Challenges and Solutions in Predictive Maintenance of Machines in Industry 4.0: Signal Analysis under Variable Operating Conditions

Read full text Summarise with VIRAT
Authors: Orlando Peña, Ignacio Gutierrez, Dammika Seneviratne, Antonio Rafael Selva, Nerea Simón
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In the context of Industry 4.0, predictive maintenance is essential for maximizing asset availability and minimizing unplanned downtime. However, traditional vibration analysis techniques struggle with non-stationary signals generated under variable operating conditions, leading to diagnostic inaccuracies. This study addresses the challenge by exploring advanced signal processing methods capable of handling variability in load and speed, aiming to improve fault detection in complex industrial systems. The research combines a comprehensive literature review with a practical case study involving a servo press operating under variable conditions. Two key techniques are evaluated: signal normalization for variable load scenarios and order tracking for variable speed regimes. These methods are tested using synthetic vibration data from a bearing with an outer race fault, sourced from the Mendeley Data repository. Both tachometer-based and tacholess approaches are analyzed, including Vold-Kalman filtering and Short-Time Fourier Transform (STFT)-based speed estimation. The study demonstrates that order tracking techniques, particularly those using angular resampling and Vold-Kalman filtering, effectively transform non-stationary signals into stationary ones, enabling traditional fault diagnosis. Tacholess methods based on STFT, and peak detection yield comparable results to tachometer-based approaches, validating their use in scenarios where direct speed measurement is unavailable. This work highlights the practical applicability of signal and order normalization techniques in predictive maintenance under variable operating conditions. It provides validated methodologies for enhancing diagnostic accuracy without relying on constant-speed assumptions or extensive instrumentation. These findings support the development of more robust, flexible maintenance strategies across diverse industrial sectors. Future research will focus on comparing multiple approaches and applying these techniques to real-world datasets and expanding their use in other domains such as robotics, automotive, and aerospace.

Fields of Study & Subjects

Machine Fault Diagnosis Techniques Fault Detection and Control Systems Reliability and Maintenance Optimization

Publication details

Journal / source
Procedia Computer Science
Volume
277
Pages
1582–1591
Publication date
1 January 2026
Language
EN