Experimental Investigation of a Low-Cost, Centrifugal Pump Cavitation Detection System

Frost, Callum (2024) Experimental Investigation of a Low-Cost, Centrifugal Pump Cavitation Detection System. [USQ Project]

[img] Text (Project – redacted)
Frost_C_Dissertation_Redacted.pdf
Restricted - Available after 21 October 2026.

Abstract

Cavitation is a symptom of centrifugal pump operation outside of design requirements and can result in damage and eventual pump failure. Cavitation is the unwanted formation of vapour bubbles, as a result of a localized drop in static pressure below the vapour pressure of a liquid, and can be costly if undetected.

In this project, a cavitation detection system was created using pressure and accelerometer signals from low cost, industry suitable sensors. Acquired data was pre-processed and machine learning methods were employed to train prediction models, providing feedback in the form of a response variable representing four cavitation classes (N, I, II, or C).

Cavitation prediction model results were evaluated for three pragmatic system use cases. The training results of the best performing models were: 100% accuracy for combined inlet, outlet pressure and accelerometer signals; 98% accuracy for combined inlet pressure and accelerometer signals; and 95% accuracy for isolated accelerometer signals only. For all presented use cases, the best performing models demonstrated 100% accuracy for the adopted benchmark cavitation, ‘NPSH3’.

Employment of industrially suited, low-cost equipment, combined with machine learning methods, were used to demonstrate feasibility, repeatability, and usefulness of an early cavitation detection system. Separate pressure and vibration methods, described in existing literature, were observed to be complementary when combined in this study. This study is but one example of how established artificial intelligence and machine learning methods can be useful as a preventative maintenance tool, with relatively low cost.


Statistics for USQ ePrint 53040
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Faculty/School / Institute/Centre: Current – Faculty of Health, Engineering and Sciences - School of Engineering (1 Jan 2022 -)
Supervisors: Saleh, Khalid
Qualification: Bachelor of Engineering (Mechanical)
Date Deposited: 09 Mar 2026 03:53
Last Modified: 09 Mar 2026 03:53
Uncontrolled Keywords: centrifugal pump; cavitation detection system
URI: https://sear.unisq.edu.au/id/eprint/53040

Actions (login required)

View Item Archive Repository Staff Only