Approaches to robust process identification: A review and tutorial of probabilistic methods

Kodamana, Hariprasad; Huang, Biao; Ranjan, Rishik; Zhao, Yujia; Tan, Ruomu; Sammaknejad, Nima

doi:10.1016/j.jprocont.2018.02.011

Cited by 52 publications

(20 citation statements)

References 94 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In relation (12) the contamination degree  is unknown. It follows from simulations that good performances of the identification algorithm are provided for [2,4]…”

Section: Generalized Identification Criterionmentioning

confidence: 99%

See 1 more Smart Citation

Robust Recursive Identification of Hammerstein Models Based on Weiszfald Algorithm

Filipović

2020

FU Aut Cont Rob

View full text Add to dashboard Cite

The Hammerstein models can accurately describe a wide variety of nonlinear systems (chemical process, power electronics, electrical drives, sticky control valves). Algorithms of identification depend, among other, on the assumption about the nature of stochastic disturbance. Practical research shows that disturbances, owing the presence of outliers, have a non-Gaussian distribution. In such case it is a common practice to use the robust statistics. In the paper, by analysis of the least favourable probability density, it is shown that the robust (Huber`s) estimation criterion can be presented as a sum of non-overlapping - norm and - norm criteria. By using a Weiszfald algorithm - norm criterion is converted to - norm criterion. So, the weighted - norm criterion is obtained for the identification. The main contributions of the paper are: (i) Presentation of the Huber`s criterion as a sum of - norm and - norm criteria; (ii) Using the Weiszfald algorithm – norm criterion is converted to a weighted - norm criterion; (iii) Weighted extended least squares in which robustness is included through weighting coefficients are derived for NARMAX (nonlinear autoregressive moving average with exogenous variable) . The illustration of the behaviour of the proposed algorithm is presented through simulations.

show abstract

“…In relation (12) the contamination degree  is unknown. It follows from simulations that good performances of the identification algorithm are provided for [2,4]…”

Section: Generalized Identification Criterionmentioning

confidence: 99%

“…The actual research in the field of identification devotes considerable attention to robust identification methods [4]. The robust identification of multi-input multi-output models using the stochastic approximation is considered in [5] and the robust adaptive prediction is presented in [6].…”

Section: Introductionmentioning

confidence: 99%

Robust Recursive Identification of Hammerstein Models Based on Weiszfald Algorithm

Filipović

2020

FU Aut Cont Rob

View full text Add to dashboard Cite

show abstract

“…Apart from industrial applications, many other areas such as economic data, biology and life sciences also require accurate identification of the underlying dynamical system model [7]. For most of these phenomenon, especially industrial, it is really difficult to synthesize the underlying models based on first-principles based learning due to their enormous complexity [8]. In such cases, the so called data driven black-box and grey-box modeling approaches have been emerged as promising and viable alternatives [9], [10].…”

Section: Introductionmentioning

confidence: 99%

Approximate GP Inference for Nonlinear Dynamical System Identification Using Data-Driven Basis Set

2020

View full text Add to dashboard Cite

We address the problem of nonlinear dynamical system identification in state space formulation using an approximate Gaussian process (GP) regression framework where the basis for GP are learned from data. Approximate GP inference is used to address the high computational cost of exact GP frameworks, and is based on basis function expansion concept. We address the design of these basis functions using data and eigenvalue decomposition framework. The need for such design arises in practical applications where the collected data has outliers, discontinuities and other non-stationary characteristics which limit not only the applicability but also the performance of traditional GP framework with inherent stationary assumptions. Using the available data set, a kernel eigenvalue problem is formulated which is solved using Monte Carlo techniques to construct basis functions. Then, a low-rank approximation to the exact GP is obtained by an approach similar to kernel principle component analysis (k-PCA). The coefficients in the basis function expansion and other unknown parameters are learned from data using sequential Monte Carlo technique. The proposed GP framework for dynamical system identification is tested and validated against fixed basis framework using simulations and experimental data. INDEX TERMS Gaussian processes regression, non-stationary kernel, kernel PCA, sequential Monte Carlo, kernel eigenvalue problem.

show abstract

“…The robustness of identification models can then be improved. The robust modeling method has been widely applied in pattern recognition [19], metallurgy [20], and other industrial fields.…”

Section: Introductionmentioning

confidence: 99%

Modeling and Synchronous Optimization of Pump Turbine Governing System Using Sparse Robust Least Squares Support Vector Machine and Hybrid Backtracking Search Algorithm

Zhang

Tian

et al. 2018

Energies

View full text Add to dashboard Cite

In view of the complex and changeable operating environment of pumped storage power stations and the noise and outliers in the modeling data, this study proposes a sparse robust least squares support vector machine (LSSVM) model based on the hybrid backtracking search algorithm for the model identification of a pumped turbine governing system. By introducing the maximum linearly independent set, the sparsity of the support vectors of the LSSVM model are realized, and the complexity is reduced. The robustness of the identification model to noise and outliers is enhanced using the weighted function based on improved normal distribution. In order to further improve the accuracy and generalization performance of the sparse robust LSSVM identification model, the model input variables, the kernel parameters, and the regularization parameters are optimized synchronously using a binary-real coded backtracking search algorithm. Experiments on two benchmark problems and a real-world application of a pumped turbine governing system in a pumped storage power station in China show that the proposed sparse robust LSSVM model optimized by the hybrid backtracking search algorithm can not only obtain higher identification accuracy, it also has better robustness and a higher generalization performance compared with the other existing models.

show abstract

Approaches to robust process identification: A review and tutorial of probabilistic methods

Cited by 52 publications

References 94 publications

Robust Recursive Identification of Hammerstein Models Based on Weiszfald Algorithm

Robust Recursive Identification of Hammerstein Models Based on Weiszfald Algorithm

Approximate GP Inference for Nonlinear Dynamical System Identification Using Data-Driven Basis Set

Modeling and Synchronous Optimization of Pump Turbine Governing System Using Sparse Robust Least Squares Support Vector Machine and Hybrid Backtracking Search Algorithm

Contact Info

Product

Resources

About