Approximate Maximum Likelihood Parameter Estimation for Nonlinear Dynamic Models: Application to a Laboratory-Scale Nylon Reactor Model

2008 ◽  
Vol 47 (19) ◽  
pp. 7274-7283 ◽  
Author(s):  
M. Saeed Varziri ◽  
Kim B. McAuley ◽  
P. James McLellan
Keyword(s):  
Parameter Estimation ◽  
Maximum Likelihood ◽  
Nonlinear Dynamic ◽  
Dynamic Models ◽  
Laboratory Scale ◽  
Reactor Model ◽  
Approximate Maximum Likelihood ◽  
Nonlinear Dynamic Models

Genetic Algorithm-Based Maximum Likelihood Parameter Estimation for Transit Buses

2001 ◽  
Author(s):  
Jie Xiao ◽  
Bohdan T. Kulakowski
Keyword(s):  
Parameter Estimation ◽  
Maximum Likelihood ◽  
Dynamic Models ◽  
Estimation Method ◽  
Optimization Techniques ◽  
Model Parameters ◽  
Parameter Estimation Method ◽  
Transit Buses ◽  
Simulation And Control ◽  
And Control

Abstract Vehicle dynamic models include parameters that qualify the dependence of input forces and moments on state and control variables. The accuracy of the model parameter estimates is important for modeling, simulation, and control. In general, the most accurate method for determining values of model parameters is by direct measurement. However, some parameters of vehicle dynamics, such as suspension damping or moments of inertia, are difficult to measure accurately. This study aims at establishing an efficient and accurate parameter estimation method for developing dynamic models for transit buses, such that this method can be easily implemented for simulation and control design purposes. Based on the analysis of robustness, as well as accuracy and efficiency of optimization techniques, a parameter estimation method that integrates Genetic Algorithms and the Maximum Likelihood Estimation is proposed. Choices of output signals and estimation criterion are discussed involving an extensive sensitivity analysis of the predicted output with respect to model parameters. Other experiment-related aspects, such as imperfection of data acquisition, are also considered. Finally, asymptotic Cramer-Rao lower bounds for the covariance of estimated parameters are obtained. Computer simulation results show that the proposed method is superior to gradient-based methods in accuracy, as well as robustness to the initial guesses and measurement uncertainty.

Sign in / Sign up

Export Citation Format

Share Document