Location:Home > Neo-confucianism > Other > System theory > Study on Dynamic Modeling Algorithm of the Sensor Based on VS-LMS and PSO

Study on Dynamic Modeling Algorithm of the Sensor Based on VS-LMS and PSO

Downloads: []
Tutor: ZhaoYuanDong
School: Nanjing University of Information Engineering
Course: System theory
Keywords: Sensor,Dynamic modeling,Adaptive LMS,PSO,Matlab
CLC: TP212
Type: Master's thesis
Year:  2013
Facebook Google+ Email Gmail Evernote LinkedIn Twitter Addthis

not access Image Error Other errors

Sensor¡¯s dynamic model is the foundation for the researching of sensor¡¯s dynamic behavior. The goal of sensor¡¯s dynamic modeling is to find a proper math model to simulate the real sensor¡¯s model by a reliable and effect algorithm.This thesis learned several primary algorithms by now such as system identification, adaptive algorithm, neural network and so on. This thesis¡¯s main achievement based on the former¡¯s work as follows:Firstly, this thesis studied adaptive filter algorithm based on Variable Step-size Least Mean Square (VS-LMS). The algorithm could process measurement data and estimate result in real time, make the error smaller by correct the weight of the model automatically. The best weight is the weight of the sensor¡¯ model. This thesis studied traditional VS-LMS algorithms specially the sigmoid VS-LMS algorithm. This thesis proposed a improved VS-LMS algorithm based on the traditional VS-LMS algorithm¡¯s shortcomings. The new algorithm produced a new step pattern by combining and transforming TANH function to maintain the efficiency in the high order system. It also use the e(n)e(n-1) as a factor in the algorithm to avoid the affect of the noise.Secondly, this thesis studied the sensor¡¯s dynamic modeling algorithm based on Particle Swarm Optimization. This algorithm takes a special neural network-wavelet neural network as a model. Combining with the PSO algorithm improved the performance of the sensor¡¯s modeling. PSO is a creature-based algorithm. The traditional PSO has some shortcomings such as low speed, convergence partly and inaccuracy. This thesis improved PSO by add in weight and learning factor changed linearly.Thirdly, this thesis do some simulations for it¡¯s algorithms by matlab. Firstly, this thesis simulate the performance of fixed step-size LMS and Sigmoid based LMS and its improved LMS in the circumstance of32order,64order and128order. It proved that this thesis¡¯s algorithm is the best by comparing former three algorithms. Secondly, the simulation also compared the performance of traditional PSO and Improved-PSO, traditional neural network and the wavelet neural network and lastly this thesis use improved-PSO and WNN to approach a real sensor¡¯s model. The result proved that improved-PSO and WNN is good for sensor¡¯s modeling.
Related Dissertations
Last updated
Sponsored Links
Home |About Us| Contact Us| Feedback| Privacy | copyright | Back to top