Unlocking Maintenance Excellence: Background on Preventive, Predictive, and Prescriptive Maintenance Across Industries — Part 1 (Theory)
Predictive, preventive and prescriptive maintenance are strategies to ensure the reliability, efficiency and longevity of machines and systems across different industries. Summarized, predictive, preventive and prescriptive maintenance is responsible for the following points:
- minimizing downtime,
- reducing costs,
- improving safety,
- optimizing the overall performance of machines and systems. [1]
The implementation of these maintenance strategies can lead to significant improvements in operational efficiency and reliability in companies across various industries. [2] The implementation of such a strategy always requires a certain amount of internal data. For example, the predictions for predictive and prescriptive maintenance are only as good as the quality of the internal data. These models are based almost exclusively on historical data and are therefore highly dependent on the level of detail and quality of the available data.
There are various models for the probability of errors occurring over the life cycle of machines and systems, the best known of which is the so-called “bathtub curve”. The bath tube curve shows three different failure phases. The first phase is dealing about the infant mortality, followed by the random failures and the last one the wear out failures. The hazard function is constant over time in the random failures phase, decreasing in the beginning and increasing again in the last phase. The most papers mainly are dealing about the initial phase (infant mortality) and describing the weak points and assumptions made for the modelling. Interestingly according to those publication this phase model is dominated by the design and manufacturing defects which cannot be completely eliminated. Something to add here is that out of the box failures could be as well linked to this initial phase. Those failures could be eliminated with stress tests such as burn-in and run-in tests.[3]
Predictive, prescriptive and preventive maintenance is coming around the corner after the infant mortality phase starting with he random failure phase. In the most publications the curve is turned around be exchanging the condition with the hazard function on the y-axis (paper [2]). Basically the the message is completely the same. In the following there is the hazard function visible showing the points where visible when we talk about the three different terms (Predictive, prescriptive and preventive maintenance). The start of failure, the ignition point is called in the following (p — precursor). The phase until this point is called “normal operation”.
Preventive maintenance typically is done during the normal operational phase (picture above). Within this period, specific maintenance windows (time windows) are established — maintenance and overhaul schedules — during which actions such as the replacement of grease, insulation materials (dry/fluid), or mechanically stressed components are undertaken as a preventive measure. These are based on past learnings which involves the usage, time, and other parameters. Consequently, based on prior studies and accumulated knowledge, the machinery or system operates for extended durations. The clear advantages of this strategy are that simplicity, the easy implementation, and the decrease of risk of breakdowns and costly repairs. On the other side, it can be quite time consuming and wasteful since the wrong parts will be exchanged or even failures could evolve during the milestones. Summarized this technique is not considering the actual condition of the machine or system (besides visual inspections from time to time).[1][2]
Predictive maintenance could be described as a proactive type of maintenance. This strategy uses data coming from various sensors (invasive and non-invasive measurements) applying analytics to check the condition of the equipment in real time. Furthermore, predictive maintenance is dealing as the name already outlines when the failure is most likely to occur. This technique helps to optimize the maintenance schedule as the schedule is based on events from the predictive maintenance algorithm. This leads to reduced downtimes, costs and help to extend the machine or system lifetime. One downside is the complexity, the higher invest in technology and the needed expertise compared to preventive maintenance. [1][2]
The third one called “Prescriptive maintenance” is the most sophisticated and innovative type of maintenance. Mainly AI is used in combination with big data to predict the failures. The goal of this technique is to prescribe the best actions to delay or even prevent failures. The clear advantages are in the improvements in maintenance efficiency, effectiveness, and quality. Furthermore, one further advantage is the decision support as well as the automation capabilities. The downsides are the very high investment in technology, data, and expertise compared to the other strategies, the trustworthiness and the explain ability of the individual results and recommendations.[1][2]
In the forthcoming chapter, we embark on an immersive journey into the area of predictive maintenance strategy implementation. We’ll meticulously explore the vital components, necessary prerequisites, and strategic data preparation techniques essential for seamless integration and optimization. Moreover, the subsequent chapter will tackle the positioning and marketing problems of these solutions, flipping the conventional approach on its head. Instead of the usual starting point, we’ll explore this topic from a fresh perspective, examining both internal and external viewpoints within the company.[1][2]
Literature
[1] Vanderschueren, Toon & Boute, Robert & Verdonck, Tim & Baesens, Bart & Verbeke, Wouter. (2022). Prescriptive maintenance with causal machine learning. 10.48550/arXiv.2206.01562.
[2] Achouch, M.; Dimitrova, M.; Ziane, K.; Sattarpanah Karganroudi, S.; Dhouib, R.; Ibrahim, H.; Adda, M. On Predictive Maintenance in Industry 4.0: Overview, Models, and Challenges. Appl. Sci. 2022, 12(16), 8081; https://doi.org/10.3390/app12168081.
[3] Maisonnier, D.. (2018). RAMI: The main challenge of fusion nuclear technologies. Fusion Engineering and Design. 136. 10.1016/j.fusengdes.2018.04.102.
