On the confidence bounds of Gaussian process NARX models and their higher-order frequency response functions
نویسنده
چکیده
One of the most powerful and versatile system identification frameworks of the last three decades is the NARMAX/NARX approach, which is based on a nonlinear discrete-time representation. Recent advances in machine learning have motivated new functional forms for the NARX model, including one based on Gaussian processes (GPs), which is the focus of this paper. Because of their nonparametric form, NARX models can only provide physical insight through their frequency-domain connection to Higher-order Frequency Response Functions (HFRFS). Because of the desirable properties of the GP-NARX form (no structure detection needed, natural confidence intervals), the analytical derivation of the HFRFs for the model is presented here for the first time. Furthermore, an algorithm for propagating uncertainty from the GP into the HFRF estimates is presented. A valuable by-product of the latter algorithm is a new test for nonlinearity, capable of detecting the presence of odd and even system nonlinearities. The new results are illustrated via two case studies; the first is based on simulation of an asymmetric Duffing oscillator. The second case study presents a validation of the new theory in the area of wave force prediction on offshore structures. This problem is one that has been considered by some of the authors before; the current paper takes the opportunity to highlight and correct a number of weaknesses of the original study in the light of modern best practice in machine learning. 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
منابع مشابه
Process Improvement of Experimental Measurements Using D-optimal Models
In this paper, the application of D-optimal models, as an alternative to response surface models (RS models) for design of experiment (DOE) was examined. Two D-optimal models for tilt-rotors in the wind tunnel experiment, as a form of quadratic functions, were generated based on a chosen optimality criterion. This optimality criterion was used to generate the optimized sampled points in the des...
متن کاملFrequency-Dependent Magnitude Bounds of the Generalized Frequency Response Functions for NARX Model
New magnitude bounds of the frequency response functions for the Nonlinear AutoRegressive model with eXogenous input (NARX) are investigated by exploiting the symmetry of the nth-order generalized frequency response function (GFRF) in its n frequency variables. The new magnitude bound of the nth-order symmetric GFRF is frequency-dependent, and is a polynomial function of the magnitude of the fi...
متن کاملPrediction of Above-elbow Motions in Amputees, based on Electromyographic(EMG) Signals, Using Nonlinear Autoregressive Exogenous (NARX) Model
Introduction In order to improve the quality of life of amputees, biomechatronic researchers and biomedical engineers have been trying to use a combination of various techniques to provide suitable rehabilitation systems. Diverse biomedical signals, acquired from a specialized organ or cell system, e.g., the nervous system, are the driving force for the whole system. Electromyography(EMG), as a...
متن کاملNew bound characteristics of NARX model in the frequency domain
New results about the bound characteristics of both the generalized frequency response functions (GFRFs) and the output frequency response for the NARX (Nonlinear AutoRegressive model with eXogenous input) model are established. It is shown that the magnitudes of the GFRFs and the system output spectrum can all be bounded by a polynomial function of the magnitude bound of the first order GFRF, ...
متن کاملHigher-order Frequency Response Functions for Nonlinear Systems from Neural Networks
algorithm [3]. A method is presented for calculating the Higher-order In this paper, the method of Way and Green [2] will Frequency Response Functions (HFRFs) of NARX neural be extended to the frequency-domain. By harmonically networks. HFRFs are the Fourier transforms of Volterra probing the network equation, the HFRFs of the network kernels and can be viewed as multi-dimensional equivawill be...
متن کامل