Service 4.0 helps spur predictive maintenance adoption

By | 2 minute read | January 14, 2019

Predictive Maintenance (PdM) is a popular topic, especially in the industrial and aerospace fields. It has been around for a while, but gained traction with IoT, digitization and Service 4.0. Yet, while many companies are exploring predictive maintenance, many haven’t tapped into its substantial business value. How can enterprises use this technology for desired results and how can technology support providers assist them?

Let’s start with a brief overview of PdM, how it works and what I seen in many client engagements.

What is predictive maintenance?

The traditional way of doing maintenance is reactive and follows this principle: Fix it when it’s broken. There are also proactive (check and maintain regularly) and predictive maintenance (check and maintain when necessary) approaches. PdM offers teams optimized maintenance and a solution to help prevent rather than react to problems.

How does it work?

With predictive maintenance, professionals determine the equipment condition to estimate when maintenance should be performed. This is achieved by observing key parameters of components such as temperature, pressure, vibration, hours used, and rotations per minute. Component data also can be enriched with service request data to incorporate operation and service activity knowledge.

Previously, PdM used multivariate statistics, defined thresholds and programmed rules to analyze data. Now, predictive maintenance tools use machine learning to understand normal behavior and detect anomalies. Through experience and predictive analytics, tools can adapt prediction models, analyze data, uncover correlations, and identify deterioration patterns prior to a malfunction. For example, I observe clients in the retail, automotive and IT industries using these capabilities.

While some instances of PdM tools act autonomously, it’s common to apply a Human-in-the-Loop (HITL) model and leave the final decision to human engineers. Supervised learning, or prediction model training, typically involves people as well.

Why use it

Predictive maintenance is mainly used to improve equipment efficiency and optimize IT maintenance spending. Increased uptime can increase asset utilization rates and ROI and helps improve the equipment lifetime. And PdM provides qualified estimates and offers cost savings over proactive maintenance, which follows simpler maintenance rules like every x hours of usage or once a quarter.

Within modern environments, many products are embedded into a process and part of a larger production system — a system of systems. PdM is applied to the whole system versus a single machine to provide the right result. Because predictive maintenance tools handle data feeds from multiple sources, they can discover unexpected correlations.

Unplanned downtime is usually more expensive than planned downtime due to labor costs and collateral damages, including the following: unrealized revenue from missed sales, website downtime, SLA penalties, or the ripple effect on other parts of the system environment. Any unintended interruption can affect customer satisfaction, lead to data loss, or violate regulatory requirements. Comparatively, planned downtime helps you avoid unwanted interruptions and optimize staff and parts resources.


PdM is not the only solution. Reactive or proactive approaches are well-suited to certain situations. However, consider PdM for mission-critical equipment.

Once you decide to implement a PdM solution, make sure you have the right staff in place, namely reliability engineers with maintenance model experience and data scientists with strong analytics skills. The nature of PdM requires a combination of IT capabilities and maintenance process knowledge. Aim to establish a culture of collaboration both internally and with external maintenance providers so you can develop a joint maintenance strategy and execution plan.