Global Optimization of a Transonic Fan Blade Through AI-Enabled Active Subspaces

Lopez, Diego I.; Ghisu, Tiziano; Shahpar, Shahrokh

doi:10.1115/1.4052136

Cited by 15 publications

(6 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…GA) when a big number of data is available (Woldermariam and Hirpa, 2019). Applications have been carried out on centrifugal impellers (Ji et al, 2021), radial turbines (Luczynski et al, 2021) and fan blades (Lopez et al, 2021). ANNs applications in advanced DA models might trigger advantages as they enable the quick processing of large amounts of data and the engineering practice showed they could be effective when it is difficult to identify, a-priori, a suitable design optimisation model.…”

Section: Context and Relevant Backgroundmentioning

confidence: 99%

Turbomachinery Design: Checking Artificial Neural Networks Suitability for Design Automation

Batini,

Becattini,

Cascini

2023

Proc. Des. Soc.

View full text Add to dashboard Cite

This paper explores the suitability of Artificial Neural Networks (ANNs) as an enabler of Design Automation in the turbomachinery industry. Specifically, the paper provides 1) a preliminary estimation of the effectiveness of ANNs to define values for design variables of reciprocating compressors (RC) and 2) a comparison of ANNs performance with traditional and more computationally demanding methods like CFD. A tailored ANN trained on a dataset composed by 350+ Baker Hughes’ RC automatically assigns values to 8 geometrical variables belonging to multiple parts of the RC in order to satisfy two target conditions linked to their thermodynamic performance. The results highlight that the ANN-assigned parameters return an optimal solution for RC also when the target values do not belong to the training dataset. Their predictive capacity for RC thermodynamic performance, with respect to CFD, are comparable (i.e. less than 2% in terms of calculated absorbed power) and the approach enables a significant gain in terms of computational time (i.e. 2 minutes vs 10 hours). Future perspectives of this work may involve the integration of this tool in an advanced DA method to lead Design Engineers (DEs) during the whole design process.

show abstract

Section: Context and Relevant Backgroundmentioning

confidence: 99%

Turbomachinery Design: Checking Artificial Neural Networks Suitability for Design Automation

Batini,

Becattini,

Cascini

2023

Proc. Des. Soc.

View full text Add to dashboard Cite

show abstract

“…The design parameters are gathered in the design vector, x, and its dimensionality is denoted by the letter m, where m = 35. The lower and upper bounds that define the ranges of the design space are prescribed based on prior experience with the parametrization, which has been used on previous optimization studies providing sufficient geometry control [28,29].…”

Section: Design Space Explorationmentioning

confidence: 99%

“…To characterise the response of the system to the geometry changes introduced by this parametrization and perform exploratory analyses, this work makes use of AI-Enabled Active Subspaces. A brief description of the methodology is presented here, for a full treatment of the topic the interested reader is referred to [28] and references therein.…”

Section: Ai-enabled Active Subspacesmentioning

confidence: 99%

Extending Highly Loaded Axial Fan Operability Range Through Novel Blade Design

Lopez

Ghisu

Kipouros

et al. 2022

Journal of Turbomachinery

View full text Add to dashboard Cite

The tip clearance size has historically been considered to be the main factor affecting stability range in axial fan and compressors. This paper reveals that the stall characteristics are defined by the axial momentum flux of the tip leakage flow and that tip clearance is primarily a strong driver for this metric. A bespoke methodology for carefully tailoing the axial momentum via three-dimensional design is presented, which enables a higher degree of control over the stability range for cases where the tip clearance responds to other considerations and cannot be defined for this purpose. The effect of the axial momentum on efficiency is also addressed and the trade-off between operability range and design point performance derived. The results show that that the conditions for optimal stability differ from those for optimal efficiency and that control over the axial momentum enables tuning the design for a desired exchange. Numerical simulations have been employed to drive the analysis through a high-fidelity computational model whose behavior is supported by rich set of experimental data. Contrary to current belief, results further indicate that an accurate characterization of stall, including onset mechanism, can be achieved through steady-state simulations, minimising the need for expensive time-accurate computations during the design phase.

show abstract

“…The ANNs were constructed for the figures of merit using information obtained from primal and adjoint solvers [25]. Figure 36 presents a trade-off between the 2 objectives, demonstrating the importance of a robust design approach [29], when compared to the traditional deterministic optimum (DGO) [30].…”

Section: Figure 35 -Vital Fan Test-case and Parameterization Of Its S...mentioning

confidence: 99%

Multidisciplinary Adjoint-based Optimizations in the MADELEINE Project: Overview and Main Results

Méheut¹

2021

Aiaa Aviation 2021 Forum

View full text Add to dashboard Cite

This paper provides an overview of the European MADELEINE (Multidisciplinary ADjoint-based Enablers for LargE-scale Industrial desigN in aEronautics) project dedicated to multi-disciplinary adjoint-based optimization in an industrial framework. MADELEINE aims at increasing the technology readiness level (TRL) and demonstrating the benefits of high-fidelity (HiFi), adjoint-based multidisciplinary optimization (MDO) to address the objectives of industry in terms of competitiveness and climate impact. The project is decomposed into two main subparts. The first one focuses on improvements of capability, efficiency and usability of adjoint-based MDO processes used in the European aeronautical industry. The second one aims at applying these methods and tools on representative test cases defined by the industrial partners. In this context, aero-structure, aero-acoustics and aerothermal problems are considered in order to optimize aircraft and engine components (wing, propeller, fan or turbine). The paper presents the overall structure of the project and the main results obtained after 3 years of collaboration.disciplinary experts and a new insight for design space exploration and the definition of better multi-disciplinary compromises.To reach this ambitious objective, the concept applied in MADELEINE focuses on synergies between enablers and demonstrators. Enablers are methods and tools that are requested to apply adjoint-based MDO approach; demonstrators are the applications of the final MDO processes on test cases representative of multi-physics industrial design problems. A. DemonstratorsIn MADELEINE, 5 relevant industrial Test Cases (TC) have been identified. TC1 and TC5 are of interest for aircraft manufacturers, TC2 to TC5 for engine manufacturers. These TCs are conducted using tools and disciplinary solvers used daily within the companies involved in the project. They are carried out by industrial partners themselves or by partners that have access to tools developed within the industrial framework. The rationale behind the TC selection is described hereafter.

show abstract

Global Optimization of a Transonic Fan Blade Through AI-Enabled Active Subspaces

Cited by 15 publications

References 27 publications

Turbomachinery Design: Checking Artificial Neural Networks Suitability for Design Automation

Turbomachinery Design: Checking Artificial Neural Networks Suitability for Design Automation

Extending Highly Loaded Axial Fan Operability Range Through Novel Blade Design

Multidisciplinary Adjoint-based Optimizations in the MADELEINE Project: Overview and Main Results

Contact Info

Product

Resources

About