Abstract
Predictive maintenance is a maintenance strategy based on monitoring the state of components to predict the date of future failure. The objective is to take the appropriate measures to avoid the consequences of this failure. For this reason, the authors determine the remaining useful life (RUL) which is the remaining time before the appearance of the failure on the component. It is an important approach that allows the prediction of aging mechanisms likely to lead components to failure. In this chapter, a new methodology for predicting the remaining useful life of components is proposed using a data-driven prognosis approach with the integration of machine learning. This approach is illustrated in a battery case study to predict the remaining useful life.
Top2 Concepts Of Maintenance
Maintenance is an essential step in the lifecycle of any complex system. According to AFNOR's definition, maintenance aims to keep or restore an asset to a specified condition so that it can provide a given service (AFNOR, 2001). There are two main categories of maintenance: corrective maintenance and preventive maintenance.
The first strategy generally incurs additional costs (Lee et al., 2006) and significant system unavailability (Palem, 2013), and it can pose safety issues. Preventive maintenance reduces the risk of unexpected failures and downtime, but it can result in high costs as some components are replaced while still in a slightly degraded operational state (Lee et al., 2006; Le, 2016).
To address these drawbacks, a new form of maintenance has emerged: predictive maintenance (Palem, 2013). It involves regular monitoring of the system's components (Bartelds et al., 2004), which enables the assessment of their proper functioning. Consequently, it becomes possible to predict the occurrence of a system failure before it happens (Lee et al., 2006; Le, 2016; Jardine et al., 2006), to plan the appropriate maintenance tasks at the right time (Mercier & Pham, 2012).
Various types of maintenance are illustrated in Figure 1.
Key Terms in this Chapter
Remaining Useful Life (RUL): It is a concept used in predictive maintenance and reliability engineering. It refers to the estimated remaining operational lifespan of a component, system, or asset before it is expected to fail or no longer perform its intended function effectively. RUL is typically determined through data analysis and predictive modelling techniques that consider factors such as historical usage patterns, environmental conditions, and degradation characteristics of the asset. By estimating the RUL, organizations can proactively schedule maintenance or replacement activities, optimize resource allocation, and minimize downtime or unexpected failures.
Predictive Maintenance: It is a proactive maintenance approach that uses data analysis and predictive modelling techniques to anticipate and prevent equipment or system failures. It involves monitoring and analysing real-time or historical data from sensors, machinery, or other sources to identify patterns, trends, and early indicators of potential issues. By predicting when equipment is likely to fail, maintenance activities can be scheduled in advance, optimizing resources, and minimizing unplanned downtime. Predictive maintenance aims to maximize the operational efficiency and reliability of assets while minimizing maintenance costs and disruptions.
Artificial Intelligence (AI): It refers to the simulation of human intelligence in machines that are programmed to think, learn, and problem-solve like humans. It involves the development of computer systems capable of performing tasks that typically require human intelligence, such as speech recognition, decision-making, visual perception, and natural language processing. AI encompasses various subfields, including machine learning, robotics, expert systems, and computer vision, and aims to create intelligent machines that can understand, reason, and interact with the world in a human-like manner.
Prognostics and Health Management (PHM): It is an interdisciplinary field that combines engineering, data analysis, and predictive modelling to assess and manage the health condition, performance, and reliability of systems and assets. It involves continuously monitoring and analysing data from sensors, diagnostics, and other sources to detect early signs of anomalies, degradation, or potential failures. By using advanced algorithms and statistical techniques, PHM aims to predict the remaining useful life (RUL), diagnose faults, and provide recommendations for maintenance or operational adjustments. PHM enables proactive decision-making, reduces downtime, optimizes maintenance strategies, and enhances the overall performance and availability of complex systems across various industries.
Random Forest Model: It is a machine learning algorithm that combines multiple decision trees to make predictions or classifications. It is an ensemble learning method that operates by constructing a multitude of decision trees during training and outputs the average or majority vote of the individual trees for prediction. Each decision tree in the Random Forest is trained on a different subset of the data, and a random subset of features is considered at each node. This randomness helps to reduce overfitting and improve the model's generalization ability. Random Forest models are widely used for tasks such as classification, regression, and feature selection, and they are known for their robustness, accuracy, and ability to handle large and complex datasets.
Prognostics: It is a field of study within predictive maintenance that focuses on estimating the remaining useful life (RUL) of a component or system and predicting its future performance and failure behaviour. It involves analysing historical data, monitoring real-time sensor data, and applying statistical and machine learning techniques to determine the health condition and expected future behaviour of the asset. Prognostics aims to provide early warnings of potential failures, enable proactive maintenance strategies, and optimize resource allocation for improved reliability, safety, and cost-efficiency.
Machine Learning: It is a branch of artificial intelligence that enables computers to learn and make predictions or decisions without explicit programming. It uses algorithms and statistical techniques to analyse data, identify patterns, and improve performance over time. There are three main types: supervised learning (using labelled data), unsupervised learning (finding patterns in unlabelled data), and reinforcement learning (learning through trial and error). Machine learning finds applications in various fields, such as image recognition, natural language processing, and fraud detection, by training models to make accurate predictions and decisions based on data analysis.