ACS6124 Assessment (Spring 2021) Part I: Multisensor and Decision Systems

ACS6124 Assessment (Spring 2021)

Part I: Multisensor and Decision Systems

This is the first of two assignments for the module ACS6124, with this one assessing the learning outcomes of the Multisensor Systems component of the module.

Assignment code: ACS6124-001
Assignment weighting: 50% of the module assessment marks
Assignment due: 23:59, 19 April 2021 (Monday, Week 8, Semester 2)
Assignment format: A report of 15 pages maximum (excluding the Appendix)
Submission mode: Report must be submitted electronically via Blackboard.

1         Assignment Processes

Report format details

The report should be in a pdf form using a top and bottom margin of 1.5 inches, a left and right margin of 1 inch, and text should be size 12 point with 1.5 line spacing.

Submission mode details

This report is to be submitted via Turnitin using the Assessment tab on Blackboard by 23:59, Monday 19th April 2021 (Week 8 of Semester 2). Please note that you will be allowed only a single submission and so the first submission made will be the one assessed.

Penalties for Submission

Work that is submitted after this deadline (without medical or other similar documented evidence unless agreed) will incur a penalty for late submission. The usual late submission penalty of 5% reduction in the mark for every working day (or part thereof) that the assignment is late and a mark of 0 for submission more than 5 days late. For more information http://www.shef.ac.uk/ssid/assessment/grades-results/submission-marking

Unfair Means

This is an individual assignment. You should not discuss the assignment with other students or work together with other students in its completion. The assignment must be wholly your own work. References must be provided to any other work that is used as part of the assignment. Any suspicion of the use of unfair means will be investigated and may lead to penalties. For more information see: http://www.shef.ac.uk/ssid/unfair-means

Extenuating Circumstances

If you have any medical or special circumstances that you believe may affect your performance on the assignment then you should raise these with the Module Leader at the earliest opportunity. You will also need to submit an extenuating circumstances form. More information at:

http://www.shef.ac.uk/ssid/forms/circs

Help

If you have any questions on the assignment, please email me at: g.konstantopoulos@sheffield.ac.uk

Feedback

Written feedback will be provided on Blackboard within 15 working days, in line with Department guidelines.

2         Assignment technical requirements

Write a report fulfilling the requirements of Problems I and II. The report should follow the specified guidlines and meet the objectives provided for each task.

Problem I

Present the findings of your work from the activities of Labs A.I and A.II. Your report should also include the additional activities:

  1. Consider the energy levels in six frequency bands and the respective filter orders and Wn as depicted in Table 1. Use these values and repeat Task II and Task III of Lab A.I activities.
Feature Name Filter Filter Order Wn
f1 Low-pass 25 Hz 7 0.05
f2 Band-pass 25 – 50 Hz 6 [0.05   0.1]
f3 Band-pass 50 – 100 Hz 9 [0.1      0.2]
f4 Band-pass 100 – 200 Hz 8 [0.2      0.4]
f5 Band-pass 200 – 350 Hz 9 [0.4      0.7]
f6 High-pass 350 Hz 16 0.7

Table 1: Features to be extracted

Present your results and make a critical comparison with the results obtained from Lab A.I. You should demonstrate an understanding of the problem formulation and include a clear

methodology for providing a solution to the problem.                                       [20%]

  1. Extend the 1-nearest neighbours algorithm developed in Lab A.II to create a k-nearest neighbours solution, where k = 1,3 and 5 and compare the classification accuracy. This is to be performed with only the four feature results that were obtained during Lab A.I. In your report you should:
    • Provide the fault classification analysis results and its critical evaluation. [20%]
    • Suggest detailed justified alternative approaches and improvements to the methods

used.                                                   [20%]

Note: Include your Matlab code corresponding to the solution of Problem I in the Appendix of the report. Your code should be clear and contain appropriate comments explaining each step of your methodology.

Problem II

Imagine you are employed in a wind turbine manufacturing company and have been tasked to design a multisensor signal estimation and health monitoring system for the blade pitching mechanism of the wind turbine. In order to measure the pitch angle ωˆ, the blade is equipped with a rotary encoder connected to the blade bearing, where the sensor noise is ν∼N(0,9).

  1. Write a Matlab script to calculate the MMSE estimator, when the prior knowledge of the angle is uniformly distributed in the range 0ωˆ ≤ 30. Consider the scenarios where:
    1. The entire measurement vector encoder.mat is provided. You can find the measurement file inside the module’s Assignment folder in Blackboard.
    2. Only the five first elements of the provided measurement vector are available.

Mathematically describe the problem formulation and the methodology leading to its solution. Compare the results and provide a critical evaluation of the outcome. [7%]

  1. Extend the script and calculate the MMSE estimator in the case when the prior knowledge is Gaussian distributed with a mean value of 15 and variance of 4. As before, consider both scenarios on the availability of measurements and mathematically describe the problem formulation and the methodology leading to its solution, while also commenting on the

results. Use the same measurement vector encoder.mat.                                    [7%]

  1. Sensor estimates, such as in the above part, are collected over k sample intervals from a strain gauge sensor that measures the vertical to the rotating axis bending moment of the blade. You can find the measurement file straingauge.mat in the module’s Assignment folder in Blackboard. Write a script to apply a two-sided CUSUM test algorithm under the assumption that the normal operation of the system is 3000 KNm with a variance of 1. You

may choose a threshold of ±20 and ignore the leakage term γ.                                    [6%]

Note: In the part of your report for Problem II, you should include parts of your Matlab code within the report to describe the methodology followed. Your code should be clear and properly commented to explain each step of your methodology.

3         Marking Criteria

This assignment is marked out of 100 and contributes to 50% of the overall module assessment. The marking criteria below provide guidance on the relationship between the quality of submission and the marks awarded.

Marking Criterion/Comments Marks
Problem I a): Clear and step-by-step procedure, evidence of understanding of the process, comments and explanation /20
Problem I b):

• Clear procedure description, evidence of understanding of the process, comments on the comparison of the results among the different k-NN applications.

/20
• Well argued alternative approaches, clear description, critical comparison with methods used /20
Problem II a): Accurate mathematical description, appropriate Matlab code used, critical analysis and comparison of the results /7
Problem II b): Accurate mathematical description, appropriate Matlab code used, critical analysis and comparison of the results /7
Problem II c): Description and implementation of a two-sided CUSUM test, appropriate Matlab code used, comments on the results /6
Report quality: Report structure and clarity of writing, clear comments in the code that show the steps followed, quality of figures and tables, properly referenced material /20