Audio Signal Processing for Machine State Monitoring
1. Assignment Aims
Acoustic signals play an essential role in machine state monitoring. The sound of the machines carries information about the machines condition. Efficient processing of real-time machine acoustic signals, to extract relevant audio features, could enable early fault detection and prevention, and avoid unexpected accidents and financial losses.
The aim of this assignment is to analyse a sound dataset for the malfunctioning valves, and design a filter to extract frequency signatures of a faulty valve.
Learning Outcomes:
Knowledge and Understanding Outcomes
2. Assess the operation of DSP systems and aspects associated with their performance.
Ability Outcomes
3. Design and evaluate digital filter design solutions using a CAE package.
4. Solve practical digital filter design problems through the development of
hardware/software solutions.
2. Assessment Brief
This is an individual assignment
Two sets of data are provided containing the sounds generated from the valves, containing normal sounds and anomalous sounds. The signals were selected from an open source dataset for malfunctioning industrial machine investigation and inspection MIMII https://zenodo.org/record/3384388#.YUuYi2LMJPa . The sounds were recorded by eight-channel microphone array with 16 kHz sampling rate and 16 bit per sample.
Tasks
1. Use MATLAB to carry out the following tasks:
• Use MATLAB command to convert the .wav files to vectors and display them with properly labelled plots and appropriate comments
• Use time, frequency and time-frequency analysis to characterise the signals
• Based on signal analysis, design a digital filter (or filters) to filter out all but the signature frequencies of a faulty valve. Use different filters or vary the parameters in your filter (passband attenuation, etc) and compare the results.
• Use time, frequency and time-frequency analysis methods, as well as MATLAB sound command, to evaluate your results. Provide appropriate plots.
• command, to evaluate your results. Provide appropriate plots.
2. Develop a Flowcode program to implement your digital filter(s) on the dsPIC.
3. Produce laboratory report. Evidence of signal analysis, filter(s) design and simulation, and evaluation, as well as digital filter implementation in the Flowcode, should be included in your laboratory report (see marking scheme).
3. Marking Scheme
The assignment work will be assessed through laboratory report, based on the practical laboratory work.
The laboratory report should have a well-defined structure similar to the following:
Organisation and content (70%)
The title and Equipment Used
1. Brief Objectives
This should state clearly what the objectives of the laboratory are.
2. Diagrams
Diagram actually used or system sketches.
3. Procedure
Brief notes on the procedure actually used, including departures from the guidance given on the lab sheet.
4. Test Results
a. Show measurements, results calculated from these and comparisons with expected results from theory.
b. Estimates of errors.
5. Graphs/plots
Sketch graphs/plots during the lab work which should be used to identify rogue or missing results
Discussion (20%)
The discussion section may need to be completed outside the timetabled session as this section shows if you understand the meaning and limitation of your results. The contents should cover the following points where appropriate.
a) What can be learned from the results?
b) Do the practical results agree with the theory or expected results? If not, suggest why?
c) What are the errors and the probable causes?
d) Could the experiment be done more accurately or in a better way?
e) Is there any further work which could be done?
Conclusion (10%)
Comments on the results obtained. A paragraph should be adequate for most sessions.
