Modelling of a helicopter-based rocket launching system

Abstract

This study describes a conceptual design of a weapon system that employs a Model Predictive Control Loop in conjunction with an Array Processor to deliver standard Wrap Around Fin rockets with high precision from a helicopter. One of the requirements of such a system is real time estimation of the main rotor downwash. The design thread created by this requirement is followed by this study, starting with the operational requirement at system level and ending with porting of code into Field Programmable Gate Arrays at component level. To prove that real time estimation of the main rotor downwash of a Helicopter is feasible a conceptual design of a Precision Rocket Launcher with integral Array Processor based on Field Programmable Gate Arrays is developed. An algorithm compatable with the Array Processor and suitable for solving a taylored version of the Navier Stokes equation in real time is developed. Two versions of this algorithm called the Concurrent Model and Pipeline Model are subsequently prepared for porting to Field Programmable Gate Arrays. The Concurrent Model is a direct implementation of the downwash algorithm in 20 bit fixed point notation. Due to its simplicity this model is capable of solving the taylored version of the Navier Stokes equation at an iteration rate of 63,4 kHz when ported to Intel device GX2800 of the STRATIX 10 family. The Concurrent Model is shown to have an unfavourable Multiplier to Fabric ratio which results in such a large component count that packaging and cooling them within the confines of the Precision Rocket Launcher is unfeasible. The Pipeline Model implements the downwash algorithm with seven pseudo-processors arranged in a pipeline configuration. It is implemented in single precision floating point notation and is capable of solving the Taylored Navier Stokes equation at a rate of 10 kHz when ported to Intel device GX2800 of the STRATIX 10 family. The Pipeline Model is shown to have a favourable Multiplier to Fabric ratio which results in a reasonable component count, making packaging and cooling them within the confines of the Precision Rocket Launcher feasible. The Pipeline Model yields a computational domain of 525x525x525 cells if it is implemented on an Array Processor based on Intel device GX2800 of the STRATIX 10 family. Throughput of this Array Processor is 0,601 petaflops (single precision) with RAM data flow rate of 0,261 petabytes/sec. The downwash models, as developed in this study, form standard building blocks which can also be used in other real time applications. The models use a relatively simple interface with any combination of system elements such as rotors, wings, fuselage, ground, gas sources/sinks and other obstructions which makes real time calculation of airflow in many configurations and conditions possible.