A Model of Pyrolysis Carbon Black and Waste Chicken Feather Using a Response Surface Method in Hot-Mix Asphalt Mixtures

Gan, Youwei; Li, Chuangmin; Chen, Anqi; Li, Yuanyuan; Wu, Shaopeng

doi:10.1061/(ASCE)MT.1943-5533.0004424

Open access

Technical Papers

Aug 18, 2022

A Model of Pyrolysis Carbon Black and Waste Chicken Feather Using a Response Surface Method in Hot-Mix Asphalt Mixtures

Authors: Youwei Gan [email protected], Chuangmin Li [email protected], Anqi Chen [email protected], Yuanyuan Li [email protected], and Shaopeng Wu [email protected]Author Affiliations

Publication: Journal of Materials in Civil Engineering

Volume 34, Issue 11

https://doi.org/10.1061/(ASCE)MT.1943-5533.0004424

PDF

Abstract

Research on waste tire pyrolysis carbon black (PCB) in bitumen mixtures indicates that it has excellent high-temperature rutting resistance; however, the degradation of low-temperature crack resistance and water stability of PCB-modified asphalt limits its wide application. Therefore, the low-temperature crack resistance and water stability of PCB-modified hot-mix asphalt (HMA) were enhanced by addition of waste chicken feather (WCF). Based on the response surface methodology (RSM), a variance analysis (ANOVA) and

P - value

test of the model response coefficient were carried out, and the performance prediction models of percentage air voids (VV) in bituminous mixtures, 30-min Marshall stability (MS), and 48-h MS were obtained. The interaction effects of different independent variables on response variables were analyzed. In addition, the optimal solutions of multiresponse variables obtained by the response optimizer were 0.20% PCB content, 0.36% WCF content, shearing time of 5.78 min for WCF, and 6.66% asphalt–aggregate ratio. The corresponding performance index prediction values were 4.00% VV, 12.80 kN 30-min MS, and 12.65 kN 48-h MS. The road performance test results demonstrated that incorporating WCF improved the high-temperature rutting resistance, low-temperature crack resistance, and water stability of PCB-modified hot-mix asphalt. This study provides a precise performance prediction model for adding PCB and WCF to HMA.

Introduction

Rutting occurs easily in areas with high temperatures and rain during summer. Therefore, modifying the bitumen or hot asphalt mixture (HMA) is necessary for improving the rutting resistance and water stability. On the one hand, previous antirutting modifiers have achieved good results, albeit with high cost, which does not promote the use of low-cost roads. On the other hand, many agricultural, industrial, and domestic wastes lack a wide range of applications, thus occupying storage space and causing environmental pollution. Nowadays, waste materials increasingly are used in asphalt pavement construction as new additives, which makes asphalt pavement composed of waste materials a popular research topic in the field of road construction (Lv et al. 2021; Mahssin et al. 2021; Niu et al. 2021). In addition, the traditional design method of HMA uses different levels of a single factor to observe the influence of variables at different levels; this requires many experiments and consumes time and resources (Yıldırım and Karacasu 2019). Therefore, it is necessary to use scientific experimental design methods to establish a performance prediction model for a HMA that uses waste materials to obtain the optimal solution of resources, costs, and performance in practice.

More than 800 million waste tires are abandoned each year globally, with an annual increase of 4%. Approximately 13 million scrap tires are produced annually in China, with an annual growth rate of 6%–8% (Li et al. 2020). The comprehensive utilization of waste tires in China includes renovation (Xu et al. 2020), reclaimed rubber (Lin et al. 2020), rubber powder (Liu et al. 2020), and pyrolysis. Compared with the other three disposal methods (Xu et al. 2020; Li et al. 2020; Lin et al. 2020), pyrolysis can degrade the organic components in rubber powder and generate volatile substances and solid coke. The main components of volatile substances are combustible gas and combustible oil, which can be used for energy supply. Thus, pyrolysis is environmentally friendly, generates no secondary pollution, and allows multiple resource recycling (Li et al. 2019; Tian et al. 2021). However, limited by the current waste tire pyrolysis technology in China, the quality of pyrolytic carbon black (PCB) produced is poor, resulting in fewer commercial PCB applications. As early as 1995, researchers used PCB to prepare modified bitumen and discovered that PCBs in bitumen had rheological properties similar to those of mineral powder (Lesueur et al. 1995). Since then, researchers have found that PCBs significantly improve the high-temperature performance and UV aging resistance of bitumen (Feng et al. 2016). In addition, the adhesion of bitumen is enhanced owing to the high free energy of the PCB surface (Tanzadeh and Shafabakhsh 2020). However, PCBs weaken the interface force between the bitumen and aggregate, which leads to a decrease in the low-temperature crack resistance (Feng et al. 2021). In addition, PCBs are not soluble in bitumen and organic solvents, which reduces the storage stability of modified bitumen, resulting in the difficulty of long-distance transportation of PCB-modified bitumen (Feng et al. 2021; Li et al. 2020).

Chicken is the most-consumed poultry in the world, constituting approximately 65 million tons/year and producing 5 million tons of waste chicken feathers (WCFs) (Mu et al. 2020). Usually, WCFs in China are converted into compost or low-characteristic animal nutrients through waste disposal stations (Casadesús et al. 2018). However, due to animal disease concerns, Chinese regulations restrict the amount of WCF in animal feed, resulting in WCF piles in slaughterhouses and environmental pollution. Researchers have studied the thermal, mechanical, and electrical properties of WCFs and found that they can be used in electrical insulation materials, geotextiles, and building materials (Tesfaye et al. 2017).

The most recommended application is to replace building materials with WCFs (Araya-Letelier et al. 2020; Šafarič et al. 2020). WCF has good heat and sound insulation when used in a composite fiberboard (Šafarič et al. 2020). The compressive and tensile strengths of cement concrete reinforced with WCFs are improved (Wahab and Osmi 2012). The study of WCF-modified bitumen shows that the addition of 2% WCF to base bitumen gives rheological properties similar to those of styrene-butadiene-styrene (SBS)-modified bitumen and has better storage stability (Rivera-Armenta et al. 2020). Furthermore, the research on WCF-enhanced HMA indicates that the addition of WCF can improve the rutting resistance and water stability of HMA (Dalhat et al. 2020). However, the effect of the WCF fiber preparation process on the performance of HMA remains unclear. Studies have shown that the length and diameter of steel fibers significantly influence the low-temperature crack resistance of HMA (Park et al. 2015).

Response surface methodology (RSM) is a combination of mathematical and statistical methods which is used to model and analyze the response, affected by multiple variables, to optimize a reaction (Montgomery 2017). Second-order equations typically are used to establish models between responses and factors. Although this model is approximate, it can be used to analyze the response value of a given variable and determine the significance of the factors and interactions between factors to predict the maximum or minimum response under certain interactions of multiple factors (Bradley 2007; Haghshenas et al. 2015; Khuri and Mukhopadhyay 2010). In the research and development tests of bitumen and HMA-modified materials, the typical independent variables of RSM are the proportion of additives, temperature, key sieve pass rate, and bitumen dosage. The dependent variable usually is the performance test results for bitumen or HMA (Bala et al. 2020; Wang et al. 2018). The Box–Behnken method (BBD) in RSM requires fewer test iterations and has good applicability to complex experiments (Hill and Hunter 1966). Many researchers have used the BBD method to study the performance prediction models of bitumen and HMA (Liao et al. 2021; Lv et al. 2020; Zhang et al. 2016). The central composite design (CCD) method has better sequentiality and prediction ability than the BBD method, but it requires more test iterations (Adnan et al. 2020; Hamzah et al. 2013; Rafiq et al. 2021; Zolgharnein et al. 2013).

In summary, two of the main obstacles to the large-scale application of PCBs in the road construction field are the decrease in the low-temperature crack resistance and decreased storage stability of PCB-modified bitumen. If the PCB is mixed with bitumen and aggregate using a dry-mixing method, a decrease in storage stability is avoided (Yucel et al. 2021); however, this method is not free of drawbacks. PCB is a type of submicron particle. The dry-mixing method struggles to disperse agglomerate PCB materials (Naseri 2021), making it difficult to enhance the performance of bitumen effectively. Moreover, PCBs filling HMA voids can easily flow out with rainwater, resulting in a decrease in water stability (Guan et al. 2021). WCF has 2.5 times more fibers than wool and silk (Mu et al. 2020). This characteristic is conducive to solving the problems of dispersion and water stability decline of dry mixing and improving the low-temperature crack resistance of HMA. However, different varieties of WCFs have different physical properties. Even when a given WCF undergoes different treatment processes, its performance varies. Therefore, it is necessary to apply the response surface method (RSM) in the mix design procedure to investigate the influence of PCB and WCF on the performance of HMA. This study examined the effects of PCB dosage, WCF dosage, WCF shear time, and asphalt–aggregate ratio on the performance of PCB- and WCF-reinforced HMA using the CCD method in Minitab version 19.1 software to provide a performance prediction model.

Materials and Experimental Methods

Materials

Aggregate and Bitumen

The aggregates were 0–3, 3–5, 5–10, and 10–15-mm limestone gravel and mineral powder, the physical properties of which are listed in Tables 1 and 2. Bitumen was 70# petroleum-based bitumen. Table 3 presents the bitumen technical performance test results. The standard citations in Tables 1–3 are from the United States [ASTM D1139 (ASTM, n.d.-a); ASTM C131 (ASTM, n.d.-k); ASTM C127 (ASTM, n.d.-c); ASTM D4791 (ASTM, n.d.-h); ASTM C128 (ASTM, n.d.-d); ASTM C1252 (ASTM, n.d.-p); ASTM D2419 (ASTM, n.d.-l); ASTM C1777 (ASTM, n.d.-j); ASTM B417 (ASTM, n.d.-b); ASTM D4318 (ASTM, n.d.-o); ASTM D5 (ASTM, n.d.-i); ASTM D36 (ASTM, n.d.-m); ASTM D113 (ASTM, n.d.-e); ASTM D92 (ASTM, n.d.-g); ASTM D70 (ASTM, n.d.-n); ASTM D2872 (ASTM, n.d.-f)], Europe [TS EN 933-9 (European Standard, n.d.)] and China [Li and Li 2011], respectively.

Table 1. Physical properties of coarse aggregates

Property	Specification limit	10–15 mm	5–10 mm	3–5 mm	Specification
Crushed stone value (%)	$\leq 26$	19.6	—	—	T0316 (ASTM D1139)
Los Angeles abrasion loss (%)	$\leq 28$	15.8	23.9	25.2	T0317 (ASTM C131)
Apparent specific density	$\geq 2.60$	2.888	2.882	2.714	T0304 (ASTM C127)
Bulk specific gravity	—	2.836	2.793	2.670	T0304 (ASTM C127)
Water absorption (%)	$\leq 2.0$	0.63	1.10	0.60	T0304 (ASTM C127)
Flat and elongated particle (%)	$\leq 15$	10.4	4.6	—	T0312 (ASTM D4791)

Table 2. Physical properties of fine aggregates

Property	Specification limit	0–3 mm	Specification
Apparent specific density	$\geq 2.50$	2.724	T0328 (ASTM C128)
Bulk specific gravity	—	2.552	T0330 (ASTM C128)
Angularity (s)	$\geq 30$	40.3	T0345 (ASTM C1252)
Sand equivalent (%)	$\geq 60$	84	T0334 (ASTM D2419)
Methylene blue value ( $g / kg$ )	$\leq 25$	1.0	T0349 (ASTM C1777)
Filler apparent density ( $t / m^{3}$ )	$\geq 2.50$	2.682	T0352 (ASTM B417)
Hydrophilic coefficient	$< 1$	0.7	T0103 (TS EN 933-9)
Plasticity index (%)	$< 4$	3	T0354 (ASTM D4318)

Table 3. Physical properties of binders

Experiment	Value	Specification
Penetration @ 25°C (0.1 mm)	68	T0604 (ASTM D5)
Softening point (°C)	48	T0606 (ASTM D36)
Ductility @ 10°C (cm)	37	T0605 (ASTM D113)
Ductility @ 15°C (cm)	$> 100$	T0605 (ASTM D113)
Flash point (°C)	294	T0611 (ASTM D92)
Specific gravity (25°C)	1.048	T0603 (ASTM D70)
Rolling thin-film oven test (RTFOT) mass loss (%)	$- 0.076$	T0610 (ASTM D2872)
RTFOT residual penetration ratio @ 25°C (%)	69.1	T0610 (ASTM D5)
RTFOT residual ductility @ 10°C (cm)	7	T0610 (ASTM D113)

PCB and WCF

The PCB was purchased from Qingdao Yikesida Technology (Qindao, Shandong, China); the physical properties of the PCB are listed in Table 4. The Changsha Poultry Slaughterhouse (Changsha, Hunan, China) provided the WCF; the physical properties of the WCF are listed in Table 5. The collected WCF was immersed in water containing detergent and stirred in a blade mixer for 2 h. Then the mixture was dried in an oven at 105°C to constant weight. The dried WCF was placed in a 39,000-rpm blade grinder and crushed for 5, 10, and 15 min. The preparation process of the WCFs with different shear times is shown in Fig. 1.

Table 4. Properties of PCB

Property	Value
Ash content (%)	14
Iodine absorption ( $mg / g$ )	81
DPB absorption ( $mL / 100 g$ )	76
Moisture content (%)	2
pH	8

Table 5. Properties of WCF

Property	Value
Specific gravity	0.493
Tensile strength (MPa)	264
Young’s modulus (GPa)	3.84

Fig. 1. Material handling process. (Images by Youwei Gan.)

Experimental Methods

Microscopic Characterization of PCB and WCF

A JSM-6010PLUS (Japan Electronics Co., Ltd., Tokyo) scanning electron microscope equipped with an energy scattering X-ray spectrometer was used. It ran at 15 kV in low vacuum mode. The average diameter of PCB, the average diameter of WCF, and the average length of WCF were analyzed using Image-Pro Plus 6.0 image analysis software.

Preparation of Asphalt Mixture Specimen

The HMA was Marshall design method AC-13 grade (Table 6). First, aggregate, WCF, and PCB were added to the blender and stirred for 45 s. The dosages of PCB and WCF were a percentage of aggregate. Bitumen then was added and stirred for 90 s. Finally, the mineral powder was added and stirred for 45 s. The HMA was removed from the mixer and placed in an oven at 165°C for 2 h to simulate the short-term aging of HMA during construction. The optimum asphalt content determined by experiments was 4.9%, the air voids (VV) was 5.821%, the Marshall stability (MS) was 9.31 kN, and the flow was 3.84 mm. Marshall and rutting specimens were formed by the compaction and wheel-grinding methods, respectively.

Table 6. Gradation curve of dense graded asphalt mixture (AC-13)

Sieve size (mm)	Upper gradation	Lower gradation	Median gradation	Synthetic gradation
16	100	100	100	100
13.2	100	90	95	94.5
9.5	85	68	76.5	74.5
4.75	68	38	53	48.5
2.36	50	24	37	33
1.18	38	15	26.5	23.5
0.6	28	10	19	15
0.3	20	7	13.5	11
0.15	15	5	10	8.5
0.075	8	4	6	6

Marshall Stability Test and Improvement Test

Marshall specimens were immersed in a thermostatic water tank at 60°C for 30 min and 48 h. After being removed, the specimens were placed into the Marshall test instrument at

50 mm / \min

, and their Marshall stability values were recorded. Owing to the different immersion times in the test process, 48-h MS values were greater than the 30-min MS values (Gao et al. 2020; Putman and Amirkhanian 2004). Subsequently, this study was aimed at improving the test method to solve the problem of water stability performance exceeding 100%. Marshall specimens were divided into sealed and unsealed samples. They were soaked in a 60°C thermostatic water tank for 48 h at a loading speed of

50 mm / \min

.

Road Performance Test

High-temperature rutting tests [T0719-2011 (Li and Li 2011)] conducted on asphalt mixture slabs with dimensions of

300 \times 300 \times 50 mm

by applying a loaded rubber tire back and forth. The loading rate was

42 times / \min

and the loading duration was 60 min. The test temperature was 60°C (Zhu et al. 2020).

Low-temperature bending tests [T0728-2000 (Li and Li 2011)] were conducted by applying loading in the center of beam specimens with dimensions of

250 \times 30 \times 35 mm

. The testing temperature was

- 10 ° C

and the loading rate was

50 mm / \min

(Zhu et al. 2020).

Freeze-thaw splitting tests [T0729-2000 (Li and Li 2011)] were carried out by applying a splitting load to Marshall specimens that were compacted 50 times on both sides. The freezing and thawing cycle temperature was

- 18 ° C

, and the loading rate was

50 mm / \min

.

Logic Map of Experimental Design

The experimental design was divided into four stages: statistical design of the experiment, model optimization according to the response coefficient of the

P - value

, testing the model applicability, and evaluating the road performance of HMA. The CCD method in Minitab software was used to determine the optimal preparation process of HMA corresponding to air voids, 30-min MS, and 48-h MS. The specific experimental plan is shown in Fig. 2. The CCD method contains an embedded factorial or fractional matrix with center points, axial points, and factorial points (Mohammed et al. 2018; Sarlak et al. 2012). The axial points of this study were on a square surface (Fig. 3). The test matrix was designed with four independent variables, including six center points, eight axial points, and 16 factorial points. Table 7 lists the independent variables and various levels of the experimental design.

Fig. 3. Central composite face-centered design.

Table 7. Variables and their levels for central composite face-centered design

Factors	Name	Low ( $- 1$ )	Center (0)	High ( $+ 1$ )
$A$	PCB dosage (%)	0.2	0.6	1
$B$	WCF dosage (%)	0.15	0.30	0.45
$C$	WCF shearing time (min)	5	10	15
$D$	Asphalt–aggregate ratio (%)	4.50	5.75	7.00

Results and Discussions

PCB and WCF Characterization

Micromorphology and Particle-Size Analysis of PCB

The morphology and particle size distribution of the PCBs are shown in Fig. 4. Scanning electron microscopy (SEM) showed that the average diameter of a single particle PCB in the sample was 15.18 μm. In addition, many tiny particles of PCB formed aggregates owing to van der Waals forces, with an average diameter of 19.90 μm. Energy-dispersive X-ray spectroscopy (EDS) showed that the main chemical elements in PCBs were carbon (C) and oxygen (O). In addition, silicon (Si), sulfur (S), and zinc (Zn) were the main components of PCB surface ash (Chaala et al. 1996; González-González et al. 2020). Parameters

D_{10}

,

D_{50}

, and

D_{90}

in dynamic light scattering (DLS) represent the particle sizes when the percentages of cumulative particle size distributions reach 10%, 50%, and 90%, respectively (Li et al. 2018). For PCB,

D_{10} = 181 nm

, and

D_{90} = 364 nm

, indicating that the particle size distribution of PCB was concentrated in the range 181–364 nm, which is a typical submicron particle [Fig. 4(c)]. The difference between the measured values of DLS and SEM was attributed to the addition of dispersant to the sample before the DLS test and ultrasonic dispersion; thus, the PCB aggregates were dispersed into a single small particle in the solvent (Pugh et al. 1983).

Fig. 4. PCB performance characterization: (a) PCB; (b) SEM and EDS of PCB at $500 \times$ ; and (c) size distribution of PCB.

Micromorphology Analysis of WCF

The WCF after washing and drying was nearly yellow (Fig. 5). The barbules on both sides of the WCF rachis were fan-shaped, and formed wool-like fibers. The morphology and size are shown in Figs. 5(b and c), respectively. EDS showed that the main elements of the rachis were carbon (C), nitrogen (N), oxygen (O), and sulfur (S). The calcium (Ca) found in barbules was the residual metal pollution after slaughterhouse pretreatment, and the proportion was only 0.84%. After the WCF was crushed, the barbules were crushed entirely and interspersed between the feather branch and broken rachis (Fig. 6). Moreover, the rachis and feather branches gradually were shortened with an increase in shear time. With the increase in shear time, the average diameter of feather branches did not change significantly. The average length decreased from 1,861.8 to 674.1 μm, which means that the length-to-diameter ratio of feather branches decreased by 63.8%. Studies have shown that the fiber length-to-diameter ratio has a significant influence on the mechanical properties, low-temperature crack resistance, and water stability of fiber-reinforced asphalt mixtures (Phan et al. 2021; Xu et al. 2010).

Fig. 5. WCF performance characterization: (a) WCF after drying; (b) SEM and EDS of WCF at $50 \times$ ; and (c) SEM and EDS of WCF at $300 \times$ .

Fig. 6. SEM of WCF at (a) $20 \times$ after shear for 5 min; (b) $500 \times$ after shear for 5 min; (c) $20 \times$ after shear for 10 min; (d) $500 \times$ after shear for 10 min; (e) $20 \times$ after shear for 15 min; and (f) $500 \times$ after shear for 15 min.

Optimization and Effect Analysis of RSM

Model Fitting

Table 8 lists the response variables for the combination of 30 factors in the CCD experiment. The optimization of the model response coefficient was carried out via ANOVA (Table 9). The significance analysis first-order, second-order, and interaction terms in the model were evaluated using a

P - value

test (

P < 0.05

was deemed significant). The first-order term was the essential component of the model hierarchy. Therefore, the terms that were not significant in the

P - value

test were discarded. The

B D

term (Table 7) of the 48-h MS value in the interaction term, although not significant in the

P - value

test, was retained because of the decrease in the corresponding model prediction after removing the term. The actual and predicted

R^{2}

values of VV, 30-min MS, and

48 - h

MS are shown in Fig. 7. The optimized model coefficients are presented in Eqs. (1)–(3). Letters

A

,

B

,

C

, and

D

in Table 9 and Eqs. (1)–(3) correspond to the factors listed in Table 7

VV = 21.05 + 2.947 A - 5.81 B + 0.1810 C - 3.742 D - 2.490 A^{2} + 0.1729 D^{2} + 0.716 B D - 0.025 C D

(1)

30 - \min M S = - 5.84 + 0.043 A + 31.97 B + 0.2805 C + 4.928 D - 46.52 B^{2} - 0.4362 D^{2} - 0.05060 C D

(2)

48 - h M S = - 10.18 + 1.287 A + 30.88 B + 0.832 C + 5.68 D - 30.81 B^{2} - 0.02573 C^{2} - 0.492 D^{2} - 5.27 A B - 0.733 B D - 0.0622 C D

(3)

Table 8. Response variable values for CCD method

Run	Independent variable values				Response variable values
Run	PCB dosage (%)	WCF dosage (%)	WCF shearing time (min)	Asphalt–aggregate ratio (%)	VV (%)	30-min MS (kN)	48-h MS (kN)
1	0.2	0.15	5	4.50	8.223	11.56	11.36
2	1.0	0.15	5	4.50	8.279	11.42	11.40
3	0.2	0.45	5	4.50	7.386	12.89	13.23
4	1.0	0.45	5	4.50	7.091	12.71	12.80
5	0.2	0.15	15	4.50	8.768	11.98	11.46
6	1.0	0.15	15	4.50	8.571	12.32	11.85
7	0.2	0.45	15	4.50	8.064	13.46	13.78
8	1.0	0.45	15	4.50	8.271	13.10	13.15
9	0.2	0.15	5	7.00	3.615	10.46	9.99
10	1.0	0.15	5	7.00	3.750	10.74	10.66
11	0.2	0.45	5	7.00	3.540	11.92	12.12
12	1.0	0.45	5	7.00	3.415	12.06	11.15
13	0.2	0.15	15	7.00	3.666	10.14	8.96
14	1.0	0.15	15	7.00	3.827	10.07	9.58
15	0.2	0.45	15	7.00	3.576	10.87	10.96
16	1.0	0.45	15	7.00	3.446	11.32	9.65
17	0.6	0.30	10	5.75	6.255	13.14	13.36
18	0.6	0.30	10	5.75	5.878	13.51	13.48
19	0.6	0.30	10	5.75	5.911	13.69	13.56
20	0.6	0.30	10	5.75	5.927	13.46	13.89
21	0.2	0.30	10	5.75	5.781	13.62	13.75
22	1.0	0.30	10	5.75	5.674	13.47	13.25
23	0.6	0.15	10	5.75	6.346	11.58	11.45
24	0.6	0.45	10	5.75	5.672	12.89	13.42
25	0.6	0.30	5	5.75	5.855	13.42	12.64
26	0.6	0.30	15	5.75	6.321	12.98	12.33
27	0.6	0.30	10	4.50	8.609	13.36	13.56
28	0.6	0.30	10	7.00	4.183	11.84	11.16
29	0.6	0.30	10	5.75	5.571	13.65	13.66
30	0.6	0.30	10	5.75	5.873	13.32	13.89

Table 9. ANOVA for response variables

Source	DF	Adj SS	Adj MS	$F$ -value	$P$ -value
VV (%)	8	93.0024	11.6253	370.59	0
$A$	1	0.0048	0.0048	0.15	0.699
$B$	1	1.1674	1.1674	37.21	0
$C$	1	0.6257	0.6257	19.95	0
$D$	1	89.9766	89.9766	2868.29	0
$A^{2}$	1	0.5466	0.5466	17.42	0
$D^{2}$	1	0.2514	0.2514	8.01	0.01
$B D$	1	0.2884	0.2884	9.19	0.006
$C D$	1	0.3906	0.3906	12.45	0.002
Lack of fit (model error)	17	0.5202	0.0306	$0.88$	$0.625$
Pure error (replicate error)	4	0.1386	0.0346	—	—
30-min MS (kN)	7	36.9129	5.27327	130.05	0
$A$	1	$5.30 \times 10^{- 3}$	$5.34 \times 10^{- 3}$	0.13	0.720
$B$	1	6.6613	6.66125	164.28	0
$C$	1	0.0491	0.04909	1.21	0.283
$D$	1	9.9458	9.9458	245.28	0
$B^{2}$	1	3.773	3.77304	93.05	0
$D^{2}$	1	1.6003	1.60028	39.47	0
$C D$	1	1.6002	1.60023	39.46	0
Lack of fit (model error)	18	$6.80 \times 10^{- 1}$	$3.78 \times 10^{- 2}$	$0.71$	$0.727$
Pure error (replicate error)	4	$2.12 \times 10^{- 1}$	$5.30 \times 10^{- 2}$	—	—
$48 - h$ MS (kN)	10	60.2298	6.023	65.8	0
$A$	1	0.2497	0.2497	2.73	0.115
$B$	1	10.2001	10.2001	111.44	0
$C$	1	0.732	0.732	8	0.011
$D$	1	18.7272	18.7272	204.6	0
$B^{2}$	1	1.364	1.364	14.9	0.001
$C^{2}$	1	1.1743	1.1743	12.83	0.002
$D^{2}$	1	1.6751	1.6751	18.3	0
$A B$	1	1.6002	1.6002	17.48	0.001
$B D$	1	0.3025	0.3025	3.3	0.085
$C D$	1	2.418	2.418	26.42	0
Lack of fit (model error)	15	1.558	0.1039	$2.29$	$0.219$
Pure error (replicate error)	4	0.1811	0.0453	—	—

Note: DF = degrees of freedom; Adj SS = sum of squared deviations; and Adj MS = mean square; F-value = ratio of two mean squares [effect term/error term]; and P-value = significance level calculated from actual statistics.

Fig. 7. Extent of agreement between actual and predicted results for: (a) VV; (b) 30-min MS; and (c) $48 - h$ MS.

Analysis of Main Effect and Interaction Effect

The main effect data are shown in Table 10. The nonlinear change terms in the main effects plot (Fig. 8) correspond to the second-order terms with significant P-values in Table 9. The VV value decreased by 54.8% when the asphalt–aggregate ratio increased from 4.5% to 7%. The value decreased by 17.1% when the WCF dosage increased from 0.15% to 0.45%. The value increased by 6.9% when the WCF shear time increased from 5 to 15 min. The effect of PCB dosage on the VV was not significant. The data changed because when the volume of the Marshall specimen is fixed, the opening gap of the aggregate and the gap between aggregates remain constant, which leads to a decrease in the VV value with an increase in the asphalt–aggregate ratio and WCF dosage.

Table 10. Main effect data

Independent variable	Level	VV (%)	30-min MS (kN)	$48 - h$ MS (kN)
PCB content (%)	0.2	5.85	11.88	11.73
	0.6	6.03	13.07	12.89
	1	5.79	11.91	11.55
WCF content (%)	0.15	6.12	11.14	10.72
	0.3	5.98	13.29	13.13
	0.45	5.07	12.36	11.90
WCF shearing time (min)	5	5.66	11.91	11.71
	10	5.97	13.13	13.06
	15	6.06	11.80	11.35
Asphalt–aggregate ratio (%)	4.5	8.12	12.53	12.51
	5.75	5.92	13.23	13.17
	7	3.67	11.05	10.41

Fig. 8. Main effect plots: (a) VV; (b) 30-min MS; and (c) $48 - h$ MS.

When the WCF dosage increased from 0.15% to 0.30%, the 30-min MS value increased by 19.3%. Compared with 0.30% and 0.15% WCF dosage, the 30-min MS values corresponding to 0.45% WCF dosage decreased by 7.0% and increased by 10.9%, respectively. When the asphalt-aggregate ratio increased from 4.5% to 5.75%, the 30-min MS value increased by 5.5%. With an increase in the asphalt–aggregate ratio to 7.0%, the 30-min MS value decreased by 16.5% compared with that for the 5.75% aggregate ratio, and was less than that for the 4.5% aggregate ratio. With an increase in the PCB dosage and WCF shear time, the 30-min MS value first increased and then decreased, but the change was not significant. The data change in Fig. 8(b) is attributed to the existence of the optimum bitumen film thickness of the aggregate. When the asphalt–aggregate ratio decreased and WCF dosage increased, bitumen could not form a sufficiently thick bitumen film on the surface of the aggregate. In contrast, excessive bitumen creates free bitumen in the voids between the aggregates. These two situations lead to a decrease in the 30-min MS.

The 48 h MS had a similar trend to that of 30-min MS, but the increase in the asphalt–aggregate ratio led to a more significant change in the 48-h MS value. With an asphalt–aggregate ratio of 7%, the 48 h MS values decreased by 16.8% and 20.9% compared with those for 4.5% and 5.75% asphalt–aggregate ratios, respectively. This was due to the combined effects of high temperature, moisture, and time. The free bitumen flow between aggregates was sufficient, which decreased the original interlocking structure strength of aggregates, resulting in a decrease in the 48-h MS value.

Fig. 9 shows the influence of the two-factor interaction effect on the response variable VV. At all levels of WCF dosage and WCF shearing time, the VV value decreased linearly with increasing asphalt–aggregate ratio. As the amount of WCF increased, the decreasing effect of VV value caused by the increase of asphalt–aggregate ratio decreased [Fig. 9(a)]. As the asphalt–aggregate ratio increased, the decreasing effect of WCF content on VV value weakened. As the WCF shear time increased, the decrease in VV value caused by the increase of the asphalt–aggregate ratio was strengthened [Fig. 9(b)]. As the asphalt–aggregate ratio increased, the effect of increasing the VV value caused by the increase of WCF shear time weakened.

Fig. 9. Response surface plots representing VV interaction: (a) $B D$ ; and (b) $C D$ .

Fig. 10. Response surface plot representing 30-min MS interaction.

As shown in Fig. 10, With the increase of WCF shearing time, the decrease of 30-min MS value caused by the increase of asphalt–aggregate ratio was enhanced. The decrease in 30-min MS value caused by the increase of WCF shear time was enhanced with the increase of the asphalt–aggregate ratio.

Fig. 11 shows the effect of a two-factor interaction on the response variable for 48-h MS. As the amount of PCB increased, the enhancement effect of increasing the amount of WCF on the 48 h MS value weakened [Fig. 11(a)]. With the increase of WCF content, the increase of PCB content resulted in a stronger decrease in 48-h MS value. With the increase of WCF content, the decrease of 48 h MS value due to the increase of asphalt–aggregate ratio was enhanced [Fig. 11(b)]. With the increase of asphalt–aggregate ratio, the enhancement effect of 48-h MS value caused by the increase of WCF content weakened. With the increase of WCF shearing time, the decreasing effect of increasing the asphalt–aggregate ratio on the 48-h MS value was enhanced [Fig. 11(c)]. With the increase of the asphalt–aggregate ratio, the increase of the WCF shear time resulted in an enhanced decrease in 48-h MS value.

Fig. 11. Response surface plots representing $48 - h$ MS interaction: (a) $A B$ ; (b) $B D$ ; and (c) $C D$ .

Multiresponse Variable Solution and Verification

The optimization of the model parameters results in an accurate prediction of each response variable model. However, the optimal solution of a single response variable in the actual use process cannot meet the complex use requirements of roads. In addition, changing the value of independent variables in a single response variable will affect the other response variable values, which makes it costly in terms of resources and time to obtain the optimal solution of multiple response variables via traditional test methods. In this study, the optimization target of VV was set to 3%–5% of the definite characteristics, and the optimization targets of 30-min MS and 48-h MS were set to the minimum value of 8 kN, and the larger the better. The RSM response optimizer obtained the optimal solution of the multiple response variables (Table 11). The predicted values, confidence intervals, and validation values of the multiresponse variables are listed in Table 12. The verification of the optimal solution proved the reliability of the model in this study within a 95% confidence interval. In addition, compared with the virgin HMA, in the optimal solution the asphalt content increased by 35.9%, the Marshall stability increased by 39.1%, and the flow decreased by 12.3%.

Table 11. Optimal solution of multiresponse variable model

Solution	PCB dosage (%)	WCF dosage (%)	WCF shearing time (min)	Asphalt–aggregate ratio (%)	VV (%)	30-min MS (kN)	$48 - h$ MS (kN)	Composite desirability
1	0.20	0.36	5.78	6.66	4.00	12.80	12.65	0.87

Table 12. Multiresponse variable prediction and verification

Response variable	Fit	SE fit	95% CI	95% PI	Verification value
VV (%)	4.00	0.09	(3.81, 4.19)	(3.59, 4.41)	3.93
30-min MS (kN)	12.80	0.10	(12.60, 13.00)	(12.33, 13.26)	12.96
$48 - h$ MS (kN)	12.65	0.16	(12.32, 12.99)	(11.94, 13.37)	12.65

Note: SE fit = standard error of fit; CI = confidence interval; and PI = prediction interval.

Road Performance

High-Temperature Performance of HMA

The rutting test of HMA was used to evaluate the high-temperature stability of pavement, which mainly reflects the ability to resist shear stress under dynamic vehicle loads in summer. The dynamic stability of the optimal solution was the largest, that is, 2.2 times that of the matrix HMA (Fig. 12). The dynamic stability of HMA with only PCB was similar to that of HMA with only WCF; both were approximately twice that of matrix HMA. Previous studies have shown that PCB and WCF can increase the high-temperature rutting resistance of HMA. The high-temperature performance was improved further after the two are incorporated into the HMA, indicating that the optimization provided by the RSM model optimal solution is significant (Dalhat et al. 2020; Li et al. 2018).

Low-Temperature Performance of HMA

A low-temperature bending test was used to evaluate the cracking resistance of HMA at low temperatures in winter. The flexural strength of HMA with only WCF was the highest (Fig. 13). This is attributed to the formation of a cross-linked network structure by the reinforcement of fiber in HMA, resulting in the enhancement of low-temperature crack resistance of HMA (Wang et al. 2021). The bending strength of HMA with only PCB was the lowest because it is difficult to disperse the agglomerated PCB effectively in a short stirring time, resulting in the presence of uneven stress interfaces in HMA. As the temperature decreases, the HMA cracks at these interfaces, resulting in cracks in the pavement structure.

Water Stability of HMA

The residual stability value represents the influence of water on the compressive HMA strength in the immersion state, which usually is less than 100%. The 48-h MS value was greater than the 30-min MS value (Table 7). The improved water stability test results are shown in Fig. 14; these results were used to study further the effect of water on the HMA. The results show that the residual stability of HMA with only PCB was lower than that of matrix HMA because the agglomerated PCB forms a barrier between the aggregate and bitumen, thereby reducing the adhesion between them. In addition, the residual stability of HMA with WCF was more than 100%, probably because the volume of WCF wool fiber increased after contact with moisture, which extended the time required for subsequent moisture to enter the aggregate and bitumen. Thus, the immersion process enhanced the characteristics of HMA doped with WCF.

Conclusion

This study improved the low-temperature performance decline of pyrolytic carbon black (PCB)-modified bitumen, and the water stability decrease of HMA in PCB dry-mixing. Waste chicken feather (WCF) was used as a fiber stabilizer, and the effects of PCB dosage, WCF dosage, WCF shear time, and asphalt–aggregate ratio on the performance of HMA were studied using the response surface method via Minitab. The research conclusions are as follows:

1.

Regarding VV, the interaction effect between the WCF dosage and asphalt–aggregate ratio, WCF shear time, and asphalt–aggregate ratio was significant. For 30-min MS, the interaction between the WCF shear time and asphalt–aggregate ratio was significant. Regarding

48 - h

MS, the interactions between the PCB dosage and WCF dosage, and between the WCF dosage and asphalt–aggregate ratio, WCF shear time, and asphalt–aggregate ratio were significant.

2.

The optimal mixture preparation values after model optimization were as follows: PCB

dosage = 0.20 %

, WCF

dosage = 0.36 %

, WCF shear

time = 5.78 \min

, and asphalt–aggregate

ratio = 6.66 %

. The predicted values of HMA performance using the optimal preparation process were

VV = 4.00 %

, 30-min

MS = 12.80 kN

, and 48-h

MS = 12.65 kN

.

3.

The

R^{2}

values of all the prediction models were greater than 90%, and the

P

-value test, due to a lack of fitting in the model, was greater than 0.05. The results of HMA performance verification based on the optimal preparation process were within a 95% confidence interval. This indicates that the fitting model has a good prediction ability, and the data in the model are significantly correlated with the model.

4.

The road performance test results show that PCB can improve the high temperature of HMA, but the low-temperature and water stability performance decreases. The incorporation of WCF improves the low-temperature and water stability performances of HMA constructed using the PCB dry method. The improved immersion Marshall test results show that the water stability of HMA significantly improved after the addition of WCF.

Data Availability Statement

All data, models, and code generated or used during the study appear in the published article.

Acknowledgments

The authors acknowledge the Science and Technology Project of the Department of Transportation of Jiangxi Province (Nos. 2020H0023 and 2021H0019), Changsha Science and Technology Plan Project (No. kq2004065), Hunan Postgraduate Research and Innovation Project (QL20210182) and the National Key R&D Program of China (2017YFE0111600), and the National Natural Science Foundation of China (51778515 and 71961137010).

References

Adnan, A. M., X. Luo, C. Lü, J. Wang, and Z. Huang. 2020. “Physical properties of graphene-oxide modified asphalt and performance analysis of its mixtures using response surface methodology.” Int. J. Pavement Eng. 23 (5): 1–15. https://doi.org/10.1080/10298436.2020.1804061.

Abstract

Introduction

Materials and Experimental Methods

Materials

Aggregate and Bitumen

PCB and WCF

Experimental Methods

Microscopic Characterization of PCB and WCF

Preparation of Asphalt Mixture Specimen

Marshall Stability Test and Improvement Test

Road Performance Test

Logic Map of Experimental Design

Results and Discussions

PCB and WCF Characterization

Micromorphology and Particle-Size Analysis of PCB

Micromorphology Analysis of WCF

Optimization and Effect Analysis of RSM

Model Fitting

Analysis of Main Effect and Interaction Effect

Multiresponse Variable Solution and Verification

Road Performance

High-Temperature Performance of HMA

Low-Temperature Performance of HMA

Water Stability of HMA

Conclusion

Data Availability Statement

Acknowledgments

References

Information

Published In

Copyright

History

Authors

Affiliations

Metrics

Citations

Download citation

Cited by

Figures

Other

Share

Copy the content Link

Share with email

Share

Request Username

Create a new account

Change Password

Password Changed Successfully

Verify Phone

Congrats!