Flow distortion reduction by blade cascades, Part 1: Adjoint optimization

The engine inlet ducts of modern combat aircraft often adopt a serpentine geometry to reduce thermal and radar signature, limit object ingestion and/or improve combined aerodynamic and engine performance. An emerging challenge in such flow conduits is flow separation and subsequent formation of vortex pairs that enter the aerodynamic interface plane (AIP) and affect engine performance. Addressing this challenge has prompted development activity that aims to improve AIP flow, typically by installing internal vanes that delay separation or modify the vortex path and rollup dynamics. The present article summarizes an effort combining computational and experimental methods to develop a vane assembly installed into the duct to condition the AIP flow and, simultaneously, serve as a support structure for an integrated heat exchanger. Numerical analysis utilizes a Cartesian grid-based flow analysis with an adjoint solver to compute design parameter sensitivities. The vane geometries are defined using conventional design variables (twist, chord, ¼ chord line, and airfoil section) whose spanwise variations are represented with B-splines and associated control points that comprise the design parameter set. The design parameters are optimized to minimize cross flow velocities at the AIP using a steepest descent algorithm and computed parameter sensitivities. These optimizations are repeated for multiple vane installations placed at different positions within the duct, thereby identifying both the most effective vane locations and associated geometry. The optimized designs achieved up to 30% reductions in maximum swirl angle at the AIP. A similar procedure was also performed to design a subscale test of the flow control device carried out at Pennsylvania State University (PSU). This test, detailed in a companion paper, PART 2 of this work, was used to confirm numerical flow predictions and demonstrated the ability to impart specified vortical flow distortions and then remove them using a second set of optimized vanes. The current article outlines the numerical procedures supporting adjoint-based vane design, describes the cost functions and design parameters employed for flow conditioning, reviews the approach adopted for vane siting, presents and explains the geometries obtained with the design process, and summarizes the use of experiment test data to corroborate computational predictions.