How predictive maintenance optimizes HVACs operations?

Introduction

Heating, ventilation, and air conditioning are all components of a well-integrated HVAC system. Such systems are installed in all large buildings, and they work non-stop 24 hours a day to cool/heat the structure and control the air quality. So a fully functioning HVAC system is crucial for a healthy and comfortable building and the smooth operation of an HVAC unit can maximize energy efficiency. While routine maintenance checks are often completed as part of an HVAC service program, they don’t always identify problems that can lead to machine failure. Many faults such as clogging the filter, pump faults, stuck valves and damper, and fouling Heat exchanger preclude a HVAC system to operate in its higher performance [1]. One way to deal with this problem is to use fault detection and diagnostics (FDD) algorithms or at a more advanced level predictive maintenance (PM) strategies that use condition monitoring tools and machine learning algorithms to detect various deterioration signs, anomalies, and equipment performance issues.

How PM and FDD work

HVAC systems are commonly equipped with different sensors to monitor data points in real time (e.g., temperatures, flows, pressures, actuator control signals, and so on) and then apply a set of rules to detect faults. To be more specific, in HVAC, there is a diagnostic or health check system that sets some error codes in the event of an emergency by setting consistent threshold and set point values. There are standards for systems that include a chiller, a boiler, air handling units which receive hot and chilled water, and terminal units which receive supply air from the air handling units, for example. If the air handling unit has staged heating and cooling or is a single-zone air handling unit, a separate set of criteria would apply. There are also varied rules for the same equipment depending on its condition. A chiller, for example, will have one set of rules when it is turned off, another set of rules when it is turned on, and yet another set of rules while it is in its steady state. If the real-time data does not follow the guidelines or the best practices, the analytics tool will detect a problem. This mechanism, however, is not able to detect all types of faults and errors.

Figure 1. An example of where and how sensors fit in an AHU

In a FDD algorithm, collected data from different heating ventilation and air conditioning (HVAC) installations using Internet of Things (IoT) devices and a building automation system (BAS) are used accompanied with a machine learning model to predict failures and generate a model of system behavior to detect and alert hidden errors [2]. With an almost similar mechanism, PM techniques reveal early problems in performance caused by anomalous behavior and predict failures before they occur, and dictate tasks that must be completed to keep the machine running at optimal levels. These two remedies can also provide a root cause for some forms of alerts.

Check how InfiTelligence provides PM and FDD

Advantages of PM and FDD

Hidden failure detection and prediction mechanisms benefit building and facility managers:

• Maintenance costs are significantly reduced because maintenance only occurs in the case of a failure or poor air quality.
• HVAC systems are one of the most energy-consuming devices in a building, and if they are operated incorrectly or inefficiently, energy waste can be significantly increased. So FDD and PM strategies can prevent wasted or excessive energy consumption by identifying failures in a timely and cost-effective manner.
• Increase HVACs’ lifetime by keeping HVACs healthy and in optimum mode.
• When an error prediction occurs in PM techniques, the administrator has enough time to coordinate with an expert and prepare the necessary parts.
• High air quality provides a good experience for customers.
• Recommend root causes of of unforeseen failures .
• Maximize asset uptime and improve asset reliability.
• Diminish operational expenses by performing maintenance only when necessary.
• Streamline maintenance costs through reduced equipment, inventory costs, and labor.

Conclusion

PM and FDD algorithms are based on monitoring and data analysis of the actual and current state of the operating equipment. Implementing these plans can extend equipment life and reduce downtime and unplanned costs by quickly providing an estimate of when equipment will fail.

References

1. Lecamwasam, L., Wilson, J. & Chokolich, D., 2012. Guide to Best Practice Maintenance & Operation of HVAC Systems for Energy Efficiency.
2. Bouabdallaoui, Y., Lafhaj, Z., Yim, P., Ducoulombier, L., & Bennadji, B. (2021). Predictive maintenance in building facilities: A machine learning-based approach. Sensors, 21(4), 1044.