Evolutionary computation methods for helicopter loads estimation

2012 IEEE Congress on Evolutionary Computation

2012

Self Cite

Abstract-The development of accurate load spectra for helicopters is necessary for life cycle management and life extension efforts. This paper explores continued efforts to utilize evolutionary computation (EC) methods and machine learning techniques to estimate several helicopter dynamic loads. Estimates for the main rotor normal bending (MRNBX) on the Australian Black Hawk helicopter were generated from an input set that included thirty standard flight state and control system parameters under several flight conditions (full speed forward level flight, rolling left pullout at 1.5g, and steady 45• left turn at full speed). Multiobjective genetic algorithms (MOGA) used in combination with the Gamma test found reduced subsets of predictor variables with modeling potential. These subsets were used to estimate MRNBX using Cartesian genetic programming and neural network models trained by deterministic and evolutionary computation techniques, including particle swarm optimization (PSO), differential evolution (DE), and MOGA. PSO and DE were used alone or in combination with deterministic methods. Different error measures were explored including a fuzzy-based asymmetric error function. EC techniques played an important role in both the exploratory and modeling phase of the investigation. The results of this work show that the addition of EC techniques in the modeling stage generated more accurate and correlated models than could be obtained using only deterministic optimization.

Section: A Full Mask Vs Reduced Mask Resultsmentioning

confidence: 44%

“…The most promising subsets were then selected and used as a base for model search. The data exploration followed in this work is described in detail in [1].…”

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Use of evolutionary computation techniques for exploration and prediction of helicopter loads

Cheung

2012 IEEE Congress on Evolutionary Computation

2012

Self Cite

“…This analysis found that 5 time lags were necessary and therefore the predictor set consisted of the 30 FSCS parameters and their 5 time lags for a total of 180 predictors [1], [2]. The Gamma test then steered the evolutionary process used for data exploration.…”

Section: Methodsmentioning

confidence: 99%

“…Preliminary work exploring the use of various computational intelligence techniques to assist in estimating helicopter loads showed that reasonably accurate and correlated predictions for the main rotor normal bending could be obtained for the flight condition forward level flight at full speed using only a reduced set of flight state and control system parameters [1], [2]. However, the main weakness of these previous efforts has been the tendency to underpredict the target signal.…”

Section: Introductionmentioning

confidence: 99%

Towards conservative helicopter loads prediction using computational intelligence techniques

The 2012 International Joint Conference on Neural Networks (IJCNN)

Cheung

2012

Self Cite

Vous avez des questions? Nous pouvons vous aider. Pour communiquer directement avec un auteur, consultez la première page de la revue dans laquelle son article a été publié afin de trouver ses coordonnées. Si vous n'arrivez pas à les repérer, communiquez avec nous à PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca. Questions? Contact the NRC Publications Archive team atPublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca. If you wish to email the authors directly, please see the first page of the publication for their contact information. NRC Publications Archive Archives des publications du CNRCThis publication could be one of several versions: author's original, accepted manuscript or the publisher's version. / La version de cette publication peut être l'une des suivantes : la version prépublication de l'auteur, la version acceptée du manuscrit ou la version de l'éditeur. NRC Publications Record / Notice d'Archives des publications de CNRC:http://nparc.cisti-icist.nrc-cnrc.gc.ca/eng/view/object/?id=10db696b-8a8c-4547-b93c-48837b7cb311 http://nparc.cisti-icist.nrc-cnrc.gc.ca/fra/voir/objet/?id=10db696b-8a8c-4547-b93c-48837b7cb311Towards conservative helicopter loads prediction using computational intelligence techniques Abstract-Airframe structural integrity assessment is a major activity for all helicopter operators. The accurate estimation of component loads is an important element in life cycle management and life extension efforts. This paper explores continued efforts to utilize a wide variety of computational intelligence techniques to estimate some of these helicopter dynamic loads. Estimates for two main rotor sensors (main rotor normal bending and pushrod axial load) on the Australian Black Hawk helicopter were generated from an input set that consisted of thirty standard flight state and control system parameters. These estimates were produced for two flight conditions: full speed forward level flight and left rolling pullout at 1.5g. Two sampling schemes were attempted, specifically k-leaders sampling and a biased sampling scheme. Ensembles were constructed from the top performing models that used conjugate gradient, LevenbergMarquardt (LM), extreme learning machines, and particle swarm optimization (PSO) as the learning method. Hybrid and memetic approaches combining the deterministic optimization and evolutionary computation techniques were also explored. The results of this work show that using a biased sampling scheme significantly improved the predictions, particularly at the peak values of the target signal. Hybrid models using PSO and LM learning provided accurate and correlated predictions for the main rotor loads in both flight conditions.

Characterization of Quantum Derived Electronic Properties of Molecules: A Computational Intelligence Approach

Artificial Neural Networks and Machine Learning – ICANN 2019: Workshop and Special Sessions

Tchagang

2019

The availability of BIG molecular databases derived from quantum mechanics computations represent an opportunity for computational intelligence practitioners to develop new tools with same accuracy but much lower computational complexity compared to the costly Schrödinger equation. In this study, unsupervised and supervised learning methods are applied to investigate the internal structure of the data and to learn the mapping between the atomic coordinates of molecules and their properties. Low dimensional spaces revealed a well defined clustering structure as defined by the measures used for comparing molecules based their atom distributions and chemical composition. Supervised learning techniques were applied on the original predictor variables, as well as on a subset of selected variables found using evolutionary algorithms guided by residual variance analysis (Gamma Test). Black and white box modeling approaches were used (random forests, neural networks and model trees and adaptive regression respectively). All of them delivered good performance, error and correlation-wise, with neural networks producing the best results. In particular white box techniques obtained explicit functional dependencies, some of them achieving considerably reduction of the feature set and expressed as simple models. Keywords: Computational intelligence • Quantum mechanics • Molecules • 3D visualization • Random forests • Neural networks • Model trees • Multivariate adaptive regression • Black box models • White box models