Auto-tuning of frequency sampling filter data analysis
Loading...
Date
2019-06
Authors
Amiruddin Bin Mustafa Kamal Iskandar
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In industrial field, a lot of companies especially those involve with process plant and manufacturing plant require a good system and process in order to make sure the outcome of products they produced are in good quality, can be ready in short time and consume minimal operating cost. Every process plant has its own measurement data that need to be collected and analysed which come from various machines and equipment. For data driven modelling, it is important to obtain a good model as that will lead to good control of the system. The purpose of this project is to study on automatic tuning for frequency sampling filter algorithm. The frequency sampling filter algorithm able to analyse data and produce step and frequency response estimates which later can be used for modelling and control purposes. The automatic tuning method is required as to avoid processing of all raw data that will lead to long computational time. The steady-state error is used as the medium for tuning and checking how much data that actually required to sufficiently represent the system. The appropriate percentage error set point must be decided at the value where the step response estimates will stop at the most stable state after going through transient state. The value must not be too small or else the graph of step response estimates will stop right after it reaches the first few steady state values, thus will resulting into having insufficient dynamics captured from the system. Besides that, the percentage error set point also must not be too big because this selection will lead to longer steady state part being captured which is not necessary. MATLAB software is used for programming process. Several set of data have been used to test the auto-tuning algorithm capability. From results, it is shown that the developed approach able to process the data and provide with step response estimates using only a smaller number of data only. The reduction of data being processed have led to reducing the overall computational time needed to analyse the data.