China
Africa
Explore chapters and articles related to this topic
Research on subsynchronous oscillation modal identification based on the FastICA–ARMA algorithm
Published in Rodolfo Dufo-López, Jaroslaw Krzywanski, Jai Singh, Emerging Developments in the Power and Energy Industry, 2019
Commonly used methods for modal parameter identification are the Prony analysis method, random subspace method, and autoregressive moving average (ARMA) analysis method (Liu et al. 2014). The Prony analysis method is very sensitive to noise. When the noise is large, it is difficult to extract the required signal matrix, and the modal parameters of the system cannot be accurately identified (Zhang & Jin 2016). The random subspace method is suitable for parameter identification of linear structural responses under stationary random excitation, but this method is prone to false modal or modal loss, and the calculation amount is relatively large, which is generally not suitable for large and complex structures (Liu 2008). The ARMA model analysis method is based on measured data to process dynamic random data through a parametric model. A linear dynamic model is widely used in system identification, prediction, and control, but the method has limited anti-noise ability (Vu et al. 2011). In He (2012), the ARMA model is applied to the modal parameter identification in time-varying systems. The effectiveness of the method is demonstrated by analyzing this example. The FastICA algorithm is a commonly used algorithm in the independent component analysis method, which can quickly extract independent components from the mixed signal. In (Zhao et al. 2018), the combination of FastICA and matrix beam algorithm is used to identify the modal parameters, which effectively improves the accuracy of the identification results of the matrix beam algorithm. The effectiveness of the algorithm is verified by an example simulation model.
Intelligent machine fault diagnosis with effective denoising using EEMD-ICA- FuzzyEn and CNN
Published in International Journal of Production Research, 2022
Hanting Zhou, Wenhe Chen, Changqing Shen, Longsheng Cheng, Min Xia
The independent component analysis (ICA) algorithm is introduced to separate the residual noise and retain the source signal from IMFs, to maximise the retention of the effective information carried by the vibration signal. The theoretical foundation of ICA is described as follows. Firstly, denoting the random observed vector whose elements are mixtures of independent elements of a random sources vector : where represents an unknown mixing matrix. The goal of ICA is to estimate the demixing matrix so that the reconstructed data matrix can be given: where the unmixing matrix is the inverse of , and is the best possible approximation of . Based on the non-Gaussianity perspective, the FastICA algorithm presents a computationally efficient, fast, and popular ICA package.
Power line interference noise elimination method based on independent component analysis in wavelet domain for magnetotelluric signal
Published in Geosystem Engineering, 2018
FastICA is an efficient and popular algorithm for ICA invented by Aapo Hyvärinen at Helsinki University of Technology (Hyvärinen, 1999; Hyvärinen & Oja, 2000). Like most ICA algorithms, through a fixed-point iteration scheme, FastICA seek an orthogonal rotation of pre-whitened data that maximizes non-Gaussianity of the rotated components. Here, Non-Gaussianity serves as a proxy for statistical independence. FastICA can also be alternatively derived as an approximative Newton iteration.
Monitoring of the UHPFRC strengthened Chillon viaduct under environmental and operational variability
Published in Structure and Infrastructure Engineering, 2020
Henar Martín-Sanz, Konstantinos Tatsis, Vasilis K. Dertimanis, Luis David Avendaño-Valencia, Eugen Brühwiler, Eleni Chatzi
PCE allows the representation of the dependence of a stochastic response on a set of input variables described by a given distribution, via a suitable projection basis. Using this tool, the evolution of uncertainty in dynamical systems, such as those pertaining to the SHM case, may be predicted. Consider a system f influenced by several independent variables with N denoting the number of input variables. For the case study of the Chillon viaducts, temperature and humidity will represent the input, with a joint probability function (PDF) while the output with finite variance, will be also random and corresponds to the structural (modal) frequencies. This output can be expressed as follows: with designating the polynomial basis functions orthonormal to d the vector of multi-indices related to the multivariate polynomial basis and θj the deterministic coefficients of projection, which are unknown. For a more detailed introduction to the method, the reader is referred to Baltman and Sudret (2010) and Spirodonakos, Chatzi, and Sudret (2016b). Since this approach requires independent inputs, a pre-processing step must be included in the analysis to de-correlate temperature and humidity. To this end, Independent Component Analysis (ICA) is applied, following the FastICA algorithm described by Hyvärinen and Oja (2000). Once this steps is completed, the estimated natural frequencies are divided into training and validation sets, in order to derive a model (surrogate) of the relationship between the Quantities of Interest (QoI), in this case natural frequencies, and the inputs (temperature, humidity).