Inverse Design based Turbomachinery Design Optimization Versus Conventional


Automatic optimization is often employed in turbomachinery design optimization process to automate exploration of the design space in a systematic way.

What is 3D Inverse Design?

Unlike traditional "direct” design methods that rely on trial and error between a given geometry and flow field predicted by CFD codes, ADT's 3D Inverse Design method starts by identifying what you want to do to the fluid flow in terms of 3D pressure field and mathematically derives the optimal geometry to achieve that outcome. This significantly reduces the time taken for each design.

Automatic optimization provides a number of benefits in terms of covering large areas of the design space, when compared to manual operations, however its potential effectiveness heavily relies on the selected blade parameterisation method.
In the direct, or conventional, design approach the main variable is the blade geometry itself, which is generally parameterised in terms of blade surface or camber line plus thickness at a number of different 2D cross sections. These sections are then stacked together to form a 3D shape.
Direct parameterisation of the blade geometry provides a number of drawbacks in the optimization process which make it more expensive, and less general, when compared to parameterisation of the blade loading as main variable, as done in the 3D inverse design approach.
Main differences between the two parameterisation methods

Direct Design

Inverse Design

Parameterization of the blade geometry As much as 30 to 100 parameters are required to describe a fully 3D blade.

Reducing the number of parameters will highly limit the covered design space.


A maximum of 8 design parameters are required for a fully 3D blade parameterization via the blade loading and as low as 4 parameters for the meridional shape.

This number already covers a significantly larger portion of the design space compared to direct design.


Blade Surface Smoothness

The generated geometries can present non-smooth surfaces in spanwise direction requiring some checking or post-processing following the geometry generation process.


The blade geometry is smooth.

The code automatically re-creates the geometry based on imposed meridional shape and blade thickness.


Specific Work of the Components

There is no guarantee that the component specific work (Head / Pressure Ratio) at the correct mass flow rate is satisfied by the generated geometry. This needs to be imposed as a constraint in the optimization.


Specific work at the required mass flow rate is an input to the code and is therefore automatically satisfied by every generated design.


Main Flow Features of the Components

To obtain performance parameters each geometry has to be analysed by CFD or other analysis code to extract basic data such as surface pressure, velocities or loss, efficiency etc.



Within 10-15 seconds, the inverse design provides the geometry along with a fully 3D inviscid solution at the design point; basic requirements can be implemented directly in the optimizer without going into CFD.

A number of correlations between surface data and flow phenomena such as Diffusion, Endwall Losses, Profile Losses, Tip Clearance and Secondary Flows have been developed to allow control of these parameters as objectives or constraints directly in the optimization process.


Simpler shape of the Objective function

The relationship between the geometry and performance parameters such as efficiency, blockage, cavitation etc is highly non-linear.


There is more direct relationship between blade loading/pressure distribution and performance parameters such as loss or cavitation. This makes shape of the objective function relating performance parameters to design parameters simpler and hence minima or maxima can be more quickly found.


Generality of Optimized Results

The result of the optimization is a blade geometry which is difficult to generalize for a different design condition or even size.


Results of the optimization is the blade loading which has generality.

The optimal blade loading for secondary flow suppression in centrifugal impellers, cavitation control in mixed-flow pumps or high efficiency and low noise fans has been found to be generally applicable across different cases.

Optimization process can be used as a KNOW-HOW generator.


Computational Cost

High number of design parameters, no direct control over the design specifications and requirements to run CFD to extract even basic data makes this approach very expensive for a day-to-day design process.


Lower number of design parameters, guaranteed satisfaction of design requirements and better correlation between design parameters and design objectives makes it a practical approach to generate design know-how and apply it to day-to-day design processes.



Coupling turbomachinery automatic optimization with 3D inverse design shows significant benefits in generating an easy to store and transfer base of design know-how. Experienced designers can also use this approach on a daily basis to quickly visualize the trade-off between contrasting design objectives (ie. efficiency vs. noise in fans, efficiency vs. cavitation in pumps or efficiency vs. stresses in compressors) and position themselves in the range required by the customer or project specifications, considerably speeding up the turbomachinery design optimization process.

“We are always constrained by development time. So, we were looking for ways to cut development time in addition to achieving the efficiency goals. TURBOdesign Suite was the only commercially available 3D Inverse Design software. We decided to use it.”

Companies across the world use TURBOdesign Suite for their turbomachinery design. 



Want to Know More?


Schedule A Personalized Demo