Nonlinear Structural Behaviour Identification using Digital Recurrent Neural Networks

Autor: Vesna RANKOVIĆ, Nenad GRUJOVIĆ, Dejan DIVAC, Nikola MILIVOJEVIĆ, Radovan SLAVKOVIĆ
Jazyk: angličtina
Rok vydání: 2012
Předmět:
Zdroj: Strojarstvo : časopis za teoriju i praksu u strojarstvu
Volume 54
Issue 3
ISSN: 0562-1887
Popis: Dynamical systems contain nonlinear relations which are difficult to model with conventional techniques. Hence, efficient nonlinear models are needed for system analysis, optimization, simulation and diagnosis of nonlinear systems. In recent years, computational-intelligence techniques such as neural networks, fuzzy logic and combined neuro-fuzzy systems algorithms have become very effective tools in the field of structural identification. The problem of the identification consists of choosing an identification model and adjusting the parameters in an way that the response of the model approximates the response of the real system to the same input. This paper investigates the identification of a nonlinear system by Digital Recurrent Neural Network (DRNN). A dynamic backpropagation algorithm is employed to adapt weights and biases of the DRNN. Mathematical model based on experimental data is developed. Results of simulations show that the application of the DRN for the identification of complex nonlinear structural behaviour gives satisfactory results.
Dinamički sustavi sadrže nelinearne veze koje se teško modeliraju konvencionalnim tehnikama. Nelinearni modeli su neophodni za analizu sustava, optimizaciju, simulaciju i dijagnostiku nelinearnih sustava. Prethodnih godina, tehnike računalne inteligencije kao što su neuralne mreže, fuzzy logika i kombinirani neuro-fuzzy sustavi postaju efikasni alati u identifikaciji nelinearnih objekata. Problem identifikacije se sastoji od izbora identifikacijskog modela i prilagođavanja parametara tako da odziv modela aproksimira odziv realnog sustava za isti ulaz.Ovaj rad proučava identifikaciju nelinearnih sustava pomoću digitalne povratne neuronske mreže. Dinamički algoritam s propagacijom pogreške unazad se primjenjuje za adaptaciju težina i pragova osjetljivosti DRNN. Matematički model se razvija na bazi eksperimentalnih podataka. Rezultati simulacija pokazuju da primjena DRN u identifikaciji kompleksnog nelinearnog strukturnog ponašanja daje zadovoljavajuće rezultate.
Databáze: OpenAIRE
Externí odkaz: