Integrating Thermal Images with HBIM for the Sustainable Evaluation of a Historic Building: Case Study of Rowheath Pavilion, Bournville

Abstract

Sustainable management of built heritage is a complex process made more difficult by the need to maintain significance and comply with any associated protections. Historic Building Information Modelling (HBIM), an information management and modelling process, has the potential to assist with this. This paper investigates how a pre-existing Historic Building Information Modelling (HBIM) model can be used to assist the energy-efficient interventions at the historic Rowheath Pavilion in Bournville, England. Two potential methods (HBIM and building energy modelling integration and HBIM and thermal image integration) are evaluated against their outputs and the available resources. Subsequently, the paper presents a case study wherein a thermal image survey was undertaken at Rowheath Pavilion and the resulting images were integrated with the pre-existing HBIM model. The apparent thermal performance of the pavilion was qualitatively evaluated. The described method was easy to apply and repeat. The integration of thermal images combined with the visualisation capabilities of HBIM resulted in the identification of energy inefficiencies and allowed the Rowheath staff to implement immediate small-scale changes to improve the sustainability of the pavilion.

1. Introduction

1.1. Rowheath Pavilion

Bournville, in Birmingham, is the village founded and built by the Cadbury brothers (of Cadbury Chocolate) with the intention of improving the quality of life of the workers and their families [1]. As well as residential dwellings, the development of Bournville village also included the provision of recreational facilities and sports grounds. The centre of these recreational facilities was Rowheath Pavilion (Figure 1), which opened in 1924 and provided amenities for sports activities and a function space for Cadbury employees [2].

The pavilion is owned by Bournville Village Trust (BVT), a charitable trust set up in 1900 to act as an ongoing custodian of the village [1]. However, over the years, the pavilion fell out of use, and its condition deteriorated. Trinity Christian Centre Limited (TCC) took over the lease of the pavilion In 1997 and has operated the site as a community centre since 2003 [4].

In 2017, Rowheath Pavilion was designated Grade II listed by Historic England. A Grade II listing provides legal protection for buildings with a special architectural or historic interest and requires additional consents to be gained before alterations can take place [5]. The justification for the designation of Rowheath was the “architectural quality” of the building (designed by John Ramsay Armstrong) and the “intact survival” of the original structure [2]. The main pavilion building has remained mostly unchanged since its construction, with the exception of a few minor extensions and alterations.

1.2. The Problem

Whilst Rowheath Pavilion is a thriving community centre, TCC is committed to ensuring the longevity of Rowheath Pavilion and is currently fundraising to improve the later additions to the building [4], which are not considered of special architectural or historical interest [2]. However, interventions are made more difficult by the Grade II listing and by the many external interests in Rowheath Pavilion (e.g., the need to gain consent for any changes from BVT, the board of trustees, the City Council, the Cadbury Family, the local Conservation Officer, etc. [6,7]).

The issue of sustainable intervention in a heritage context is of increasing interest to the global community, and the sustainability of cultural heritage is noted as a constituent part of the United Nations’ Sustainable Development Goals (UN SDG 11.4 [8]). Furthermore, when considering the environmental sustainability of buildings, it has been argued that environmental sustainability (encompassed by UN SDG 13 [9]) can be better achieved by improving existing buildings as opposed to constructing new, more efficient ones [10].

Consequently, Rowheath Pavilion has been the subject of three unpublished MSc projects at the University of Birmingham, two in 2022 [6,7] and one in 2024 [3], which have provided the foundation of the work outlined in this paper. Nguyen [6] and Jiang [7] carried out their investigations simultaneously. They conducted interviews with key individuals involved with Rowheath to establish the history and current situation of Rowheath, both in terms of the building fabric and in regard to its past and present use.

The work carried out by Nguyen [6] investigated the application of Historic Building Information Modelling (HBIM) to Rowheath Pavilion. BIM is a 3D modelling and information management technique commonly employed in the construction of new assets. BIM models typically consist of parametric modelling objects semantically enriched with additional information. The application of BIM to heritage was first proposed in 2009 [11], and interest in applying BIM to existing buildings has grown over time [12,13,14,15,16]. Nguyen [6] carried out a Terrestrial Laser Scanning (TLS) survey of the site and subsequently used Revit 2025 to create bespoke parametric families for an HBIM model of Rowheath Pavilion. The parametric families depicted the actual surface geometry of the real objects (according to the geometry provided by the TLS survey) but did not depict object-specific variation caused by localised degradation, e.g., the effects of weathering. The objects were manually aligned with the point cloud. Where known, the families were semantically enriched with information regarding the building materials, obtained via interviews with building staff [6,7]. The resulting model was demonstrated to the architects working on the proposed renovation project in the Fulmax Wide BIM Cave (Fulmax Wide OS v2024.2) [17], a 3D visualisation environment, at the University of Birmingham. The model was subsequently updated in response to their feedback. Whilst the model at this stage lacks detailed information on the construction techniques and material makeup, the intention is that future work undertaken by TCC and University of Birmingham students will result in additional information being added to the HBIM model. The HBIM model is depicted in Figure 2.

Jiang [7] investigated sustainable renovation approaches for Rowheath encompassing environmental, economic, social and cultural sustainability. They conducted further surveys with stakeholders of Rowheath Pavilion (primarily the local community and active users/employees of the pavilion) to gather additional perspectives. The surveys revealed a strong desire, from all those surveyed, to implement sustainable energy solutions at Rowheath Pavilion. Therefore, two potential sustainable energy solutions were investigated (solar panels and geothermal heat pumps), and potential impacts were evaluated. The proposed schemes were considered theoretically worthwhile by Rowheath staff due to their wish to reduce the carbon footprint of Rowheath and the fact that, even though minimal financial savings were anticipated, the reduction in energy costs would provide TCC with additional funds to support community activities.

The work carried out by Puri and Davies [3] investigated the integration of thermal performance information with HBIM and provides much of the content of this article. The work was founded on qualitative evidence (interviews with building staff [3,7]) of thermal issues with Rowheath Pavilion. These issues can be briefly summarised as follows:

• The café, lower office rooms, and staff room are found to overheat in the summer and be too cold in the winter.
• Drafts are experienced by occupants of many rooms.
• A lack of thermostatic controls in rooms leads to inefficient energy use and uneven heating, with some rooms being too cold and some too hot.
• Energy bills are increasing at a considerable rate.

Al horr et al. [18] and Zare et al. [19] state that buildings that do not provide appropriate thermal comfort for occupants result in non-optimal energy usage and solutions. For instance, some small-scale solutions have been adopted by Rowheath Pavilion to improve occupant comfort, including using shrink wrap on windows as temporary insulation and providing personal heaters to staff in cold rooms. However, the window wraps have not been effective, and the personal heaters are not viable as a long-term solution due to their effect on operating costs [3].

It became clear that, to make long-lasting, substantial improvements to the thermal performance of Rowheath Pavilion, larger interventions are required. However, qualitative, anecdotal evidence alone was considered insufficient to support the business case for any interventions and to gain the required consents for proposed changes. Retrofit is known to be more complex in historic structures due to additional constraints imposed by the need to maintain the significance of the structure [20] and the need for any interventions to be the minimum required to achieve the required result [21,22,23]. Thus, a more evidence-based approach that can be easily shared with others was required.

It was proposed that the existing HBIM model could assist with this purpose. Since its conception, a large proportion of HBIM research has focused on increasing the efficiency and accuracy of model creation [13,24,25,26,27,28,29]. However, as the field of HBIM has matured, some research focus has shifted to applications for the HBIM models after their creation, perhaps facilitated by increased access to pre-existing HBIM models. These applications are often related to routine conservation tasks such as assessing the physical condition of assets and planning subsequent interventions [30,31,32]. However, more pertinently, HBIM is also used for large-scale energy performance evaluation and intervention planning, often referred to as ‘Green Retrofit’ or ‘Energy Retrofit’ [20,33,34] among other terms.

Therefore, this article aims to evaluate how the existing HBIM model can be used to support energy efficient interventions at Rowheath Pavilion.

Two potential approaches (HBIM and Building Energy Model (BEM) integration and HBIM and thermal image integration) were considered as potential solutions. Section 1.3 provides an overview of the two approaches as well as an evaluation of their suitability for Rowheath Pavilion. Subsequent sections will detail the application of the chosen method. Future work is discussed in Section 5.

1.3. Potential Approaches

This section will evaluate the two approaches that were originally proposed: HBIM and BEM integration and HBIM and thermal image integration. Existing case studies, prerequisites and possible outputs will be discussed and then compared with the available resources at Rowheath Pavilion. The intention of the study was to assist TCC with the long-term sustainability of Rowheath Pavilion. Whilst the investigators would implement and test the chosen method in the first instance, it was intended that TCC staff should be able to repeat the method independently in the future. Hence, evaluation of the approaches will also consider the capabilities and resources of Rowheath staff.

Whilst some studies investigating HBIM to aid with energy retrofitting of heritage assets have only used HBIM for assessing the visual impact of proposed interventions [32], most recent studies have investigated the integration of HBIM and BEM, a term often used interchangeably with Building Energy Simulation (BES). These are static simulations that can be used to assess the current performance of a building or the impact of any proposed changes, the latter of which is the most frequent use case [20,35,36]. Dynamic systems that integrate BIM and live energy monitoring (digital twins/shadows) have been the subject of research [37] for many years but are only beginning to achieve useful results.

HBIM and BEM integration studies [34,35,38,39] for historical structures typically investigate the alteration of existing heating, ventilation and air conditioning (HVAC) systems, as this is considered more feasible than altering the building fabric in a historical context. A notable exception to this is provided by Iskandar et al. [40], who investigated the impact of altering the physical fabric of a historic dwelling on its energy performance. However, it should be acknowledged that the feasibility of the alterations from a planning perspective was not considered.

Etemad et al. [38] is a good example of HBIM integrated with BEM. They created an HBIM model of a museum in Tehran that included all building information, including material properties and all details of Mechanical, Electrical and Plumbing (MEP) properties (e.g., HVAC systems, pipework, ducts, etc.). They utilised Design Builder to carry out BEM simulations of the current building performance as a base case. The simulation was validated using temperature and weather data collected from on-site sensors, and they found a good degree of similarity between the real measurements and simulation. With the validated base case, they were able to evaluate the performance of proposed new HVAC solutions.

It is important to note here that a pre-existing HBIM model is not a prerequisite for carrying out BEM simulations, and there are existing case studies for historic buildings that have successfully investigated heritage interventions without a pre-existing HBIM [20]. In fact, several simulation software experience interoperability issues with HBIM [41,42], and consequently, HBIM and BEM integration remains an infrequent occurrence [41]. However, Etemad et al. [38] noted two key benefits of having an HBIM model: (i) it was possible to discount solutions that would not fit in the available space or that would damage the historical surface using the clash detection features of BIM software; (ii) having a pre-existing HBIM model streamlines the BEM process, a benefit also identified by Gao et al. [43].

It is an acknowledged truth that the as-designed energy performance of a building varies from its operational performance [19]. This is due to differences between the assumed and real behaviour of both the building and its occupants [18]. Therefore, whilst assumed base cases can be created and used for simulating proposed solutions [39], the most successful examples of BIM-BEM simulations [10,36,38,40,41,43] have relied on not only a high degree of knowledge about the building (e.g., details of building materials, all HVAC systems, and locations of ducts and pipework) but also site data for validation (e.g., exact weather and temperature data collected by sensors or exact energy bills).

No site data is currently available at Rowheath, and whilst previous energy bills were provided by the Rowheath staff, there was no way of distinguishing the energy used in the original building and the unlisted later addition (for which no HBIM exists and the construction differs [3,6,7]).

Furthermore, like with many historic structures, the exact wall composition for Rowheath is unknown, so only assumed material characteristics (e.g., U-values) can be used. This knowledge gap is difficult to fill since the historical significance of the building and the legal protection provided by the Grade II listing mean that destructive exploratory investigations are not feasible [22,23]. Likewise, whilst non-destructive methodologies [44], such as Raman spectroscopy [45], ground penetrating radar (GPR) [46] and infrared thermography [47,48], are gaining prominence for material characterisation for built heritage, these methodologies often require a high degree of technical knowledge for their execution and interpretation. They would also entail a considerable financial investment to purchase the equipment or commission the survey. Hence, since the suggested methodology was intended to be repeatable for Rowheath staff, they are not feasible at this time.

Sofronievska et al. [10] overcame uncertainty in material properties by placing temperature sensors on either side of an external wall and calculating the u-values. If further funds were made available to purchase sensors, a similar approach could be utilised in the future at Rowheath. However, this was not possible at the time of writing. It is possible to calculate u-values using the outputs of thermal images, but this still results in some uncertainty [49]. Without reliable validation data at Rowheath for potential BEM outputs, this uncertainty was deemed too great. Therefore, whilst BEM simulations have proven applicability in a heritage setting, they were deemed unsuitable for Rowheath Pavilion at this time.

Since BEM simulations were discounted, it was instead proposed that insight into the thermal performance of the building could be obtained by utilising thermal imaging cameras. Thermal imaging has been utilised in previous case studies investigating energy performance of structures [19] and is suggested as an appropriate methodology by the Royal Institute of British Architects (RIBA) ‘Conservation Guide’ [22]. However, there are concerns about the accuracy of the obtained images in several studies [50,51,52,53,54]. However, the authors do not feel these inaccuracies are significantly detrimental to the outputs of the current study. The resulting thermal images could then be integrated with the Revit model, leveraging the improved visualisation capabilities provided by HBIM [55].

There is considerable justification for the proposed approach. Depicting condition results visually within HBIM is a known use case, having been applied for several existing case studies [31,56,57], and there are existing examples [55,58] where the visualisation capabilities of HBIM have been used to identify the sources of maintenance issues that were not otherwise clear. Depicting information in a visual way has also been shown to aid the dissemination of the results to a non-expert audience and consequently enable proactive decision making [59]. Furthermore, the use of thermal imaging cameras and the integration of images within Revit require minimal technical knowledge and digital proficiency. This is desirable since Agliata et al. suggest [31] solutions should not be complex and should not rely on a high amount of human skill. Likewise, Lovell et al. [55] identified perceived digital literacy as a key barrier to wider HBIM implementation.

Table 1. Comparison of the two proposed approaches and the available resources (“x” indicates not required for method or available at Rowheath; “✓” indicated required for method or available at Rowheath).

Resources Required/Available	BEM (Required)	Thermal Images (Required)	Rowheath (Available)
3D dimensions	✓	✓	✓
Validation data (energy bills or in situ sensor data)	✓	x	x
Knowledge of building fabric	✓	x	x
Software/hardware	BEM software Optional HBIM model	HBIM model Thermal camera	HBIM model Thermal camera
Technical capability and/or knowledge needed to repeat method	High	Low	Low: Staff have no prior experience with HBIM models, thermal imaging or BEM. Training would be required for all approaches.
Outcome	Achievable: Current and future performance of both building fabric and HVAC systems	Achievable: Current performance of physical fabric only. No evaluation of HVAC systems.	Desired: Evidence to support qualitative reports of issues experienced. Evidence to support design of any interventions.

summarises the discussion by comparing the prerequisites of the two proposed approaches to the data and resources available at Rowheath. It is evident that, solely based on the available resources, the integration of thermal images was the most appropriate method. The remaining sections of this article detail the application of the chosen method at Rowheath and evaluate its effectiveness.

2. Methodology

2.1. Methodological Constraints

Since the aim of this article is to evaluate how the existing HBIM model can be used to support the identification of energy-efficient interventions at Rowheath Pavilion, the methodology was designed to be repeatable by all Rowheath staff, who had no prior experience in thermal analysis and minimal HBIM experience. Digital literacy and technical capability are major perceived barriers to the adoption of digital technologies in the heritage sector [55]. Therefore, the investigators aimed to make the chosen methodology as simple to implement as possible. Therefore, as far as possible, the automatic or default settings were used for all data collection, exportation and processing. Furthermore, whilst quantitative evaluation of thermal performance, such as u-value calculations, may have been possible with additional validation activities (e.g., using contact thermometers to validate the measured temperature), it was decided that quantitative analysis would make the methodology less accessible to the intended end user whilst requiring additional time and financial investments (e.g., to purchase contact thermometers). For instance, a quantitative u-value is likely to have less meaning for an everyday user than a qualitative colour-coded image where a red area indicates a warmer surface.

Therefore, this article will only discuss the qualitative results obtainable from the proposed methodology as these are deemed most useful for the intended end user with their current expertise. As Rowheath staff become more familiar with the process, technology and technical aspects, they will be able to refine the results to reduce inaccuracies and conduct more quantitative analysis.

2.2. Thermal Image Collection

The investigators conducted an external and internal thermal image survey. Whilst all accessible rooms were surveyed, time and access constraints meant it was not possible to gather thermal images of the entire internal environment. Consequently, the internal thermal images are not discussed in detail in the current study. Thus, all details of image acquisition refer to the external survey exclusively. Future work may seek to complete the internal image survey.

The thermal images were collected using a handheld FLIR E8 WiFi thermal camera (Teledyne FLIR, Wilsonville, OR, USA) [60], which captures radiometric images and has a thermal sensitivity of 60 mK, capable of detecting minute temperature shifts as slight as 0.06 °C, and an infrared resolution of 320 × 240 pixels, enabling detailed thermal imagery capture [3]. The MSX mode was chosen, allowing the camera’s internal processing to overlay visible light data with the infrared data so that it is easier to distinguish features in a photo [61]. Within the camera settings, the distance to the object was set at 10 m, and the emissivity was assigned as a ‘matt’ surface. The reflected apparent temperature was not entered.

Image collection took place on 7 February 2024 between 1 and 3:30 pm. The camera was switched on upon arrival at Rowheath and allowed to reach equilibrium whilst the capture locations were planned [52]. The approximate outside temperature during image collection was 1–2 °C (recorded on site and input to the camera), and the weather was mostly overcast with intermittent sunny periods [3]. The ambient air temperature was recorded externally and for all accessible internal rooms (data available upon request). Internal temperature was recorded approximately 2 h after image acquisition began. However, the internal temperature was assumed to be reasonably consistent throughout as the heating conditions were not altered. The environmental conditions created a clear disparity between the interior temperature of the building and the outside temperature. Default humidity values were used.

There was a minor amount of solar loading (increased temperature due to the intermittent effect of sunlight) evident on the roof of the single-storey sections of the Pavilion. However, the time of day, the trees to the west of the site (providing some shade to the western façade) and the fact that the remainder of the day had been primarily overcast meant that this was minimal. Furthermore, since the roof was not intended to be included in the thermal HBIM at this time, this was not detrimental to the outcomes of the survey. All instances of suspected thermal loading were noted in the report provided to TCC [3].

The camera was handheld, and the operator was positioned at a distance of approximately 10 m from the pavilion, to be suitable for capturing the whole height of the façade in one image without excessively angling the camera. The height above ground was approximately 1.5 m. Image capture spacing was planned to allow a small amount of overlap (approximately 20%) between adjacent images to avoid accidental data voids. The camera also recorded a visual image at the same instance to ensure the location of thermal scans can be accurately determined. Image capture was instantaneous.

The thermal camera automatically suggests a scale for each thermal image, with the user being able to manually change these. In order to select an appropriate, consistent scale, initial thermal images were gathered externally and then separately internally to enable the determination of a pragmatic fixed temperature scale and minimise saturation in the images. The temperature scale for external images was then set to between 3 °C and 13 °C, and that for internal images to between 3 °C and 25 °C. This ensured the same temperature scale on all external images and internal images.

A total of 98 thermal images were collected, including 32 internal and 66 external images. Figure 3 details the number of images captured for each façade. The quantity of images varied according to the complexity of the façade; e.g., the rear façade required additional images due to occlusions caused by the first-floor balcony.

The images were exported from the camera using the default USB export to FLIR Tools v5.X [62]. No additional post-processing was performed for the images in FLIR Tools.

Figure 4a,b are examples of a thermal image and a corresponding visual image, respectively. For ease of interpretation, a temperature scale is displayed on the right-hand side of Figure 4a with approximate temperatures. The same temperature scale is applicable to all thermal images in this article. Figure 5 indicates the approximate corresponding location of image capture. The whole dataset of images and individual image analysis were shared exclusively with TCC. A corresponding visual image is available for all thermal images.

As previously mentioned, it is possible to calculate approximate u-values from thermal images [49]. However, it was decided that including the calculation at this stage would reduce the accessibility of the method. Moreover, since the internal thermal image survey was incomplete, it would have only been possible to conduct this calculation for some façades. Furthermore, whilst the image provided by FLIR Tools is an absolute temperature map, reportedly displaying the exact temperature for a pixel, there are several potential sources of error in the temperature measured by the thermal camera. For instance, Wan et al. [52] found that a distance to an object of greater than 2 m resulted in the measured temperature of the object being lower than the actual temperature. Therefore, given the distance was approximately 10 m, it is likely that the recorded thermal image temperatures are lower than reality. In their study, Playà-Montmany and Tattersall [63] reported an underestimation in temperature measurements of approximately 6 °C, corresponding to approximately 20% of the actual temperature. Furthermore, the actual distance to the building may have varied slightly between images due to terrain constraints. Likewise, whilst the surface was assigned a ‘matt’ emissivity parameter to account for the largely rendered surface, the exact emissivity was not known, and this did not account for the emissivity of more reflective surfaces such as the glass doors. Wanderley Neto et al. [64] found that an error in emissivity can result in an error in measured temperature of greater than 10%. However, they do note that if the values are only being used for a qualitative, comparative study, which this research is, then the error caused by incorrect emissivity is not a problem.

Moreover, whilst the air temperature was measured on site (measured temperatures available upon request), no spot temperature checks were undertaken to validate the FLIR surface temperature readings. The reflected apparent temperature was not entered into the camera parameters, which would have also impacted the measured temperature [51,52]. The thermal camera reports an accuracy of ±2 °C [61], but the exact accuracy obtained in the Rowheath survey is unknown. In their study, Adán et al. [51] found that uncorrected raw temperature values from a thermal image required up to 8% correction (an error of approximately 1.5 °C) and that this resulted in a u-value error of 94%.

Given the potential sources of uncertainty and their suggested influence on the measured temperature, the exact certainty of the results cannot be quantified. Therefore, given the qualitative nature of the thermal evaluation, the manufacturer’s reported accuracy of ±2 °C will be assumed for all temperature measurements [61], as this should account for potential errors of between 10 and 20% of the actual temperature. However, the measured temperature can consequently only be referred to as the ‘apparent’ temperature [50]. Therefore, it would not be possible to quantify the accuracy of the u-value calculation. It would also have provided minimal benefit to the Rowheath staff, as it was also already known to the investigators at the survey stage that a future project was planned to investigate the quantitative thermal performance of Rowheath Pavilion using sensors.

For the current study, it was determined that qualitative evaluation was sufficient for TCC’s requirements, namely easily shareable evidence of potential thermal issues. Whilst there is known uncertainty in the measured temperature values of the thermal images, the relative temperature difference depicted by the thermal images was sufficient for this purpose. For instance, in Figure 4a, the apparent temperature difference between the bottom of the door and the wall, likely caused by a poor seal, is greater than 5 °C. Taking the reported accuracy of ±2 °C, this difference is still notable.

2.3. Integration with HBIM

The thermal images were then integrated with the existing HBIM model in Revit 2024. Several existing methodologies were considered. For instance, Solla et al. [57] integrated multiple non-destructive testing results with HBIM by creating ‘fictitious’ wall elements for each test and applying the resulting images as wall cladding. Likewise, Zhuo et al. [65] and Aricò et al. [66] also utilised fictitious objects to map damage and other condition surveys. Other authors [44,67] have instead chosen to create special parametric objects (e.g., cubes) that can then be interrogated to access the results of the condition surveys. In these instances, the results are not usually visually integrated with the HBIM and thus may not improve the interpretability for non-experts. For all approaches, the additional objects could then be turned on and off as needed. Whilst these methods are undoubtedly successful, they could result in unexpected errors for the end user. For instance, duplicate walls in Revit would have resulted in modelling clash error, and adding 3D objects to a wall surface may have impacted the results of any volume calculations. Whilst these potential errors can be overcome, Rowheath only required thermal analysis, and since the end users of the model had limited to no experience with Revit, adding additional ‘fictitious’ elements may have caused undue confusion.

Instead, decals were chosen as the most suitable method for Rowheath. Decals in Revit are visible in the rendered view only [68], so they do not negatively impact the model’s appearance for other users. They are not a material characteristic of the object families, instead being a purely visual feature, and they also have no effect on the model geometry, so they will not negatively impact any volumetric analysis undertaken. They are also easy to update, requiring only the decal image file to be changed, a workflow that is easy to teach to new users. To the best of our knowledge, decals have not been applied for this purpose elsewhere.

The thermal images were edited in the Microsoft Photos app in Windows 11 [69]. The photo app was chosen due to its familiarity and ease of use for the general public. Using native features of the app, the images were manually edited to adjust the alignment, reshape the images, and crop and straighten them as needed. The limitation of the chosen software was that the images were not fully rectified, meaning there was some minor distortion caused by perspective. However, since the images were only intended to be used for general evaluation of the structure, and the main architectural features were visually identifiable from the images, this distortion was deemed insignificant for the purposes of the study. In the future, further image rectification would increase the accuracy of the decals and enable their use for more detailed analysis [70]. The edited images were applied to the views as decals. Decals were applied in Revit by navigating to the ‘insert’ tab and selecting the decal drop-down [68]. The size of the decal was then manually adjusted to fit the chosen surface, and the corresponding image file was then assigned to the decal. The standard decal parameters (e.g., brightness, reflectivity, etc.) were used. The chosen image was automatically fit to the decal in Revit. Where required, manual adjustments to the decal location were then made in Revit to improve the visual alignment with known architectural features.

During the creation of the HBIM model for Rowheath Pavilion [6], an online data store was created to store additional documents related to the Pavilion. The software used was Microsoft OneDrive [71] due to its familiarity for Rowheath staff and its low implementation cost. The online data store was linked by URL to the Revit model. The data store was by default read-only with additional access consents, a native feature of OneDrive, distributable by the Building Manager. The thermal images were added to the same data store. The thermal survey file consisted of the thermal orthophotos, the MSc report [3] detailing the survey acquisition process and interpreted findings, and a sub-folder containing the 103 thermal images as well as the corresponding visual images.

3. Results

Figure 6 depicts the resulting thermal HBIM model of Rowheath Pavilion. The model and accompanying thermal images were evaluated, and all results were shared with TCC.

The key findings from the evaluation of the thermal HBIM data were that it is evident that the external walls of Rowheath Pavilion were of low temperature (typically less than 8 °C), whilst the internal walls were of higher temperature (room-dependent) which is indicative of reasonable performance of the main wall structures; e.g., internal heat is not being majorly distributed to external façades. However, within the model, areas of higher temperatures were visible. Some of these can be attributed to a number of common factors, primarily (1) leaks from windows that are not sealed/closed at the top (or unable to be closed), (2) drafts at the bottom of external doors, (3) radiators on walls that do not have reflective insulation on the external wall behind them, and (4) poor-quality insulation on external doors—in particular the external ‘church office’ door and the café walls.

In addition, the thermal images also provide indications of how the internal temperature and heating process of the rooms may affect the building fabric. For instance, the main hall of the pavilion (central section in Figure 6) had a recorded internal air temperature of approximately 17 °C, whereas the café (right, setback section of Figure 6) had a recorded internal air temperature of 23 °C. It can be seen from Figure 6 that the rendered sections of the café façade appear significantly warmer than the rendered section of the main hall. Repeating the thermal image survey of the façade whilst maintaining a consistent internal temperature may provide different results. However, a uniform internal temperature is currently difficult to achieve with the current heating equipment for Rowheath Pavilion.

Since the images were automatically fit to the decal in Revit, the accuracy of the image positioning is dependent on the scaling and adjustment of the images. Therefore, there are a few minor image location issues present. The majority of these are on the first floor of the pavilion, since the ground height image capture inherently results in some distortion. Whilst the exact location error is unknown, comparing the first-floor features (namely windows) in the thermal images to the HBIM suggests a potential error of up to 20 cm in places. Conducting additional image rectification would resolve these errors. Alternatively, repeating the survey with a drone would enable images to be captured perpendicular to the surface, reducing angular distortion. This must be completed before the thermal HBIM can be used for planning destructive surveys. However, for the identification of thermal inefficiencies, such as drafts from windows, the location error is insignificant as the MSX setting of the camera means the features are visually clear in the image. Likewise, the warm patches caused by radiators are large enough that the error is insignificant.

4. Discussion

The thermal images themselves provide immediate justification for some small-scale changes that may improve the energy efficiency of Rowheath Pavilion. For instance, investing in draft excluders for external doors could well be justified with the images alone. Notably, the heat loss evident at the bottom of doors may explain the drafts described by Rowheath staff (see Section 1.2). Hence the proposed methodology does, as required, provide evidence that can be shared with the other stakeholders to support the lived experience of the community that uses Rowheath. Repeating the thermal image survey in the summer would potentially provide evidence to support the summertime issues described by staff, e.g., overheating. However, the time constraints of the MSc project meant that this was not possible at the time of writing.

However, the key benefit of the thermal HBIM model compared to the thermal images on their own is the additional contextualisation provided by the HBIM environment. In the bottom right corner of the thermal HBIM (Figure 6), red patches, indicating a higher temperature, are visible. Figure 7a shows the isolated thermal image of the corresponding area, and Figure 7b shows the corresponding visual image. By only referring to Figure 7a, it may not immediately be clear what is causing the temperature variation.

However, within the 3D environment of the thermal HBIM, it can be easily interpreted that the heat is being emitted from a radiator on the internal face of the wall. As a result, Rowheath staff were able to justify the installation of reflective foil behind the radiators by providing clear visual evidence of the issue and its cause.

The proposed approach is easy to apply and can be simply taught to others, overcoming perceived barriers relating to digital literacy in the heritage community [55]. Furthermore, the approach has good long-term viability, a key concern associated with digital solutions for asset management [72]. If additional thermal imaging surveys are conducted or previous surveys are updated to monitor changes over time, the Revit model can be easily updated with the new images whilst retaining previous images in the online data store. This would enable Rowheath staff to assess the impact of any implemented changes.

One limitation of the proposed method is that it is largely qualitative, only indicating relative temperature differences and potential problem areas. However, it can be used to inform future quantitative evaluation, which will require financial investment from TCC (e.g., commissioning of non-destructive testing and the installation of sensors). For instance, from the thermal HBIM, it is apparent that the exposed brick arches around the doors and the unrendered brick elements of the wall are typically warmer than their rendered surroundings, which may indicate poorer thermal performance. Future testing could seek to quantify the variation in performance of the different construction and material types.

Another limitation of the method is that the thermal decals were manually aligned and, whilst slightly adjusted, were not fully rectified. This method was chosen to make the method as accessible as possible to the intended user. Consequently, there is some deviation between the thermal image positioning and the real positioning. Conducting further image rectification and aligning the images via georeferenced control points would increase the accuracy of the study. However, the error in image positioning was not considered detrimental for the purposes of the study, as the intention was to conduct a general performance evaluation, and the MSX thermal camera mode allowed architectural features to be clearly visible within the thermal photos, allowing an acceptable estimate of the actual position.

At the moment, it is only possible to apply decals to planar or cylindrical surfaces in Revit. This posed no issue for Rowheath Pavilion, which has largely regular architecture, but may limit the methodology’s applicability to certain heritage structures. This limitation was also present in the other considered methodologies [65], so it is not unique to the chosen method and is instead a result of current technological limitations with regard to irregular architecture.

As previously discussed, there is some uncertainty in the measured temperature values (see Section 2.2), so only a qualitative evaluation is possible. However, the relative apparent temperatures depicted in the thermal HBIM suggest that the main building fabric may have good thermal performance since internal heat was not being obviously, excessively transmitted to the external surfaces. Further validation and correction of the measured temperature values would enable quantitative evaluation of the building’s performance and validation of this assumption.

Based on the qualitative evaluation, it is suggested that any future large-scale energy retrofit at Rowheath Pavilion prioritises alternative heating or energy solutions as opposed to major material interventions. This will not only provide tangible efficiency benefits but is also more feasibly allowable given Rowheath’s Grade II listed status: TCC had previously gained planning permission to install solar panels on the new structure at Rowheath [3], and previous investigations have found both public and TCC sentiment in favour of more sustainable energy and heating solutions. Since the thermal HBIM is reasonably geometrically accurate, it could be used in the future to assist with the design of any proposed changes, by allowing the simulation of the proposed changes to assess their impact on the historical significance of Rowheath as well as their buildability.

5. Conclusions and Future Work

This article serves as proof of concept for the proposed methodology. It demonstrates the feasibility of the proposed methodology and provides preliminary evidence of its potential benefits. The recommendations made to Rowheath by the investigators at the conclusion of the work enabled them to make immediate small-scale interventions and plan for future changes.

Future work may, if funding and regulatory compliance allow, seek to complete the thermal image survey, potentially by using drone technology to capture images of the pavilion roof and by completing the interior thermal image survey. There is also an ongoing project investigating the installation of temperature sensors, which is intended to enable more detailed thermal analysis in the future, including the integration of the HBIM model and BEM solutions. Another potential area of future study that would enrich the output of the BEM simulations could be the investigation of non-destructive techniques, such as Raman spectroscopy or infrared thermography, to analyse wall compositions.

In addition, BIM and HBIM models are currently demonstrated in the University of Birmingham, School of Engineering, BIM cave [17] as part of Undergraduate and Postgraduate degree teaching and external outreach to schools and colleges, interactively exhibiting how BIM connects to heritage. In the future, the demonstration of the thermal HBIM model created as part of this research could be further developed for specialised teaching and outreach in terms of the sustainable management and/or retrofit of heritage structures.