ACS323 Assignment
Intelligent Systems
Academic Year: 2024/2025
Module: ACS323
Title: Intelligent Systems
Date Set: Friday 06/12/2024 at 15:00
Date for Submission: Monday 23/12/2024 at 23:59
Task: Produce written solutions to all the assignment questions. Where numerical answers are required, show your full procedure, including details of the fuzzy rules in terms of grid- tables, fuzzy membership functions, scaling factors, plots and diagrams. THERE IS NO LIMIT TO THE NUMBER OF PAGES YOU WISH TO SUBMIT.
Notes:
• This work is worth 40% of your overall mark for this module
• PARTS A & B carry equal weighting of 50 Marks each
• You are reminded that the submitted work must be personal.
PLEASE READ ALL OTHER ANSCILLARY MATERIAL CAREFULLY THAT SUPPORTS THIS ASSIGNMENT SHEET TO ENSURE THAT YOU ARE SUBMITTING ALL WORK MATERIAL IN THE REQUIRED WAY:
“guide notes for acs323 assignment 2024 2025.pdf”
“instructions on how to provide acs323 assignment files 2024 2025.pdf”
PART A- FUZZY DECISION-MAKING
Consider the process relating to muscle relaxation, as seen in your Laboratory Session PART A, represented by the following Equations [PLEASE NOTE THE NEW VALUE RELATING TO THE PREDOMINANT TIME-CONSTANT- CHANGE IT IN YOUR MATLAB-SIMULINK MODEL]:
where Y is the overall output of the system (muscle relaxation), U is the input to the system (amount of drug infused).
The main objective is to design a closed-loop control strategy which should maintain a steady level of muscle relaxation (output ‘Y’) by manipulating the level of drug infused (input ‘U’), given a reference target of muscle relaxation (‘Ref’)- see Figure A1- Use a reference target of 0.8 all throughout and a simulation time of 300 minutes:
a) Write a two-page survey (a critic) about three (3) applications of fuzzy logic based technologies (choose one or several applications areas: control, prediction, classification, fault monitoring, etc.) for real-world problems which you would have identified from the open literature (include full details of all references hence used- these can be research papers, general articles, and/or weblinks). [8 MARKS]
b) Design a Fuzzy PI-type controller, as shown in Figure A1, via a fuzzy rule-base which should include 25 fuzzy rules with Gaussian Membership Functions.
[The candidate is expected to use: 1. the fuzzy 3D surface to tune the rules; 2. their knowledge of tuning the PID tuning factors, all in order to obtain the best possible outcome for the output response in terms of minimum overshoot, fast rise-time and fast settling-time; include simulations with disturbances]. [13 MARKS]
c) Transform the fuzzy rule-base designed in Part A)b) into a Fuzzy PD-type controller
which should also include 25 fuzzy rules with Gaussian Membership Functions.
[The candidate is expected to use: 1. the fuzzy 3D surface to tune the rules; 2. their knowledge of tuning the PID tuning factors, all in order to obtain the best possible outcome for the output response in terms of minimum steady-state error, minimum overshoot, fast rise-time and fast settling-time; include simulations with disturbances].
Figure A1- Fuzzy-PI Control of Muscle Relaxation
[12 MARKS]
d) Using the closed-loop data from Part A)c) derive an ANFIS based controller for the process, which is described by Equations (A. 1) and (A.2), to achieve a similar control performance as that in c). What would be the advantages of such a new system? [12 MARKS]
e) Compare the controllers in Parts A)b), A)c), and A)d) in terms of: flexibility in the structure (controller type) and performance (accuracy). For the latter you can rely on one or more performance indices, e.g. Mean Absolute Error (MAE), Mean Square Error (MSE), and Root-Mean Square Error (RMSE). [5 MARKS]
PART B- FUZZY PREDICTIVE MODELLING
On Black-Board, you will find one (1) file named “acs323assignmentdata.mat” (in the folder “ Module Assignment”) relating to industrial data. This data set should reflect a system with four (4) inputs (Input 1, … , Input 4) and one (1) output. The minimum and maximum values for all inputs and outputs in the provided data can be found using the “min” and “max” MATLAB commands.
Upload this file onto your local drive which you will subsequently use to carry-out tasks Parts
B)a)-B)e) in MATLAB.
Once you have uploaded this file in MATLAB and double-clicked on it, it will collapse into two (2) files: one file, “acs323assignmentdata”, contains the actual quantitative data, and the second file, “explanation”, provides details on the names for each of the five (5) data features.
a) Use the ANFIS tool in MATLAB to obtain a fuzzy TSK-type model, with 3 membership functions for each input. The fitness of the model should be assessed with the use of a quantitative index (or indices) such as the RMSE, MSE or MAE to establish the validity of the model. You can use one performance index or more than one all throughout.
[The candidate will partition the data accordingly, select the most appropriate type of fuzzy MFs, output function and the number of learning epochs which will lead to the best outcome in the least-square sense]. [18 MARKS]
b) Using the model derived in Part B)a) find the values of the output for the following input vectors:
Input 1 = 0.15; Input 2 = 0.22; Input 3 = 1; Input 4 = 0.011; Input 1 = 0.06; Input 2 = 0.28; Input 3 = 0.4; Input 4 = 0.012; [4 MARKS]
c) Extend the fuzzy modelling exercise conducted in Part B)a) to include 4 membership functions for each input, then 5 membership functions for each input. Here also, the fitness of the models should be assessed with the use of a quantitative index (or indices) such as the RMSE, MSE or MAE to establish the validity of the models.
[The candidate will partition the data accordingly, select the most appropriate type of fuzzy MFs, output function and the number of learning epochs which will lead to the best outcome]. [18 MARKS]
d) Using the models derived in Part B)c) find the values of the output for the following input vectors:
Input 1 = 0.15; Input 2 = 0.22; Input 3 = 1; Input 4 = 0.011;
Input 1 = 0.06; Input 2 = 0.28; Input 3 = 0.4; Input 4 = 0.012; [4 MARKS]
e) Compare the models derived in Parts B)a) and B)c) and draw your own conclusions with respect to model accuracy and generalisation properties as far as: 1. Data partitioning between training and testing; 2. The number of fuzzy MFs, are concerned. [6 MARKS]