Micro-sensor networks for heritage fire protection: designing early warning systems without compromising historic buildings

This article examines distributed IoT-based sensor networks, referred to herein as “sensor networks.” These systems, which typically comprise low-cost micro-sensors, environmental sensors, and communication nodes, facilitate early warning and condition monitoring within cultural heritage environments.

The discussion centers on sensor typologies, communication architectures, alarm threshold configurations, and integration with video analytics and emergency response protocols.

While sensor networks offer significant early warning capabilities, they must be evaluated within the context of the site’s comprehensive fire safety strategy. Furthermore, systems requiring life-safety alarm functions must adhere to applicable local regulations and standards, and their design must be executed by qualified professionals.

1. Context and rationale

Historic buildings often feature:

• Complex spatial layouts (high naves, crypts, attics, light wells).
• Hidden combustible materials (wooden structures, straw, hard-to-reach storage areas).
• Aesthetic and structural constraints that limit drilling, cable routing, and the visibility of devices.

Conventional point-detector systems can be effective when properly designed, but in complex heritage environments they may leave blind spots if detector spacing, mounting height, smoke stratification, and architectural constraints are not carefully considered.

A well-designed distributed sensor network can:

• Reduce initial detection times (early warning).
• Pinpoint the origin of the anomaly more precisely.
• Monitor environmental parameters linked to risk (humidity, temperature, CO, VOCs).

However, there is a risk of turning the building into a “technological laboratory” that is difficult to maintain, prone to false alarms, and incompatible with heritage preservation constraints. 

The challenge is therefore one of design, not just technology.

2. Sensor types relevant to heritage sites

An effective network combines different types of sensors, selected based on the specific environments and risks involved.

2.1 Smoke detection

• Optical (photoelectric): suitable for smoke from smoldering combustion (cellulosic materials, furnishings, storage areas). Less sensitive to coarse dust than some older technologies, though calibration is still required in dusty environments.
• Ionization: more sensitive to invisible smoke from rapid combustion, but increasingly restricted or prohibited in some jurisdictions due to environmental regulations and disposal issues. Local rules must be checked.
• Aspirating smoke detection (ASD): samples air from multiple points via concealed piping. It is well suited to high-ceiling spaces, storage areas, and attics because it can provide excellent coverage with limited visual impact. However, ASD involves higher cost and complexity, requires careful design and maintenance, and may involve invasive piping; it is therefore not automatically a minimal-impact solution.
• Sensor performance depends on installation height, airflow, dust, humidity, and maintenance.

2.2 Heat detection

• Rate-of-rise detectors trigger when the temperature rises rapidly; they are useful in kitchens, utility rooms, and areas with dust or vapors that would trigger false smoke alarms.
• Resistance temperature detectors (RTDs) / thermistors: measure absolute temperature; can support dynamic threshold logic (e.g., alarm if T > seasonal threshold).

2.3 Gas and air quality

• CO (carbon monoxide): an early indicator of incomplete combustion, useful in environments with fireplaces, stoves, or heating systems. However, CO can also be produced by normal heating, candles, or other non-fire sources, so thresholds must be based on local baselines and event context.
• CO₂ (carbon dioxide): mainly an occupancy and ventilation indicator rather than a direct fire parameter. It can support contextual logic, for example “high occupancy + rising CO + rising temperature.”
• VOCs (volatile organic compounds): can indicate material overheating or off-gassing from glues, paints, resins, or restoration materials. VOC sensors are often less specific and may respond to cleaning products, adhesives, or normal environmental changes; they should therefore be used as contextual indicators rather than standalone fire detectors.

2.4 Environmental parameters

• Relative humidity: crucial for preservation, as well as for assessing electrical risks and the spread of certain types of fire.
• Vibration/tilt: these sensors may support structural condition monitoring in timber structures or suspended elements, but they should not be relied upon as primary fire detection devices.

Selection should be based on “homogeneous zones” (storage area, exhibition hall, nave, crypt, attic), taking into account:

• Materials present.
• Plausible ignition sources.
• Typical environmental conditions (dust, humidity, occupancy, use of incense, etc.).

3. Communication and power supply architectures

Communication technology and power supply are often the real bottlenecks in heritage buildings. 

3.1 Wired vs. wireless

• Wired (dedicated buses, conventional loops): high reliability and strong resistance to interference, but require cable routing, drilling, and invasive work; often incompatible with heritage constraints.
• Wireless: Wireless solutions drastically reduce structural impact, but they also expand the building’s digital attack surface. The design should include network segmentation, secure authentication, encrypted communication, controlled firmware updates, and clear procedures for loss of connectivity or gateway failure. Wireless systems also require attention to:
• radio coverage in thick masonry, vaulted ceilings, and crypts;
• interference with other systems such as Wi-Fi, audio systems, and security systems;
• battery management and maintenance.

In many cases, a hybrid solution is optimal: a wired backbone for accessible zones and wireless for areas with strict constraints or difficult access.

3.2 Wireless protocols

• LoRaWAN: long range, low power consumption; well suited to sensors distributed over large areas (monumental complexes, extensive archaeological sites). Requires dedicated gateways.
• NB-IoT / LTE-M: utilize cellular networks; good coverage and simplified management, but involve dependence on network operators and recurring costs.
• Wi-Fi: easy integration with existing infrastructure, but higher energy consumption and variable coverage in historic buildings.
• Bluetooth Mesh: suitable for dense networks in confined spaces (museums, storage areas), with effective management of numerous nodes.

The choice depends on: site size, available infrastructure, cybersecurity policies, and operating budget.

3.3 Power supply

• Battery: the most common solution. Battery sizing should consider sensor type, transmission frequency, environmental temperature, and maintenance access. Sleep modes and event-driven transmission strategies can extend autonomy. A sustainable battery replacement plan is essential, especially for inaccessible areas.

• Energy harvesting:
  • Photovoltaic (including small cells placed near windows or skylights).
  • Vibrations (in areas with machinery, airflow, or foot traffic).
  • Thermal differentials (between indoor/outdoor environments or near building systems).
  • Hybrid: battery + harvesting to extend service life and reduce maintenance. In heritage buildings, the ability to avoid installing electrical wiring is often a decisive advantage, provided there is a clear and sustainable battery replacement plan.

4. Defining alarm thresholds and data fusion logic

One of the major risks associated with a distributed sensor network is an excess of false alarms, which can lead to system deactivation, ignored warnings, or a loss of confidence in the system.

4.1 Typical sources of interference in heritage sites

Typical sources of interference in heritage sites include:

• dust from construction work, restoration, or cleaning;
• fluctuating humidity in non-climate-controlled churches, crypts, and underground spaces;
• use of incense, candles, or open flames during liturgical or ritual activities;
• seasonal crowding from events, temporary exhibitions, or pilgrimages.

These factors can trigger:

• False positives from optical smoke detection (dust, steam).
• CO/CO₂ fluctuations linked to ventilation and occupancy rather than fire.
• Temperature spikes caused by solar radiation, heating systems, or temporary events.

4.2 Static vs. dynamic thresholds

• Static thresholds: easy to implement but lack adaptability to changing conditions; risk of false alarms during specific seasons or times of day.
• Dynamic thresholds: based on:
  • Recent sensor history (local baseline).
  • Season/time of day (e.g., different temperature and humidity thresholds for summer vs. winter).
  • Type of event in progress (mass, concert, exhibition, maintenance work).

A response zone is a defined area mapped to a specific emergency response action, such as staff verification, localized alarm, partial evacuation, or full evacuation. An effective logic might combine:

• a pre-alert triggered when a parameter exceeds a moderate threshold for a prolonged period;
• a confirmed alarm when two or more independent sensors in the same response zone report correlated anomalies within a defined time window, preferably using different sensing principles where possible. For example, smoke plus CO, or smoke plus rapid temperature rise, may provide stronger confirmation than two identical sensors responding to the same local disturbance.

The allowed verification time for a pre-alert should be risk-based. A period of 2–3 minutes may be acceptable in low-risk, unoccupied areas, but it may be too long for occupied spaces or high-value collection areas.

4.3 Integration with video analytics

Video analytics can support verification and reduce false alarms when integrated with a sensor network. The logic may include:

• A smoke/CO alarm triggers an automatic video check of the affected area (scanning for visible smoke, flames, or unusual movement).
• Video analytics detecting smoke or flames can carry “greater weight” in the alarm decision if accompanied by a simultaneous rise in sensor readings.
• In the event of a conflict (sensors in alarm state, but no visual confirmation), the system can generate an “intermediate-level alarm” requiring rapid human verification, rather than an immediate evacuation alarm.

Video analytics can support verification but should not be treated as a standalone fire detection system. Its use must consider privacy requirements, data minimization, access control, and the possibility of false positives caused by steam, dust, incense smoke, lighting changes, or movement.

5. Design criteria for a reliable network

Designing a sensor network for heritage sites requires a scenario- and zone-based approach, rather than simple adherence to standard regulatory mapping.

5.1 Density and positioning

General guidelines (to be adapted to the specific case):

• Storage areas and technical rooms: higher sensor density (smoke + CO + T), given the presence of combustible materials and reduced continuous visual monitoring.
• Exhibition halls: discreet sensors, potentially integrated into furnishings, spotlights, or existing elements; priority given to optical smoke and temperature (T) sensors, with CO sensors included if technical systems are present.
• High-ceiling spaces (naves, atriums): high-level sensors (smoke, T) + optional ASD for multi-level sampling; attention must be paid to hot smoke stratification.
• Underground spaces (crypts, hypogea, tunnels): smoke, CO, and humidity sensors; assess radio communication at depth (a repeater or local gateway may be required).

Positioning must take into account:

• Likely smoke and heat propagation paths.
• Air stagnation zones.
• Strategically visible points for visual or video verification.

5.2 Redundancy and fault management

• Incorporate a degree of redundancy: multiple sensors per critical zone, ensuring that the failure of one does not leave the area unprotected.
• Implement “fault/low battery” signals managed as maintenance priorities.
• Define clear procedures: who receives the fault notification, the timeframe for intervention, and documentation requirements.

5.3 Integration with emergency and evacuation plans

The sensor network is not an end in itself; it must feed into the emergency plan. Map each sensor to a specific “response zone” within the emergency plan. Define standard scenarios:

• Pre-alert: verification by staff, without evacuation. Confirmed alarm: activation of partial or total evacuation procedures; call to the fire department.
• Link sensor network data to:
  • Alert systems (voice announcements, lights, staff apps).
  • Procedures for securing structures and collections (where applicable).

This approach should align with established emergency evacuation and safety planning principles for cultural heritage collections.

6. Hypothetical case study: small church with nave, sacristy, and storage room

The following case study is intentionally simplified and does not fully consider typical occupancy levels, opening hours, use of candles or incense, existing cameras, available power and network infrastructure, local fire regulations, response time objectives, or conservation constraints.

To illustrate this practically, let us consider a typical case: a small-to-medium-sized church featuring:

• A single vaulted nave, ~10 m high.
• An adjacent sacristy containing cabinets, vestments, and minor electrical systems.
• A basement storage room used for furnishings, liturgical materials, and boxes.

6.1 Proposed zoning and sensors

The number of sensors must be confirmed by a site-specific risk assessment and applicable detector spacing requirements. In this hypothetical case, an indicative layout would be:

• Nave: 2–3 combined wireless sensors (optical smoke + temperature) mounted at height and integrated into existing elements, such as light fixtures or non-protected decorative features. One CO sensor near heaters or fireplaces, if present. Because hot smoke may stratify in a 10 m high nave and may not reach low-level sensors quickly, multi-level sampling or ASD should also be considered.
• Sacristy: 1 smoke + temperature + CO sensor—preferably wired, if possible, given fewer heritage constraints. 1 humidity sensor to monitor conditions for preserving vestments.
• Basement storage: 2 smoke + temperature + CO sensors, given the higher density of combustible materials. 1 humidity sensor and, if relevant, a VOC sensor (for glues, varnishes, organic materials). In any case, a radio coverage survey is required for the basement. If LoRaWAN cannot reliably reach the basement, a local gateway, repeater, wired link, or alternative protocol could be adopted.
• Communication: a LoRaWAN gateway located in the sacristy or utility room, subject to radio coverage verification. Wireless LoRa sensors for the nave and storage room; sacristy sensors wired or Wi-Fi.

6.2: Alarm logic

Pre-alert: a single sensor exceeds the moderate smoke or CO threshold for a risk-based period. For example, 2–3 minutes may be acceptable in unoccupied storage areas, while occupied spaces or high-value collection areas may require a shorter verification time. The system sends a notification to the person on duty, via app or SMS, and activates a fixed camera, if present, for rapid verification.

Confirmed alarm: two or more independent sensors in the same response zone report correlated anomalies within a defined time window, preferably using different sensing principles where possible. For example:

• smoke + CO; or
• smoke + rapid temperature rise.

Alternatively, a confirmed alarm may be triggered by a sensor alarm combined with video analytics detecting visible smoke. The system activates a localized voice alarm, or a general alarm depending on the emergency plan, and initiates the fire department call-out procedure.

This logic limits false alarms caused by dust, incense, or temporary fluctuations while maintaining high sensitivity to actual fires. Attention should also be given to:

• a baseline monitoring period before final threshold tuning;
• event tagging for incense, candles, maintenance, cleaning, exhibitions, or occupancy peaks;
• periodic sensor calibration and cleaning;
• review of false alarm causes after each event;
• maintenance priorities for contaminated or drifting sensors;
• avoiding automatic system deactivation due to alarm fatigue.

7. Operational checklist for designers and managers

Before sizing a sensor network in a historic building, it is useful to ask the following key questions.

7.1 Contextual questions

• What are the main plausible ignition sources (systems, candles, incense, maintenance work, events)?
• Which areas are less visually monitored (storage rooms, attics, underground spaces)?
• What aesthetic and structural constraints limit installations and cabling routes?

7.2 Technology choices

• Which sensor types are priorities for each zone (smoke, heat, gas, environmental parameters)?
• Which communication architecture is most suitable (LoRaWAN, NB‑IoT, Wi‑Fi, hybrid)?
• How can a sustainable power supply be ensured (battery, energy harvesting, hybrid)?
• How will cybersecurity, firmware updates, and device authentication be managed?
• How will camera data and sensor data be governed in terms of privacy, retention, and access?

7.3 Operational management

• Who receives and manages pre-alerts and alarms (roles, shifts, on-call availability)?
• How are false alarms and system faults handled (procedures, response times, documentation)?
• How is the sensor network integrated into the existing emergency and evacuation plan?
• What is the commissioning process, including baseline monitoring before final threshold tuning?
• How often will sensors be calibrated or cleaned?
• What is the procedure for reviewing false alarms after each event?
• How will staff be trained to interpret pre-alerts and confirmed alarms?

Conclusion: an outlook and future developments

Distributed sensor networks for heritage sites are set to evolve in three main directions:

• greater integration with building management platforms and digital twins, providing a unified view of safety, conservation, and building usage;
• artificial intelligence algorithms for detecting anomalous patterns, such as combinations of temperature, humidity, CO, and VOC levels, to reduce false alarms;
• standardization of guidelines specific to heritage sites, which currently remain fragmented compared with those for conventional buildings.

In this context, the designer’s role is not merely technical but also acts as a bridge—or “translator”—between preservation requirements, fire safety needs, and the opportunities offered by new technologies.