A fault detection strategy for an ePump during EOL tests based on a knowledge-based vibroacoustic tooland supervised machine learning classifiers

Aug 5, 2026

Upload by Asher

Fault detection for electric pumps during end-of-line (EOL) testing is addressed using acceleration and pressure pulsation signals to train supervised machine learning models. The approach reduces the risk of non-conforming units reaching the assembly line, where defects can still pass quality control yet later cause premature failure or abnormal noise during field operation. A knowledge-based vibroacoustic tool and a Python library generate informative features from the measured signals, then an ensemble learning algorithm combines multiple classifiers. Experimental validation on eighty ePumps reports above 95% accuracy, with comparisons supported by principal component analysis and a sensor sensitivity study.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages26

Reading Time1 h 6 min

AccessFree to download

Introduction

Fault impact and need for early anomaly detection

Electric pumps and limitations of manufacturing quality control

Methodology

Feature generation from acceleration and pressure pulsation data

Knowledge-based vibroacoustic tool integration

Supervised machine learning ensemble classifiers

Experimental setup and evaluation

Dataset and testing campaign on electric pumps

Accuracy results

Dimensionality reduction using PCA

Sensor sensitivity study

Meccanica [https://doi.org/10.1007/s11012-024-01754-w](https://doi.org/10.1007/s11012-024-01754-w) A fault detection strategy for an ePump during EOL tests based on a knowledge‑based vibroacoustic tooland supervised machine learning classifiers Pasquale Borriello · Fabrizio Tessicini · Giuseppe Ricucci · Emma Frosina · Adolfo Senatore Received: 27 July 2023 / Accepted: 3 January 2024 © The Author(s) 2024 Abstract This paper presents a methodology for identifying faulty components in an electric pump during the end-of-line test based on accelerationsand pressure pulsation data used to train an ensemble learning algorithm based on supervised machine learning classifiers. Despite various quality control measures in pump manufacturing, some out-of-tolerance components can pass through and end up on the assembly line, potentially leading to premature failure or abnormal noise during real-field operation. Because of the high impact, it is very important to put in place actions to mitigate the risk of delivering non-conform units, even if properly working in terms of pressure-flow rate performances. In this paper, an innovative knowledge-based vibroacoustic tool together with a machine learning built-in Python® library have been used to post-process acceleration and pressure pulsations data to generate features, which are then used to train, and test several supervised machine learning algorithms. The ensemble P. Borriello (*) · A. Senatore Department of Industrial Engineering, University of Naples Federico II, Via Claudio, 21, 80125 Naples, Italy [e-mail: pasquale.borriello@unina.it](e-mail: pasquale.borriello@unina.it) F. Tessicini · G. Ricucci Fluid-O-Tech S.R.L., Via Leonardo da Vinci, 40, 20094 Corsico, Milan, Italy E. Frosina Department of Engineering, University of Sannio, Piazza Guerrazzi, 21, 82100 Benevento, Italy learning algorithm combines the best classifiers to identify healthy electric pump units with high accuracy, achieving above 95% accuracy in an experimental test campaign carried out on eighty electric pumps. Results are compared using principal component analysis for dimensionality reduction, and a sensor sensitivity study is conducted. Keywords External gear pump · Fault detection · The end of line (EOL) tests · Condition monitoring · Knowledge-based vibroacoustic model · Ensemble machine learning List of symbols Δp Delta pressure σ Standard deviation γ Regularization parameter AI Artificial intelligence AMR Amplitude of modulation at rotation frequency AOR Amplitude of orders of rotation ApEn Approximate entropy BLDC Brushless direct current C Penalty parameter cm Confusion matrix DOE Design of experiments DL Deep learning DM Dummy model DT Decision tree DUT Device under test DWT Discrete wavelet transform 1 3 EGM External gear machine EOL End of line test ePump Electric pump FFT Fast Fourier transformation HP High pressure KNN K-nearest neighbors LLE Locally linear embedding LP Low pressure ML Machine learning MLP Multilayer perceptrons MODWPT Maximal overlap discrete wavelet packet transform Mtry Number of the random input variables NB Naïve Bayes NIDAQ National instruments data acquisition Ntree Number of trees NVH Noise, vibration, and harshness PCA Principal component analysis PE Permutation entropy PACF Partial autocorrelation function PHM Prognosis and health management Q Volumetric flow rate rbf Radial basis function RF Random forest R&D Research and development RMS Rooth mean square RUL Remaining useful life SDG Stochastic gradient descent SVM Support vector machine t-SNE T-distributed stochastic neighbor embedding TDV Tooth defect vibration X Input variable Time series average of input variable 1 Introduction Faulty components within a pump increase the risk of premature failure or unsatisfactory performance. In the context of an electrically driven external gear machine

FAQs

Why is fault detection during EOL testing important for electric pumps?

Even when manufacturing quality controls are applied, some out-of-tolerance components can pass and reach the assembly line. Such defects may later lead to premature failure or abnormal noise during real-field operation, making early identification critical.

What data and tool are used to build the fault detection strategy?

The method uses acceleration and pressure pulsation data from EOL tests. A knowledge-based vibroacoustic tool post-processes these signals to generate features for supervised learning.

How does the strategy combine machine learning models and how is performance assessed?

An ensemble learning algorithm combines the best supervised classifiers to identify healthy ePump units. Results are compared using principal component analysis for dimensionality reduction and a sensor sensitivity study, achieving above 95% accuracy on eighty pumps.

Aug 5, 2026

Upload by Asher