Inverse modelling of multi-objective thermodynamically optimized turbojet engines using the Group Method of Data Handling-type neural networks and evolutionary algorithms

Inverse modelling of multi-objective thermodynamically optimized turbojet engines using the Group Method of Data Handling-type neural networks and evolutionary algorithms

0.00 Avg rating0 Votes
Article ID: iaor20062138
Country: United Kingdom
Volume: 37
Issue: 5
Start Page Number: 437
End Page Number: 462
Publication Date: Jul 2005
Journal: Engineering Optimization
Authors: , , , ,
Keywords: optimization, programming: multiple criteria, heuristics, neural networks
Abstract:

A novel approach is presented in this article for obtaining inverse mapping of thermodynamically Pareto-optimized ideal turbojet engines using group method of data handling (GMDH)-type neural networks and evolutionary algorithms (EAs). EAs are used in two different aspects. Firstly, multi-objective EAs (non-dominated sorting genetic algorithm-II) with a new diversity preserving mechanism are used for Pareto-based optimization of the thermodynamic cycle of ideal turbojet engines considering four important conflicting thermodynamic objectives, namely, specific thrust (ST), specific fuel consumption (SFC), propulsive efficiency (ηp), and thermal efficiency (ηt). The best obtained Pareto front, as a result, is a data table representing data pairs of non-dominated vectors of design variables, which are Mach number and pressure ratio, and the corresponding four objective functions. Secondly, EAs and singular value decomposition are deployed simultaneously for optimal design of both connectivity configuration and the values of coefficients, respectively, involved in GMDH-type neural networks which are used for the inverse modelling of the input–output data table obtained as the best Pareto front. Therefore, two different polynomial relations among the four thermo-mechanical objectives and both Mach number and pressure ratio are searched using that Pareto front. The results obtained in this paper are very promising and show that such important relationships may exist and could be discovered using both multi-objective EAs and evolutionarily designed GMDH-type neural networks.

Reviews

Required fields are marked *. Your email address will not be published.