Experimental study on variable frequency control and air leakage control of multi-fan combination in subway tunnels

Abstract

In centralized smoke exhaust emergency ventilation systems for metro tunnels, axial supply fans are often located far from the tunnel exit due to complex terrain constraints. This configuration creates a parallel-duct airflow path, causing a portion of the fresh airflow to bypass the traffic lane and escape directly through the exit, significantly reducing smoke exhaust efficiency. To address this leakage issue, this study takes the Wawuzhuang–Guizhou Road section of Qingdao Metro Line 1 as the engineering prototype and establishes a 1:10 scaled-down experimental platform based on similarity criteria, with a variable cross-section exit simulating a typical single-side leakage scenario. Four fan combination modes—comprising axial supply and axial exhaust fans—and 80 variable-frequency operating conditions were designed. Key parameters, including static pressure distribution, average airflow velocity in the traffic lane, and fan power consumption, were systematically measured. The effective smoke exhaust efficiency (η_e) was introduced as the core quantitative metric, and an equivalent length model was employed to parameterize leakage effects by converting abnormal operating states (e.g., tunnel exit leakage) into equivalent resistance increments in the ventilation system. Experimental results show that the configurations “dual-supply zero-exhaust” and “dual-supply dual-exhaust” exhibit higher leakage rates and lower η_e compared to “dual-supply single-exhaust”. Further comparison between “dual-supply left-exhaust” and “dual-supply right-exhaust” reveals similar leakage rates under identical frequency settings; however, when the exhaust fan frequency exceeds 18.5 Hz, the “dual-supply right-exhaust” configuration achieves significantly higher η_e than its left-exhaust counterpart, with η_e increasing by 3–5% per 1 Hz rise in exhaust frequency. This highlights the critical influence of exhaust fan placement on system performance. The findings indicate that, in centralized smoke exhaust metro tunnels, increasing the frequency of the axial exhaust fan positioned on the exit side can effectively mitigate leakage and enhance smoke exhaust efficiency, supporting the recommendation of the “dual-supply right-exhaust” fan configuration. This work provides crucial experimental evidence for optimizing fan combinations and frequency settings in long tunnels under extreme emergency scenarios—such as single-side exit failure—and offers practical insights for improving emergency smoke control and operational safety in tunnel systems.

Introduction

With the acceleration of global urbanization, the safe operation of metro tunnels faces severe challenges. According to the International Tunnel Association (ITA), between 2010 and 2024, tunnel fire incidents directly attributed to ventilation system failures accounted for one-third of all tunnel fire accidents¹. The longitudinal pressurized ventilation system, renowned for its high efficiency and energy-saving features, is widely adopted in tunnel engineering. However, in tunnels employing this method, the necessity to locate supply fans far from the tunnel exit due to spatial constraints often leads to significant air leakage at the exit, reducing smoke exhaust efficiency by 40–60%. This issue becomes particularly acute in specialized geological projects such as cross-sea tunnels. During emergency smoke evacuation, air leakage through a single-side tunnel exit can delay smoke removal and significantly increase the risk of smoke spread, as shown in Fig. 1^2,3.

Fig. 1

Schematic of air leakage at the outlet of a centralized exhaust metro tunnel.

To improve ventilation and smoke exhaust efficiency, scholars have conducted extensive research on supply ventilation tunnels through small-scale physical modeling and numerical simulations. Yue Kun et al.⁴ studied the Zhuhai Xingye Express Line (West Line) tunnel project and concluded that supply ventilation should be implemented during construction to maintain sufficient airflow rate near the working face. Regarding supply fan layout, Zhao Ningyu et al.⁵ developed a length correction formula for supply ventilation in high-altitude tunnels based on the principle of flow conservation and empirical data from low-altitude projects. Li Yuefeng⁶ proposed an integrated ventilation method combining light-steel ducts with supply ventilation, drawing on construction experience from the Gansu–Qinghai tunnel on the Xining–Chengdu Railway. Huang Qiang⁷ analyzed numerous intermediate ventilation shaft projects across China and, based on system design requirements, proposed six optimized tunnel ventilation configurations. Chen Shiqiang et al.⁸ developed a tunnel ventilation model based on the section between Dayang Station and Qingdao North Station on Qingdao Metro Line 8. Through parallel fan experiments, they identified an optimal frequency-matching range, providing guidance for variable-frequency control during operation and effectively mitigating air leakage at the tunnel exit. Chen Fangxing et al.⁹ quantified the influence of various fan combinations and operating parameters on tunnel flow field characteristics. By applying dimensionless criteria and Euler number analysis, they compared the smoke exhaust efficiency of different configurations and found that activating exhaust fans significantly enhanced both ventilation and smoke removal capacity. Miho Seike¹⁰ constructed a small-scale tunnel model to investigate the delaying effect of fixed barriers on smoke propagation during fires and the flow structure of smoke and fresh air in the upper tunnel space. Feng Tao^11,12 established a scaled-down metro tunnel model based on similarity criteria and conducted fire experiments, revealing the longitudinal attenuation characteristics of smoke temperature. Tao et al.¹³ examined the temperature distribution and control strategies for tunnel smoke using numerical simulations and small-scale model experiments. To meet emergency smoke exhaust requirements, supply fans are often combined with exhaust fans. Lei Zhenlin et al.¹⁴ integrated large-scale model experiments with numerical simulations to study internal flow fields under different fan combination modes, characterized smoke flow behavior, and optimized operating conditions based on effective smoke exhaust efficiency. Cheng Xueyou et al.¹⁵ experimentally evaluated two fan combinations—dual-supply and dual-supply/single-exhaust—using a physical tunnel model to clarify the relationship between fan power consumption and ventilation capacity. Focusing on fan arrangement and spacing parameters, Xu Zhisheng et al.¹⁶ found that a lateral spacing of three times the fan diameter yielded optimal flow field distribution and pollutant control performance.

Existing studies have primarily focused on ventilation optimization in mid-tunnel sections^17,18,19, while research on the unilateral air leakage at tunnel exits induced by the offset placement of supply fans and the interaction effects of multi-fan combinations remains insufficient^20,21,22. Current achievements primarily address overall ventilation efficiency in highway tunnels, whereas localized ventilation challenges in metro tunnels—such as spatially constrained fan layouts—lack systematic experimental validation²³. This study employs the Wawuzhuang–Guizhou Road section of Qingdao Metro Line 1 as a prototype, utilizing a 1:10 scale model with a variable-section outlet to simulate air leakage scenarios under offset supply fan configurations. It reveals the effects of multi-fan combinations and variable-frequency control on static pressure, average wind speed in the traffic lane, and air leakage rate, and proposes an optimal configuration using air leakage control and effective smoke exhaust efficiency as evaluation criteria. The findings provide critical experimental data to inform the revision of ventilation design standards for metro tunnels.

Research objectives and methods

This study addresses the issue of fresh air leakage—wherein supply air bypasses the traffic lane and escapes directly through the tunnel exit—caused by the excessive distance between the supply fan and the tunnel exit in centralized smoke exhaust systems for metro tunnels. The research aims to quantify the impact of different fan combinations and variable-frequency control strategies on effective smoke exhaust efficiency, identify optimal fan placement and operational parameters, and thereby enhance emergency smoke removal performance—particularly by providing optimized solutions for scenarios involving single-side tunnel exit failure (e.g., due to blockage or structural constraints). The research methodology and experimental procedure are summarized in the flowchart in Fig. 2.

Fig. 2

Research methodology and experimental procedure.

Engineering overview

The tunnel section between Wawuzhuang Station and Guizhou Road Station on Qingdao Metro Line 1 has a total length of approximately 8.1 km, including a 3.49 km undersea section, making it one of the longest metro tunnel segments in China. Compared with shorter tunnels, the potential consequences of accidents in this section are significantly more severe. Due to the infeasibility of installing intermediate ventilation shafts, overhead smoke extraction ducts were installed in the mined tunnel section. The schematic plan view of the undersea section is shown in Fig. 3a, while the engineering geological profile is presented in Fig. 3b.

Fig. 3

Engineering layout and geological profile of the Wawuzhuang Station–Guizhou Road Station section, Qingdao Metro Line 1.

Experimental model platform

This study employs a 1:10 scale experimental model to simulate the ventilation and smoke exhaust flow characteristics in the tunnel section between Wawuzhuang Station and Guizhou Road Station on Qingdao Metro Line 1, based on Reynolds number similarity and the fully rough turbulent flow similarity principle established by Nikuradse. At sufficiently high Reynolds numbers, the flow enters the fully rough regime, where the Darcy friction factor depends only on relative roughness. Consequently, strict Reynolds number similarity is not required, provided that the model and prototype share the same relative roughness. To satisfy this condition, the tunnel model’s internal surface roughness was scaled proportionally. The geometric scale ratio is 1:10 for length, resulting in a cross-sectional area ratio of 1:100. Under these conditions, dynamic similarity in the fully rough regime allows the use of identical airflow velocities and pressures (scale ratio 1:1) in the model and prototype, ensuring consistent aerodynamic behavior¹⁴.

This scaled model effectively supports experimental investigations into air leakage control and effective smoke exhaust efficiency under various fan combinations and variable-frequency operating conditions. The primary components of interest include the traffic lane, axial supply fans with their associated duct segments, and the variable-cross-section tunnel exit. All other non-essential features were simplified to construct the scaled tunnel model platform. A plan view of the tunnel ventilation test rig is shown in Fig. 4, where the central red section represents the overhead smoke outlet, the left red section denotes axial supply fan #2, and the right red section indicates axial supply fan #1.

Fig. 4

Plan view of the tunnel ventilation experimental model platform.

Based on the schematic diagram in Fig. 4, the experimental model platform was constructed as shown in Fig. 4. Plan view of the tunnel ventilation experimental model platform, with the main functional sections detailed as follows:

• (1) Power system, consisting of four explosion-proof axial-flow fans. Fans #1 and #2 are supply fans with a rated power of 11 kW, installed at the model inlet to generate tunnel airflow. Fans #3 and #4 are exhaust fans with a rated power of 5.5 kW, connected to the overhead smoke exhaust duct. All fans are axial-flow type, employed for centralized ventilation and distinct from jet fans that rely on momentum transfer. Fans #1 and #2 deliver fresh air into the tunnel by generating positive pressure, while fans #3 and #4 extract smoke through the overhead smoke outlet by creating negative pressure.

• (2) Supply air inlet section, comprising the main inlet ducts and wind speed measurement devices for the ventilation system.

• (3) Traffic lane section, comprising duct segments equipped with wind speed and static pressure sensors. The traffic lane is horizontally divided into left and right lanes at the measurement cross-section; this study focuses on the right traffic lane.

• (4) Overhead smoke outlet, which connects the traffic lane below to the smoke exhaust duct above, facilitating smoke extraction from the tunnel.

• (5) Variable cross-section outlet, featuring a sudden contraction–expansion circular geometry, internally connected to the traffic lane and externally open to ambient air. A throttle ring assembly is installed at the outlet, enabling adjustment of the effective opening area by adding or removing rings. In addition to the fully open configuration, four distinct outlet area ratios are achievable. To simulate severe air leakage scenarios observed in real engineering projects, the outlet is set to the smallest opening area, as shown in Fig. 5.

• (6) Electrical control system, comprising a control cabinet and an operation console. The cabinet adjusts fan speed through variable-frequency drives, while the console displays and records data from all sensors and transmitters in real time, with measurements logged every second.

Fig. 5

Ventilation experimental model platform of the Wawuzhuang–Guizhou Road section, Qingdao Metro Line 1.

The experimental measurement system included wind speed sensors, differential pressure sensors, and power sensors installed in the traffic lane, along with an anemometer positioned at the tunnel outlet. Detailed specifications of these instruments are provided in Table 1.

Table 1 Experimental measuring instrument and its main parameters.

Variable frequency control principle and its influence on experimental system parameters

The principle of variable frequency control for fans is based on the fan affinity laws. At its core, it involves adjusting the rotational speed of the motor by varying the supply frequency, thereby dynamically regulating the fan’s airflow rate, pressure, and power consumption. The specific principles and their impacts on experimental parameters are as follows:

• (1) Fundamental principle of variable frequency control.

The rotational speed n of an AC motor is directly proportional to the supply frequency f (i.e., n∝f ). A variable frequency drive (VFD) enables continuous, stepless speed control by modulating the input frequency. Provided the fan’s geometric configuration remains unchanged, its performance parameters vary with speed (and thus frequency) according to the fan affinity laws:

• (a) Airflow rate: Q∝f.

• (b) Total pressure: P∝f²

• (c) Shaft power: N∝f³

This implies that even a modest increase in frequency results in a significant rise in both pressure and power consumption.

• (2) Influence of variable frequency control on experimental parameters.

• (a) Average airflow velocity in the traffic lane: Airflow increases linearly with frequency, directly enhancing air velocity within the tunnel and thereby affecting smoke exhaust capacity and overall airflow organization.

• (b) Static pressure distribution: Since pressure scales with the square of frequency, adjusting the frequency substantially alters the static pressure gradient—particularly near the fans and at the tunnel exit—consequently influencing the magnitude and spatial distribution of air leakage.

• (c) Fan power consumption: Power consumption varies with the cube of frequency, making it a critical metric for evaluating system energy efficiency. Although higher-frequency operation improves effective smoke exhaust efficiency, it leads to a sharp increase in energy demand, warranting further investigation to identify optimal trade-offs.

Theoretical derivation of tunnel model equivalent length and air leakage

Due to geological and topographical constraints, metro tunnels are often deeply buried, limiting the placement of ventilation equipment. In this study, the supply fan is located at a significant distance from the tunnel exit—a common scenario in real-world projects. By integrating real-world air leakage scenarios with a 1:10 scale tunnel model, we simulated the flow field between the supply fan—positioned away from the tunnel exit—and the outlet. The resulting parallel bypass flow path around supply fan #1 is illustrated in Fig. 3.

The fresh airflow from supply fan #1 splits into parallel paths. Since fan #1 is located approximately 1 m from the tunnel model exit (i.e., L_c≈1 m, as shown in Fig. 6), a portion of the airflow bypasses the traffic lane and leaks directly through the outlet.

With the smoke generator activated, supply fans #1 and #2 deliver airflow rates Q₁ and Q₂, respectively, into the tunnel. Exhaust fans #3 and #4 extract airflow at rates Q₃ and Q₄, drawing smoke through the overhead smoke outlet. Under steady-state conditions, the air leakage rate at the tunnel exit, Q_leak, is determined by the imbalance between total supply and total exhaust, as expressed in Eq. (1):

$${Q_{{\rm{leak}}}}{\rm{ }} = {\rm{ }}{Q_1} + {Q_2} - {Q_3} - {Q_4}$$

(1)

Fig. 6

Schematic of the parallel airflow paths formed by axial supply fan #1.

To quantify the impact of outlet air leakage on the overall performance of the ventilation and smoke exhaust system, this paper introduces the effective smoke exhaust efficiency (η_e) as the core evaluation metric. This metric is defined as the ratio of the airflow that effectively contributes to smoke removal to the total exhaust capacity of the system.

Specifically, the numerator is taken as Q₂, the airflow rate delivered by supply fan #2. Due to its strategic placement near the smoke exhaust zone, Q₂ serves as a representative indicator of the fresh air driving smoke toward the exhaust outlet. It is calculated from the wind speed measured by an anemometer at the traffic lane cross-section near fan #2 multiplied by the cross-sectional area. The denominator, Q₃ + Q₄, represents the total exhaust airflow from exhaust fans #3 and #4, obtained by multiplying the wind speeds measured in their respective exhaust ducts by the corresponding cross-sectional areas. Both numerator and denominator are derived directly from experimental measurements and can be cross-validated against airflow data within the exhaust ducts.

Accordingly, the effective smoke exhaust efficiency is given by Eq. (2):

$${\eta _e}=\frac{{{Q_2}}}{{{Q_3}+{Q_4}}} \times 100\%$$

(2)

The airflow rate supplied by fan #1 into the tunnel is Q₁. Air leakage at the outlet creates a parallel bypass flow path among fan #1, the traffic lane, and the tunnel exit, causing a portion of Q₁ to escape directly through the outlet. This leakage reduces the airflow available in the traffic lane for smoke transport, thereby decreasing the smoke-laden airflow captured by exhaust fans #3 and #4, and ultimately weakening the system’s overall smoke exhaust capability. To quantitatively assess the comprehensive influence of outlet leakage on ventilation and smoke exhaust performance, key experimental parameters were measured, including the average airflow velocity in the traffic lane, static pressure distribution, and air leakage rate at the outlet.

Simultaneously, to emulate tunnel length and realistic leakage scenarios, a throttle ring is installed at the outlet of the experimental platform, forming a sudden contraction–sudden expansion geometry with a variable cross-section. This configuration effectively controls the equivalent tunnel length of the traffic lane section and thereby adjusts the axial distance between the supply fan and the outlet, enabling deliberate regulation of the air leakage rate.

Based on the equivalence principle between frictional (major) losses and local (minor) losses, the additional pressure drop induced by the variable cross-section outlet is equivalently represented as an added tunnel length. According to the empirical relationship established in Reference²⁴—derived from CFD simulations and scaled experimental data—the equivalent length L is related to the outlet area ratio x as given in Eq. (3).This relationship exhibits a strong linear correlation (R² > 0.96 ) within the interval x∈[0.133,0.735], which is therefore adopted as the valid range of application²⁴.

$$L=5736.853 \cdot {e^{ - \frac{x}{{0.03787}}}}+484.963 \cdot {e^{ - \frac{x}{{0.12905}}}}+21.863$$

(3)

The area ratio x is defined as the ratio of the throttle ring opening area to the maximum outlet area. The range 0.133 ≤ x ≤ 0.735 is jointly determined by the physical realizability of the experimental setup and typical engineering operating conditions. The lower bound (x = 0.133 ) corresponds to the most severe leakage scenario observed in field conditions, where the leakage area is 0.0113 m² relative to a traffic lane cross-sectional area of 0.085 m². The upper bound (x = 0.735 ) is constrained by the minimum feasible size of the throttle ring in the experimental model—beyond which airflow blockage or hydrodynamic instability may occur. Preliminary experiments confirmed that this value represents the maximum area ratio at which stable operation is achievable. This range thus encompasses typical emergency ventilation scenarios, from “severe leakage” to “nearly leakage-free,” ensuring engineering representativeness.

Taking the minimum area ratio case as an example: when the supply fan delivers air toward the tunnel portal with a leakage area of 0.0113 m², the corresponding area ratio is x = 0.133. Substituting into Eq. (3) yields an equivalent full-scale tunnel length of 1968.05 m. According to the 1:10 geometric scale, the physical model platform has a total length of 20 m. Given that the axial distance between the supply fan and the outlet in the model is approximately 1 m, the corresponding full-scale distance is scaled proportionally to 98.4 m. This value falls well within the 300 m effective monitoring range recommended in¹⁵, further validating the engineering applicability of both the defined area ratio range and the equivalent length model.

Experimental procedures and operating conditions design

In actual emergency ventilation scenarios, air leakage at metro tunnel exits primarily consists of fresh supply airflow that bypasses the traffic lane, rather than smoke-laden gases. To simulate a range of realistic leakage conditions, the experiment employed a variable cross-section outlet, with the minimum leakage configuration (corresponding to the largest area ratio) serving as a baseline for comparison. In this setup, the axial distance between supply fan #1 and the outlet represents the equivalent tunnel length, which is determined by the cross-sectional area of the variable-geometry outlet structure. This critical distance was precisely controlled by assembling or disassembling throttle rings at the outlet.

The experimental design established two key independent variables: fan combination and frequency adjustment. Four distinct fan combinations were systematically investigated: dual-supply zero-exhaust, dual-supply left-side exhaust, dual-supply right-side exhaust, and dual-supply dual-exhaust.

Fan operational control, including on/off switching and frequency modulation, was executed through a centralized control console. The spatial distribution of measurement cross-sections and monitoring points is presented in Fig. 7.

During the experiments, sensors were deployed to monitor parameters including the average wind speed in the tunnel model, static pressure, and outlet wind speed. Multiple datasets meeting the critical smoke exhaust conditions were obtained through experimental observations. Following fan activation, the anemometers continuously collected data at 1-s intervals, with 120 s of continuous data extracted for analysis. To minimize systematic errors caused by significant flow field variations and uneven wind speed distribution across cross-sections, the original wind speed data were corrected using the area-weighting method. The steps are as follows:

• (1) Average wind speed measurement: Equivalent wind speeds were first obtained for each cross-section; The average value of wind speeds from three cross-sections was calculated as the mean wind speed in the tunnel driving lane;

• (2) Detailed 120-s data from the differential pressure sensor at Measurement Cross-Section 1 were exported; The mean value of these data was adopted as the static pressure in the tunnel.

• (3) Export wind speed measurement: The fluid motion at the driving lane outlet exhibited jet flow characteristics. After stabilizing the fan operation at set frequencies; A calibrated anemometer measured the axial wind speed of the outlet jet; Measurements were recorded at 1-s intervals for 20 s. The instrument automatically calculated the 20-s average as the outlet wind speed;

• (4) System exhaust air volume was determined by: Acquiring airflow rates from each ventilation shaft using the installed anemometers; Calculating the total volume based on shaft cross-sectional areas; Cross-verifying with the exhaust duct airflow measurements.

Average wind speed measurement: Equivalent wind speeds were first obtained for each cross-section; The average value of wind speeds from three cross-sections was calculated as the mean wind speed in the tunnel driving lane; Detailed 120-s data from the differential pressure sensor at Measurement Cross-section 1 were exported; The mean value of these data was adopted as the static pressure in the tunnel. For outlet measurements: The fluid motion at the driving lane outlet exhibited jet flow characteristics; After stabilizing the fan operation at set frequencies; A calibrated anemometer measured the axial wind speed of the outlet jet; Measurements were recorded at 1-s intervals for 20 s; The instrument automatically calculated the 20-s average as the outlet wind speed. System exhaust air volume was determined by: Acquiring airflow rates from each ventilation shaft using the installed anemometers; Calculating the total volume based on shaft cross-sectional areas; Cross-verifying with the exhaust duct airflow measurements.

Fig. 7

Schematic of measurement cross-sections and sensor locations.

The experimental protocol defined four distinct fan combination operating conditions:

Condition 1: Activation of both axial supply fans #1 and #2, i.e., a dual-supply zero-exhaust fan combination (abbreviated as DSZE);

Condition 2: Based on Condition 1, activation of axial exhaust fan #4, i.e., a dual-supply left-side exhaust fan combination (abbreviated as DSLE);

Condition 3: Based on Condition 1, activation of axial exhaust fan #3, i.e., a dual-supply right-side exhaust fan combination (abbreviated as DSRE);

Condition 4: Based on Condition 1, activation of axial exhaust fans #3 and #4, i.e., a dual-supply dual-exhaust fan combination (abbreviated as DSDE).

Fig. 8

Schematic of the four fan combination operating modes.

The four fan combination operating modes are illustrated in Fig. 8. Taking the DSZE case as an example, the frequency adjustment procedure was as follows: axial supply fans #1 and #2 were initially activated at 6.0 Hz and 7.8 Hz, respectively. The frequency of fan #1 was then fine-tuned to achieve a critical smoke exhaust state at the outlet. Once stabilized, all experimental parameters were recorded.

The frequency adjustment protocol was designed to maintain airflow symmetry by synchronously adjusting the frequencies of both supply fans. In each experimental run, the frequencies were incremented by approximately 1 Hz. Smoke extraction performance was documented via video recording for post-analysis. A total of 20 experimental runs were conducted, with frequency pairs ranging from (6.0 Hz, 7.8 Hz) to (25.2 Hz, 26.8 Hz). The specific settings for each fan combination and frequency condition are listed in Table 2.

Fan combination refers to the number and operational status (ON/OFF) of fans #1, #2, #3, and #4. For example, the dual-supply zero-exhaust combination denotes the operation of only axial supply fans #1 and #2, with axial exhaust fans #3 and #4 turned off.

The experimental run numbers (1–20) on the right denote, for a given fan combination, the specific frequency settings assigned to the active fans. Within the same fan combination, each run number represents a unique fan frequency combination. For instance, under the dual-supply zero-exhaust combination:

Run 1 corresponds to fan #1 at 6.0 Hz and fan #2 at 7.8 Hz;

Run 2 corresponds to fan #1 at 6.9 Hz and fan #2 at 8.8 Hz.

Table 2 Experimental ventilation fan frequency combination scheme (Hz).

Experimental phenomena

During the pre-experiment phase, the tunnel traffic lane outlet was kept closed, the smoke generator was activated, and supply fan #1 was operated at 15 Hz. The measured data showed positive and stable values. The airtightness of the traffic lane was verified through visual observation and data analysis, as shown in Fig. 9a.

When the experimental platform was sealed, the critical smoke exhaust state served as the evaluation criterion. This state is achieved when the longitudinal airflow velocity is sufficient to prevent smoke back-layering, confining the smoke to the fire-affected zone and directing it toward the overhead smoke outlet—commonly referred to as the critical velocity condition⁹.

To apply this concept to the outlet air leakage scenario, the DSZE fan combination was employed. The smoke generator in the left traffic lane was activated to simulate a tunnel fire, while both supply fans were simultaneously turned on. Following the frequency scheme in Table 2, fan frequencies were adjusted sequentially, with intervals exceeding 3 min to ensure airflow stability. Experimental phenomena were observed and recorded, along with data from all sensors.

Fig. 9

Experimental phenomena of the DSZE fan combination at the exhaust outlet. (a) Check the airtightness of the roadway, (b) Initial status, (c) Left flue gas leakage, (d) critical exhaust, (e) flue gas recirculation.

Figure 9b–e show the experimental phenomena below the exhaust outlet, using a dual-supply zero-exhaust fan combination as an example. This is because the dual-supply zero-exhaust (DSZE) configuration is the only scenario without active mechanical extraction. The movement of smoke within it is governed by the balance between mechanical air supply and tunnel resistance (with negligible thermal buoyancy in air medium). Defining the “critical smoke exhaust state” under this most unfavorable condition yields the most conservative and universally applicable criterion, providing an objective baseline for evaluating the performance enhancement of more complex fan combinations.

Figure 9b depicts the initial state without smoke release; when the frequencies of fans #1 and #2 are 12.6 Hz and 14.8 Hz, respectively, the experimental phenomenon shown in Fig. 9c occurs, with smoke overflowing to the left; when the frequencies of fans #1 and #2 are 12.8 Hz and 14.8 Hz, respectively, the experimental phenomenon is as shown in Fig. 9d, reaching the critical exhaust state; when the frequencies of fans #1 and #2 are 13.0 Hz and 14.8 Hz, respectively, the experimental phenomenon is as shown in Fig. 9e, with smoke backflow occurring. It can be concluded that to achieve the critical exhaust state, the frequency of the supply fan needs to be adjusted within a range of approximately 1.0 Hz. For example, when the critical exhaust state is achieved, the frequencies of fans #1 and #2 in Group 8 of the DSZE fan combination are 12.8 Hz and 14.8 Hz, respectively, while the frequencies of fans #1 and #2 in Group 9 are 13.9 Hz and 15.8 Hz, respectively, and for Group 10, the frequencies of fans #1 and #2 are 14.8 Hz and 16.8 Hz, respectively. Similarly, the frequency for each group must be adjusted based on the experimental phenomena, with the adjustment difference fluctuating around 1.0 Hz.

Experimental data analysis

Through the above experiments, when the experimental phenomenon in Fig. 9 reaches the critical smoke exhaust state, the combination parameters of the ventilation and smoke exhaust system that provide air volume are selected, and a qualitative study of their smoke exhaust effect is conducted.

The specific steps are as follows: Taking the DSDE system as an example, the first step is to control the smoke generator in the left lane to release smoke; the second step is to simultaneously activate four ventilation fans; third, observe the movement of the smoke flow within the traffic lane. Fan #2, operating at a frequency of 7.8 Hz, provides part of the airflow, forcing the smoke to move from the left lane to the right; fan #1, operating at a frequency of 7.4 Hz, also provides part of the airflow. It can be observed that the airflow exiting from supply fan #1 forms a parallel flow path within the tunnel traffic lane. The airflow flows to the left side of the ventilation fan and the outlet, respectively. The airflow toward the left-hand traffic lane interacts with the airflow from fan #3, causing the smoke to move from the right-hand traffic lane toward the left. The two airflows then converge and rush toward the ceiling exhaust outlet, where they are extracted through the exhaust duct by fans #3 and #4, which operate at a frequency of 6.5 Hz. Continuous observation of smoke flow yielded experimental data for each of the following configurations: dual-supply zero-exhaust, dual-supply left-exhaust, dual-supply right-exhaust, and DSDE, with 20 sets of data for each.

Based on the measured data, flow field parameter curves were plotted for the four fan combinations and different frequencies, analyzing flow field characteristics and parameter patterns. The average wind speeds across the left and right traffic lanes under each fan combination are shown in Fig. 10. Using the exhaust outlet as the symmetry axis, the tunnel traffic lanes are divided into left and right sides, with the left side being the smoke-generating side (i.e., the fire occurrence location) and the right side being the evacuation side (i.e., the pedestrian evacuation side).

As shown in Fig. 10, for the DSZE fan combination, in the first group of fan frequency categories, due to the lower frequency of the supply fan, the leakage air volume at the outlet is smaller, resulting in an average wind speed on the right-hand lane section being greater than that on the left-hand lane section; between the first and second groups of frequency categories, the average wind speeds on the left and right lanes converge; for the second and subsequent fan frequency groups, the average wind speed on the left lane remains greater than that on the right lane. As the fan frequency increases, the average wind speeds on both lanes gradually converge, meeting the general escape conditions, where the wind speed on the smoke-affected side (fire side) is greater than that on the fresh air supply side (escape side).

In the four subplots of Fig. 10, the horizontal axis represents the fan frequency combinations (Groups 1 to 20) for a specific fan configuration, while the vertical axis shows the corresponding physical parameter varying with fan frequency. For instance, the top-left subplot illustrates the relationship between average wind speed and fan frequency for the dual-supply zero-exhaust configuration; the remaining subplots follow analogously for the dual-supply left-side exhaust, dual-supply right-side exhaust, and dual-supply dual-exhaust configurations, respectively. The specific fan frequency values for each group are listed in Table 1.

DSLE fan combination: When fans #1, #2, and #4 are activated simultaneously, the frequencies of fans #1 and #2 remain unchanged, while the frequency of fan #4 starts at the first group and increases by 1.0 Hz per group. As the frequency of fan #4 gradually increases, the average wind speeds on the left and right lanes first diverge and then gradually converge, but the average wind speed on the left lane remains higher than that on the right lane.

DSRE fan combination: When fans #1, #2, and #3 are operated simultaneously, the frequencies of fans #1 and #2 remain unchanged, while the frequency of fan #3 increases by 1.0 Hz per group starting from the first group. As the frequency of fan #3 gradually increases, the average wind speeds on the left and right lanes gradually converge, but the average wind speed on the left lane remains consistently higher than that on the right lane.

DSDE fan combination: When fans #1, #2, #3, and #4 are all activated simultaneously, the frequencies of fans #1 and #2 remain constant, while the frequencies of fans #3 and #4 increase by 1.0 Hz per group starting from the first group. As the frequencies of fans #3 and #4 increase, the average wind speeds on the left and right lanes gradually converge, but the wind speed on the left lane remains consistently higher than that on the right lane, with the two nearly equal by the 20th group.

Fig. 10

Average wind speed in the roadway section of each ventilation fan combination.

Figure 11 shows the changes in static pressure on the left and right lanes of the roadway as a function of fan frequency when four multi-fan combinations are used, with the exhaust outlet as the axis of symmetry. Analysis of Fig. 11 shows that for the DSZE fan combination, in the first group of fan frequencies, due to the smaller frequency of the supply fan, the leakage volume at the outlet is smaller, resulting in higher static pressure on the right-hand lane than on the left-hand lane. For the second group and subsequent groups of fan frequencies, the static pressure on both sides of the lanes is approximately equal, and as the fan frequency increases, the static pressure on both sides of the lanes also increases.

DSLE fan combination: compared to the second group, the static pressure on the left and right lanes of the first group first decreases and then increases. Starting from the second group frequency, the static pressure on the left and right lanes continues to increase until the fifth group. Starting from the fifth group, the static pressure on the left and right lanes gradually decreases as the frequency of fan #4 increases, reaching a minimum value in the 13th group. From the 13th group onwards, the static pressure on both sides of the roadway increases as the frequency of fan #4 increases, ending at the 17th group; from the 17th group to the 20th group, the static pressure on both sides of the roadway gradually decreases as the frequency of fan #4 increases.

DSRE fan combination: from Group 1 to Group 8, the static pressure in the left and right lanes decreases as the frequency of fan #3 increases. From Group 8 to Group 11, the static pressure in the left and right lanes increases as the frequency of fan #3 increases. From Group 11 to Group 20, the static pressure in the left and right lanes decreases as the frequency of fan #3 increases. As can be seen, compared to the DSZE combination, the DSLE and DSRE fan combinations do not exhibit linear changes in static pressure on the left and right lanes as the fan frequency increases, but rather exhibit oscillations. This indicates that interference from multiple fan combinations is the primary cause of these oscillations. Whether such oscillations affect, and how they affect, the leakage air volume at the outlet will be further explored in the following sections.

DSDE fan combination: in the first seven groups of fan frequency, the static pressure of the left and right lanes occasionally oscillates, but basically shows a downward trend. From the eighth group onwards, the static pressure of the left and right lanes continues to decrease. At this point, the static pressure is mainly affected by the fan frequency rather than the fan combination.

In summary, the static pressure oscillation in the traffic lane of the DSLE fan combination and the DSRE fan combination is primarily influenced by the fan combination, specifically the position and number of the supply fans and exhaust fans. When the number of exhaust fans equals that of supply fans, the static pressure in the traffic lane is primarily influenced by the fan frequency, such as in the DSZE and DSDE fan combinations.

Fig. 11

Static pressure of each ventilation fan combination on the left and right lanes.

Figure 12 illustrates the power consumption of individual fans and total system power under different fan combinations, as group number increases (corresponding to incrementally higher fan frequencies). All power curves exhibit an upward trend with increasing group number, and the number of activated fans correlates positively with power growth—confirming that the observed behavior aligns with the previously mentioned fan affinity laws (N ∝ f³).

Under the dual-supply zero-exhaust (DSZE) combination, activating additional supply fans significantly enhances airflow delivery capacity, but power consumption increases proportionally. In the absence of exhaust fans, the system operates in a “no-load” state, resulting in relatively low efficiency.

Under the dual-supply left-side exhaust (DSLE) combination, the blue triangular curve (representing the addition of exhaust fan #4) exhibits a reduced slope between Group 10 and Group 15, suggesting that the airflow may have reached a locally stable state or that synergy between supply and exhaust fans has improved. This indicates that incorporating the left-side exhaust fan improves smoke extraction performance while inducing only a marginal increase in power consumption. The system enters an optimal operating range within the medium-frequency interval (Groups 10–15).

Under the dual-supply right-side exhaust (DSRE) combination, the green triangular curve (representing the addition of exhaust fan #3) displays a slight plateau between Groups 8 and 12, possibly due to flow redistribution or weakening of local vortices. This suggests that the right-side exhaust fan imposes slightly greater resistance on the system than its left-side counterpart, leading to marginally higher total power consumption. However, beyond Group 15, the power growth trend converges with that of the DSLE configuration, indicating system stabilization at high frequencies.

Under the dual-supply dual-exhaust (DSDE) combination, the total system power is highest among all configurations. However, whether this corresponds to a similarly elevated effective smoke exhaust efficiency requires further analysis.

In summary, during the medium-frequency range (Groups 10–15), both the DSLE and DSRE combinations consume comparable power levels and generally conform to the cubic power-frequency relationship (N ∝ f³). In emergency scenarios, however, power consumption alone should not be the sole criterion for evaluating fan combination performance; smoke exhaust efficiency must also be considered. To determine the optimal smoke exhaust efficiency and compare the relative performance of DSLE versus DSRE across different frequency ranges, further analysis of effective exhaust airflow and air leakage volume under each fan combination is required.

Fig. 12

Individual fan power and total power under each ventilation fan combination.

Air leakage rate is a relative indicator of air leakage volume, reflecting the overall severity of air leakage in the system. The air leakage rate is calculated as the ratio of air leakage volume to total air volume, where total air volume is the sum of air leakage volume and exhaust air volume. The air leakage rates for the four combinations of fans were calculated separately, and the results are shown in Fig. 13.

In Fig. 13, the left subplot illustrates the variation of air leakage with fan combination and frequency across the four combinations. The horizontal axis represents progressively increasing fan frequency; however, as each curve corresponds to a different fan configuration, the generic label “fan frequency group” is adopted. Here, “group” emphasizes the incremental escalation of operating frequency across experimental runs—rather than absolute frequency values, which differ among combinations. The same labeling convention is applied in Figs. 14 and 15.

As shown in Fig. 13, the DSZE combination results in an increase in air leakage volume as the frequency of the tunnel ventilation fan increases, while the air leakage rate remains nearly constant, fluctuating around 15%. The air leakage rate is higher than that of DSLE and DSRE, but lower than that of dual-supply dual-exhaust. The reason for this is that as the fan frequency increases, the static pressure rises sharply, leading to an increase in the pressure difference across the air leakage opening. Under constant system resistance, the excess air inevitably escapes through the path of least resistance (the air leakage opening). Additionally, except for the first two groups, the air leakage volume is significantly higher than that of the other three combinations.

Dual-supply single-exhaust combination: air leakage fluctuates slightly with the increase in the frequency of the exhaust fan, but the air leakage volumes across different groups are not significantly different. Additionally, the air leakage volume of the DSLE combination is consistently higher than that of the DSRE combination. When comparing air leakage rates, both the DSLE and DSRE combinations exhibit a significant decrease in air leakage rates as the frequency of the exhaust fan increases. When dividing the exhaust fan frequency groups into intervals, from Group 1 to Group 7, the leakage rate of the DSLE is greater than that of the DSRE; from Group 8 to Group 10, the leakage rates of the two are almost identical; from Group 11 to Group 20, the leakage rate of the DSRE is greater than that of the DSLE.

DSDE fan combination: the frequencies of fans #1 and #2 remain constant, and the input airflow is fixed. Since fan #1 is located near the air leakage opening, between Groups 1 and 10, the air leakage volume fluctuates up and down as the frequencies of fans #3 and #4 increase, influenced by the airflow Q₁ of the supply fan. Starting from the 10th group, individual oscillations occur, followed by a gradual decrease. At this point, the frequencies of fans #3 and #4 become the primary influencing factors, and the higher the frequencies of fans #3 and #4, the smaller the air leakage rate. Analysis reveals that when the frequency of the exhaust fan increases, the input airflow is larger, and the airflow Q₃ and Q₄ of fans #3 and #4 may suppress air leakage. Under the DSDE fan combination, the leakage rate changes gradually with increasing fan frequency, remaining around 47%.

The air leakage rates under four combinations of multi-fan are ranked from highest to lowest, with three scenarios: when the frequencies of fans #3 and #4 are between 6.5 Hz and 12.5 Hz, the following order applies: DSDE > DSZE > DSLE > DSRE; when the frequencies of exhaust fans #3 and #4 are between 13.5 Hz and 15.5 Hz, the sequence is: DSDE > DSZE > DSRE ≈ DSLE; when the frequencies of exhaust fans #3 and #4 are between 16.5 Hz and 25.5 Hz, the sequence is: DSDE > DSZE > DSRE > DSLE.

A comparison of the air leakage rates of DSZE and DSDE fan combinations shows that when the number of supply fans and exhaust fans is equal and their layout positions are symmetrical, although there are differences in the air leakage rates of the two systems, their trends are almost identical. As the fan frequency increases, the air leakage rate remains almost unchanged.

Fig. 13

Air leakage volume and leakage rate at tunnel exits of each ventilation fan combination.

Since the air leakage of the DSLE and DSRE fan combinations is significantly affected by the fan frequency, the reason lies in the fact that the fan frequency influences the system exhaust airflow, thereby affecting the air leakage volume and air leakage rate. Based on experimental data such as the average wind speed under each fan combination, the system exhaust airflow is calculated. A relationship between the fan frequency group and the system exhaust airflow is plotted, as shown in Fig. 14.

Analysis of Fig. 14 shows that, under the DSZE fan combination, except for the first frequency group where the experimental data is not yet stable, the system exhaust airflow of the remaining groups increases as the fan frequency increases.

Fig. 14

Exhaust air volume of each ventilation fan combination.

DSLE fan combination: the system exhaust airflow increases as the frequency of fan #4 increases, reaching a maximum value of 1.20 m³/s. Under the DSRE fan combination, the system exhaust airflow decreases as the frequency of fan #3 increases, reaching a minimum value in the 18th group, followed by an upward trend, with the maximum total system exhaust airflow reaching 1.17 m³/s. It can be concluded that when the frequencies of fans #3 and #4 are equal, the exhaust airflow of the DSLE fan combination system exhibits an opposite trend compared to that of the DSRE fan combination system.

Air leakage reduces exhaust airflow, thereby affecting ventilation and smoke exhaust efficiency. To further determine the relationship between air leakage rate, exhaust airflow, and ventilation and smoke exhaust efficiency, we quantified the relative sizes of effective smoke exhaust efficiency under four different fan combinations. Effective smoke exhaust efficiency is the ratio of smoke exhaust volume to exhaust airflow, where smoke exhaust volume refers to the average volume of smoke near the experimental smoke generator.

Fig. 15

Smoke exhaust volume and effective smoke exhaust efficiency of each ventilation fan combination.

Plot the smoke exhaust volume and effective smoke exhaust efficiency curves for different fan combinations and fan frequency groups, as shown in Fig. 15. Analysis of Fig. 15 shows that under the DSZE fan combination, the smoke exhaust volume increases with the increase in fan frequency, while the effective smoke exhaust efficiency decreases; under the DSLE fan combination, as the fan frequency increases, the smoke exhaust volume reaches its minimum value at the 16th fan frequency group, then gradually decreases after recovering, with the effective smoke exhaust efficiency changing synchronously; under the DSRE fan combination, as the fan frequency increases, the exhaust volume reaches its minimum value at the 7th fan frequency group, then gradually increases after recovering, with the effective smoke exhaust efficiency changing synchronously; under the DSDE fan combination, as the fan frequency increases, the exhaust volume gradually increases, while the effective smoke exhaust efficiency decreases.

A comparison chart of effective smoke exhaust efficiency under four different fan combinations is shown in Fig. 16. Analysis of Fig. 16 reveals that the effective smoke exhaust efficiency for both the DSLE and DSRE fan combinations is higher than that for the DSZE fan combination. From Group 1 to Group 12, the experimental data for effective smoke exhaust efficiency under the DSLE fan combination and DSRE fan combination exhibit a significant alternating pattern, demonstrating a cyclical alternation between high and low values within the experimental groups. Specifically, in Groups 1, 3, 5, 7, 9, and 11, the DSRE fan combination values are higher than the DSLE fan combination values, while Groups 2, 4, 6, 8, 10, and 12 show the opposite trend, until Group 13 where the two values converge. Subsequently, the effective smoke exhaust efficiency of the DSLE fan combination begins to decrease, while the effective smoke exhaust efficiency of the DSRE fan combination continues to increase, reaching a maximum in the 18th group, after which it shows a decreasing trend. Except for the 1st group, the effective smoke exhaust efficiency of the DSZE fan combination was higher than that of the DSDE, with both showing a relatively smooth trend.

Fig. 16

Comparison of effective smoke exhaust efficiency of each ventilation fan combination.

In general, when comparing the effective smoke exhaust efficiency under four different fan combinations, the patterns of DSZE and DSDE fan combinations are similar, both decreasing gradually with the fan frequency group, further validating the similarity of flow field patterns under symmetrical fan distribution; when the input frequency of the fans is the same, the effective smoke exhaust efficiency of the DSRE fan combination is higher than that of the DSLE fan combination. Considering the installation positions of the fans and the smoke generator, this indicates that when the frequency is between 18.5 Hz and 25.5 Hz, the effective smoke exhaust efficiency of the exhaust fan on the side of the parallel flow path is higher. Additionally, for every 1 Hz increase in the frequency of the exhaust fan, the effective smoke exhaust efficiency increases by approximately 3–5%.

Conclusion

This study is based on a 1:10 scale experimental model of the tunnel section between Wawuzhuang Station and Guizhou Road Station on Qingdao Metro Line 1. By measuring wind speed, static pressure, fan power and outlet airflow under various multi-fan combinations and frequency groups, the flow field characteristics and effective smoke exhaust efficiency (η_e) were systematically analyzed, with a focus on outlet air leakage and system performance. The main conclusions are as follows:

• (1) Flow field analysis reveals that, in the dual-supply zero-exhaust (DSZE) configuration, supply fan #1 generates a parallel bypass flow path near the tunnel exit, causing significant air leakage and reduced smoke exhaust efficiency. Introducing a single exhaust fan (DSLE or DSRE) establishes a more symmetrical static pressure distribution; however, flow field oscillations arise due to fan interaction, with the DSRE configuration exhibiting lower air leakage than DSLE. The dual-supply dual-exhaust (DSDE) configuration synergistically suppresses leakage by enhancing airflow capture, thereby significantly reducing fresh air escape through the exit.

• (2) Blindly increasing the number or operating frequency of fans does not guarantee improved ventilation performance. Under identical supply and exhaust frequencies, the DSDE configuration consumes the highest total power, while DSLE and DSRE show nearly equivalent power demands. Although the DSZE configuration has the lowest power consumption, its efficiency is poor due to a no-load operational state. Notably, the dual-supply single-exhaust configurations (DSLE/DSRE) achieve higher effective smoke exhaust efficiency than DSDE while consuming less energy, demonstrating superior energy efficiency. Rational coordination of fan combinations and frequency settings—particularly to avoid high-frequency competing airflow—effectively optimizes cross-sectional velocity distribution and enhances overall system performance.

• (3) The effective smoke exhaust efficiency of DSLE and DSRE is comparable when the exhaust fan frequency ranges from 6.5 to 18.5 Hz. However, in the high-frequency range (18.5–25.5 Hz), the DSRE configuration shows a marked improvement: each 1 Hz increase in exhaust fan frequency yields a 3–5% rise in η_e. This trend arises because increasing the frequency of the exit-side exhaust fan strengthens local negative pressure, which reduces the proportion of fresh air bypassing the traffic lane and leaking through the tunnel exit. Consequently, a greater fraction of the supplied airflow contributes to smoke transport, thereby improving effective smoke exhaust efficiency. Therefore, in similar ventilation systems requiring high fan operating frequencies, the dual-supply right-side exhaust fan combination is recommended—specifically, the axial supply fans positioned on both the smoke-source side and the opposite side, with the axial exhaust fan located on the side opposite the smoke source. This configuration achieves a low air leakage rate, high effective smoke exhaust efficiency, and relatively low power consumption.

• (4) Although the 1:10 scale tunnel model employed herein is validated based on similarity principles, this study has certain limitations. It may not fully reproduce all turbulent and thermal effects present in full-scale fire scenarios. Furthermore, the experiments assume steady-state airflow and do not account for transient fire development or buoyancy-driven smoke transport. To enhance both scholarly rigor and the scientific basis of emergency ventilation design, future work will focus on integrating thermal and fire-source modeling—through coupled computational fluid dynamics (CFD) simulations and medium-scale fire tests—to investigate key phenomena such as smoke stratification and critical velocity under air leakage conditions.