Chapter 3

Experimental Studies with Transit Buses

Methods

Generation and dispersion of COVID-like test aerosol

Three aerosol generators were designed, manufactured, and evaluated (Figure 1). The continuous generator produces a continuous stream of NaCl particles. The size of the particles is determined by the NaCl concentration in water. The cough generator produces “puffs” of particles at a frequency and duration set by the user. The third generator is not shown. It produces a high concentration of particles using an ultrasonic technique.

Acquisition and Preparation of the Test Bus

The team acquired two buses for this project. Bus A, in Figure 2 is an in-house bus which was borrowed from CE-CERT’s previous research on a hybrid bus project. Bus A was used to conduct baseline testing. Bus A was modified to conduct Task 2 specifically for the parallel air ventilation system. Bus B, in Figure 3 was borrowed from the LA Metro for this project. This bus was used to conduct a portion of Task 2. It was especially used for a test with passengers on-board. Bus B was used for on-road testing.

Table 1 and Table 2 show specifications of both buses.

Table 1: Bus A Test Vehicle Specifications.

Table 2: Bus B Test Vehicle Specifications.

The bus HVAC system includes filters that impact the movement of airborne particle concentrations inside the cabin and thus virus transmission. The goal is to reduce airborne particle concentrations (i.e., airborne COVID-like aerosol concentrations) via multiple approaches. The experimental approaches used in this study are divided into two groups: Retrofit and redesign as shown by the flowchart in Figure 4. The

Effect of Door Opening

To determine the effect of door opening on air exchange rates and particle concentrations, five different scenarios were tested. The first scenario type includes a baseline test with both the rear and front doors in a closed configuration. The second scenario tests the effect from both doors being in an open configuration to simulate passenger loading and unloading. To determine the effect from each door independently, the next two scenarios test the front door open while the rear door is maintained in a closed configuration and the rear door open while the front door is maintained closed. Lastly, to replicate the effect of air flow while the bus is driving throughout city routes, a Hartzell Series 23 utility air circulating fan was faced towards the bottom of the front door while both doors were closed (Figure 7). As there is a door gap with the floor and air is viscous, the wind may pull cabin air out of the bus cabin increasing air exchange rate.

To compare air changes per hour (ACH) a.k.a. AER (Air Exchange Rate) between each scenario, CO₂ canisters were released throughout each bus cabin while ten TSI AirAssure sensors measured CO₂ gas decay over time. The rate was calculated by fitting the measured CO₂ decay to Equation 1 (Jung et al. 2017). Coefficients A, B, and C were determined by fitting to the measured CO₂ levels reported from each sensor. Variable A represents the CO₂ concentration at t = 0, B is the background CO₂ concentration in ppm, and C is represented as the inverse of the time constant in Equation 2 (based from Jung et al. 2017). C is equal to the AER represented in hr^-1_. The Matlab2023 Curve Fitter Toolbox application was used to perform a non-linear regression fit to determine each constant. The constants’ constraining bounds are described in Table 3. The constant A was limited to values between 0 to 3000 ppm to represent an initial concentration value for each test. Constant B had a lower bound of 300 ppm based on sensor sensitivity from estimated background concentrations of 420 ppm up to 500 ppm. Constant C was allowed to vary without bounds to obtain the best fit line. The best fit was determined using the selected baseline sensor #4 (refer to Appendix A: Laboratory Tests) and used the Least Absolute Residual method. A fit with the lowest sum of squared estimate of errors and highest adjusted r² value was selected for the evaluation of the remaining sensors.

C_cabin(t) = (A-B)exp(-Ct)+B _{(Equation 1)}

$C=\frac{1}{\tau }*60$ _{(Equation 2)}

Table 3: Coefficient Constraints of the non-linear CO₂ regression analysis using MatLab2023.

Effect of Cabin Air Filtration Improvements

Each bus HVAC system was equipped with a pre-used MERV 13 filter across the inlet. To evaluate the effectiveness of implementing cabin air filtration improvements a baseline test was conducted to compare results using the existing filter with no modifications. In this phase of testing, the aerosol generator was powered on to increase PM concentration within the cabin and then powered off to allow for the PM concentration to decay through filtration and particle deposition loss to the walls. The aerosol generator was placed in the front, middle, and back seats of the cabin to test varying source locations inside the cabin. Particle count and size distributions were measured using a TSI 3080 electrostatic classifier and 3776 CPC. An eACH (Equivalent ACH) value is calculated by fitting the exponential PM decay over time after the aerosol generator is powered off. The time constant represented by Equation 2 represents the particle removal rate and is used to compare between tests. A new electrostatically charged ThermoKing MERV 13 filter (46 in x 20.6 in) was installed to compare to the in-use pre-used MERV 13 filter and review the efficiency of PM removal.

Effect of On-board Air Cleaners

Prior to replacement of the pre-existing MERV 13 filter equipped in each bus HVAC system, the effect of on-board standalone air cleaners was tested. The modifications included the use of two Honeywell HEPA air purifiers (model HPA300) with a flow output of approximately 320 CFM air flow on the “turbo” setting, in tandem with the bus AC system after the aerosol generator is powered off. This method affects the rate of PM decay which can be compared to the baseline scenario. One of the HEPA air purifiers was placed near sensors #1-4 located at the front section of the bus cabin, while the second HEPA air purifier was placed near sensors #5-8 located at the back of the cabin. The aerosol generator was placed in the front, middle, and back of the cabin to also test varying source locations within the cabin.

Effective of Ventilation Design Changes

A standard transit bus has a ventilation system which operates by supplying air into the bus cabin through two vents which run along the corners of the ceiling of the bus’s cabin, supplying air throughout most of the bus’s length. As seen in Figure 8, air inside the cabin is first removed through an extraction fan in the rear of the bus, causing the need for air to move across the entire length of the bus prior to being removed. If someone in the front seat coughs, then the virus aerosol can be transmitted to passengers in the back seat easily following this type of ventilation air flow.

The proposed parallel ventilation system, depicted in Figure 9, combines the standard upper ventilation system with a lower system which replaces the function of the extraction fan located at the rear to remove air from the bus. Figure 10 shows this lower ventilation system which consists of 17 vents that branch into each row of seats removing air throughout the length of the bus. Each branch is capped with a register box.

A dual blower system, consisting of two three-phase 5HP motors and PB15A Cincinnati fans, creates a closed loop system in which air is being cycled from the outlet of each blower to the upper ventilation system while air is being removed through the lower ventilation system which is led into the blower inlets. A variable frequency drive is what controls the motor and fan, enabling flow rate adjustments in the lower

Determine Effect of “Clean Air Shower System”

This task is functionally the same as the task described above (2.2.5). An air shower technique in stationary test used only ventilation slots from roof vents located directly above floor suction vents. Specifically, the roof ventilation contains multiple air slots of 50.51 mm, width of 9.87 mm, and straight length of 44.62 mm (Figure 14) throughout the length of the vent duct. The slots that were not directly above the floor suction vents were sealed.

On-road Testing

The use of on-road testing provides realistic driving conditions experienced by bus operators and a traveling public. On-road testing incorporates the identical sensor and aerosol generator placement set up used in the stationary testing. A baseline test was performed without the use of the aerosol generator to determine backline conditions. A map of the on-road test pathway is shown in Figure 15 and measures approximately 1.11 miles in perimeter and is composed of 35 mph to 45 mph roads to mimic the driving pattern of a public transit bus. Driving patterns for each test include 2-3 continuous laps around the test pathway, showing repeatable driving patterns throughout the speed profiles (Figure 16). Ambient conditions during two testing days included an average temperature of 81.65 °F and 77.89°F and maximum wind speeds of 14 mph and 17 mph.

Test Matrix

Table 4 shows the test matrix for ACH/AER testing using CO₂ canisters. Decay rates of CO₂ concentrations were used to determine ACH at different test conditions. All windows and roof latches were maintained in a closed position for all tests in this study. CO₂ canisters were released throughout the bus cabin and the concentration was allowed to decay within each test. The opening and closing of the bus doors were used to test different air exchange conditions including opening both doors, opening only one door, and both doors closed. In addition, a Hartzell Series 23 utility air circulating fan was used to simulate driving conditions with air flow outside the cabin.

Table 5 shows the test matrix of the eACH testing using a particle generator during stationary testing with the regular bus HVAC system. The eACH value for particles removed were determined from the decay rates of particle concentrations at multiple locations of the bus cabin for different test conditions. The aerosol generator was placed at three locations: at the front of the bus in the seat facing perpendicular to the center aisle, middle of the bus near sensors 5 & 6, and center of the back row seat against the back wall of the cabin, see Figure 17. The generator was powered to allow the particle mass concentration to reach an approximate plateau before turning off allowing for the concentration to decay. A variety of experimental conditions used include installing plexiglass barriers to create partitions inside the cabin, and the use of standalone HEPA air purifiers.

Table 6 shows the test matrix of eACH testing during stationary testing with the parallel ventilation system. The aerosol generator is also placed at the same three locations throughout the bus, and experimental conditions include the use of barriers, and an air shower method described later.

Table 7 shows the test matrix during stationary testing using a new MERV 13 filter replacement to determine particle removal and filter efficiency.

Table 8 provides a summary of the test matrix used for on-road testing. The stop-and-go experimental condition signifies minimal traffic roads containing stop signs and traffic signals.

The aerosol generator was powered on until the particle mass concentration stabilized to a plateau, replicating stationary testing. The bus AC system was set to high and in a stationary position. CO₂ canisters

The filter efficiency curve provides insight into the fraction of entering particles that are retained by the filter. Therefore, filter efficiency was determined using Figure 18 by calculating the difference between measured particle number counts when the aerosol generator was on, and the particle number counts when the aerosol generator was off. Values were then divided by the aerosol generator particle on count. Figure 19 shows the filter efficiency throughout the measured particle sizes for all test conditions equipped with a MERV 13 filter. The minimum efficiency near 10% occurs at a particle size of .07 µm (MPPD, most penetrating particle diameter) for Bus B bus. The MPPD is very close to the smallest COVID virus size of 100nm.

Higher efficiencies are observed with Bus A parallel and new filter testing. Higher efficiency is also observed for Bus A regular system, however, similar efficiencies of 40-80% particle removal match that of Bus B for particles larger than 0.2 µm. It was assumed the higher filtration efficiency could be caused either by the difference of the in-use filters or by the larger ACH of the Bus A. Larger ACH brings in outside air which contains lower background aerosol concentrations leading to dilution effect. The high efficiency of the new MERV 13 filter was attributed to the electrostatic effects. The in-use filters may have lost the electrostatic effects long time ago. It is encouraging that the parallel system also showed high efficiency. It

Air Change per Hour: Bus B

To calculate ACH, CO₂ gas was released within the bus cabin while the AC system was turned on and the AC fans set to high. The temperature within the cabin was maintained near a range of 15.5 to 17 °C.

Air Change per Hour: Bus A

Non-linear regression fit models were also computed for Bus A tests #10-13. Figure 28 shows the exponential decay of CO₂ measured by sensor #4 when both doors are closed. ACH rate when in this test with the AC system set to high has an average value of 1.64 ± 0.10 hr^-1. This is higher when compared to Bus B by a value of 1 meaning the volume of air inside the cabin will be replaced one additional time within the same hour. This was attributed to the better door seals in Bus B reducing the space for air to enter/escape.

Figure 29 shows decay curves of a repeated doors closed scenario followed by both doors opening. The repeated test showed a higher average ACH rate of 2.54± 0.21 hr^-1, however, this is attributed to outside air influence as the test conditions remained the same. Average wind speeds on this day were 8 to 10 mph. Average ACH rates are also expected to be significantly larger when both doors are open as outside air is cycled throughout the cabin at 36.48 ± 6.86 hr^-1 with both doors open.

Test #12 and #13 in Figure 30 and Figure 31, respectively, incorporate the use of a Dyno blower aimed at the floor of the front door to simulate driving conditions. This is followed by turning the blower off and opening either the front or back door. ACH rates are comparable between repeated tests with simulated driving at 3.15 ± 0.36 to 3.78 ± 0.2 hr^-1. Opening one of the doors does not differ greatly between the front and rear door and showed values at 12.81 ± 0.55 hr^-1 and 12.23 ± 0.82 hr^-1, respectively. This is a larger ACH value for the same case for Bus B bus. This was attributed to a shorter body length of Bus A facilitating better air exchange when a door is open.

Table 10 gives a summary of ACH rates determined by each sensor for tests discussed #10-13. In Shinohara et al. (2022), they tested a combination of opening windows rather than doors. However, the overall trend showed an increase in ACH as number of windows opened increased.

Air Change per Hour: On-Road Results

Non-linear regression fit models were computed for the on-road tests 2-4 and 6. Figures 32-35 show the exponential decays of CO₂ measured by TSI AirAssure sensor #4 with the fitted curve. Test #2 had the aerosol generator at the front of the bus prior to driving, while for test #3 the generator was placed at the center of the bus cabin on the seat next to sensor 5. For each test, the CO₂ concentration profile along with the bus driving speed profile is shown in Figure 32 and Figure 33, respectively. The average ACH rates are similar between the tests with 7.51 ± 0.52 hr^-1 for test #2 and 8.46 ± 0.79 hr^-1 for test #3 as expected due to similar set up.

Speed profiles show the bus cycled between 5 to 30 mph. In Shinohara et al. (2022), the test bus was driven at speeds of 10, 20, and 30 kph, which is equivalent to 6, 12, and 18 mph, respectively. Their results showed ACH values of 1.9 hr^-1 for the lowest speed and up to 2.4 hr^-1 for the highest speed. These values are lower than those reported in this study. There can be multiple reasons such as driving patterns, wind conditions, and bus dimensions/design for that, but overall Shinohara et al.’s (2022) results suggest their test bus was better air sealed that the test bus of this study. The higher ACH results in this study is also likely due to the difference in air pressure between the inside of the bus cabin and the outside that assists ventilation rate.

When comparing between on-road driving tests and the simulated driving using dyno fans, the ACH rates are about twice larger compared to the blower test results. This means the size of the fan did not replicate the pressure experienced by realistic driving conditions. The use of driver fans was also incorporated in test #4 (Figure 34), however, the average ACH did not vary significantly to tests without these fans. It is assumed the dynamic pressure difference allows outside air to penetrate the bus cabin regardless of the driver fan setting at this typical city driving speed.

Figure 35 shows the CO₂ decay and driving speed profile for the stop-and-go test. In this study, the stop- and-go means that the bus doors were open for approximately 20 seconds at the end of each lap to simulate air exchange conditions during passenger loading and unloading with both doors open. This is similar to Shinohara et al. (2022) as their stop-and-go tests also opened the doors for 20 seconds, however, it was after moving 500m (0.3 miles) and driving 30kph (48mph). The ACH rate is 27.07 ± 0.86 hr^-1 which is approximately three times more than that reported by Shinohara et al. (2022). They also tested stop-and-go

Particle Arrival Time

The aerosol generator was placed at different locations inside each bus cabin to determine the PM arrival time per sensor. The PM arrival time is defined as the time in minutes after the aerosol generator is powered on at which the PM concentration is measured at a value that is equal to ½ of the maximum concentration. The PM concentration plateaus after 10 to 15 minutes of the aerosol generator being on, a 5-minute average window within the plateau is used to calculate the maximum concentration per sensor (Figure 35). The ½ max PM concentration is the rise value in µg/m³ which is expected to change depending on the proximity between the generator and each sensor location. The elapsed time until the PM concentration meets or exceeds the rise value measurement is identified as the PM arrival time. This was the most reasonable choice because of the time resolution of the TSI AirAssure sensors. The sensors have a time resolution of 1 min as a default but was later updated to a time resolution of 10 second measurement. This calculation provides a measure of when the aerosol reaches each sensor throughout the cabin. To compare maximum concentrations between sensors and within each test a relative max value was obtained. This is calculated using the upper and lower quartile of each sensors’ recorded maximum and minimum PM concentration during baseline testing with the aerosol generator at the front, middle, and back. Bus B bus data used a relative min of 0 µg·m^-3 and relative max of 1165 µg·m^-3.

Particle Arrival: Stationary Bus

Table 12 is a summary of particle arrival times inside Bus B cabin for tests #1-#3. In baseline test #1, the AC system is off, and the aerosol generator is located at the front of the bus. Sensor #1 has the first arrival time at 2 minutes as it is the closest to the generator location. Remaining sensors have arrival times between 6-8 minutes as expected due to their increased distance away from the aerosol generator. Comparing results to baseline test #2, this time with the AC system on, sensors #2-10 have much faster arrival times of 4 and 5 minutes likely due to air flowing towards the AC inlet at the back of the cabin. In test #3, the generator is moved to the middle of the cabin next to sensor #5. Sensor #6 recorded a 1.1 relative max concentration and had the fastest arrival time showing that the flow of particles reaches across the center aisle. This behavior is expected as air is cycled towards the back panel and through the filtration medium before it reenters the cabin from the top vents. Lastly, the longest particle arrival time of 6 minutes is measured by sensor #1.

Particle Arrival: Barriers Using Conventional and Parallel Ventilation

Table 13 provides the summary of particle arrival times inside Bus A cabin for tests #4 -9 to compare the effect of barriers to baseline results. The two buses cannot be used as a direct comparison as their cabin volumes and lengths are different, therefore, a new relative minimum (1 µg·m^-3) and maximum (581 µg·m^-3) were calculated. With the aerosol generator placed at the front of the cabin, there is no major difference between the use of barriers in test #4 and the baseline test #9 without barriers. Particle arrival times were approximately 2 to 3 minutes throughout each sensor for both tests. This is likely due to the aerosol cloud immediately reaching the center of the aisle from the orientation of the aerosol generator (Figure 17). The baseline test #8 with the aerosol generator in the middle had a 1-minute arrival time for the sensor directly behind it. The nearby sensors at the middle and back of the cabin had faster arrival times by one minute compared to the sensors towards the front.

When barriers are present in Bus A, a faster speed of dispersion is seen for all sensors with 2-3 minute arrival times compared to 2-5 minutes without barriers. On the other hand, lower rise value concentrations are observed with barriers suggesting some particles were removed through deposition. When the aerosol generator is placed at the back of the cabin, arrival times are approximately ~3 minutes and rise value

Stationary Testing with Passengers

A stationary test using ten passengers was conducted with Bus B. The AC system was powered on to high and set to 60-63 °F inside the cabin. Sensors #1-#10 recorded CO₂ concentration levels throughout the testing. The bus doors were open prior to passenger loading during installation of sensors. Once all passengers boarded the bus, the doors were closed. Sitting location was random and selected individually by the passenger (Figure 43). The CO₂ concentrations, seen in Figure 44, shows a steady increase from an average of 668 ppm to 1291 ppm during the first 25 minutes. Sensors #3 and sensor #7 recorded higher CO₂ levels than other sensors during this period likely due to those nearby passengers continuously speaking more compared to passengers near sensors #5 and #6. At minute 25 the driver fans and blower on the left and front control panel respectively were set to high. This caused a decrease in overall CO₂ concentrations throughout the cabin. At minute 33, one passenger was unloaded from the bus where both doors were opened and closed within one minute. The CO₂ concentration is observed to decrease over a 3-minute period from an average of 1004 ppm to 790 ppm followed by an increase again to an average value of 832 ppm after 16 minutes. Both doors were opened at minute 53 and passengers exited the bus. This activity caused a rapid decrease in CO₂ levels to 561 ppm. Lastly, the bus engine and AC system was turned off and the CO₂ concentration depleted to a background level of 464 ppm while the doors remained closed.

Bus cabin air quality is a balance among internal sources of emissions, filtration, and dilution. This parking lot test with passengers in this section demonstrates the importance of bringing outside air into the cabin by fan. As shown previously when a vehicle is in motion the dynamic pressure difference causes outside air to penetrate into the bus cabin. This helps suppress the increase of CO₂ exhaled by passengers. In the real world, especially on the busy arterial roads, outside air can contain high concentrations of pollutants (CO₂, CO, VOC, particulates, and NOx). In that case, it is required that the ventilation inlet of the bus has a filtration system to remove the outside pollutants before getting into bus cabin because otherwise passengers could be exposed to fresh pollutants coming from outside.