Analysis of the ensemble of regression algorithms for the analog circuit parametric identification. (August 2020)

• Record Type: Journal Article
• Authors: Bilski, Piotr
• Is Part Of: Measurement. Volume 160(2020)
• Publication Date: 2020-08
• Subjects: Regression -- Diagnostics of analog systems -- Parameter identification; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2020.107829
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Introducing ensemble of regression algorithms for the analysis of analog systems. Tests of the ensemble on the analog filter and verification of analysis accuracy. Comparing accuracy of the ensemble and the Support Vector Regression. Abstract: The paper presents the application of the combined group of regression algorithms for the parameter identification of the analog circuit's state. The fusion of regression machines is a new approach aimed at obtaining the high accuracy in the diagnosis of parametric faults determined in the presence of noise. The ensemble consists of multiple approaches, mainly based on variants of the linear regression techniques. Because the methods are simple, it is easier to build the accurate module than for the typical heuristic approach, such as Support Vector Machines (SVM). The methodology consists in preparing the ensemble architecture, selecting computational methods, optimizing features extracted from the diagnosed system and testing the module. It was tested on the 5th order lowpass filter and compared with the single regression algorithm, treated as the reference method. Obtained results show the usefulness of the proposed framework for the accurate identification of analog system parameters.