Figure 1 illustrates a schematic of the brake wear particulate matter test setup. The brake pad wear test is conducted using the TS102B Brake Inertia Test Stand to simulate multiple braking modes of a vehicle. The Brake Bench can accurately control the brake speed and torque, and accurately monitor the brake disk temperature so that each test condition can be accurately controlled. In order to reduce the influence of particles in the ambient air on the experiment, a sealed Plexiglas chamber (500 mm × 600 mm × 1,000 mm, W × D × H) was added to the brake (including pads, discs, and calipers) with a high-efficiency particulate air (HEPA) filter at the air inlet, following the method of Iijima et al. (2008). The pump (with constant flow controller) is connected to the sealed chamber, and the air flow is filtered by HEPA into the sealed chamber, and the particles emitted from brake wear (Filtered air with dust) are carried into the pipeline (Tunnel), and at the same time, the monitoring and sampling instruments connected to the pipeline are used to test and collect the particles in the air flow. The monitoring instrument uses the Electrical low-pressure impactor (ELPI+) manufactured by Dekati, Finland, which is based on the principle of inertial impact separation and real-time measurement of airborne particulate matter particle size spectral distribution and mass concentration. The ELPI + impinger classifies 6 nm–10 μm particles into 14 stages, the first 13 stages measure 17 nm–10 μm particles, and the last stage is a filtering section to collect 6 nm–17 nm particles. According to the study of Hagino et al. (2015), when the air flow rate of sampling tunnel is between 0.5 and 5 m³/min, the flow rate does not have a significant effect on the emission rate of particulate matter emitted from brake wear, so in this study, the air flow rate of the sampling tunnel is set to 2.8 m³/min.

Since different brands of automobiles are equipped with different types of brake pads, and different brake pads contain large differences in the components, and different types of brake pads yield slightly different particle size distribution characteristics, this study selected five small buses (including domestic brands and imported brands) with a large domestic ownership in the middle and low price range, and purchased five sets of original matching right front brake assemblies for testing and sampling. The brand affiliation, brake pad type, brake pad appearance and brake pad area of each sample are shown in Table 1.

Brake wear test conditions are determined by combining the initial braking speed (V, km/h), deceleration speed (D, m/s²), and the initial braking temperature (T, °C) (the temperature of the brake disc at the beginning of braking). The common speed limits are 40 km/h and 60 km/h for urban roads, 80 km/h for expressways, and 120 km/h for highways. In order to have a comprehensive understanding of the particulate emissions from braking under various working conditions, five initial braking speeds (V = 40, 60, 80, 100, and 120 km/h), five deceleration speeds (D = 1, 2, 3, 5 and 7 m/s²), and five initial braking temperatures (T = 100, 150, 200, 250°C and 300°C) were selected for experiments. Each experiment involves starting from a certain speed and decelerating to zero, resulting in a total of 29 working conditions, and the test was repeated four times for each working condition. Detailed information of the 29 conditions is shown in Table 2. For each new braking system or set of brake pads, at least 200 braking break-in tests need to be conducted before performing the braking wear particulate emission testing experiment. This is to remove the relatively rough surface friction material and reduce its impact on the experimental results. The experimental conditions are set as follows: initial braking temperature of 25°C, initial speed of 40 km/h, final speed of 0, and deceleration of 1 m/s². And the test conditions were only started when the brake disk temperature had risen to the target temperature. This means that the speed is adjusted to 40 km/h, braking is started (the brake fluid applies pressure to the piston and presses the brake pads against the brake disk, generating a friction torque) until the speed is reduced to 0, and the second test is carried out when the temperature of the brake disk again reaches 100°C. In order to reduce experimental errors, each condition was repeated four times, and the whole simulation test process was controlled by computer.

In order to prevent the resuspension of particles remaining on the brake discs and pads during each braking process from interfering with the ELPI + test data, we defined the amount of particulate matter emitted during the entire braking process, i.e., the amount of particulate matter emitted from a single braking (PE, mg/braking/wheel). The PE is the cumulative amount of the entire emission process, i.e., the cumulative amount during the time period between the beginning of the rise of the particle concentration in the compartment (t₀) and the return of particle concentration to the background concentration level (t₁), which was calculated by the following formula: P E = V ′ × S × ∫ t 0 t 1 C t d t V' is the duct wind speed (6 m/s), S is the cross-sectional area of the duct (0.01131 m²), and C_t is the corresponding concentration at each moment (μg/m³).

To improve the visualization effect, we performed a meticulous cluster analysis on the evolving patterns of particulate matter across 14 distinct particle size categories, utilizing component linkage and cosine similarity metrics. And we used a dendrogram (Figure 3) to show the clustering process, which was completed by IBM SPSS Statistics. The horizontal axis of the dendrogram represents the relative distance between clusters, while the vertical axis represents distinct test conditions. Notably, test conditions within the same cluster exhibit similar patterns in particulate matter mass size distribution trends. The 25 working conditions of Pad.1 were divided into three categories, namely,: ➀15–25-2-6-1–20–13–14-7-8–24–23-4–5; ➁9–10-3; ➂ 16–21-11–17-12–18-22–19.25 working conditions of Pad.2 were divided into two categories: ➀1 15–17-18–19-11–14-12–13-2-1–16–20–22–23–24-3-4-5–21; ➁6-7-9–10–8.29 working conditions in Pad.3 were divided into two categories: ➀12-3–26-7-8-6-4–27-5-9–10–12–14–15–11–13; ➁28–29-21–23-16–20-25–17-24–22-19–18. The 29 working conditions of Pad.4 are divided into three categories: ➀1-2-3-4-7–26-8-5–28–29-6–27–13–14–12; ➁9–10-11; ➂24–25-23–21-22–20-16–17-18–19-15.29 working conditions of Pad.5 are divided into three categories: ➀1-3-2-4-5-6-9–10-7-8–26–13–14–12–15–11; ➁16–17-18–19-20–24-25–21-22–23; ➂28–29-27.

The particle size distribution of each sample mass is shown in Figure 4. From the figure, it can be seen that although the systematic clustering classifies the particle size distribution of abraded particles into two to three categories, when the particle size is less than 0.50 μm, the distribution trend is very similar among the samples, and all of them are remain basically unchanged with the increase of the particle size. When the particle size is larger than 0.50 μm, there are some differences in the emission trends. For all samples and conditions, the emissions start to increase significantly at a particle size of about 0.75 μm. In the range of 0.01–8.11 μm, for Pad.1, there were two peaks in the distribution of category ➀ and ➁, which occurred between 2 and 3 μm and >8.11 μm, respectively, while there was only one peak in the distribution of category ➂, which occurred between 2 and 3 μm. Sample Pad.2 has a peak emission value in the particle size distribution of category ➀ at around 5 μm, and the peak concentration of category ➁ is at >8.11 μm. Sample Pad.3 has a peak emission value in the particle size distribution of categories ➀ and ➁ at around 3 μm, but the emission of category ➁ does not decrease significantly at particle sizes of 3–8.11 μm and maintains a high level of emission. Pad.4 samples have only one peak particle size for all conditions, but the peak size fluctuates, the peak size is around 5 μm in the distribution of category ➀ and ➁, and between three and 5 μm in the distribution of category ➂. in the distribution of particle size in category ➀, ➁, and ➂of sample Pad.5, there is only one peak particle size for all conditions, and the peak size is around 5 μm in the distribution of categories ➀ and ➁, and the peak size is > 8.11 μm in the distribution of category ➂. From the above analysis, it can be seen that the mass of particles emitted from all five samples is dominated by particles with larger particle sizes (Figure 5). Particles with particle size less than 0.5 μm accounted for 0.1%–0.7% of the total particulate mass, with an average of 0.38%; particles with particle size between 0.5 and 3.0 μm accounted for 33.11%–60.1% of the total particulate mass, with an average of 49.1%; particles with particle size 3.0–8.11 μm accounted for 39.5%–66.8% of the total particulate mass, with an average of 49.7%; particles with particle size 0.5–8.11 μm accumulated more than 99% of the total particles. This particle size distribution characteristic allows brake wear particles to persist in the atmosphere for longer periods of time and exacerbate their health risks. This is consistent with previous research conclusions, such as Sanders et al. (2003)tested the PM₁₀ emissions of brake pad wear under typical urban driving conditions for low-metal, semi-metal, and NAO brake pads and found that the mass-weighted average diameter was 5–6 μm; Iijima et al. (2008) tested the mass-weighted average diameter of NAO brake pads to be 3–6 μm. Kukutschova et al. (2011)tested low-metallic brake pads and found that the mass exhibited a unimodal distribution with a peak concentration appearing at 2–4 μm; von Uexküll et al. (2005) tested the front and rear brake pads of trucks and found that the peak concentration appeared at 2–3 μm. Harrison et al. (2012) collected air particles of different sizes and used the particle size distribution of specific marker elements to estimate the contribution of brake wear to particle mass. They found that the PM₁₀ mass of brake wear exhibited a unimodal distribution with a peak located at 2–3 μm.

Garg et al. (2000) found that PM_2.5 accounted for 63% and PM₁₀ 86% of total particulate matter emitted by wear by testing different brake pads, and 33% of the particulate matter had a particle size of less than 0.1 μm; Sanders et al. (2003) had similar results, with PM₁₀ accounting for 63%–85% of the total particulate matter; Iijima et al. (2008) found 56%–70% PM_2.5 and 95%–98% PM₁₀. EMEP/CORINAIR emission inventory guidelines state that brake wear 39% PM_2.5 and 98% PM₁₀ (Thorpe and Harrison, 2008), which is more similar to the results of this study. On the other hand, brake wear was also found to be a higher percentage of PM₁₀ than PM_2.5 in the source resolution of particulate matter performed by receptor modeling (Abu-Allaban et al., 2003).

The number size distribution of brake wear emission particles is also very important because a large portion of the particulate matter produced by brake wear is distributed in the smaller size range. Figure 6 shows the number size distribution of brake wear particles for some conditions, and the number size distribution trends are similar. Within the measurement range of the ELPI+, it can be seen from Figure 6 that for each sample and condition, the maximum number of particles was found at a particle size of 0.01 μm, the number of particles showed a decreasing trend in the range of 0.01–0.2 μm, and the number of particles showed an increasing trend in the range of 0.2–1.0 μm, with another peak at about 1 μm. In this study, the number of brake wear particles was bimodally distributed, one in the nucleation mode (Dp < 10 nm) and the other in the accumulation mode (1 μm).

From the above analysis, it can be seen that the number of brake wear particles is mainly dominated by ultrafine particles (Dp < 0.1 μm) (Figure 7). Among them, the amount of nucleated modal particles (Dp < 0.05 μm) accounted for 46.52%–73.50% of the total number of PM₁₀, accounting for 60.11% on average; the amount of particles with a particle size of 10 nm–0.1 μm accounted for 12.21%–39.79%, averaging 22.04%; the amount of particles with a particle size of 0.1–1.0 μm accounted for 1.23%–29.58%, averaging 14.04%; and the amount of particles with a particle size of 1.0–8.11 μm accounted for 14.04%, averaging 14.04%. The number of particles with particle size of 1.0–8.11 μm accounted for 0.45%–9.27% and 3.81% on average. The number of ultrafine particles (Dp < 0.1 μm) accounted for 61.1%–98.3% of the total PN₁₀(Number of particles with a diameter less than 10 μm), and the average accounted for 82.15%.

Most of the studies found that BWPs have at least one peak size in the ultrafine fraction, and tests conducted by Garg et al. (2000) found that the largest number of brake wear particles was less than 30 nm. A road simulation study conducted by Mathissen et al. (2011) found that the number of brake wear particles produced by a vehicle decelerating from 100 km/h to 0 was bimodally distributed with peaks at 10 nm and 30–50 nm, respectively, and when decelerating from 30 km/h to 0, the number was unimodally distributed with peaks at 70–90 nm, compared to 100 km/h. Kukutschova et al. (2011) found that although <500 nm particles were negligible at low disk temperatures, as the disk temperature increased (up to 340°C), particles smaller than 100 nm were generated. They suggested that submicron particles may be produced by evaporation and condensation of initially produced nanoparticles rather than abrasive wear. Wahlstrom et al. (2010)also pointed out that there is a peak in the particle number distribution at approximately 100 nm, and they found that when testing the front brakes (compared to the rear brakes), the PN peak particle size distribution shifted towards smaller particle sizes. Riediker et al. (2004) tested six different passenger car brake lining materials under controlled environmental conditions and found concentration peaks at 80 nm (depending on the vehicle and braking behavior) and 200–400 nm. Compared to normal deceleration, braking from initial speed to a complete stop resulted in the generation of more nanoparticles. However, in some studies, the number of BWPs peaked at larger particle sizes. Wahlstrom et al. (2010) tested the brake emission processes of low-metal and non-asbestos organic linings and found a bimodal distribution with peaks of about 280–350 nm, which may be caused by the limitation of the instrumental measurement particle size range (>0.30 μm). Mosleh et al. (2004) tested a semi-metallic brake lining for commercial trucks and found a bimodal number concentration distribution with one peak at about 350 nm and another larger peak at about 2.0 μm. Sanders et al. (2003) conducted wear tests on three different types of brake lining materials using both a wear test bench and actual road tests, and found that the peak particle size was approximately 1.0 μm. Iijima et al. (2007) used an APS spectrometer (measuring particles >0.5 μm) to study the number concentration distribution of wear particles from non-asbestos organic brake linings and found a peak at 1.0 μm. They also observed that as the temperature increased, the peak distribution gradually shifted towards larger particle sizes. Overall, the number of particulate matter for brake wear fine particle modes into a bimodal distribution.