Comprehensive Errors-invariables-based Model Identification (Part 1)
Source: https://wsai.iitm.ac.in/projects/comprehensive-errors-invariables-based-model-identification-part-1/ Parent: https://wsai.iitm.ac.in/projects/
Comprehensive Errors-invariables-based Model Identification (Part 1)
Investigators
Arun Tangirala Shankar Narasimhan
Tags
process control model identification data science
Learning or developing dynamic models from experimental data is crucial in monitoring and control of processes. In general, measurements of both inputs and outputs of a processare subject to noise. Model development from such data is treated under the broad banner of errors-invariables (EIV) identification. In the preceding two years, we have developed a dynamic iterative principal components (DIPCA)-based approach to identify linear dynamic models for the EIV case, which has the ability to estimate the delay, order, error variances and model parameters based on a rigorous theoretical formulation and without requiring any prior knowledge. We have successfully demonstrated the use of our proposed approach to develop models for single-input single-output systems, for a variety of model structures (auto-regressive exogenous, output-error, Box-Jenkins, etc.). This project aims to extend the above approach for developing models for multi-input single output (MISO) systems and further to multi-input multi-output (MIMO) systems. We also propose to derive the theoretical properties of the estimates, in particular the consistency of the estimates, to provide a strong theoretical basis. Our ultimate goal is to develop a comprehensive EIV-based model identification approach that is envisaged to become a standard technique and be a part of MATLAB’s System Identification Toolbox.