Predictive maintenance examples: how it works in practice

Predictive maintenance is the future of facilities management – a shift driven by converging challenges of aging infrastructure, ongoing skilled labor shortages and increasing pressure to reduce energy consumption. Facility teams know predictive maintenance is important and that they can no longer afford maintenance strategies that rely on reactive repairs or calendar-based inspections.

However, the challenge that remains is understanding what this shift looks like in their building, and how those insights translate into business value.

As researchers point out, two identical AHUs are no longer the same once installed in different commercial buildings. Unlike manufacturing, where the same equipment operates under almost identical conditions, every commercial building operates differently. That’s why having multiple building-specific examples provides practical context for facility and operations leaders who want to move beyond abstract theory and understand how predictive maintenance applies to the systems they manage.

Each section focuses on both specific systems and building environments where predictive maintenance pays off – ranging from HVAC systems to central utility plants. While the systems and environments differ, the same logic applies to each: monitor data, detect the right signals and act before failure occurs.

HVAC and chiller systems

Space heating and cooling account for 21% of commercial energy use, making HVAC systems one of the highest-value targets for predictive maintenance. That’s because these systems often operate for extended hours and are critical to building performance, making failures costly. Crucially, they also show subtle early warning signals that analytics can detect before issues occur, such as temperature drift.

Of all HVAC equipment, chillers are at the top of the list. They can account for up to 40% of a building’s energy use and are key to occupant comfort and safety, making them high risk. By monitoring chiller performance, analytics can detect energy-efficiency drift and mechanical stress to help keep equipment operating under optimal conditions.

How predictive maintenance identifies issues

As an example of predictive maintenance in action, a hypothetical commercial building’s chilled water supply temperature was gradually drifting upward. The fault detection and diagnostic (FDD) spotted this in raw building automation system (BAS) data and checked it against water flow rate, which was dropping outside the expected range.

The analytics layer diagnosed the issue as a suspected failure of the condenser water pump. A technician confirmed the issue, replaced the pump and restored flow. Without analytics, the drift might not have been noticed until the issue escalated into a fault code or – worse – a chiller failure during peak demand.

Central utility plants

A central utility plant (CUP) generates and distributes hot and chilled water across a large site, such as multiple buildings on a campus or a hospital site. These interdependent systems are best managed as a unified system with a single view on performance. This allows teams to spot and catch issues in one area of the plant before they spread.

Example of this in action

Analytics detects subtle vibration changes from dedicated IoT condition-monitoring sensors attached to a chilled-water pump bearing in a university building. The failing pump supplies chilled water to several downstream buildings, so catching it protects the entire distribution system – not just a single asset. The alert generated a work order in the CMMS and scheduled a replacement during an expected maintenance window before the issue caused a drop in water pressure across campus.

Building automation systems

Every predictive maintenance example in this article ultimately depends on data generated by a building automation system (BAS). This serves as the foundation for equipment monitoring, analytics and fault detection across a smart building ecosystem. A BAS continuously generates thousands of data points across a building's equipment by covering air handlers, pumps, chillers and VAV boxes.

On its own, that raw stream is just numbers. FDD software reads it and compares signals against each other – and against expected baselines – to surface developing issues before they trigger a fault code or affect occupant comfort.

Unlike other examples in this article, a BAS is not a standalone asset. It acts as the data foundation that powers predictive maintenance across building systems.

Example of this in action

In a commercial building, FDD software tracked an air-handling unit's cooling coil and found that the chilled-water valve was progressively opening over several weeks to maintain the same supply-air temperature.

Each reading looked normal on its own, but the trend in their relationship indicated declining heat transfer at the coil – consistent with early fouling. A work order went out to clean the coil before it forced the valve fully open, leaving the AHU unable to hold setpoint on a hot day.

Healthcare facilities

In mission-critical environments like hospitals, HVAC and other systems must function without interruption. Otherwise, they risk impacting patient safety and breaching strict regulatory requirements. Predictive maintenance enables hospitals to continuously monitor critical building systems to maintain uptime and help organizations meet strict compliance requirements.

Example of this in action

One hospital adopted predictive maintenance to achieve real-time visibility into conditions through IoT sensors. Analytics tracked a reheat coil valve and the air handling unit (AHU) serving an operating theatre. It detected a slow decline in the coil's heat-transfer performance over several days, consistent with fouling or a sticking valve. A work order was placed during a scheduled turnover to keep the theatre in service, and the compliance log was captured automatically.

Large campuses and multi-building portfolios

Predictive maintenance is essential for large portfolios because it allows small teams to manage assets across multiple buildings remotely. When energy drift is detected across the portfolio, it shifts maintenance intelligence into a strategic financial and sustainability lever.

Example of this in action

A commercial real estate group used centralized monitoring across its portfolio. When BAS data showed a growing gap between commanded and actual fan speeds on an AHU in a building, FDD diagnosed developing fan-belt wear.

Analytics ran a proactive check on other AHU equipment across the portfolio and identified others showing very early signs of degradation. The alert was entered into the CMMS, which prioritized work orders by urgency. A technician addressed each AHU before failure occurred, helping to extend asset life and avoid unplanned downtime.

Data centers

In data centers, thermal management is critical. Even the slightest HVAC failure can trigger an equipment shutdown and an SLA breach. For this reason, cooling uptime is non-negotiable – something predictive monitoring can support.

Data center energy efficiency is under increasing scrutiny, too. Power usage effectiveness (PUE) is one of the primary metrics by which data center operators are measured, and cooling systems are the biggest lever. Keeping them operating at peak efficiency is a constant business pressure.

Example of this in action

At a data center, continuous chiller monitoring detected a downward trend in suction pressure. Analytics flagged the pattern as consistent with early-stage refrigerant loss.

The alert generated a work order in the CMMS, dispatched a technician to inspect the unit and confirmed the issue as caused by a leak from a loose valve fitting. Catching the issue early kept the data center fully operational and energy efficiency on track.

The operational logic behind every predictive maintenance example: monitor, detect, act

For predictive maintenance to be successful in any environment or situation, the process requires seamless end-to-end integration. The six examples in this article involve different building types, equipment and faults. However, the underlying predictive maintenance logic is consistent: monitor equipment closely, detect anomalies and patterns, and act promptly before failure occurs.

• Monitor: Everything begins with continuous data collection. Whether that’s IoT vibration sensors on a pump bearing or existing BAS data on HVAC equipment, reliable real-time data is the key to predictive maintenance.
• Detect: The analytics layer processes the data. Issues can be caught using FDD software to identify specific fault patterns – such as an AHU and a VAV box heating and cooling simultaneously. It can also compare against expected baselines, such as a chiller underperforming in a data center.
• Act: After an issue is flagged by analytics, the CMMS automatically generates a work order. The technician will be given relevant information and details of the suspected fault to review. Work is scheduled according to the asset’s priority level.

Each step should be highly automated to accelerate response and reduce manual effort while still allowing appropriate human validation when needed. This flow setup should form part of the core predictive maintenance strategy and be incorporated at the implementation stage.

Each predictive maintenance example listed above highlights how the monitor-detect-act strategy works in practice.

Predictive maintenance works across every system you manage

Most organizations start with HVAC, but predictive maintenance is not limited to a single system or asset type. It’s an operational approach that applies to a full range of systems – chillers, central utility plants, building automation networks, healthcare environments, large campuses, and data centers. It delivers measurable value in each.

Its value stems from the same three-step process used across every environment: monitor equipment continuously, detect early warning signals before they become failures and act while intervention is still planned (rather than reactive). Predictive maintenance can prevent unplanned downtime, extend equipment lifespan and reduce energy waste.

What makes predictive maintenance successful is the ability to connect building data, analytics and maintenance workflows into a single operational process. Johnson Controls brings these capabilities together through its building automation expertise, advanced analytics and connected service ecosystem to help organizations move from reactive maintenance to proactive performance management.

FAQs

1. What does predictive maintenance look like in practice?

Predictive maintenance follows a clear operational logic: continuously monitor equipment, detect anomalies and developing issues, and act before failure occurs. The examples in this article each follow that same logic across different building systems.

For this process to work seamlessly, there need to be end-to-end automated workflows from initial alert to work order.

2. What are common maintenance schedule mistakes?

The most common mistake is treating maintenance as a fixed calendar, rather than according to equipment condition. Predictive maintenance analyzes real-time condition data and uses advanced analytics to predict issues and equipment failures before they become reality.

Preventive maintenance, on the other hand, is based on fixed calendar scheduling. While this is a step up from reactive ‘break-fix’ maintenance, it does mean all equipment gets serviced at intervals – regardless of condition. This could lead to overservicing or issues developing between service intervals.

3. Which building systems benefit most from predictive maintenance?

The building systems that benefit most from predictive maintenance are those that run continuously, are critical to operations, carry high replacement or failure costs, and have sufficient data to detect early warning signs.

For these reasons, HVAC systems are the ones most organizations start with. Chillers provide the most logical starting point, as they can account for up to 40% of a building’s energy consumption.