ECE 551-B Problems in Machine Learning

PROJECT TRACKER

PROJECT PROPOSAL

Purpose Questions [1/3]

[-] Solving What Problem(s)? [1/1]
- [X] [#1] Save Money $

Reduce number of unexpected days downtime and limit degree of damage due to uncaught catastrophic events such as: debris intake small, debris intake large, fibre / organic material wrapping or binding, cavitation initiation (impeller coating worn off), etc.

[X] [#2] Save Money $$

Provide real-time, non-control channel, degree of variance from baseline operation in form of relative ‘grading’ or ‘scoring’ of each pump present in a given station (group of pumps). Higher pump id’s would receive grade or score based on last known performance upon shutdown. Prior to startup of station, pump scores are polled and depending on how many units need to be activated the choice of pump and sequence of start up order will be chosen on a highest score wins case / switch. Baselining and grading would be unique to a given pump station. Different pump station locations have different system characteristics that make comparing pumps across multiple stations unlikely.

[ ] Deliver report & presentation for UNM ECE-551

Project Questions [2/13]

[X] Q/A: ML / Detection vs Recogntion?

A:DETECTION ONLY

[X] Q/A: Remote vs. On-site implementation?

A:REMOTE

[ ]Q/A: Algorithms: SVM, NN?

should focus on locally runable algorithms, SVM?

[ ] Q/A: Datasets: ToyADMOS, ToyAMOS2, DCASE, PRONOSTIA, ?
[ ] Q/A: Processing Architecture: x86, ARM, FPGA, SoC, etc?
[ ] Q/A: Sensors: ultrasonic vs. audible range? / air-gapped vs. surface mounted?
[ ] Q/A: Intended budget level: affordable?
[ ] Q/A: Utilize existing client network?
[ ] Q/A: Utilize existing client control sys infrastructure?
[ ] Q/A: Utilize existing power / electrical supply vs. mobile solution?
[ ] Q/A: ML processing Real-Time vs Off-line // local vs. cloud?
[ ] Q/A: Input data format (wav, mp3, fft, etc)
[ ] Q/A: Output data format (text based logs, messages?)
[ ] Q/A: Bandwidth limitations; microwave, radio, fiber, etc.
[ ] Q/A: Working prototype or Proposed Design?

ROADMAP

[ ] Standard Report Outline (abstract, … , citations)
[ ] Abstract
[ ] Index Terms
[ ] Introduction (problem to solve, state of the art, available means)
[ ] Theory (hardware, software, & algorithms)
[ ] Methods and Materials (experiment)
[ ] Results (output)
[ ] Conculsion (did we solve a problem, future thoughts)
[ ] References
[ ] Present Final report in person / live presentation

STATE OF THE ART

KEYWORDS:

acoustic anamoly detection, one class SVM, novelty detection, unsupervised anomaly detection, real-time, acoustic emission, ultrasonic,

Search Results:

Academic Papers:

A Machine Learning Approach for locating Acoustic Emissions

Acoustic emmision used for detection of crack generation in propellers of turbine-pump units

The application of acoustic emission for detecting incipient cavitation and the best efficiency point of a 60kW centrifugal pump: case study

Machine-Learning-Based Methods for Acoustic Emission Testing: A Review

Acoustic Anomaly Detection in Additive Manufacturing with Long Short-Term Memory Neural Networks

Acoustic Anomaly Detection of Mechanical Failures in Noisy Real-Life Factory Environments

Validation of artificial neural networks to model the acoustic behavior of induction motors

A comparison between psychoacoustic parameters and condition indicators for macinery fault diagnosis using vibration signals

Psychoacoustic approach for cavitation detection in centrifugal pumps

Remote acoustic analysis for tool condition monitoring

A low engery FPGA platform for real-time event-based control

Several Approaches for Anomaly Detection From Sound

An FPGA Platform Proposal for Real-Time Acoustic Event Detection

FPGA Implementations of Support Vector Machine Using Ising Model for AI on Things

FPGA-Embedded Anomaly Detection System for Milling Process

A machine sound monitoring for predictive maintenance focusing on very low frequency band

Design and Implementation of Acoustic Sensing System for Online Early Fault Detection in Industrial Fans

FPGA based monitoring platform for condition monitoring in cylindrical grinding

The development of surface acoustic wave sensors (SAWs) for process monitoring

One class svm for link

Anomalous Sound Detection link

Discussion of Features for Acoustic Anomaly link

Potential Data Sets

MIMII Dataset: Sound Dataset for Malfunctioning Industrial Machine Investigation and Inspection

ToyADMOS dataset

ToyADMOS dataset 2

PRONOSTIA Bearing Dataset

Existing Kaggle, Jupyter Notebooks, GitHub Repo’s

DCASE2020 Challenge Task 2 baseline variants

Federated Learning for Autoencoder-based Anomaly Detection in the Industrial Iot

Self-Supervised Acoustic Anomaly Detection…

https://github.com/Armanfard-Lab/AADCL

Unsupervised Anomaly Detection…

https://www.kaggle.com/code/victorambonati/unsupervised-anomaly-detection

Commercial, Industry Links:

Webinar: Automated Acoustic…

https://www.youtube.com/watch?v=-6B_XsEN2Q4

YT: One class SVM detection…

https://www.youtube.com/watch?v=0dngOGhv5Mc

Files

ECE551b.org

Latest commit

History