Predictive Maintenance with BMS: How Fault Detection Prevents HVAC Failures

An HVAC system failure in a Dhaka commercial building in July is not an inconvenience — it is a crisis. Temperatures can exceed 38°C outdoors and climb rapidly inside. Productivity collapses, server rooms overheat, and healthcare facilities face patient safety risks. Emergency repair crews charge premium rates, critical spare parts may take days to source, and the reputational damage of a facility failing its occupants lingers long after the equipment is repaired.

Yet in most Bangladesh commercial buildings, HVAC maintenance is reactive: problems are discovered when equipment fails, not before. A Building Management System with Fault Detection and Diagnostics (FDD) capability changes that equation fundamentally. This article explains how BMS-based predictive maintenance works, what faults it detects, and what it saves.

What Is Fault Detection and Diagnostics (FDD)?

Fault Detection and Diagnostics, commonly abbreviated as FDD, is an analytical capability that continuously monitors BMS sensor data and control parameters to identify patterns that indicate abnormal equipment behaviour. FDD does not wait for a piece of equipment to stop working. It looks for the early warning signs — the subtle changes in performance that precede failure — and alerts your maintenance team while there is still time to intervene.

The word "diagnostics" is as important as "detection." It is not enough to know that something is wrong. Your maintenance team needs to know what is wrong and why. A good FDD system not only raises an alert but also identifies the probable root cause, the likely consequence if the fault is unaddressed, and the recommended corrective action. This actionable intelligence is what separates FDD from simple BMS alarming.

FDD operates by comparing actual equipment behaviour against a model of how the equipment should behave under current operating conditions. Deviations from expected behaviour trigger fault flags. The sophistication of that comparison model determines whether the FDD is rule-based or AI/ML-based — two distinct approaches with different strengths.

Rule-Based FDD: Logic That Never Sleeps

Rule-based FDD encodes the knowledge of experienced HVAC engineers into logical rules that the BMS evaluates continuously. Each rule describes a fault condition — a specific combination of sensor readings, control signals, or operational parameters that indicates a particular problem.

How Rules Are Constructed

A rule might state: "If the chilled water supply temperature is more than 2°C above the chiller setpoint for more than 15 minutes during compressor operation, flag a Chiller Capacity Degradation fault." Another might state: "If AHU supply fan is commanded ON but supply air static pressure is less than 20 Pa for more than 2 minutes, flag a Supply Fan Failure fault."

These rules are built from first principles of HVAC engineering and from empirical knowledge of how particular faults manifest in sensor data. The ASHRAE Research Project 1020-RP (now embodied in ASHRAE Guideline 36) provides a library of standardised FDD rules for common HVAC equipment types, giving BMS vendors and system integrators a validated starting point.

Advantages and Limitations of Rule-Based FDD

Rule-based FDD is transparent and explainable — you can see exactly why an alert was raised and verify the rule logic against your engineering knowledge. It is also computationally simple, running within standard BMS controllers without specialised hardware.

The limitation is that rules must be written for each fault type. A fault pattern that was not anticipated when the rules were written will not be detected. Rules also require tuning for each specific installation: a rule written for a centrifugal chiller does not apply to an absorption chiller, and threshold values that work well in a well-insulated building may generate false alarms in a less insulated one.

AI/ML-Based FDD: Learning from Your Building's Own Data

Machine learning-based FDD takes a fundamentally different approach. Instead of encoding expert rules, an ML model learns the normal operational patterns of your specific equipment from historical BMS data and then flags deviations from those learned patterns.

The Training Phase

During an initial training period — typically 4 to 8 weeks of normal operation — the ML model observes thousands of variables simultaneously: supply temperatures, return temperatures, valve positions, fan speeds, outdoor conditions, occupancy patterns, and energy consumption. It builds a multi-dimensional model of how all these variables relate to each other under normal, healthy operation.

Once trained, the model monitors incoming sensor data in near-real time. When it detects a pattern that deviates from its learned normal — even if that pattern does not match any pre-defined rule — it flags it for investigation. Because the model has learned the relationships between many variables simultaneously, it can detect complex, multi-variable faults that would be nearly impossible to capture in a simple rule.

AI/ML FDD in Practice

An ML-based FDD system might detect that a chilled water pump is consuming slightly more power than usual, while simultaneously the differential pressure across the evaporator is slightly lower than normal and the chilled water flow is slightly reduced — a pattern consistent with impeller wear or partial blockage in the pump strainer, but none of the individual deviations large enough to trigger a rule-based alarm independently.

Cloud-based ML FDD platforms can also aggregate data across many buildings, learning from faults detected in hundreds of HVAC systems globally and applying that collective knowledge to detect rare fault types that might never appear in a single building's data.

Common HVAC Faults That BMS FDD Detects Early

Refrigerant Leak

A refrigerant leak is one of the most consequential HVAC faults because it degrades cooling capacity progressively and, if undetected, can cause compressor failure — the most expensive component in any chiller or DX system. Early indicators visible in BMS data include: rising discharge pressure relative to condenser water temperature, decreasing evaporator superheat, falling chiller COP, and increased compressor current for the same cooling load.

Without FDD, a refrigerant leak may go undetected until the system can no longer maintain setpoint on a hot day. With FDD, the BMS detects the performance drift weeks earlier, allowing a planned refrigerant charge and leak repair before the compressor is stressed.

Belt Wear and Failure in Air Handling Units

Many older AHUs in Bangladesh commercial buildings use belt-driven fans. Belt wear is gradual: as the belt stretches and its surface degrades, it slips slightly on the pulleys. This slip wastes motor energy, reduces airflow below the designed rate, and eventually leads to belt failure and complete loss of ventilation.

BMS FDD detects belt wear by monitoring the relationship between motor current (or speed command), supply fan speed (via duct static pressure or airflow measurement), and the expected static pressure for that fan speed. A widening gap between commanded speed and achieved airflow indicates increasing slip. The BMS can alert the maintenance team to inspect and replace the belt before it breaks.

Fouled Cooling Coils and Heat Exchangers

Over time, cooling coils accumulate dust, biological growth, and scale — particularly the condenser coils of cooling towers in Bangladesh's humid, dusty environment. A fouled coil has higher thermal resistance, meaning more temperature difference is required across it to transfer the same amount of heat. This forces compressors to work harder, increasing energy consumption and mechanical stress.

BMS FDD detects coil fouling by monitoring the approach temperature — the difference between the condenser water temperature leaving the cooling tower and the outdoor wet-bulb temperature. As the tower fills or the coil fouls, the approach temperature rises. An increasing approach temperature trend indicates it is time for cleaning, weeks or months before performance degrades to the point of failure.

Simultaneous Heating and Cooling

In HVAC systems with both heating and cooling coils (common in hospitals, laboratories, and high-specification office buildings), a controller fault or actuator failure can cause the heating valve and cooling valve to both open simultaneously. This wastes enormous amounts of energy — the chiller cools air that the heater immediately re-heats — but often does not cause an obvious symptom because the zone temperature may remain on setpoint.

BMS FDD detects this by monitoring the positions of heating and cooling valves together. If both are open by more than a threshold amount simultaneously, it flags a Simultaneous Heating and Cooling fault. Studies in comparable climates suggest this fault is present in 15–30% of HVAC zones at any given time in buildings without FDD, wasting significant energy silently.

Variable Frequency Drive (VFD) Faults

VFDs control pump and fan motor speeds and are critical components in energy-efficient HVAC systems. Common VFD faults include overheating, voltage sags causing underperformance, and capacitor degradation. The BMS monitors VFD output frequency, motor current, and drive temperature. Deviations from normal operating envelopes — particularly increasing drive temperature or unexpected current spikes — indicate developing faults before a VFD trips on protection.

Stuck Dampers and Control Valves

Pneumatic and electric actuators on dampers and valves can fail in partially open or fully closed positions. A stuck outside air damper may leave a building with inadequate ventilation (a health issue) or excessive infiltration (an energy waste). A stuck chilled water valve may over-cool a zone, cause occupant discomfort, and waste chiller energy.

BMS FDD detects stuck actuators by comparing the commanded position with the expected system response. If a cooling valve is commanded to 0% but the chilled water differential temperature across that coil is still high, the valve is probably not closing properly.

Alarm Management: Prioritising by Severity

An FDD-equipped BMS can generate a very large number of alerts. Without intelligent alarm management, facility teams face alarm fatigue — so many notifications that critical alerts are missed in the noise. Effective alarm management is therefore as important as fault detection capability.

ANSI/ISA-18.2, the international standard for alarm management (applicable to building systems as well as industrial processes), defines a hierarchy of alarm priorities:

• Critical (Priority 1): Immediate response required. Equipment failure that causes safety risk, loss of cooling to critical loads (server rooms, ICUs), or fire/life safety impact. Response target: 10 minutes.
• High (Priority 2): Urgent response. Equipment failure causing significant occupant discomfort, energy waste above defined threshold, or imminent risk of consequential damage. Response target: 1 hour.
• Medium (Priority 3): Planned response. Developing fault, early-stage performance degradation, or maintenance indicator. Response target: within 24 hours.
• Low (Priority 4): Information. Trend approaching threshold, minor setpoint deviation, scheduled maintenance reminder. Response target: within 1 week.

A well-configured BMS FDD system ensures that a stuck damper in a storage room generates a Low or Medium alarm, while a chiller tripping on high discharge pressure in August generates a Critical alarm. The facility team sees the right level of urgency for each situation.

Maintenance Cost Reduction: What the Numbers Show

Independent studies and field data from BMS installations consistently report maintenance cost reductions of 30–50% when rule-based or ML-based FDD is implemented and acted upon. These savings come from several sources:

• Elimination of emergency call-outs: Emergency repairs typically cost 3–5 times more than planned maintenance for the same work scope. Converting emergency interventions to planned ones is the single largest source of maintenance savings.
• Extended equipment life: Equipment that runs closer to its design operating envelope lasts longer. A chiller that avoids compressor stress from refrigerant leaks and fouled condensers may reach 25–30 years of service life rather than 15–18 years.
• Reduced spare parts consumption: When faults are caught early, the damage is typically limited to a single component (a bearing, a belt, a refrigerant charge). Left undetected, the same fault may cascade to failure of the primary equipment and secondary components.
• Labour optimisation: When the BMS tells your technician exactly what fault has occurred and its probable cause, the technician arrives prepared with the right tools and parts. Diagnostic time is reduced, and the repair is completed correctly the first time.

For a typical Dhaka commercial building spending BDT 20–30 lakh per year on HVAC maintenance, a 35% maintenance cost reduction represents BDT 7–10 lakh in annual savings.

Bangladesh Context: Why Hot Climate Amplifies HVAC Fault Risk

Bangladesh's tropical climate places exceptional stress on HVAC equipment compared to temperate climates. Outdoor air temperatures frequently exceed 36°C during the pre-monsoon period (April–June), and humidity remains high throughout the year. These conditions mean:

• Chillers run at or near full load for extended periods, leaving no spare capacity to mask developing faults
• Condenser systems (cooling towers and air-cooled condensers) operate in high wet-bulb conditions, reducing their heat rejection efficiency
• Outdoor air intakes draw in hot, humid, dusty air that accelerates coil fouling
• High-load operation accelerates wear on compressors, bearings, and belts

This climate context makes predictive maintenance not merely an economic preference but a practical necessity. A chiller failure in April or May — when every HVAC technician in Dhaka is overwhelmed with emergency calls — means days of outage rather than hours. The BMS FDD capability that gives you 2–4 weeks of advance warning of a developing chiller fault is the difference between a planned repair during a cool period and an emergency breakdown at the worst possible time of year.

Getting FDD Right: Implementation Essentials

FDD is only as good as the data it receives. Several implementation factors determine whether your BMS FDD delivers its promised value:

Sensor coverage and quality: FDD requires sensors at all the measurement points the rules or ML models need. An AHU without supply air temperature, return air temperature, and static pressure sensors cannot be monitored for the most common faults. Investing in comprehensive sensor coverage at the design stage is far cheaper than retrofitting sensors later.

Commissioning and verification: BMS sensors must be calibrated and their readings verified against reference measurements during commissioning. A temperature sensor that reads 2°C high will generate false alarms and cause the FDD to miss real faults. Proper commissioning is non-negotiable.

Tuning alarm thresholds: Default FDD rule thresholds are starting points, not final settings. During the first months of operation, thresholds should be tuned to match your building's specific equipment, operating patterns, and climate exposure. A threshold that generates too many false alarms will cause your team to stop responding; one set too conservatively will miss real faults.

Maintenance follow-through: FDD only delivers value if maintenance teams respond to and close out alerts. A BMS work order system that links FDD alerts to maintenance tasks, tracks response times, and records the outcome of each inspection ensures that your FDD investment translates into actual equipment protection.

Start Protecting Your HVAC Investment Today

Every month that your HVAC system operates without FDD is a month when developing faults can progress undetected. In Bangladesh's climate, with HVAC assets worth crores of taka, that is a risk you do not need to carry.

SSAL designs and commissions BMS installations with FDD capability built in from the start. Our engineering team configures fault rules appropriate for your specific equipment, tunes alarm thresholds during the commissioning period, and trains your facility management team to interpret and respond to FDD alerts effectively.

Get Free Consultation — Contact SSAL to discuss how BMS-based predictive maintenance can protect your HVAC assets, reduce your maintenance costs, and prevent the next equipment failure before it happens.