation with a 2-D blade-to-blade solver setting of blade thickness setting of suction side curvature distribution monotone and smooth fixing of blades in aerodynamic block increase choke margin up to the required percentage by moving the stagger create all required interpolated sections radial smoothing (if necessary) parametric stacking
25
creation of parametric CAD blade with class A high quality surfaces The parametric CAD model can be represented in a standard CAD solids format like ACIS or Parasolids. This particularly improves the interface to platform/foot design, stress calculation etc. The above described process is sufficient for the design of 2-D blades, not particularly designed for 3-D effects such as passage vortex and tip clearence. However, it gives a fully parametrized description of the 3-D blade shape and hence allows optimisation towards all of these 3-D aerodynamic effects. The parametric blade design systems forms a bais for the design process. How it is used for different design tasks depends on the design philosophy of the individual cases.
Software quality and integration A production tool to support the blade design process has to fulfil some fundamental software conditions. In this sense, the parametric design system has to support all necessary workflow and process functionality for the user.:
Workflow information: Where am I now? What am I doing here? Which section do I edit? Which sections are displayed? user interface features: Use of Scroll-Bars Manual input of values Error-Management at array limits Rounding error checking for input values read/write access control and security for concurrent simultaneous engineering life cycle and database information: Which sections are in work/release? Which projects are running? How many blades are released already?
Testing and introduction as project standard This new parametric blade design system is based on a fundamentally different approach than existing geometry systems like the ones used currently at RR. There are technical as well as process-oriented advantages with the new system leading to a better design within less amount of engineering time. But - as always - using the system and exploit the advantages of the new technology requires user training, learning and acceptance of the differences. This is the critical success factor for the return of the investment into this new technology. A company will only get the real benefit out of it if they decide to go for it, use it and incorporate the necessary changes in the interfacing processes.
26
Aerodynamic Standard profile properties with the Parametric Blading system

To meet the requirements of a standardised profile geometry generation process standard profile properties have been introduced. The user now has the choice of selecting standard curvature distributions This option offers two different options for standard suction side curvature. Both options will create a convex decreasing curvature on the suction side. Fixed parameters: all angles stagger, chord radius leading + trailing edge pressure surface free parameters tangent length, curvature + stiffness suction side leading + trailing edge thickness correction This option lets the user enter a desired profile thickness. This thickness is obtained by changing the leading edge portion of the pressure side profile fixed parameters all angles stagger, chord radius leading + trailing edge suction surface tangent length, curvature + stiffness pressure side trailing edge free parameters tangent length, curvature + stiffness pressure side leading edge With these new standard profile options the new 2-D profile design process can be sketched as follows:
27
read throughflow & annulus definition
av. Time required to perform action:
Run Rules to create initial profile
1 min
Check profile properties in b2d
5 min
Adjust major parameters if necessary (angles, chord, stagger)
Use standard options for fine tuning of curvature distribution, choke margin, thickness, PMXC File parameters
2 min
Total less than 10 minutes File stream section
Abbildung 15 Profile design process
Throughout this process the aerodynamic engineer only modifies meaningful aerodynamicrelated parameters. The new standard options free the engineer from using the complex B-spline parameters tangent, curvature and acceleration with limited transparency for aerodynamic effects. This approach is a combination of a fully CAD-compliant spline-profile representation and a simple standardised user interface to speed up the design process. The parametric blade design system enables the user to design a 100%-Blade section geometry as a smooth spline parameter profile with the required aerodynamic properties for a given set of inlet- and exit throughflow conditions and a preferred suction side curvature distribution within less than 10 minutes.
Curve and profile representation types

Within the parametric blade design system different representations for the geometry are used for the different tasks. The conversion from one format to another is carried out with a set of programs and subroutines. The figure below outlines the main interrelationships and transfer routines.
28
p - Program Geometric x,y,,tang,curv,stiff App.f (atan,curvature,stiff2) Hermite spline x,x,x, y,y,y Generate_spline_coefs_p(p) quintic_spline_to_geometric_trafo(p) HrmBsp
Profile_to_geom(p) geom_to_profile(p) B-spline polygon xcoeff(6),ycoeff(6) Approx.: app.f (unlsf) Thickness correction, standard curvature distribution Read_hparameter Profile parameters (20) mui, muo, bia, boa, radl, radt, stag, chord, 4x(curv, tang, stiff) points (x,y)
Abbildung 16 Profile representation types and interfaces
Cloning
A frequently repeated task in design projects is the adoption of the blade geometry to a slight change in the throughflow aerodynamics, e.g. to meet changed loading or thrust requirements. This process is called cloning. It is estimated, that 80% of the design work is in fact incremental change of existing already designed parts. One objective of the attempts to improve efficiency of engineering work for this tedious and long-lasting task is to reduce manual work and (at least partially) support it with automated features. This chapter proposes a way to get there by means of the parametric blade description.
Problem description
29
The workflow in the rules-based aerodynamic design is a sequence of rules applications and manual adjustments:
rules
manual adjustment of single sections
section 1...nsec: throughflow parameter

Abbildung 17 Design workflow
Profile-Parameter p1,...,p20
modified ProfileParameter p1*,...,p20*
If during the design phase a new slightly changed throughflow data set appears, all manual adjustments to the individual sections have to be repeated. This is not feasible. To solve the problem, the use of an offset-matrix was introduced. All individual changes beyond the automatically created blade sections are stored in an offset-matrix. These offset values represent the distance of each individual blade parameter from the rules-based automatic value. This is applied to all sections and all blades of a given design. Since we can assume that the rules already create a reasonable geometry, offset values will usually be small and the disturbances represented by the offset matrix can be expected to behave linearly. In case of non-linear effects or intentionally tweaked designs it is very difficult to trace and find correlations. For this cases a 2-stage process is proposed: First, the best statistical correlation should be derived and applied to the design task. As a second step, the perturbations from the correlation formula are added to the rules-based design in form of an offset-matrix. This 2-stage process ensures that the advantage of a rules-based process is kept and the necessary flexibility of designing individual features for certain sections is prevented. The basic idea of parametric cloning is to carry design intent information from one blading project to the next by means of an offset-matrix.
Definition of the offset matrix:

The offset matrix is defined by section i=1,...,nsec, parameter j=1,..,20: offset(i,j):=p(i,j)-p*(i,j),
where p(i,j) denotes the automatically created AutoBlading parameter j for section i and p*(i,j) denotes the manually adjusted parameter value for the same section for a reference design.
With that matrix in hand, we get a new design process.
30
In case of a new throughflow coming up, the following process is required: New throughflow:
Section i
Manual correction p Database sample profiles p p New rule
Profil parameter p
AutoBlading rule
First throughflow
Throughflow parameter
Abbildung 18 Offset matrix and design
New (modified) throughflow
1. application of rules to the new throughflow 2. increment the AutoBlading section parameters with the offset matrix 3. further manual adjustment of sections may lead to a new updated offset matrix In case of a new rule, the user has to 1. take all offset information an throughflow data 2. create new correlations 3. create new offset-dataset (out of reference throughflow + previous offset dataset This process guaranties an optimal fit of the blading to new throughflow data. Moreover, the information from the manual modifications is kept and can even be enhanced. With that technology in hand a step-by-step improvement of the design based on varying throughflow data is possible with minimal effort. With the traditional process every change in the throughflow design data had to be manually upgraded to the new boundary conditions. Implementation & layout
The implementation of the offset matrix feature in a graphical user-interface is a straight forward rapid prototyping activity. New software-engineering technologies are used to achieve a user friendly and standardised stable software solution. The principal setup of the graphical user interface in described in the figure below.
31
Graphics area
Sections coloured according to state: rules-based modified not yet created interpolated Graphical selection of sections with mouse
S1 R1
R2 Compressor
AutoBlading Interpolation of parameters Cloning Manual Modify (b2d) Parametric stacking
MISES
<results, information etc.>
Text area
Abbildung 19 Integrated system blading - layout of user interface
Concept for the application of AutoBlading for design work

The AutoBlading system & technology has a high potential for the improvement of the design process. In order to have the full benefit out of this the system has to be integrated into the complex design environment. This also requires changes in the affected working practices. The users will experience a learning curve till the point, where the parametric approach is fully integrated. For in use of the system we have different alternatives depending on the focus of the project. This may be automation of procedures variability application to different types of engine components
32
individual adjustments
Automation
The parametric rules based design approach is clearly not flexible enough to cover all possible compressors or turbines. I.e. for a new design project the rules need to be adapted. They are not so general to cover the whole design space. We will have individual rules for high speed compressors, low speed compressors etc. Once the rules have been adapted to the individual case, the fine tuning is a relatively small task. That means that the main advantage of the system is in the speed-up of the iteration process in the design. Cloning The main advantage of cloning is that we re-use knowledge of previous experience. We apply cloning rules from an aerodynamically similar previous blade which once has been optimised for a particular task to get a new blading configuration for new aerodynamic conditions. Thus we have a kind of expert system with a knowledge base in form of parametrised data and correlations. This advantage is only partially valid for the AutoBlading system itself, because a single rule is originally based on statistical interference, where all previous blades are equally weighted. Thus the individual previous blades are not weighted with respect to a distance for the new design task. On one hand this keeps a required generality of the rules, since they are not designed for a singe parent blade set. On the other hand this causes difficulties to set correlations for some parameters with high variance. To combine the advantages of both aspects a so-called AutoCloning has been adapted: 1. 2. 3. 4. A Clone-Parent (Clone-profile+Clone-throughflow) is parametrised for the Clone-parent AutoBlading parameters are detected out of Clone-throughflow AutoBlading parameter are generated from the new throughflow the new blade is new determined as a combination of: Blade inlet+outlet angle, chord from AutoBlading all other Parameters from the Clone-parent
Example: Determination of Incidence 1. Clone BIA=55, Air inlet=40, Rule-BIA=52 2. Rule for throughflow new: BIA=65, throughflow-Air inlet = 45 3. new Blade inlet = 45 + (55-40) 0 60 With that we have the following procedure:
33
DT-3 DT-3 Throughflow data Throughflow Throughflow Project 1 Version a Project 1 Version a Project 1 Version a
Project 1
Rule Rule Rule Rule
Profile 1 a
Project 2
AutoBlading
Profile 1 a
Profile 2 a Project n Rule Rule
Profile2 b
Picture 20 Design process with AutoBlading
Cloning section DT-3 DT-3 (DT-6 + DT-3) Throughflow Throughflow Project 1 Version a Project 1 Version a Project 1 Version a
DT-3 DT-3 Throughflow data Throughflow Throughflow Project 1 Version a Project 1 Version a Project 1 Version a
Project 1
Rule Rule Rule Rule
Profile 1 a
Project 2
AutoBlading-Cloning
Profile 1 a
Profile 2 a Project n Rule Rule
Profile2 b
Picture 21 Design process with AutoBlading+Cloning
From the process point of view, the cloning states a new requirement. Additional to the rules and the throughflow data, for each individual section a clone-parent ( blade section and throughflow) are necessary as input data.
34
This work can simplified by providing a set of optimal blades and let the AutoCloning process search for the nearest neighbour by means of a given distance function. Application to Fan, Booster, HPC, HPT, LPT The requirements for the different components are fundamentally different. The fan blade has to be designed individually in each section with large 3-D effects and major radial blade profile changes. For a typical booster and HP compressor the large number of blades is the major problem. They should all be designed according to similar rules and with the same blading philosophy. Turbines require a different parameter set, especially for low-pressure turbines the cloning in combination with an inverse design strategy seems to be the best choice. The system supports an easy way of defining a large group of profiles with similarities for a complete component, to find design correlations and to correct them manually afterwards. This manual work is reduced to a very low effort. Usually only a small number of profiles has to be adjusted manually and these adjustments are restricted to a low number of parameters. Most of the standard actions can be carried out automatically, such as incidence and deviation setting, adjustment to aeroblock, setting of thickness and position of maximum thickness and standard linear suction side curvature distribution.
35
Blade design system integration

To get the real benefit out of the parametrics in blade design, the integration into the existing design procedures and systems has to be implemented. Traditionally, there are 2 different ways to do that, and we call them here the aerodynamiccentric and the CAD centric integration. The main differences of the two alternatives are outlined in the following table.
Aerodynamic centric versus
CAD-centric CAD-centric Blades defined as CAD-surfaces right from the beginning Aerodynamics creates intersections with stream surfaces for their purposes advantage: better integration into CAD (virtually no interface required, no data translators etc.) CAD license required to do aerodynamic design
Aerodynamic centric Aerodynamics defines blade shape with internal geometry definition blade geometry is then transferred to CADDS (with some approximations) advantage: very short iteration loops for Aerodynamic design disadvantage: design loops with stress, CADDS very time consuming, multiple geometry data approximation leads to geometry errors
solution:
- identical parametric Blade geometry representation for Aerodynamics and CAD - standard Hermite splines of degree 5 - bi-directional data exchange between Aerodynamics, stress, design, CFD etc.
parametric integrated blade design system
What can be saved with the new process? What improvements do I have? design time cycle time reductions from 6 weeks (per HPC) down to 5 days repeatability of design work due to parametric representation improved quality and uniformity of profiles common geometry representation throughout the whole design process
36
Integrated Blade design system
Traditional process
Preliminary design
Mean line prediction, surge line etc.
throughflow
Streamline curvature
aero-design system
Blading, stacking, blade-toblade
Smooth spline fitting
Legende
Transformation with approximation Nonparametric representation uniqueTransformation without approximation (1-to-1 mapping) no blade geometry present Application program
37
New process
Preliminary design
throughflow
Mean line, surge line, etc.
streamline curvature
Automatic Blade generation AutoBlading 90% blade
enhanced system b2d, b3d, parametric 3-D stacking, radial smoothing
b2d, b3d, stack3d, MISES parametric
Legende
Transformation with approximation
Stress, thermal analysis

uniqueTransformation without approximation (1-to-1 mapping)
Spline representation
Nonparametric representation
no blade geometry present
Application program
38
Combined process
Preliminary design
Mean line, surge line, etc. Strea mline curvat ure
Blading
throughflow
Automatic Blade generation AutoBlading 90% blade
aero-design Stacking system Blade-toBlade
Smooth spline fitting
CAD
enhanced parametric b2d, b3d, stack3d, system MISES b2d, b3d, parametric 3D stacking, radial smooting
Stress, thermal analysis
Legende
Transformation with approximation
Spline representation
Nonparametric representation
uniqueTransformation without approximation (1-to-1 mapping) no blade geometry present Application program
39
Parametric profile editor b2d

As a result of the experiences from the AutoBlading rules-based system and the requirements for a stream-section profile modification tool a new profile editor b2d has been developed for geometric shape definition and modification of turbine and compressor blades. Features of b2d The profile editor comprises the following features: 1. 2. 3. 4. 5. 6. 7. 8. 9. 100% compatibility to the AutoBlading system automatic optimum fit parametrisation of existing profiles smooth blade surfaces 2-D design capabilities improved easy-to-use graphical user-interface improved program reliability parametric curve description with Hermite curves of degree 5 read & write to the BladeFile database, to a parametric database and to external geometry can be used in conjunction with other design tools
Program description The profile editor uses the same parameter representation of the presentation for the section geometry as the autoblading system. The user is referenced to part I of the presentation for more details.
Abbildung 22 Blade profile editor b2d

40
Basically what the user modifies are parameters like blade inlet & exit angle, blade wedge angles, stagger, chord, tangential length, curvature and stiffness of suction and pressure side curve on both leading and trailing edge. This sums up to a total of 20 parameters. Experience shows that this should be sufficient to meet design criteria for all compressor and LP turbine blades in engine projects relevant for BRR. The will be further investigation whether the same approach can be applied for the more "kinky" HP turbine blades as well. The main window The main window consists of a menu bar for read & write options of section geometry, a single scrollbar and a display panel for the parameter values of all 20 section geometry parameters. The use can select an arbitrary parameter by clicking with the mouse on the respective radio button an modify the value by using the slider.
Abbildung 23 Parameter variation in a common scrollbar
The graphics window The main graphics window displays the original section geometry (blue), the new modified section geometry. Below that tree additional graphics windows display tangent angle, relative angle and curvature distribution of the original and the modified profile for both suction and pressure profile. Automatic fitting The input profiles are automatically fitted with an optimum fit procedure based on a non-linear least squares optimisation routine. This means that each input profile (e.g. From DT6 of the BladeFile) will automatically transformed into the parametric description used within b2d. If the system is used in conjunction with the AutoBlading, no fitting is necessary, since the AutoBlading is 100% compatible in the geometry description (it uses the same parameter set).
41
Abbildung 24 Example of profile fitting
Stacking information The program b2d treats the section as a translation invariant geometry entity. This means that all geometry parameters are invariant under any form of axial or circumferential shift. The program does not manipulate stacking information of the original DT6 stream section profile on the BladeFile. This means that the stacking point of the original profile will not be moved for any changes of any of the given parameters. E.g. if the chord length is changed and the original blade was stacked on the leading edge, the resulting blade will have the same leading edge stack (=position in 3-D space), but a different shape of the trailing edge. Use of the scrollbar All parameters can be varied in real time via a single scrollbar. In order to modify a single section parameter, click with the mouse on the respective radio button to activate the slider for the parameter. With each parameter is associated a definition interval. The user is only allowed to modify this parameter within the given definition interval. In order to map the different definition intervals for the different parameters a linear transformation is applied, cf. Table 1.
Parameter
Definitionsbereich
Lineartransfor mation f=a*x+b: (a,b)
mui
number(14,7),
[-5..180]
(10,50)
42
muo bia boa radl radt stag cord curvlss curvlps curvtss curvtps tanglss tanglps tangtss tangtps stifflss stifflps stifftss stifftps
number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7), number(14,7)) ;
[-5..180] [-90..90] [-90..90] [0.0002..0.01 ] [0.0002..0.01 ] [-90..90] [0..0.5] [-1..1] [-1..1] [-1..1] [-1..1] [0..2] [0..2] [0..2] [0..2] [-1..1] [-1..1] [-1..1] [-1..1]
(10,50) (5,450) (5,450) (50000,0) (50000,0) (5,450) (1000,0) (500,500) (500,500) (500,500) (500,500) (500,0) (500,0) (500,0) (500,0) (500,500) (500,500) (500,500) (500,500)
The slider has an definition interval of [0,1000]. The definition interval of the parameter is mapped onto the definition interval of the slider via the linear transformation of table 1. This mapping can be modified by means of an coefficients file trafo.dat. If this file does not exist on the users current directory, it is created with default values valid for compressor cases. Turbine design requires a slightly different transformation file, available from the group directory.
Parametric database To enable concurrent simultaneous engineering with the parametric blade design system, a database connection for the design data is necessary. It was decided to use latest & best technology for this database. A web-based Oracle implementation of the parametric database was developed for that purpose. The data can be accessed via any Internet-Browser from any machine in the computer network. The solution developed here is based on Oracles latest technology and can be used as a prototype solution for the BRR/RR functional database.
43
Abbildung 25 Parametric database login window
Details of the specification of the database, the implementation and a user guide are contained in reference 24.
Abbildung 26 Query in parametric database
The database is part of the BRR EDM/PDM solution based on Windchill. It can be viewed as departmental data-mart and part of the companys data-warehouse.
44
Abbildung 27 The result of the query can be viewed following a hyperlink
The database is fully implemented in Internet-technology and allows easy navigation, search, collection and e-mail forwarding of information. Integration into the 3-D parametric design system The way the 2-D design tasks for individual sections and the 3-D geometry features are combined and implemented into user-friendly software determines the overall efficiency of the designers work. We can see, that for the streamwise and the radial direction different types of parameters have been used in order to implement the users desired options most realistically. We have now the task to integrate the different philosophies into a single tool with a reasonable user-interface. This interface supports the selective working on the stream section and radial stacking layer.
Conclusions
Blade design technology should no longer be viewed as the companies secret core technology. Parametric representations of profiles and blade stacking enable various optimisation procedures and establish bi-directional data exchange to standard CAD systems on the basis of spline curves, surfaces and solid models. Future trends will very likely show a systematic commercialisation of the standard techniques. Similar to the development in the FEM and CFD area, the basic technologies are relative mature now. Their application is mandatory to maintain design process efficiency and keep leading in the cost-, time and innovation competition world-wide.
References
/1/ AGARD Lecture Series No. 167: Blade design for Axial Turbomachines, 1989. /2/ Dunham, J. A parametric method of Turbine Blade Profile design, ASME Paper 74-GT119, 1974 /3/ Farin, G.: Curves and Surfaces for CAGD, Academic Press, Inc., 1993. /4/ Hausenblas, H.: Profilfamilien fr Turbinenteile von Gasturbinen. MTZ, Jahrgang 22, Heft 1, Januar 1991
45
/5/ Hourmouziadis, J., Buckl, F. and Bergmann, P.: The development of the profile boundary layer in a turbine environment, ASME Journal of Turbomachinery, Vol. 109, pp. 286-295, 1987. /6/ Hourmouziadis, J.: Sonderprobleme bei der Auslegung von Strmungsmaschinen, UNIBWM Institut fr Strahlantriebe, 1988. /7/ Korakianitis, T.: Design of Aerofoils and cascades of Aerofoils. AIAA Journal, Vol. 27, No.4, pp.455-461, 1989. /8/ Pierret, S., Van den Braembussche, R.A., Turbomachinery blade design using a navierStokes Solver and artificial neural Network, ASME Paper 98-GT-4, 1998. /9/ Korakianitis, T.: Prescribed-curvature-Distribution Airfoils for the preliminary geomertic design of axial turbomachinery cascades. ASME Journal of Turbomachinery, Vol.115, pp.325-333, 1993. /10/ Korakianitis, T. and Pantazoupulos, G.I.: Improved Turbine-Blade Design techniques using 4th order paramertic spline segments, Computer Aided design, Vol. 25, No.5, pp.289299,1993. /11/ Leonard, O. and Van den Braembussche, R.A.: Design method for subsonic and Transsonic Cascade with prescribed mach number distribution. ASME Journal of turbomachinery, Vol. 114, pp.553-560, 1992. /12/ Mortenson, M: Geometric Modelling, 1985. /13/ Newman, William M. and Sproull, Robert F.: Gr7undzge der interaktiven Computergrafik, Hamburg: McGraw-Hill Book Company, 1986. /14/ Pritchard, L.J.: An eleven parameter axial turbine airfoil Geometry Model. The American Society of mechanical Engineers, Bericht Nr. 85-GT-219, 1985. /15/ Pugathofer, Werner, Grafische Datenverarbeitung. Wien, New York: Springer 1985 /16/ Ross, Hubert R.: Entwicklung von Verfahren zur Erzeugung von Schaufelprofilen fr Axialturbinen. Technische Hochschule Darmstadt, 1988.
/17/ Haarmeyer, J., D. Steinebach, J. Anders, AutoBlading - Results of Pilot phase, TR78096(IR)ISS00, 22.01.1997. /18/ Anders, J., Revision of the b3d BladeDesign tool, short message, 5.12.1996. /19/ Wenger, U., Compressor Methods Brainstorming , ET-2/M0001/97 DW, 8.1.1997. /20/ Specification of the parametric 3-D stacking, SP004/97, BMW Rolls-Royce 1997. /21/ Parametric 3-D stacking manual, BMW Rolls-Royce 1997.
46
/22/ Specification of the Radial blade smoothing and interpolation tool, BMW Rolls-Royce 1998. /23/ Specification of the BladeOpt profile optimisation tool, BMW Rolls-Royce 1998. /24/ Specification of the parametric database, BMW Rolls-Royce 1998.
47
Contents
INTRODUCTION ................................................................................................................................................ 2 BLADE DESIGN & GAS TURBINE BUSINESS ........................................................................................................... 2 CAD DEVELOPMENTS.......................................................................................................................................... 3 AERODYNAMIC BLADE DESIGN PROCESS ............................................................................................................. 5 ASPECTS OF COMMERCIALISATION ...................................................................................................................... 8 THE DESIGN TASK............................................................................................................................................ 9 CONSTRAINTS AND DESIGN TARGETS................................................................................................................... 9 THE BRR PARAMETRIC BLADE DESIGN SYSTEM ................................................................................ 10 PARAMETRIC BLADE REPRESENTATION.............................................................................................................. 10 RULES-BASED DESIGN ....................................................................................................................................... 11 DERIVATION AND EDITING OF RULES - SOFTWARE AUTO_BLADER ................................................................. 12 RULES GENERATION PROCESS ............................................................................................................................ 13 LINEAR & STANDARD SUCTION SIDE CURVATURE ............................................................................................. 15 MATHEMATICS OF STANDARD CURVATURE DISTRIBUTION.......................................................... 15 SUCTION SIDE CURVATURE DISTRIBUTION, FORM FACTOR H AND LOSS LOOP .................................................. 16 RESULTS AND DISCUSSION ................................................................................................................................. 18 THICKNESS CORRECTION ................................................................................................................................... 18 DISTANCE FUNCTION ......................................................................................................................................... 19 DEFINING TARGET FUNCTIONS - GENERAL APPROACH ....................................................................................... 20 PARAMETRIC 3-D STACKING ............................................................................................................................. 20 RADIAL BLADE SMOOTHING AND INTERPOLATION ............................................................................................. 22 PROFILE OPTIMISATION ..................................................................................................................................... 23 DESIGN PROCESS............................................................................................................................................ 25 SOFTWARE QUALITY AND INTEGRATION ............................................................................................................ 26 TESTING AND INTRODUCTION AS PROJECT STANDARD ....................................................................................... 26 AERODYNAMIC STANDARD PROFILE PROPERTIES WITH THE PARAMETRIC BLADING SYSTEM............................................................................................................................................................... 27 CURVE AND PROFILE REPRESENTATION TYPES................................................................................. 28 CLONING ........................................................................................................................................................... 29 PROBLEM DESCRIPTION ..................................................................................................................................... 29 DEFINITION OF THE OFFSET MATRIX: ................................................................................................................. 30 CONCEPT FOR THE APPLICATION OF AUTOBLADING FOR DESIGN WORK ............................................................ 32 AUTOMATION .................................................................................................................................................... 33 CLONING ........................................................................................................................................................... 33 APPLICATION TO FAN, BOOSTER, HPC, HPT, LPT ........................................................................................... 35 BLADE DESIGN SYSTEM INTEGRATION.................................................................................................. 36 AERODYNAMIC CENTRIC VERSUS CAD-CENTRIC ......................................................................................... 36 WHAT CAN BE SAVED WITH THE NEW PROCESS? WHAT IMPROVEMENTS DO I HAVE? ........................................ 36 INTEGRATED BLADE DESIGN SYSTEM ................................................................................................................ 37 PARAMETRIC PROFILE EDITOR B2D ....................................................................................................... 40 FEATURES OF B2D .............................................................................................................................................. 40 PROGRAM DESCRIPTION..................................................................................................................................... 40 PARAMETRIC DATABASE.................................................................................................................................... 43 INTEGRATION INTO THE 3-D PARAMETRIC DESIGN SYSTEM ............................................................................... 45 CONCLUSIONS ................................................................................................................................................. 45
48
REFERENCES ................................................................................................................................................... 45 CONTENTS ........................................................................................................................................................ 48
49
This document was created with Win2PDF available at http://www.daneprairie.com. The unregistered version of Win2PDF is for evaluation or non-commercial use only.

A Parametric Blade Design System - Anders

Uploaded by

Copyright:

Available Formats

A Parametric Blade Design System - Anders

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Parametric Blade Design System - Anders

Uploaded by

Copyright:

Available Formats

What is the overall purpose of the parametric blade design system being presented?

What is the overall purpose of the parametric blade design system being presented?

What are some of the main advantages offered to users by the design system?

What are some of the main advantages offered to users by the design system?

A Parametric Blade Design System (Part I + II)

Jrgen M. Anders and Jrg Haarmeyer

Abbildung 1 BR715 Turbine Blades

Abbildung 2 The new Boing717 twinjet with BR715 engines

Abbildung 3 The dressed BR715 engine

Abbildung 4 Sketch of the BR715 engine

Aerodynamic blade design process

VKI LECTURE SERIE1.doc

Abbildung 5 Aerodynamic performance prediction for LP turbine blade section

VKI LECTURE SERIE1.doc

VKI LECTURE SERIE1.doc

VKI LECTURE SERIE1.doc

The design task

3-D stack (positioning of blade sections along in the radial direction)

object passage area, throat area

stagger camber position of maximum thickness

3-D blading tip leakage flow casing & hub treatment

VKI LECTURE SERIE1.doc

Parametric blade representation

VKI LECTURE SERIE1.doc

Abbildung 7 Blade parameter definition

VKI LECTURE SERIE1.doc

Physical Throughflow values

Abstract Throughflow parameters

Streamsection profile geometry

Abbildung 8 Rules-based AutoBlading process

4. High lift turbine

VKI LECTURE SERIE1.doc

Linear & standard suction side curvature

The definition of a standard curvature may be described as: (2)

(3) Now denoting: (4)

(1), (2) may be rewritten as: (5a) (5b) (5c) (5d) y1 ( x ) = y 2 ( x ) y2 ( x) = y3 ( x) 1 + y2 ( x ) 2 y3 ( x) = y4 ( x) y 4 ( x ) = w( x )

Suction side curvature distribution, form factor H and loss loop

Style 1 Style 2 Suction side Curvature

Suction side surface length x/c

original design redesign original design redesign

Modification of the 3rd knot point on the pressure side

VKI LECTURE SERIE1.doc

Abbildung 10 TecPlot display of stacked blade

Abbildung 11 GUI for the parametric 3-D stacking program

VKI LECTURE SERIE1.doc

Abbildung 12 Standard views for parametric stacking

VKI LECTURE SERIE1.doc

Abbildung 13 Radial Blade smooting & interpolation tool

Ref. 22 contains a description of the rit tool.

VKI LECTURE SERIE1.doc

A detailed description of BladeOpt is given in Ref. 23.

VKI LECTURE SERIE1.doc

Aerodynamic Standard profile properties with the Parametric Blading system

VKI LECTURE SERIE1.doc

read throughflow & annulus definition

av. Time required to perform action:

Run Rules to create initial profile

Check profile properties in b2d

Adjust major parameters if necessary (angles, chord, stagger)

Total less than 10 minutes File stream section

modified ProfileParameter p1,...,p20