Content area
Full text
Introduction
Lifecycle engineering aims to optimize the performance and sustainability of products, processes, and services throughout their lifecycle, promoting environmentally sustainable practices among manufacturers. A critical strategy for achieving this sustainability is equipment maintenance, which helps extend their life expectancy, thus reducing the need for frequent replacements of components or the equipment itself and minimizing waste and the environmental impact associated with manufacturing components [1, 2].
Maintenance ensures the proper functioning and efficiency of equipment; essentially, it prevents equipment failures. There are three types of maintenance: (1) Corrective maintenance is performed when equipment starts failing and requires urgent repair or replacement. It’s also done when a decline in performance is detected, scheduled interventions, in this case, can wait; (2) Preventive maintenance aims to reduce machine downtime due to breakdowns through scheduled maintenance; (3) Predictive maintenance relies on monitoring mechanisms that track parameters related to equipment condition to diagnose or predict anomalies if parameters change [3, 4].
Recent studies focus on predictive maintenance, diagnosing equipment conditions using continuously collected operational data. This data is analyzed using statistical methods like Principal Component Analysis (PCA), K-Means, Isolation Forest, Gaussian Naive Bayes (NB), k-Nearest Neighbors (k-NN), or deep learning algorithms such as Convolutional Neural Networks (CNN) or Long Short-Term Memory networks (LSTM) [5, 6] to detect any anomalous operation early, which could lead to failure. Continuous monitoring can also predict when maintenance is necessary, optimizing resources by replacing components only when an anomaly is detected. This reduces repair costs and downtime, avoids unexpected failures and unscheduled shutdowns. Predicting part failures enables better inventory and logistics management for spare parts, resulting in financial optimization. It also aids maintenance workshops in solving issues by providing more information and knowledge about failures [6].
Anomaly detection is used in various fields, including manufacturing, healthcare, fraud detection, and cybersecurity, among others. In fields like manufacturing or healthcare, the detection must be effective and reliable since it is the most viable practice for decision-making processes to prevent machines from causing catastrophic resource or life losses [7]. The rapid development of information technologies and IoT has become an excellent option for developing intelligent and real-time predictive maintenance solutions. Deep learning (DL) algorithms have recently become increasingly useful in these domains [8].
Data-based machine learning (ML) methods...