Abstract
Civil defense projects, designed as wartime underground spaces, often lack effective natural ventilation and have considerable depth, which complicates their use as public spaces in peacetime. However, the application of passive ventilation technologies can create effective airflow channels within these structures, significantly enhancing ventilation efficiency and thus improving the overall thermal comfort level. For this study, air age, along with average wind speed, temperature, and relative humidity as stipulated by the “Requirements for Environmental Sanitation of Civil Air Defense Works during Peacetime Use” (GBT 17216-2012), were selected as evaluation metrics. This paper compares the ventilation effectiveness between single ventilation shafts and multiple ventilation shafts under positive and negative pressure conditions in underground civil defense structures. The results indicate that negative pressure ventilation in multiple shaft configurations performs optimally across various ventilation approaches. Subsequently, the Response Surface Methodology (RSM) was utilized to further optimize the positioning of multiple ventilation shafts. The study examined the impact of three ventilation shaft locations on average wind speed, temperature, relative humidity, and air age, leading to an optimized design. Specifically, the optimal positions are 54.76 m for Shaft A, 51.45 m for Shaft B, and 79.85 m for Shaft C, achieving an average wind speed of 0.222 m/s, a temperature of 26 °C, a relative humidity reduction to 85.47%, and an average air age of 10.57 s. This research provides practical insights for the optimization of ventilation in underground civil defense facilities.
Supplementary Information
The online version contains supplementary material available at 10.1038/s41598-024-73059-7.
Keywords: Tunnel-type underground spaces, Ventilation efficiency, Response surface methodology, Civil defense underground spaces, Design optimization
Subject terms: Civil engineering, Urban ecology
Introduction
Civil defense projects are underground protection structures constructed specifically for the sheltering of personnel and materials, civil defense command, and medical aid during wartime1,2. Depending on their construction methods, these projects can be categorized into excavated underground civil defense structures and tunneled underground civil defense structures. The former is subdivided into standalone and integrated constructions, whereas the latter includes tunnel-type and gallery-type underground structures3. In the context of “long-term preparedness, focused construction, and peacetime-wartime integration,” numerous civil defense structures are being incorporated into urban development, impacting everyday life1. Common examples of civil defense utilization in peacetime include underground public transport facilities, underground commercial streets, and air-raid shelters4. China has established a series of standards for the use and construction of civil defense works, among which the “Requirements for Environmental Sanitation of Civil Air Defense Works during Peacetime Use (GB/T 17216-2012)” stipulates environmental standards concerning temperature, humidity, wind speed, and air volume during peacetime conversion.
As a type of underground construction, civil defense structures share similar physical environmental characteristics with other underground buildings. Due to the presence of soil, these structures exhibit significant thermal stability, typically having an average radiant temperature about 1 °C lower than the above-ground air temperature5. Furthermore, the encasing soil results in higher internal humidity levels compared to the exterior, with measurements in underground mines showing relative humidity ranging from 75 to 98%6. In high humidity conditions, even a slight increase in temperature can make the environment feel uncomfortably warm and damp7. Most civil defense projects initially did not consider passive measures to enhance thermal conditions; therefore, tunnel-type underground civil defense spaces, with their extensive depth, tend to have poor air circulation. During wartime, these areas may become crowded, exacerbating air circulation difficulties and potentially leading to oxygen deficiencies that can threaten human life8. In such structures during summer, internal humidity levels approach saturation9, with significantly elevated levels of dust, radon, and microbes compared to normal conditions10. The cost of retrofitting these structures for dual-use purposes is considerable, and the severe internal conditions initially designed for wartime significantly impair thermal comfort, making them unsuitable for regular public use in peacetime.
To enhance the thermal comfort in underground civil defense facilities and improve the overall physical environment, significant research has been conducted on these types of non-air-conditioned underground spaces. Mukhtar et al. analyzed the thermal environment of underground shelters in hot and humid regions such as Malaysia, showing that optimizing the positions of air inlets and outlets significantly enhances thermal comfort within these structures11. Jin et al. studied the thermal environment under natural convection in mine refuge chambers deeper than 2 m, finding that natural ventilation promotes a more uniform and stable indoor thermal environment12. Lin et al. optimized the ventilation systems for sleeping spaces in underground shelters to improve indoor physical parameters and thermal comfort13. Additionally, studies on spaces with physical environments similar to underground civil defense structures, such as mines6,14,15, underground tombs16,17, wine cellars18–20, and tunnels21,22, have shown that ventilation in these underground spaces is often inadequate for achieving suitable levels of thermal comfort.
These studies indicate that natural ventilation is crucial for diluting moisture in the air23, enhancing air quality24, and improving thermal comfort. Vent shafts, as a common means of passive ventilation, are utilized in many underground constructions11. However, research on the improvement of the overall thermal environment in deep underground civil defense structures through passive ventilation via vent shafts remains limited.
Therefore, this study focuses on the tunnel-type underground civil defense space of Yunlong Mountain in Xuzhou. Under the premise of integrating peacetime and wartime functionality, the research constructs a realistic terrain model to simulate the internal environment of the tunnel-type space during summer and evaluates its thermal characteristics. The study then compares the changes in the overall thermal environment of the tunnel under positive and negative pressure conditions, considering both single and multiple vent shafts, to identify the most effective ventilation strategy. Response Surface Methodology (RSM) is employed to analyze how varying the positions of vent shafts influences ventilation efficiency under optimal conditions (Fig. 1). Design optimization is performed using Design Expert 13, and numerical simulations are conducted with Cradle CFD. Figures 7, 10, and 20 were generated using the Launch Postprocessor in Cradle CFD 2022.1, while Figs. 16a and 19 were created using Design-Expert 13. By investigating thermal environmental changes across different ventilation scenarios, this research provides design references for the practical integration of tunnel-type underground civil defense spaces for both peacetime and wartime uses.
Fig. 1.
Technology roadmap.
Methodology
Research location
Xuzhou, located in the southeastern part of the North China Plain, spans from 116°22′ to 118°40′ East longitude and from 33°43′ to 34°58′ North latitude. The city experiences a temperate monsoon climate with distinct seasons. The average annual temperature in Xuzhou is 16.16 °C, with significant monthly variations. July is the hottest month, averaging 31.4 °C, while January is the coldest, with an average temperature of 0.7 °C. The average annual relative humidity is 68.58%, with an average water vapor pressure of 7.40 mmHg. Xuzhou receives an average of 2220.9 h of sunshine annually, with the majority concentrated between April and August. The average annual wind speed is 2.19 m/s, showing uniform variation throughout the year. Xuzhou comprises five districts: Yunlong, Gulou, Quanshan, Tongshan, and Jiawang. The focus of this study, the Yunlong Mountain tunnel civil defense project, is located in Quanshan District (Fig. 2).
Fig. 2.
Location of the study object.
Model background information
Field surveys of the Yunlong Mountain tunnel and civil defense project, combined with the technical blueprints, provided the structural dimensions for the Yunlong Mountain civil defense project. The entire project was supported by the Xuzhou Municipal Civil Defense Office, and through simplification of the provided blueprints, the floor plan was derived as shown in Fig. 3. The corresponding model was developed based on these blueprints.
Fig. 3.
Underground space distribution of civil air defense.
The model consists of three main components: the external mountain body measuring 330 m by 240 m; the internal tunnel, which is 343 m long, 20 m wide, and 6 m high; and the internal civil defense structure, which extends 180 m in length, 72 m in width, and 3.6 m in height. The internal civil defense area primarily comprises an outer hall and the internal tunnel. The average vertical distance from the internal tunnel to the mountain peak is approximately 40 m. The main air inlets for the civil defense area are two openings, with areas of 28.8 m2 and 30.42 m2, respectively. The specific dimensions of the civil defense structure are depicted in Fig. 3, and real-life images of the project entrance and interior are shown in Fig. 4.
Fig. 4.
The real scene map of the entrance and the interior of the research object.
Environmental monitoring
Given that July is the hottest month in Xuzhou, this study selected a representative week in the summer (July 5–11, 2023) for environmental monitoring at six specific locations within the external tunnel and internal civil defense spaces. These locations included two points within the tunnel and four within the civil defense structure. The physical parameters measured were temperature (°C), relative humidity (%), and wind speed (m/s). Details of the monitoring equipment are provided in Table 1.
Table 1.
Monitoring equipment.
| Measuring physical variables | Instrument model | Measurement range | Accurate |
|---|---|---|---|
| Air temperature | Testo174-H1 | −20 to 70 ℃ | 0.5 ℃ |
| Relative humidity | Testo174-H1 | 0–100% | ± 3% |
| Air velocity | JT-IAQ | 0.05 –5 m/s | 0.03 m/s |
Monitoring Point 1, located outdoors, was designated to capture the thermal and humidity conditions in the urban area of Xuzhou during the study period. Monitoring Point 2, situated within the Yunlong Mountain Tunnel, was employed to evaluate the thermal environment of the tunnel itself. Monitoring Points 3 through 6 were strategically placed at 400-meter intervals along the entrance to monitor the internal thermal conditions of the civil defense structure. The locations of these monitoring points are depicted in Fig. 5.
Fig. 5.
Measurement point arrangement.
Temperature, relative humidity, and wind speed are critical indicators affecting comfort levels. To minimize the impact of extreme weather conditions, the study also incorporated historical average air temperature, relative humidity, and wind speed for July at Monitoring Point 1 for comparison. The results and comparative analysis are presented in the Table 2.
Table 2.
Measurement point 1 measurement data average and comparison.
| Parameters | Measurement point 1 | Historical average | Misalignment |
|---|---|---|---|
| Air temperature | 29.9 ℃ | 31.4 ℃ | 5% |
| Relative humidity | 73.9% | 80% | 8.2% |
| Air velocity | 2.14 m/s | 2.30 m/s | 7.6% |
It can be seen that the deviation between the corresponding data in July and the historical average is within the normal range. On the premise of ensuring the reasonable and normal data, the data of the corresponding measurement points are summarized as shown in Table 3.
Table 3.
Summary of average data values for each measurement point.
| Parameters | Measurement point 2 | Measurement point 3 | Measurement point 4 | Measurement point 5 | Measurement point 6 |
|---|---|---|---|---|---|
| Air temperature | 27.7 ℃ | 27.1 ℃ | 23.5 ℃ | 21.8 ℃ | 21.8 ℃ |
| Relative humidity | 80.2% | 82.8% | 92.2% | 94.8% | 95.2% |
| Air velocity | 1.72 | 0.92 | 0.31 | 0.16 | 0.11 |
Value simulation and settings
Numerical model
Computational Fluid Dynamics (CFD) relies on solving the Navier–Stokes equations, which include a three-dimensional continuity equation and equations controlling the momentum transfer components. Typically, these nonlinear equations are solved numerically, requiring appropriate boundary conditions. CFD analysis enables detailed examination of pressure and the distribution of three velocity vector components across the space25. Common CFD models include Reynolds-Averaged Navier–Stokes (RANS), Large Eddy Simulation (LES), and Direct Numerical Simulation (DNS). In practical engineering applications, the RANS model is favored due to its computational efficiency and applicability26.Computational domain As shown in Fig. 5.
This study employs the Reynolds-Averaged Navier–Stokes model and the Re-Normalization Group (RNG) turbulence model to simulate turbulence. Here, air is considered a Newtonian fluid—meaning it is incompressible and turbulent—with constant fluid properties assumed. To account for the impact of temperature differences on the airflow, the Boussinesq approximation is used to describe air density variations27. The governing equations—Eqs. (1)–(3)—represent the continuity equation, momentum equation, and energy equation, respectively, while Eq. (4) models the turbulent kinematic viscosity µt in the RNG k−
model.
 |
1 |
 |
2 |
 |
3 |
 |
In this study, 
represents the fluid density in kg/m3;
and 
denote the average fluid velocity components in the x, y,z directions, respectively, in m/s; 
is the average fluid pressure in Pascals; µ and µt are the dynamic viscosity and turbulent dynamic viscosity, respectively, in m2/s; β is the thermal expansion coefficient in /°C; 
is the fluid temperature in °C; 
is the reference temperature in °C; α and αt are the thermal diffusivity and turbulent thermal diffusivity, respectively, in m2/s.
For solving the humidity inside the tunnel, studies28,29 have indicated that water vapor should be considered a passive scalar that diffuses with air movement. This characteristic is described by solving the convection-diffusion equation, where 
represents the partial pressure of water vapor; 
is the molecular diffusion coefficient, in m2/s; 
is the turbulent Schmidt number.
 |
5 |
In terms of setting boundary conditions, the inflow boundary for wind is simulated using a power-law wind profile. The governing Eqs. (6)–(9) calculate wind speed data at specific heights from meteorological data, allowing for the calculation of wind speed, turbulent kinetic energy, and turbulent dissipation rate at specific heights30. In simulations, the boundary at Zmax is set as a free-slip boundary, outflow boundaries are also set to free-slip, and other walls are set as no-slip.
 |
6 |
 |
7 |
 |
8 |
 |
9 |
is the wind speed at height, in (m/s); 
is the wind speed at reference height R, in (m/s); 
is the ground height, in (m); m,
is the reciprocal of the power-law exponent, with a value of 5.0 for environmental category III (urban areas composed of small buildings); 
is the turbulent kinetic energy, in (m2/s2); 
is the turbulent dissipation rate, in (m2/s3); 
is the turbulence intensity at height, in (%); Cµ is a model constant, 0.09; 
for roughness category III is 450 m.
The thermal outer boundary is adiabatic, and the flow-solid boundary is set with a logarithmic law of heat conduction, while solid-solid boundaries are set to conductive. Humidity input units are in relative humidity, with the power-law wind set to 75% relative humidity, and the humidity boundary for civil defense set to a constant humidity flux of 1.6e-7 kg/(m2·s). Additionally, using TMY data, the “Kusuda-Achenbach” method is applied to calculate soil temperatures31,32. Civil defense and tunnel materials are set to rock, with a density of 2800 kg/m3, specific heat capacity of 840 J/(kg·K), and thermal conductivity of 3.5 W/(m·K). Based on the aforementioned parameters, the computational domain for this study is defined as shown in Fig. 6.
Fig. 6.
Computational domain.
Grid independence and numerical validation
This study utilizes Cradle scSTREAM V14 for CFD simulation, due to its advantages in mesh division and computational speed33. To enhance the accuracy of the simulations, a sensitivity analysis of parameters including grid resolution, discretization schemes, and convergence criteria is conducted before simulation, as detailed in Tables A1–A3. The predefined error standard is set below 10%. Ultimately, the simulation uses a mesh count of 9,036,720, a Quick discretization scheme, and a convergence criterion of 10-4. Validation details are provided in Tables A1 to A9.
Evaluation standards
For the conversion of civil defense engineering for peacetime use, the “Environmental Sanitation Requirements for Civil Air Defense Engineering in Peacetime Use” (GBT 17216-2012) establishes standards for various physical parameters, including relative humidity, wind speed, and temperature. According to the standard, the acceptable range for relative humidity is 30-80%, summer temperature should be maintained between 26 °C and 28 °C, and wind speed should exceed 0.2 m/s. In addition to these parameters, the standard also imposes limits on pollutant levels within civil defense spaces. This study employs Air Age (AOA) as a key evaluation metric. AOA was first introduced by Sandberg in the 1980s and represents the average time it takes for air particles to travel from the inlet point to the measurement point. AOA can be used to assess ventilation efficiency, air freshness, and air distribution in both single-zone and multi-zone systems34. By measuring AOA, the overall airflow organization within a room can be evaluated. Common methods for calculating AOA include tracer gas techniques, CFD simulations, and statistical inference35.
Results and discussion
Summer simulation evaluation of civil defense shelters
Airflow organization
The airflow within the civil defense shelter can be categorized into two main paths: the airflow through the main hall and the airflow within the internal tunnel. The majority of the air flows through the main hall and exits through openings at both ends, while a minor fraction enters the internal tunnel. Figure 7 presents airflow organization maps at 0.5 m, 1.5 m, and 2.5 m heights within the civil defense space under summer ventilation conditions. As warm air enters through the upper openings due to its lower density compared to cold air, the organization of the airflow varies at different heights.
Fig. 7.
(a) Airflow organization at different elevations; (b) Physical environment at different elevations; (c) Physical environment at profiles A and B.
Firstly, at the 2.5 m elevation, the airflow enters through the upper inlet with most of it passing directly through the hall and exiting through the lower outlets. For the air that enters the internal civil defense tunnel, the majority is diverted at X = 95 m, flowing towards the X-axis direction and against the Y-axis direction, eventually meeting at X = 185 m where they interact and produce a downward vertical airflow.
Secondly, for the 1.5 m elevation intake airflow, the entry of warmer upper air causes the internal air to sink and meet the hall’s warm air at this elevation, creating two significant vortexes. The sinking airflow at 2.5 m still maintains some momentum at this 1.5 m height, resulting in the airflow at X = 185 m being largely unaffected.
Lastly, at the 0.5 m elevation, due to compression by the upper airflow, the cooler air tends to converge at the bottom, while the warm air from the upper entrance meets the cooler air inside almost immediately upon entry, forming a vortex at X = 35 m. The originally sinking airflow at X = 185 m also diverges horizontally at the 0.5 m height due to the exhaustion of vertical momentum, flowing outward towards the tunnel exits.
In terms of overall airflow speed, the air in the main hall of the civil defense shelter, being closer to the intake and traveling a shorter distance, maintains an average speed of 0.3 m/s. Conversely, within the internal civil defense tunnel, due to the longer distance, the average speed is below 0.1 m/s.
Vertical physical environment distribution
Given that underground civil defense spaces must function in both peacetime and wartime scenarios, the typical postures of individuals inside these spaces can generally be categorized as sitting or standing. The height for a sitting posture is considered to be 0.9 m, while the height for a standing posture is 1.5 m. By appropriately extending the statistical range, the average physical environment indicators were ultimately calculated for the height range of 0.5 m to 1.9 m.
Figure 8 reveals that with increasing height, the average temperature linearly increases by 1.2 °C. In contrast, the average humidity shows a linear decrease of 2.2%. The average wind speed initially decreases then increases, indicating that above 1 m height, the presence of warm air and thermal pressure drives the airflow within the shelter, generally directing it towards the tunnel. Below 1 m, cooler air dominates the internal environment, pushing the overall airflow outward. The primary airflow organization within the shelter is characterized by high entrance and low exit, with the age of air showing a linear decreasing trend.
Fig. 8.
Vertical physical environment changes in human defense.
Horizontal physical environment distribution
Horizontal airflow enters through the shelter’s entrances, where its kinetic energy rapidly dissipates. To visually capture this phenomenon, this study analyzes cross-sectional average physical measurements at equidistant points from the entrance. The sections from 10 m to 90 m cover the external hall area, 90 m to 110 m the transition area, and 110 m to 210 m the internal tunnel area.
As shown in Fig. 9, the average temperature generally decreases with increasing depth. In the external hall, where cold and warm air currents converge, the temperature experiences a moderate decline due to the formation of vortices. The temperature within the internal tunnel demonstrates a clear linear relationship with depth. Conversely, the overall change in average humidity is inversely related to the average temperature, with humidity levels approaching 100% at the deepest part of the shelter. The average wind speed shows a slight increase in the middle of the hall (between 30 m and 50 m) due to vortex formation; however, as the airflow enters the internal tunnel, the wind speed exhibits a clear linear decline. Additionally, the age of air significantly increases beyond 190 m, highlighting the challenges of maintaining effective natural ventilation in the deeper sections of the tunnel.
Fig. 9.
Changes in the horizontal physical environment for human defense.
Discussion on ventilation strategies
The simulation results from the previous section reveal that the air quality within the civil defense hall up to 70 m is satisfactory due to effective air convection and higher wind speeds, with air age, relative humidity, and temperature all reaching appropriate levels. However, deeper areas of the shelter, due to their extended length, struggle to form effective natural convection paths, resulting in suboptimal physical conditions. Past research indicates that traditional passive ventilation not only conserves energy but also significantly improves underground ventilation36,37. Passive ventilation can generally be categorized into two types: positive pressure ventilation, which actively introduces external natural air using devices like wind catchers, and negative pressure ventilation, which employs passive devices like wind caps to create a suction effect that exhausts indoor air.
The diameter and the mode of pressure (positive or negative) of the ventilation shafts directly influence the volume of air ventilated. In this study, the diameter of the ventilation shafts was set at 4 square meters based on relevant literature38; positive pressure ventilation was established by placing wind catchers 2 m high in all four directions at the top of the shafts39, while negative pressure was set at a constant low wind speed of 0.4 m/s based on practical usage of wind caps40.
Single ventilation shaft airflow organization
As the simulations indicated, air speed within the three tunnels of the shelter is low, humidity is high, air age is prolonged, and the average temperature is low. This section discusses the airflow organization under both positive and negative pressure by setting ventilation shafts in the middle of the three tunnels.
Figure 10 illustrates the airflow organization at a 1.5 m height under negative pressure conditions. The setup of negative pressure ventilation shafts at points A, B, and C resulted in average wind speeds of 0.150 m/s, 0.155 m/s, and 0.160 m/s, respectively, with speeds increasing the further the shafts are from the air inlet. However, shafts B and C, being far from the air inlet, though somewhat effective, do not provide adequate ventilation for the rear sections of the shelter due to their limited airspeed and narrow ducts.
Fig. 10.
Negative pressure single ventilation shaft airflow organization and air velocity diagrams.
Figure 11 shows the airflow organization at a 1.5 m height under positive pressure conditions, where the settings at A, B, and C correspond to average wind speeds of 0.138 m/s, 0.143 m/s, and 0.145 m/s, respectively. Shaft A tends to suppress overall airspeed while shaft C enhances it slightly. Overall, single shaft positive pressure provides minimal improvement in internal airflow.
Fig. 11.
Positive pressure single ventilation shaft airflow organization and air velocity diagrams.
Multiple ventilation shafts airflow organization
In scenarios with multiple ventilation shafts set up in the three tunnels of the shelter, negative pressure ventilation acts as a relay, enhancing the air’s momentum as it passes each shaft, ensuring it reaches the next, as illustrated in Fig. 12.
Fig. 12.
(a) Comparison of positive and negative pressure ventilation airflow organization at different elevations; (b) Comparison of positive and negative pressure ventilation physical environment at different elevations; (c) Comparison of positive and negative pressure ventilation physical environment at different profiles.
The specific airflow patterns are as follows: At an elevation of 2.5 m, the negative pressure ventilation shaft increases the wind speed at the air inlet, while some of the air also forms an airflow field with ventilation shaft C due to the negative pressure effect. However, at the deepest point inside the civil defense area, the opposing airflows create a blockage. At an elevation of 1.5 m, ventilation shafts A and C handle the majority of the airflow, while shaft B forms an airflow field with the remaining air. The external hall experiences an overall increase in wind speed, which transforms the previously disordered airflow into a large vortex. At an elevation of 0.5 m, the three ventilation shafts alleviate the previous issue of cold airflow compressing hot airflow at the entrance. As a result, hot air can still form convection currents in the civil defense atrium, while the corresponding cold air is expelled outdoors through the ventilation shafts.
For positive pressure ventilation, the arrangement of the three shafts not only suppresses air intake at the entrance but also enhances the outflow at the exit, as detailed in Fig. 11. At 2.5 m, the introduced natural ventilation meets the original airflow at the transition area, causing a change in the direction of the initial airflow. At 1.5 m, the organization of the airflow in the shelter’s hall is disrupted by the newly introduced air, pushing the intersection of the two air streams to 50 m from the entrance, with the main airflow now running from the ventilation shafts to the exit. At 0.5 m, the airflow is entirely directed from the shafts to the exit, and due to internal wind pressure, the original air intake struggles to enter the shelter’s hall.
Longitudinal physical environment distribution with multiple ventilation shafts
Under the conditions of multiple ventilation shafts, the trends in other physical parameters are generally similar to those before optimization, as shown in Fig. 13. Firstly, in the vertical space between 0.5 m and 1.9 m, the average temperature shows the following trend: negative pressure ventilation > positive pressure ventilation > unoptimized conditions. Compared to the unoptimized state, negative pressure ventilation raises the temperature by 2.6 °C, while positive pressure increases it by 1.2 °C. The average relative humidity follows the trend of negative pressure ventilation < positive pressure ventilation < unoptimized, with positive pressure reducing the relative humidity by 4.2% and negative pressure by 8.3%. The average wind speed shows negative pressure ventilation > positive pressure ventilation > unoptimized. Additionally, due to conflicts with the original airflow organization, positive pressure ventilation increases the overall wind speed by only 0.01 m/s, whereas negative pressure increases it by 0.06 m/s. Air age is seen as negative pressure ventilation < positive pressure ventilation < unoptimized, with positive pressure reducing the air age in the vertical space from 0.5 m to 1.9 m by 7.6 and negative pressure by 7.7, indicating that both ventilation forms similarly improve air age in the vertical space (Fig. 13).
Fig. 13.
Longitudinal physical environment comparison of positive and negative pressure ventilation.
Horizontal physical environment distribution with multiple ventilation shafts
The horizontal space is significantly affected by the location of the ventilation shafts, which are positioned at 130 m, 150 m, and 190 m, respectively. From the results, it can be seen that the overall trend of changes in negative pressure ventilation is consistent with unoptimized ventilation, while positive pressure ventilation shows certain fluctuations in some areas due to the influence of new airflows introduced by the ventilation shafts, as detailed in Fig. 14.
Fig. 14.
Comparison of lateral physical environment changes in positive and negative pressure ventilation.
Regarding temperature, aside from the areas where the ventilation shafts are located, positive pressure ventilation does not significantly alter the internal ventilation of the shelter. The change in relative humidity is more pronounced in the internal tunnel of the shelter and less so in the main hall. In terms of wind speed, because of increased internal wind pressure, airflow at the entrance is restricted, reducing wind speed in the main hall but improving it in the internal tunnel. Lastly, regarding air age, since the ventilation shafts can directly transport external air, changes in air age around both types of ventilation shafts are significant. Additionally, positive pressure ventilation effectively improves air age conditions at the deepest parts of the shelter.
Optimization of tunnel civil defense ventilation using response surface methodology (RSM)
The main challenge in practical engineering optimization lies in defining the optimization function, as the commonly used trial-and-error method does not cover all scenarios effectively41. Common engineering optimization methods include Response Surface Methodology (RSM), stochastic (Kriging), and Artificial Neural Network (ANN)42. RSM, consisting of a series of statistical and mathematical techniques, is particularly effective for multi-parameter problems43 and is widely used in engineering design optimization due to its lower computational resource demands44.
Given the comparative analysis of positive and negative pressure ventilation efficiencies, it appears that negative pressure ventilation coordinates better with the original airflow organization. With the premise of three ventilation shafts, some physical quantities already meet the regulatory requirements. Hence, this section, based on the findings from the previous section regarding the three ventilation shafts, optimizes the parameters related to the locations of ventilation shafts 1 (A), 2 (B), and 3 (C). Considering the demand for regular utilization of underground spaces in peacetime-war transitions, this optimization calculates physical parameters at the average walking height in public buildings (1.5 m). Using RSM, the parameters are optimized to achieve a better physical environment for the civil defense purposes. Specific variable conditions and their meanings are shown in Table 4; Fig. 15.
Table 4.
Design variable.
| Typology | Variant | Title | Unit | Range |
|---|---|---|---|---|
| Importation | A | Ventilation shaft 1 location | m | [0,80] |
| B | Ventilation shaft 2 location | m | [4,68] | |
| C | Ventilation shaft 3 location | m | [0,120] | |
| Exports | Average wind speed | m/s | ||
| Average temperature | ℃ | |||
| Average relative humidity | % | |||
| Age of air | - |
Fig. 15.
Variable scope display.
Experimental design of the RSM
Based on the corresponding variables, the experimental design was conducted using Design Expert 13, and 17 experimental cases were analyzed by using the Box-Behnken (BB) design method. The numerical results are shown in Table 5.
Table 5.
Designed experimental matrix and corresponding results.
| Sequences | Position 1 (m) | Position 2 (m) | Position 3 (m) | Response value | |||
|---|---|---|---|---|---|---|---|
| Average wind speed (m/s) | Average temperature (°C) | Average relative humidity (%) | Average age of air (s) | ||||
| 1 | 0 | 4 | 60 | 0.188 | 25.58 | 86 | 13.33 |
| 2 | 80 | 4 | 60 | 0.204 | 25.84 | 86.1 | 11.96 |
| 3 | 0 | 68 | 60 | 0.199 | 25.80 | 85.6 | 11.44 |
| 4 | 80 | 68 | 60 | 0.228 | 26.21 | 85.5 | 10.21 |
| 5 | 0 | 36 | 0 | 0.195 | 25.49 | 86.4 | 13.20 |
| 6 | 80 | 36 | 0 | 0.2 | 25.65 | 85.9 | 13.34 |
| 7 | 0 | 36 | 120 | 0.201 | 25.85 | 85.3 | 11.60 |
| 8 | 80 | 36 | 120 | 0.219 | 26.09 | 85.2 | 10.45 |
| 9 | 40 | 4 | 0 | 0.176 | 25.41 | 86.3 | 13.84 |
| 10 | 40 | 68 | 0 | 0.203 | 25.69 | 86.0 | 11.82 |
| 11 | 40 | 4 | 120 | 0.198 | 25.93 | 85.4 | 11.52 |
| 12 | 40 | 68 | 120 | 0.208 | 26.18 | 85.2 | 10.44 |
| 13 | 40 | 36 | 60 | 0.22 | 25.79 | 85.7 | 11.25 |
| 14 | 40 | 36 | 60 | 0.221 | 25.81 | 85.8 | 11.56 |
| 15 | 40 | 36 | 60 | 0.217 | 25.80 | 85.6 | 11.53 |
| 16 | 40 | 36 | 60 | 0.216 | 25.86 | 85.8 | 11.25 |
| 17 | 40 | 36 | 60 | 0.214 | 25.76 | 85.6 | 11.22 |
Response surface model analysis
Following the numerical simulation based on the Response Surface Methodology (RSM), the results were further analyzed using Design Expert 13 software.
Response surface model for average wind speed
The statistical significance of average wind speed was investigated using the Analysis of Variance (ANOVA) method, with a quadratic model employed for fitting, as shown in Table 6. During the fitting process, the presence of some quadratic term coefficients led to an overfitting issue, with an R-squared adjusted (R2Adj) fit of 0.9259, but a prediction accuracy (R2Pred) of only 0.6420. Therefore, to generalize the model and enhance its predictive capability, non-significant quadratic terms (P < 0.05) such as AB, AC, and A2 were removed. Consequently, the R2Adj was adjusted to 0.8842, and the R2Pred improved to 0.7773. The results indicate that the impact on average temperature, in terms of significance, is in the order of the location of ventilation shaft 2 > ventilation shaft 1 > ventilation shaft 3, all showing significant effects.
Table 6.
Mean wind speed regression model ANOVA.
| Source | Equation of squares | Degrees of freedom | Mean square | F-value | P-value |
|---|---|---|---|---|---|
| Model | 0.0027 | 6 | 0.0004 | 21.37 | <0.0001 |
| A | 0.0006 | 1 | 0.0006 | 27.61 | 0.0004 |
| B | 0.0006 | 1 | 0.0006 | 30.96 | 0.0002 |
| C | 0.0003 | 1 | 0.0003 | 16.15 | 0.0024 |
| BC | 0.0001 | 1 | 0.0001 | 3.45 | 0.0928 |
| B2 | 0.0004 | 1 | 0.0004 | 21.46 | 0.0009 |
| C2 | 0.0005 | 1 | 0.0005 | 25.86 | 0.0005 |
| Residual | 0.0002 | 10 | 0.0000 | ||
| Lost proposal | 0.0002 | 6 | 0.0000 | 3.54 | 0.1209 |
| Pure error | 0.0000 | 4 | 8.300E-06 | ||
| Aggregate | 0.0029 | 16 | |||
| R2 | 0.9277 | ||||
| R2Adj | 0.8842 | ||||
| R2Pred | 0.7773 |
Considering the individual effects of factors A, B, and C (Fig. 16), the average wind speed is directly proportional to the position of A, shows a quadratic relationship with the position of B, and also a quadratic relationship with the position of C. Hence, positioning A at the far end, and B and C towards the middle to rear sections of the tunnel could effectively improve the overall average wind speed. The specific relationship equation between average wind speed and the positions of the ventilation shafts is given by Eq. (10):
Fig. 16.
(a) B and C factor-mean wind speed response surfaces and contours; (b) single-factor impact curve for mean wind speed..
 |
10 |
A represents the position of ventilation shaft A, in meters (m); B represents the position of ventilation shaft B, in meters (m); C represents the position of ventilation shaft C, in meters (m).
Response surface model for average temperature
A linear model significantly represents the results for average temperature, as displayed in Table 7. The F-value indicates the level of influence on average temperature, with the position of ventilation shaft C having the greatest impact, followed by shaft B and then shaft A. A direct proportional relationship is evident between average temperature and the positions of these ventilation shafts (Fig. 17), suggesting that openings at the deepest corners of the shelter most effectively improve the average temperature inside the shelter.
Table 7.
Mean temperature regression model ANOVA.
| Source | Equation of squares | Degrees of freedom | Mean square | F-value | P-value |
|---|---|---|---|---|---|
| Model | 0.7094 | 3 | 0.2365 | 70.47 | <0.0001 |
| A | 0.1431 | 1 | 0.1431 | 42.65 | <0.0001 |
| B | 0.1568 | 1 | 0.1568 | 46.73 | <0.0001 |
| C | 0.4095 | 1 | 0.4095 | 122.04 | <0.0001 |
| Residual | 0.0436 | 13 | 0.0034 | ||
| Lost proposal | 0.0383 | 9 | 0.0043 | 3.20 | 0.1374 |
| Pure error | 0.0053 | 4 | 0.0013 | ||
| Aggregate | 0.7530 | 16 | |||
| R2 | 0.9421 | ||||
| R2Adj | 0.9287 | ||||
| R2Pred | 0.8857 |
Fig. 17.
One-way influence curves for mean temperature.
The specific relationship equation for average temperature in relation to the positions of the ventilation shafts is given as follows (Eq. 11):
 |
11 |
A represents the position of ventilation shaft A, in meters (m); B represents the position of ventilation shaft B, in meters (m); C represents the position of ventilation shaft C, in meters (m).
Response surface model for average humidity
The results from a linear model for average humidity are significant, as shown in Table 8. The positions of ventilation shafts B and C significantly affect the average relative humidity (P < 0.05), while the impact of shaft A is limited. For relative humidity, all three positions show a negative correlation (Fig. 18), indicating that ventilation shafts placed deeper within the shelter effectively improve the internal relative humidity.
Table 8.
Mean humidity regression model ANOVA.
| Source | Equation of squares | Degrees of freedom | Mean square | F-value | P-value |
|---|---|---|---|---|---|
| Model | 1.86 | 3 | 0.6192 | 40.69 | <0.0001 |
| A | 0.0450 | 1 | 0.0450 | 2.96 | 0.1092 |
| B | 0.2812 | 1 | 0.2812 | 18.49 | 0.0009 |
| C | 1.53 | 1 | 1.53 | 100.64 | <0.0001 |
| Residual | 0.1978 | 13 | 0.0152 | ||
| Lost proposal | 0.1578 | 9 | 0.0175 | 1.75 | 0.3090 |
| Pure error | 0.0400 | 4 | 0.0100 | ||
| Aggregate | 2.06 | 16 | |||
| R2 | 0.9038 | ||||
| R2Adj | 0.8816 | ||||
| R2Pred | 0.8194 |
Fig. 18.
One-way influence curves of mean relative humidity.
The specific relationship equation for average humidity in relation to the positions of the ventilation shafts is presented below:
 |
12 |
A represents the position of ventilation shaft A, in meters (m); B represents the position of ventilation shaft B, in meters (m); C represents the position of ventilation shaft C, in meters (m).
Response surface model for air age
A quadratic linear model holds significant results for air age, as indicated in Table 9. The presence of non-significant quadratic terms AB and B2 (P < 0.05) led to model overfitting, with an R2Pred of only 0.6152. By removing these terms, the predictive accuracy improved to 0.8154, where the difference between R2Adj and R2Pred is less than 0.2, indicating a high correlation between actual and predicted values, affirming the model’s validity. Single-factor analysis shows that air age is nearly linearly related to the position of ventilation shaft B and exhibits a quadratic relationship with positions A and C (Fig. 19).
Table 9.
Mean air age regression model AOA.
| Source | Equation of squares | Degrees of freedom | Mean square | F-value | P-value |
|---|---|---|---|---|---|
| Model | 17.79 | 7 | 2.54 | 40.41 | <0.0001 |
| A | 1.63 | 1 | 1.63 | 25.90 | 0.0007 |
| B | 5.68 | 1 | 5.68 | 90.30 | <0.0001 |
| C | 8.38 | 1 | 8.38 | 133.33 | <0.0001 |
| AC | 0.4160 | 1 | 0.4160 | 6.62 | 0.0301 |
| BC | 0.2209 | 1 | 0.2209 | 3.51 | 0.0937 |
| A2 | 0.4089 | 1 | 0.4089 | 6.50 | 0.0312 |
| C2 | 0.9776 | 1 | 0.9776 | 15.55 | 0.0034 |
| Residual | 0.5660 | 9 | 0.0629 | ||
| Lost proposal | 0.4533 | 5 | 0.0907 | 3.22 | 0.1402 |
| Pure error | 0.1127 | 4 | 0.0282 | ||
| Aggregate | 18.36 | 16 | |||
| R2 | 0.9692 | ||||
| R2Adj | 0.9452 | ||||
| R2Pred | 0.8154 |
Fig. 19.
(a) Single-factor influence curves for air age; (b) Air age response surfaces and contours for factors A and C; (c) Air age response surfaces and contours for factors B and C.
The specific relationship equation for average air age in relation to the positions of the ventilation shafts is outlined below:
 |
13 |
A represents the position of ventilation shaft A, in meters (m); B represents the position of ventilation shaft B, in meters (m); C represents the position of ventilation shaft C, in meters (m).
Design optimization
Optimal parameters for the shelter’s internal environment were determined using Design Expert 13 software. The optimization targets were set to maximize average wind speed, set the average temperature to 27 °C, minimize average relative humidity, and minimize average air age. As this is a multi-objective optimization problem, the solution set is a Pareto set. The software provided a feasible set of values, from which the first set was chosen for validation. Design Expert 13 suggested the following positions: A = 54.76 m, B = 51.45 m, C = 79.85 m. The response values were as follows: average wind speed = 0.222 m/s, average temperature = 26 °C, average relative humidity = 85.47%, and average air age = 10.57s. Compared to the standards set by “Civil Air Defense Engineering Environmental Health Requirements” (GBT 17216-2012), which specify average wind speed (greater than 0.2 m/s), average temperature (26–28 °C), and average relative humidity (less than 80%), the results show that average wind speed and temperature meet the standards, while relative humidity shows significant improvement but remains slightly above the standard, and air age is effectively reduced.
Cradle CFD simulation results are as follows: average wind speed 0.219 m/s with a 1.4% error, temperature 25.96 °C with a 0.02% error, average relative humidity 85.41% with a 0.07% error, and average air age 10.78s with a 2% error. The errors in the physical quantities are within a reasonable range. Figure 20 displays the physical distribution at a height of 1.5 m after optimization.
Fig. 20.
Comparison of 1.5 m plane physical environment before and after optimization.
Conclusions
This study simulated the physical environment inside the Yunlong Mountain tunnel shelter in Xuzhou City. By comparing single-shaft and multi-shaft ventilation systems under positive and negative pressures and optimizing the position of the multi-shaft negative pressure ventilation, the following conclusions were drawn:
Natural ventilation alone is insufficient in the deeper parts of the tunnel, with weak airflow that fails to form an effective ventilation path. At a height of 1.5 m, the average temperature was 23.4 °C, average humidity 93.8%, average wind speed 0.143 m/s, and average air age 18.4s, with the first three indicators failing to meet regulatory standards.
Compared to single-shaft ventilation, multi-shaft negative pressure ventilation enhances ventilation in stages, while positive pressure ventilation shows an accumulation effect in air volume.
Negative pressure ventilation significantly improves all parameters within the shelter’s tunnels, while positive pressure ventilation conflicts with the original airflow organization, making positive optimization difficult. In the center of the three internal tunnels, setting 4 m2 negative pressure shafts at a uniform speed of 0.4 m/s outward at a height of 1.5 m increases the average temperature by 9.8%, decreases relative humidity by 8.6%, increases average wind speed by 55.2%, and reduces air age by 39%.
Based on RSM analysis, the positions of ventilation shafts A, B, and C significantly affect average wind speed and temperature, while the positions of B and C significantly impact average relative humidity and air age.
RSM identified an optimal arrangement of ventilation shafts, validated by Cradle CFD software simulation, showing a maximum physical quantity deviation of 2%, providing a reference for practical engineering applications.
This paper studied the distribution of the physical environment inside the civil defense tunnels, providing methods and references for the transformation and integration of underground civil defense spaces. However, this study has limitations: even with passive optimization, the internal relative humidity of the shelter remains above the regulatory value, and mechanical ventilation equipment may still be needed in actual modifications. The number of ventilation shafts also significantly affects the internal physical environment; this paper only discussed the scenario of three shafts distributed in three tunnels, not considering other numbers of shafts. Some RSM linear models only achieve about 80% prediction accuracy; although they can predict to a certain extent, there is still room for optimization. Future research could consider changing the prediction model to more advanced algorithms like Random Forest or Deep Neural Networks to further explore the effects of random positions of ventilation shafts in the underground shelters on improving the thermal environment.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Author contributions
Y.J. analyzed the data and wrote the paper; J.L. participated in the revision of the paper; X.H. and J.D. designed the research framework and analyzed the data; F.H. participated in the revision of the paper. All authors have read and agreed to the published version of the manuscript.
Funding
This research was funded by Jiangsu Vocational Institute of Architectural Technology, Jiangsu Collaborative Innovation Center for Building Energy Saving and Construct Technology, Major Research Fund Project (No. SJXTZD21051), 2023 Jiangsu Province Industry University Research Cooperation Project (No. BY20231316); Research Project on Teaching Reform at Yangzhou University in 2023 (No. YZUJX2023-d9); Subproject of National Key R&D Project “Research on Comprehensive Planning and Renovation Technology of Underground Space” (No. 2018YFC0704903-03).
Data availability
Data is provided within the manuscript or supplementary information files.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Renwu, D. Design of urban underground space: considering human defense facilities. J. Civil Eng. Urban Plann.5(10), 1–5 (2023). [Google Scholar]
- 2.Lei, S., Shen, Y., Xiao, Q., Xi, J. & Gu, L. Current situation and future development concept of urban underground space development and utilization. J. Undergr. Space Eng.15, 965–979 (2019). [Google Scholar]
- 3.Chao, W. Y., Peng, Z. Z., Fu, Q. H. & Jun, C. Q. Research progress of blast performance evaluation for typical urban underground air defense structures under external explosion. Sci. Technol. Eng.23(32), 13633–13648 (2023). [Google Scholar]
- 4.Li, Y., Yuan, Y., Li, C., Han, X. & Zhang, X. Human responses to high air temperature, relative humidity and carbon dioxide concentration in underground refuge chamber. Build. Environ.131, 53–62 (2018). [Google Scholar]
- 5.Li, Y., Geng, S., Zhang, X. & Zhang, H. Study of thermal comfort in underground construction based on field measurements and questionnaires in China. Build. Environ.116, 45–54 (2017). [Google Scholar]
- 6.Luo, S., Liang, W. & Zhao, G. Likelihood-based hybrid Oreste method for evaluating the thermal comfort in underground mines. Appl. Soft Comput.87, 105983 (2020). [Google Scholar]
- 7.Marshall, G. Safety Engineering: Brooks/Cole Engineering Division (1982).
- 8.Wang, Y. Study on the smoke flow regulation in underground space based on the Cfd simulation. Appl. Mech. Mater.638, 858–863 (2014). [Google Scholar]
- 9.Zhang, H., Liu, J. & Li, C. Long-term investigation of moisture environment in underground civil air defence work. Indoor Built Environ.26 (6), 744–757 (2017). [Google Scholar]
- 10.Tao, H., Li, Y. F. X. & Zhang, Z. Investigation of formaldehyde and TVOC in underground malls in Xi’an, China: concentrations, sources, and affecting factors. Build. Environ.85, 85–93 (2015). [Google Scholar]
- 11.Mukhtar, A., Ng, K. C. & Yusoff, M. Z. Passive thermal performance prediction and multi-objective optimization of naturally-ventilated underground shelter in Malaysia. Renew. Energy123, 342–352 (2018). [Google Scholar]
- 12.Jin, T. et al. Experimental investigation on thermal performance of underground refuge chamber under natural convection and ventilation. Case Stud. Therm. Eng.52, 103637 (2023). [Google Scholar]
- 13.Lin, J., Kong, Y. & Zhong, L. Optimization of environment control system for narrow sleeping space in underground shelters. Energy Build.263, 112043 (2022). [Google Scholar]
- 14.Roghanchi, P., Kocsis, K. C. & Sunkpal, M. Sensitivity analysis of the effect of airflow velocity on the thermal comfort in underground mines. J. Sustain. Min.15 (4), 175–180 (2016). [Google Scholar]
- 15.Sasmito, A. P., Kurnia, J. C., Birgersson, E. & Mujumdar, A. S. Computational evaluation of thermal management strategies in an underground mine. Appl. Therm. Eng.90, 1144–1150. (2015).
- 16.Xiong, J. et al. Probing the historic thermal and humid environment in a 2000-year-old ancient underground tomb and enlightenment for cultural heritage protection and preventive conservation. Energy Build.251, 111388 (2021). [Google Scholar]
- 17.Zhang, J. et al. Automatic relative humidity optimization in underground heritage sites through ventilation system based on digital twins. Build. Environ.216, 108999 (2022). [Google Scholar]
- 18.Mazarron, F. R. & Canas, I. Seasonal analysis of the thermal behaviour of traditional underground wine cellars in Spain. Renew. Energy34(11), 2484–2492 (2009).
- 19.Tinti, F. et al. Experimental analysis of shallow underground temperature for the assessment of energy efficiency potential of underground wine cellars. Energy Build.80, 451–460 (2014). [Google Scholar]
- 20.Cañas, I. & Fernando, R. M. The effect of traditional wind vents called Zarceras on the hygrothermal behaviour of underground wine cellars in Spain. Build. Environ.44(9), 1818–1826 (2009). [Google Scholar]
- 21.Liu, Wei, S., Liang & Shi, C. Risk modelling and simulation of thermal safety in underground railway tunnel surrounding. Accid. Anal. Prev.168, 106620 (2022). [DOI] [PubMed] [Google Scholar]
- 22.Guo, J., Li, A., Zhang, C., Li, J., Che, J., Xiong, J. & Jiao, X. Modeling effective heat transfer and ventilation in deeply buried Underground tunnels. Int. J. Therm. Sci.184, 107949 (2023). [Google Scholar]
- 23.Klauss, A. K., Tull, R. H., Roots, L. M. & Pfafflin, J. R. History of the changing concepts in ventilation requirements. ASHRAE J.53(2), 34–41 (2011).
- 24.Fisk, W. J. How home ventilation rates affect health: A literature review. Indoor Air28(4), 473–487 (2018). [DOI] [PubMed]
- 25.Kato, S. Review of airflow and transport analysis in building using CFD and network model. Japan Archit. Rev.1(3), 299–309 (2018). [Google Scholar]
- 26.Wen, Y., Wei, J., Lau, S. K., Gu, Z. & Leng, J. A two-level optimisation approach for underground natural ventilation based on CFC and building energy simulations. Energy Build.310, 114102 (2024). [Google Scholar]
- 27.Ren, C. & Cao, S.-J. construction of linear temperature model using non-dimensional heat exchange ratio: towards fast prediction of indoor temperature and heating, ventilation and air conditioning systems control. Energy Build.251, 111351 (2021).
- 28.Qiaoyun, H., Yi, Z., Shenghua, Z. & Li, K. Computational evaluation of cooling system under deep hot and humid coal mine in China: A thermal comfort study. Tunn. Undergr. Space Technol.90, 394–403 (2019). [Google Scholar]
- 29.Ma, H., Huang, J. & Zhou, X. Effect of ventilation on thermal and humidity environment of the underground utility tunnel in the Plum rain season in Southern China: field measurement and CFD simulation. Undergr. Space13, 301–315 (2023). [Google Scholar]
- 30.Tominaga, Y. et al. AIJ guidelines for practical applications of CFD to pedestrian wind environment around buildings. J. Wind Eng. Ind. Aerodyn.96, 1749–1761 (2008). [Google Scholar]
- 31.Kusuda, T. & Achenbach, P. R. Earth Temperature and Thermal Diffusivity at Selected Stations in the United States Vol. 71 (National Bureau of Standards, 1965).
- 32.Labs, K. & Oak Ridge National Laboratory. Regional Analysis of Ground and above-Ground Climate (Oak Ridge National Laboratory, 1981).
- 33.Wen, Y., Lau, S. K., Leng, J., Zhou, K. & Shi-Jie, C. Passive ventilation for sustainable underground environments from traditional underground buildings and modern multiscale spaces. Tunn. Undergr. Space Technol.134, 105002 (2023). [Google Scholar]
- 34.Federspiel, C. C. Air-change effectiveness: theory and calculation methods. Indoor Air9, 47–56 (1999). [DOI] [PubMed] [Google Scholar]
- 35.Van Buggenhout, S. et al. Data-based mechanistic modelling approach to determine the age of air in a ventilated space. Build. Environ.41, 557–567 (2006). [Google Scholar]
- 36.Heidari, A. & Peyvastehgar, Y. Using multi objective optimization in term of position and area of the air outlet shaft in Showadan. Tunn. Undergr. Space Technol.132, 104778 (2023). [Google Scholar]
- 37.Porras-Amores, C., Mazarrón, F. R., Cañas, I. & Sáez, P. V. Natural ventilation analysis in an underground construction: CFD simulation and experimental validation. Tunn. Undergr. Space Technol.90, 162–173 (2019).
- 38.Hazbei, M., Nematollahi, O. & Behnia, M. Reduction of energy consumption using passive architecture in hot and humid climates. Tunn. Undergr. Space Technol.47, 16–27 (2015). [Google Scholar]
- 39.Montazeri, H. Experimental and numerical study on natural ventilation performance of various multi-opening wind catchers. Build. Environ. 46(2), 370–378 (2011).
- 40.Zhou, K., Li, S., Cao, R. & Ma, X. Optimal technology innovation of teaching buildings for primary and middle schools based on compound negative pressure ventilation. Appl. Nanosci.13(3), 2037–2048 (2023). [Google Scholar]
- 41.Simpson, T. W., Poplinski, J. D., Koch, P. N. & Allen, J. K. Metamodels for computer-based engineering design: survey and recommendations. Eng. Comput.17, 129–150 (2001).
- 42.Shen, X. & Zhang, G. Assessments of experimental designs in response surface modelling process: estimating ventilation rate in naturally ventilated livestock buildings. Energy Build.62, 570–580 (2013). [Google Scholar]
- 43.Huang, X., Yao, S., Yang, X. & Zhou, R. Melting performance assessments on a triplex-tube thermal energy storage system: optimization based on response surface method with natural convection. Renew. Energy188, 890–910 (2022). [Google Scholar]
- 44.Alizadeh, M. & Sadrameli, S. M. Indoor thermal comfort assessment using PCM based storage system integrated with ceiling fan ventilation: experimental design and response surface approach. Energy Build.188, 297–313 (2019). [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data is provided within the manuscript or supplementary information files.