Design of low chloride salt deicing filler based on SCD and optimization of dosage using NSGA-II algorithm

Abstract

Winter road snow melting is an important issue for traffic safety in cold regions. Traditional snow-melting materials, due to their high chloride content, cause severe corrosion to the environment and infrastructure. To solve this issue, a new type of Low Chloride Salt Snow-Melting Filler (LCSMF) was developed in this study, and its optimal preparation process and dosing amount were determined through system optimization and performance evaluation. First, the best carrier and the best hydrophobic modifier were determined by material comparison, and the modification conditions were optimized: modification temperature, modification time, and modifier dosage. Secondly, the active ingredient of snow melting was optimized by orthogonal experiment, and the optimal ratio of the active ingredient was determined by Simplex Centroid Design (SCD) test using Minitab software. The optimized salt-storing filler was incorporated into the asphalt mixture by isovolumetrically replacing the mineral powder, and its comprehensive performance was evaluated by snow melting performance (ice-road bond test) and pavement performance (high temperature stability, low temperature crack resistance, water stability test). The results showed that the salt-storing filler significantly deteriorated the low-temperature crack resistance and water stability, which could be defined as performance shortcomings. Based on this finding, snow-melting performance, low-temperature crack resistance, and water stability were selected as key indicators to construct a multi-objective optimization model. The NSGA-II algorithm was used to generate the Pareto front, and the ideal point method (TOPSIS) was employed to select the solution closest to the ideal point from the Pareto front. The optimal dosage was finally determined: in the AC gradation, the salt-storing filler could replace 99.9% of the mineral powder, and in the SMA gradation, it could replace 86.7% of the mineral powder. This study provides an efficient and environmentally friendly new material for winter road snow melting and offers a scientific basis for its practical application.

Introduction

By the end of 2023, China’s total highway mileage had reached 5.4368 million kilometers, with a highway density of 56.63 km per hundred square kilometers¹, marking a rapid increase that has significantly contributed to the country’s socio-economic growth and improving people’s living standards. However, approximately 75% of China’s territory falls within the winter snowfall and frost zone². The harsh temperature conditions in seasonal frost regions often lead to snow accumulation and icing on asphalt pavements during winter, severely impeding road traffic and resulting in frequent traffic accidents, posing a significant threat to the lives and property of Chinese citizens. Current de-icing technologies are categorized into active control and passive control types, among which chemical frost-inhibition paving techniques are noted for their simplicity, ease of application, effectiveness, and economic benefits³. The basic principle involves partially or fully replacing the mineral powder in asphalt mixtures with ice-melting and snow-inhibiting fillers, which are then laid on the surface layer of asphalt pavements. During winter rainy and snowy weather, the asphalt pavement layer undergoes repeated friction and compaction from vehicles, causing the effective snow-melting components in the porous salt-storage fillers to slowly release through capillary action, osmotic pressure, and vehicle pumping effects. This actively reduces the freezing temperature of the pavement to between − 20℃ and − 3℃, achieving an ice-inhibiting and snow-melting effect that “melts light snow and makes heavy snow easier to clear"⁴. Fillers such as Anti-Freeze Filler⁵ and Mafilon⁶ used in this type of paving have received widespread recognition from researchers. Research on such fillers in China began relatively late, with ice-inhibiting and snow-melting materials first being applied in asphalt mixtures in 2008. Corresponding products developed include ICB(Ice-Ban), Zhong-Jiao Anti-Dark Ice Filler, IGD (IceGuard), and other materials. While the application of salt-storage asphalt mixtures can significantly enhance the road’s snow and ice removal capabilities, the effective snow-melting components in the fillers, primarily chlorides, can induce changes in the performance of the mixture when aqueous solutions precipitate from the pavement during rainy and snowy weather. This may even affect the surrounding environment, causing severe corrosion damage to nearby structures, such as reinforced concrete structures, and resulting in substantial die-off of surrounding vegetation^7,8. The U.S. Department of Commerce has confirmed that the economic impact of chloride-based snow melts on bridge corrosion is considerable, with direct costs for annual bridge repair estimated at 6.43to10.15 billion, and indirect losses (such as traffic delays and production disruptions) being ten times those direct losses⁹. Given the significant risk of chloride-based snow melts corroding bridge structures, this type of salt-storage filler is also unsuitable for use in the paving surface layer of bridges. Therefore, the development of a low chloride salt-storage filler for the paving layer of urban viaduct decks holds great application value and economic benefits, addressing the issues of ice inhibition and snow melting on bridge decks and ensuring road traffic smoothness during winter rainy and snowy weather.

In recent years, scholars from both domestic and international communities have conducted extensive research on the performance optimization of salt-storage fillers from the perspectives of material modification, formulation design, and environmental friendliness. In terms of material modification, Tan et al.¹⁰ and Liu et al.¹¹ experimentally verified the performance enhancement of asphalt mixtures containing antifreeze fillers in low-temperature environments, but did not systematically address the issue of synergistic optimization across multiple performance indicators. Zou¹², in this study, utilized Verglimit-260 ice-inhibiting additive in combination with rubber particles and self-developed salt-storage fillers in high-elasticity/salt-storage asphalt mixtures, primarily exploring their de-icing performance and impact on pavement performance, while neglecting the consideration of environmental corrosion risks. In the realm of formulation design, Kale and Ravi¹³ revealed the variation patterns of physicochemical properties of bentonite through thermal load experiments, providing a theoretical basis for the selection of salt-storage carriers. Xu et al.¹⁴ developed a steel slag-based composite phase change salt-storage aggregate, but their research did not involve the balancing mechanism of multi-objective performance. Yuan¹⁵ designed a novel salt-storage carrier using magnesium oxychloride cement to prepare salt slow-release ice-inhibiting materials, proposing a preparation method for road-use slow-release salt-storage materials to enhance the salt storage capacity and slow-release performance of ice-inhibiting materials in asphalt mixtures; however, their formulation design still relies on empirical trial-and-error methods. In terms of environmental friendliness, Terry et al.¹⁶ reviewed the impact of traditional de-icing agents on metals and infrastructure, pointing out that low-chlorination and the use of acetates as alternative de-icing agents in road applications represent future development trends. Meng et al.¹⁷ developed an environmentally friendly snow-melting agent, but their research did not address the issue of multi-performance synergistic optimization. Furthermore, in the aspect of performance evaluation, Zou et al.¹⁸ experimentally studied the impact of salt-storage snow-melting and ice-inhibiting materials on asphalt mixture performance, but their evaluation methods still rely on a few performance indicators as references. Liu et al.¹⁹ systematically summarized the performance and evaluation methods of salt-based materials and their mixtures, yet did not propose a quantitative optimization solution. In the field of algorithm application, Marler and Arora²⁰ systematically reviewed the application of multi-objective optimization methods in engineering, providing theoretical support for algorithm selection. Li et al.²¹ applied multi-objective optimization algorithms to asphalt mixture design, but their research did not involve the performance optimization of salt-storage fillers.

Although previous studies have made positive progress in the development and application of salt-storage materials, existing methods still have three significant limitations: First, most studies continue to use high-chloride formulations and fail to fully respond to the urgent need for low corrosion in bridge deck pavements; second, performance optimization is often based on single indicators or empirical trial-and-error, lacking systematic balance among snow-melting performance, road performance, and environmental impacts; third, although multi-objective optimization theories are applied in general engineering, their targeted integration with low-chloride formulations in AC and SMA gradations has not been seen.To systematically advance this field, this study innovatively integrates low-chloride formulation design with data-driven multi-objective optimization methods to achieve scientific design of salt-storage fillers in two typical gradation systems, AC and SMA. Specifically, this research first develops a low-chloride composite filler, then constructs a multi-objective response system covering snow-melting performance (ice-pavement bonding tests) and road performance (low-temperature crack resistance, water stability), and introduces the NSGA-II algorithm for systematic optimization to generate a Pareto front; finally, the ideal point method is used to select the optimal content scheme from the non-inferior solution set.The results show that the content schemes optimized by the integrated algorithm (replacing 99.9% of mineral powder in AC gradation and 86.7% in SMA gradation) ensure efficient ice inhibition and snow melting while significantly reducing the corrosion and ecological risks caused by chloride ion release. This work not only confirms the feasibility of combining low-chloride formulations with systematic multi-objective optimization methods but also provides new materials, new methods, and new paths to solve the engineering problem of safe wintering of bridge decks in seasonally frozen areas, which has important theoretical and practical significance for promoting the development of green and intelligent roads.

Development process of the filler

Figure 1 illustrates the primary research workflow for the development of salt-storing fillers in this study. This process encompasses several key stages: the comparison and selection of carrier materials, the comparison and selection of modification materials, the exploration of the most efficient modification conditions, and the determination of the optimal ratio of deicing salt to be adsorbed by the carrier.

Fig. 1

Development process of salt storage filler.

Selection of carrier materials

In ice-inhibiting and snow-melting asphalt mixtures, the LCSMF serves as a substitute for mineral powder. This filler is required to perform core functions within the asphalt mixture, including filling voids, enhancing bonding strength, facilitating the formation of asphalt mastic, and regulating overall performance. Additionally, it should possess the capability to continuously release snow-melting salts, thereby lowering the freezing temperature of the pavement surface. This ensures the preservation of the strength and stability of the asphalt concrete pavement while simultaneously reducing the resistance imposed by ice and snow on vehicles, improving vehicle traction on icy and snowy roads, and ultimately enhancing road safety.The carrier is the primary component of anti-icing and snow-melting fillers, and ideal carrier materials should possess a large specific surface area and abundant pore structure²² to achieve efficient adsorption and storage of salt solutions, reducing the dissipation rate of deicing salt.

When considering the choice of carriers for LCSMF, several materials exhibit distinct advantages. Shell powder^23,24, owing to its porous structure, possesses excellent adsorption capabilities. Bentonite²⁵, with its microporous structure, can adsorb and fixate salt, thereby preventing rapid salt loss in the environment and ensuring the long-term efficacy of the salt-storing fillers. Diatomite²⁶ is characterized by its porosity, low density, and large specific surface area. Zeolite²⁷ has attracted much attention due to its unique adsorption and ion exchange properties. Steel slag^14,28, which is known for its high hardness and good stability, also has a porous structure conducive to salt adsorption and fixation. Therefore, these materials have extensive application prospects in the field of carriers for salt-storing fillers.

Comparative tests and result analysis of carrier materials

Crush and grind the above five materials until they can pass through a sieve with a hole size of 0.075 mm. The specific surface areas of the above five materials were detected by using a specific surface area and pore size analyzer. The test results are plotted as shown in Fig. 2 below.

Fig. 2

Test results of specific surface area and SEM images.

Specific surface area is one of the crucial factors in determining carrier materials. The five aforementioned materials (denoted as A ~ E, representing shell powder, diatomaceous earth, bentonite, zeolite, and powder) were crushed and ground until they could pass through a 0.075 mm sieve. Samples were then taken to measure their specific surface areas, and the statistical results are presented in the bar chart in Fig. 2. Upon careful observation of these statistical results, it is readily apparent that bentonite has a distinct advantage over the other four materials. To further delve into its microstructural characteristics, this study conducted an analysis in conjunction with scanning electron microscopy (SEM) images. From the SEM images, it is easy to observe that the surface of bentonite exhibits a highly rough morphology and is endowed with a rich pore structure. In contrast, the surface of steel slag is extremely flat, with a negligible number of pores. This microstructural difference revealed in the SEM images corroborates the results obtained from the specific surface area tests. Moreover, since the acidic components in asphalt can react chemically with the alkaline components in aggregates, and the resulting substances can enhance the interfacial bonding strength, aggregates with alkaline properties exhibit superior adhesion to asphalt compared to those with acidic properties¹³. Therefore, this study leans more towards selecting weakly alkaline mineral powders. Bentonite typically demonstrates alkaline characteristics in aqueous media, with its pH usually ranging from 8.29 ~ 8.47²⁹. Considering factors such as the acid-base properties of the carrier, the magnitude of the material’s specific surface area, and cost, this study selects bentonite as the optimal carrier.

Selection of hydrophobic materials

In the process of preparing salt-storage fillers, the hydrophobic capacity of the carrier surface plays a significant role in the long-term stability of the filler and the controlled release of salt. To enhance this crucial performance, the present study necessitates surface modification treatment of the carrier. Selecting an appropriate surface modifier is of paramount importance for optimizing the hydrophobic properties of the salt-storage filler. Therefore, several commonly used hydrophobic modification materials were selected for this study: stearic acid, sodium stearate, sodium oleate, and silane coupling agent KH-550. These modifiers can bond with the surface of the carrier material to form a hydrophobic layer that prevents the penetration of water molecules, thereby achieving the goal of slowly releasing the effective snow-melting components (Figs. 3 and 4).

Fig. 3

Flowchart for the surface modification process of saturated adsorbent carriers.

Comparative test on the performance of hydrophobic materials

• 1. Surface Modification of Saturated Carrier: The selected surface modifiers, including stearic acid, sodium stearate, sodium oleate, and silane coupling agent, were placed at room temperature (25 °C) with a mass ratio (modifier/MgCl₂-saturated carrier) set at 20%. An excess amount of anhydrous ethanol was then added, and the mixture was continuously stirred using a magnetic stirrer for 20 min to ensure complete mixing and reaction. The slurry was subsequently dehydrated to constant weight in a 150 °C oven. The process of crushing and grinding was repeated, and the four types of salt-storage fillers that passed through a 0.075 mm sieve were stored in a ventilated, dry place at a constant temperature.

• 2. Hydrophobic Performance Experiment: Identical masses of the ground and qualified salt-storage fillers were weighed and added to deionized water. The conductivity of the mixed solution was rapidly tested. Since the salt-storage fillers were saturated with snow-melting agents, upon contact with water, they would continuously release the chemical components from the porous carrier. Consequently, the conductivity of the solution would gradually increase until all the chemical components in the carrier were released and the conductivity reached a stable value. In this study, 1 g of the prepared salt-storage filler was added to a beaker containing 100 ml of deionized water, and the growth rate of conductivity was used to measure the hydrophobicity and controlled-release effect of the material.

Fig. 4

Conductivity tester.

Comparison results of hydrophobic material properties

The hydrophobic test results, based on the aforementioned experimental requirements, are shown in Fig. 5.

Fig. 5

Conductivity diagram.

Controlling variables such as test temperature and the amount of test material incorporated, this study set the soaking time of snow-melting fillers as the independent variable and conductivity as the dependent variable. Conductivity measurements were taken every 10 min within the first 0–120 min, followed by observations every 30 min in the subsequent 60 min. Test data from six groups of materials were recorded and used to plot Fig. 5 above. The figure clearly shows that the blank group (carriers without surface modification) rapidly released the adsorbed chemical components upon contact with deionized water, reaching a stable state within approximately 20 min. The initial modification effects of the coupling agent and sodium oleate were comparable; however, as time progressed, differences in their hydrophobic effects became apparent. The conductivity of the coupling agent stabilized around 40 min, whereas the effect of sodium oleate lasted longer, extending to 80 min. Sodium stearate and stearic acid exhibited significant sustained-release effects, with both continuing to release for over 110 min. However, stearic acid showed a higher initial sustained-release effect compared to sodium stearate. This study also conducted conductivity tests using commercial snow-melting filler Mfalion for comparison. As seen in the Fig. 5, Mfalion’s sustained-release time reached 110 min, while stearic acid’s conductivity had not yet stabilized at 120 min. Therefore, stearic acid was selected as the optimal component for the sustained-release agent in this study. Additionally, the high conductivity of Mfalion was mainly due to the use of a non-saturated MgCl₂ solution adsorbed on the test carriers, resulting in slightly lower stable conductivity values for the self-made salt-storage fillers.

Fig. 6

Macroscopic effect of hydrophobic modification.

Figure 6 illustrates the macroscopic hydrophobic effects of various modified materials after modification. Labels a to f represent modification with KH−550, sodium oleate, sodium stearate, stearic acid, a blank control group, and Maflion, respectively. It can be clearly observed that in the blank control group, only a small amount of powder floats on the water surface. The material modified with KH−550 mostly aggregates in the center of the water surface, exhibiting poor dispersion. In contrast, the material modified with stearic acid, which shows the best modification effect, uniformly floats on the water surface with excellent dispersion. This difference in macroscopic hydrophobic effects is also corroborated by the results of the electrical conductivity tests. In this filler, the porous bentonite plays the role of a “sponge,” fully adsorbing the snow-melting components. Stearic acid, in the modified structure, acts as a “valve.” With its hydrophobic properties, it covers the surface of bentonite particles and the openings of pores, effectively delaying the penetration of water into the interior and the rate of salt precipitation. Combined with the electrical conductivity tests, it is further verified that the slow-release effect of this material is comparable to, or even slightly superior to, that of commercial fillers. Based on the content discussed in this section, it is found that stearic acid exhibits excellent hydrophobic properties. Additionally, since the carrier modified with stearic acid needs to be incorporated into asphalt mixtures, this modifier should possess certain thermal stability. Relevant research indicates that the thermal decomposition temperature of this material can reach 287.7°C³⁰, which is significantly higher than the temperature during the mixing of the mixture. Therefore, this study deems it highly reasonable to select stearic acid as the surface hydrophobic modifier for ice-inhibiting and snow-melting fillers.

Screening of effective components for snow-melting

CaCl₂, MgCl₂, KCl, and NaCl are four traditional snow-melting salts. NaCl is relatively inexpensive and has good snow-melting efficiency, while KCl has a high unit price. Considering the strong corrosion that chloride salts can cause to cement and metals, high-efficiency chloride salts are required as practical snow-melting components. Considering their snow-melting efficiency, environmental impact, cost-effectiveness, and other application characteristics, CaCl₂ and MgCl₂^16,31,32,33 can be considered as better options. Potassium acetate (CH₃COOK), magnesium acetate Mg(CH₃COO)₂, and calcium acetate Ca(CH₃COO)₂ are commonly used as raw materials for environmentally friendly snow-melting agents. Compared to traditional chloride-based snow-melting agents, acetate-based snow-melting agents have a minor environmental impact during the snow-melting process, protecting the ecological environment³⁴.

Screening test for active ingredients in snowmelt

15 ml of deionized water were added to each 50 ml beaker and then frozen in an environment at -5 °C ± 1 °C for 2 h. The initial mass of the beaker was recorded as m₀. Thirty milliliters of different concentration solutions were poured into the beaker containing ice cubes under constant temperature conditions of -5 °C ± 1 °C. After 30 min, the saturated carrier and melted liquid were quickly poured out, and the mass of the ice and beaker was recorded as m₁. The ice-melting rate formula is :

$$\eta =\frac{{\left( {{m_0} - {m_1}} \right)}}{{15}}$$

(1)

Analysis of screening test results

Based on the five optimized snow-melting salts selected as the effective components for ice-inhibiting and snow-melting fillers, multiple sets of solutions with designed concentrations were prepared. In the Table 1 and 1 represents a solution with a concentration of 0, 2 represents a solution at 1/3 saturation, 3 represents a solution at 2/3 saturation, and 4 represents a saturated solution. With reference to the ice-melting experiment outlined in Sect. 2.3.1, the orthogonal test results for the ice-melting rates under different concentrations of the snow-melting salt formulations are presented in the following Table 1.

Table 1 Orthogonal test results.

Based on the experimental data in Table 1, a range analysis was conducted and the ranges were ranked as follows: R₃ > R₂ > R₁ > R₅ > R₄. Analysis of the range sizes revealed that potassium acetate and calcium chloride exhibited significant ice-melting effects, while sodium acetate had the worst snow-melting efficiency. Further consideration was given to the fact that magnesium chloride causes less corrosion to cement concrete than calcium chloride³⁵. However, magnesium chloride is prone to deliquescence in the air, forming MgCl₂·6H₂O, which releases a large amount of heat upon contact with water and is difficult to transport and store. Therefore, taking into account all these factors, magnesium chloride and sodium acetate were excluded, and calcium chloride, potassium acetate, and magnesium acetate were selected as the effective components for the ice-inhibiting and snow-melting agent.

The determination of the proportion of snow-melting components

Following the identification of three effective ice-melting components, it is necessary to further explore their optimal formulation. To this end, prior to conducting mixture experiments for precise ratio design, pre-experiments were carried out to determine the minimum effective dosage (lower limit) of each component, thereby establishing a reasonable experimental domain for subsequent optimization. The experimental setup included an ice mass of 150 g and a solution volume of 300 mL, with the test duration extended to 60 min to allow for an in-depth analysis of the influence of snow-melting materials at varying mass fractions on deicing efficiency.

Fig. 7

Melting effect of snow melting active ingredients.

The above experimental results can further verify the snow melting capacity of the three materials. Figure 7 shows that the ice melting quality is significantly improved with the increase of mass fraction. The mass fraction of CaCl₂ solution increases slowly in the range of 0%-15%,the ice-melting efficiency improves significantly in the range of 15%-30%, and then the melting effect slows down gradually. When the mass fraction of Mg(CH₃COO)₂ solution is low, the melting ability is weak, and the melting effect begins to appear when the mass fraction of the solution is 10%-15%. The melting ability increases rapidly in the range of 15%~25%. With the increase in mass fraction, the snowmelt capacity of CH₃COOK solution consistently increased significantly, and the snowmelt growth rate began to slow down when the mass fraction of CH₃COOK solution reached 45%. Considering the significant growth interval of the ice melting capacity of the three materials, this study took the ice melting 2.5 g as a reference and specified that the lower limit of the mass fraction of CaCl₂ was 15%, the lower limit of the mass fraction of Mg(CH₃COO)₂ was 20%, and the lower limit of the mass fraction of CH₃COOK was 15%.

Analysis of the proportion of active ingredients in snow melting

Experiment and analysis method

• 1. Ice melting test.

  In 200 ml identical beakers, 150 ml of deionized water was added separately, and then the beakers were placed in an environment at -5℃ ± 1℃ for 2 h of freezing. The initial mass of the ice and the beaker was recorded as m0. Ten grams of saturated adsorption carriers prepared, using solutions of different ratios, were poured into the beakers containing ice under constant temperature conditions of -5℃ ± 1℃. The carriers were evenly spread over the surface of the ice, and then 5 ml of deionized water at 0℃-5℃ was used to moisten the surface of the saturated adsorption carriers. After 60 min, the saturated carriers and the melted liquid were quickly poured out, and the mass of the ice and the beaker was weighed as m1. The mass of ice melted was calculated as m₁-m₀.

• 2. Analysis method of optimal ratio of effective components of snowmelt.

  A mixture experiment model was designed using Minitab software and optimized via the Simplex Centroid method. This approach arranges experiments at centroid points of the simplex to explore the effects of different component ratios on ice-melting performance. A total of 10 experimental points (n = 10) were selected in this study, ensuring model significance and predictive capability while also controlling experimental costs and timeline. Although this sample size has certain limitations in estimating higher-order interaction effects, it remains sufficient to identify main component effects and key low-order interactions, thereby meeting the research objectives for trend analysis and preliminary optimization of the formulation. The average ice-melting mass at 60 min was used as the evaluation metric, and the specific design ratios are presented in Table 2.

Table 2 Experimental design table of simplex centroid method.

Analysis of the ratio results of effective components for snow melting

In previous sections, an optimization of effective snow-melting components was conducted, leading to the selection of three materials: CaCl₂, CH₃COOK, and Mg(CH₃COO)₂. To ensure that the combination of these materials achieves optimal performance in practical applications, this paper employs the “Simplex Centroid Design Method.” This method involves adjusting the volume proportions of the three saturated solutions such that their percentage sum equals one. Consequently, when the carrier adsorbs the snow-melting agent solution, it does so in accordance with this proportion, ensuring the rationality of the material mixture. Table 2 presents the results of the simplex centroid experimental design, with the ice-melting capacity per unit weight as the response value. Experiments were conducted using ten different mixtures ratios set up according to the experimental design. The mean values of the responses were calculated, and subsequent analysis using Minitab software yielded the contour plot (Fig. 8a) and surface plot (Fig. 8b) of the responses.

Fig. 8

Ice-melting response map.

It can be observed from the above melting mixed surface diagram that the melting effect of Mg(CH₃COO)₂ is weak among the three materials. With the increase of CaCal₂ and CH₃COOK, the response value is rapidly improved and gradually approaches the maximum value. Further observation of the contour map makes it easy to see that this figure is approximately symmetrical. Among them, the symmetry axis is the vertical line with magnesium acetate as the apex to the opposite side. Magnesium acetate reduces the overall effect when the three materials are mixed to melt ice, and CH₃COOK and CaCal₂ have significant snow-melting effects and similar snow-melting abilities. The Minitab software was run to fit the curve between the response and the snowmelt variable, and the following Table 3 was obtained.

Table 3 Results of fitting curves.

Analyzing Table 3, it can be seen that the P-value of the linear term is 0.045, which means that the linear term is statistically significant, but the significance level is relatively low, close to the usual significance level of 0.05. The p-value of the quadratic term is 0.003, which indicates that the quadratic term is statistically significant and has a significant effect on the response variable. The P values of the secondary interaction terms AB, AC and BC were 0.008, 0.002 and 0.006, respectively, which were all less than 0.05, indicating that these interaction terms were statistically significant. The p-value of the ternary interaction term ABC is 0.145, which is not statistically significant, and we do not have sufficient evidence to consider these terms as having an effect on the response variable. This analysis of variance shows that the linear and quadratic terms (including their interaction terms) have statistically significant effects on the response variable, while the special cubic and ternary interaction terms are not statistically significant. When constructing a prediction model, it is necessary to consider the inclusion of linear and quadratic terms, rather than the inclusion of ternary interaction terms (It is obtained based on the regression curve, y_max = 14.423 and the corresponding independent variable x₁ = 0.390; x₂ = 0.2; x₃ = 0.410).

LCSMF preparation process

Research on modification conditions of LCSMF

Optimum conditions for surface modification test

Based on the research data presented in Fig. 5, the stabilization time is defined as the time required for the electrical conductivity to reach 5000 µS/cm. In the experiments, the modification temperature was set between 20 °C and 60 °C. This range covers the typical temperature interval suitable for common organic modification reactions, ensuring effective adsorption and reaction of stearic acid on the carrier surface while avoiding side reactions or structural changes in the material due to excessively high temperatures. The modification time was set from 5 to 60 min, aiming to investigate the entire process from initial adsorption to reaction stabilization, thereby providing a basis for identifying an economically viable process window.

Surface modification optimal dosage test

Fixed reaction conditions: modification time 30 min, reaction temperature 30℃. The dosage was 2%-7% (modifier mass/carrier after saturated adsorption). Twenty-five samples were prepared under each dosage group to reduce the data errors. The dosage of stearic acid was increased step by step, and the time point reached 5000 μm/cm under each dosage was recorded.

Modification results and discussion

Fig. 9

Optimal modification condition test diagram.

Figure 9a presents the relationship curve between the modification reaction temperature and the stabilization time. In this study, other variables were controlled, and the modification temperature was gradually increased from 20 °C with a gradient of 5 °C. Conductivity tests were conducted on the materials after modification. The figure indicates that the modification temperature has a minor impact on the stabilization time. Starting from 20 °C, the stabilization time gradually increases and exhibits a turning point at 30 °C. As the temperature continues to rise, the stabilization time changes slightly, displaying a horizontal trend. This is because stearic acid can dissolve in ethanol solution at room temperature to form an ethanol solution of stearic acid. Stirring and reacting allow it to fully contact the bentonite carrier, resulting in high reactivity. An increase in temperature can slightly enhance the reactivity between stearic acid and the carrier, but the effect is minimal. Considering the sustained-release effect, when the temperature is above 30 °C, the hydrophobic effect is comparable. To reduce unnecessary energy consumption, the optimal modification temperature should be set at 30 °C.

Figure 9b depicts the relationship curve between the modification time and the stabilization time. In this study, other variables were controlled. The figure shows that the reaction time significantly affects the stabilization time. Within 0 ~ 30 min, the stabilization time rapidly increases with the modification time. When the modification time exceeds 30 min, the stabilization time remains basically unchanged, stabilizing within the range of 125–130 min. This is because at this point, stearic acid and the carrier have fully reacted, and the hydrophobic modification effect has reached its optimum. Therefore, the optimal reaction time for modification is determined to be 30 min.

In Fig. 9c, we can observe that the incorporation level of stearic acid has a significant impact on the stabilization time. Combining the experimental data, it is evident that as the stearic acid content increases from 2% to 5%, the hydrophobic effect grows remarkably, with the stabilization time extending from 80 min to 120 min. However, after the stearic acid content reaches 5%, further increasing its incorporation results in minimal enhancement in the modification effect. This phenomenon can be attributed to the fixed surface area of the bentonite carrier. When the stearic acid content is low, the carrier surface can fully undergo modification by stearic acid without reaching saturation, and the hydrophobic duration extends as the stearic acid content increases. When the stearic acid content reaches 5%, the carrier surface is nearly fully coated by stearic acid, approaching saturation, resulting in excellent hydrophobic performance. Considering that the stabilization times at 6% and 7% incorporation levels are basically the same and slightly higher than that at 5%, we have determined the optimal stearic acid incorporation level to be 6%.

The preparation flowchart of LCSMF

For ease of reading, based on the research results in the previous text, the preparation process of LCSMF is presented in Fig. 10 below in this paper. Particular attention needs to be paid to the saturation adsorption of snow-melting components by the carrier, which must meet the condition m₁ = m₂ to ensure that the carrier pores have maximally adsorbed the snow-melting salt. (m₁ is the mass after saturated adsorption of the deicing agent solution under negative pressure and then drying. m₂ is the mass after repeating the above steps. If m₁ < m₂, the adsorption and drying process needs to be repeated again until the mass m_q=m_q+1 for two consecutive times) .After meeting the above process, the material needs to be crushed and ground until it can pass through a 0.075 mm sieve. Subsequently, the powder is subjected to surface modification under the specified time, temperature, and other conditions in Fig. 10. The crushing and grinding process is then repeated to achieve the specified particle size. Once all the above processes are completed, the salt-storing filler is prepared and can be stored in a constant temperature, dry, and ventilated place.

Fig. 10

Preparation process of salt storage filler.

Research on the mix proportion of salt-stored asphalt mixture

Experimental materials and methods

There are two methods for replacing mineral powder: equal-mass replacement and equal-volume replacement. Due to the difference in density between mineral powder and LCSMF, with LCSMF having a lower density than mineral powder, using equal-mass replacement would result in an increase in the total volume of the filler, thereby affecting the porosity. This could reduce the flowability of the asphalt mixture, potentially leading to agglomeration and increasing the difficulty of mixing and paving. Additionally, a decrease in the residual porosity of the asphalt mixture would impair the high-temperature stability of the pavement and the effectiveness of freeze inhibition in winter. Therefore, we adopt the method of partially or fully replacing mineral powder with LCSMF on an equal-volume basis for the design of salt-storage mixtures³⁶.

Experimental materials

The indicators of asphalt, coarse aggregate, fine aggregate, LCSMF, mineral powder and asphalt required for the study of the mix ratio of salt storage fillers are shown in Appendix 1 ~ 5. This study used continuous dense gradation AC−13 and discontinuous gradation SMA−13, and their gradation curves are shown in Fig. 11,the detailed table is shown in Appendix 6 and Appendix 7.

Fig. 11

AC-13 and SMA-13 grading curve; (a) AC-13 (b) SMA-13.

After the design of the asphalt mixture gradation curve was completed, a comprehensive evaluation was conducted on the stability, flow value, void ratio and other indicators of AC-13 and SMA-13, and the optimal asphalt aggregate ratios were determined to be 4.9% and 5.9% respectively.

Experimental methods

A. Anti-icing performance experiment

In this study, the Anti-icing performance of salt-stored asphalt mixtures was evaluated using the Ice-road bonding force test, as outlined in Appendix A of T0719 within the Technical Standards for Design, Construction, and Inspection of Ice-Resistant and Snow-Melting Asphalt Pavements in Cold Regions (DB23/T 3087 − 2022).

B. Road performance test

• 1. High temperature stability

  The high-temperature stability of asphalt mixtures was assessed in this study through the asphalt mixture rutting test specified in T0719 of the Test Methods of Bitumen and Bituminous Mixtures for Highway Engineering (JTG-E20-2011).

• 2. Low-temperature crack resistance

  To evaluate the low-temperature crack resistance, the study employed the beam bending test outlined in T0715 of the Test Methods of Bitumen and Bituminous Mixtures for Highway Engineering (JTG-E20-2011). This test determines the deformation resistance of asphalt mixtures under specified temperatures and loads.

• 3. Water stability

  The water stability of asphalt mixtures was evaluated in this study using the asphalt mixture water immersion Marshall test and the freeze-thaw splitting test, as specified in T0709 and T0729 of the Test Methods of Bitumen and Bituminous Mixtures for Highway Engineering (JTG-E20-2011).

Results and discussion

Ice inhibition performance experiment of the mixture

Fig. 12

Ice-road bonding force test.

Fig. 13

Ice-road bonding force test results.

The Ice-road bonding force test is designed to evaluate the bond strength between asphalt pavements and ice layers, which is crucial for achieving the goals of “melting light snow quickly and facilitating snow removal for heavy snow” while maintaining traffic safety during winter conditions (Fig. 12). In this study, we utilized a Universal Testing Machine (UTM) system in conjunction with a custom-made fixture to investigate the level of ice-pavement adhesion at various LCSMF replacement rates. The conclusion drawn from Fig. 13 is that as the replacement rate of LCSMF increases, the ultimate tensile failure force at the ice-pavement interface decreases significantly. A control group (filler: Mafilon, gradation: SMA-13) was specifically incorporated in this experiment. The test results clearly demonstrate nearly identical outcomes between the control group and the SMA-13 experimental group. Furthermore, the findings presented in Sect. 2.2.2 indicate that LCSMF and Mafilon exhibit comparable sustained release effects. Based on the collective evidence from these experimental components, it can be concluded that the self-developed LCSMF performs equally to the commercially established product Mafilon in terms of both short-term and long-term release efficacy. Furthermore, upon detachment, the ice-pavement interface remains smooth (as shown in Fig. 12), and only trace amounts of residual ice remain in the cracks at the center of the asphalt mixture’s exterior. This observation is of great significance for maintaining skid resistance on pavements after mechanical ice removal following heavy snowfall. Additionally, this situation indirectly indicates that LCSMF can be uniformly dispersed during the asphalt mixture mixing process, resulting in excellent ice-inhibiting effects.

At zero LCSMF incorporation rate, the interface adhesion force of gap-graded asphalt is higher than that of continuous-graded asphalt. However, as the incorporation rate increases, the ice-inhibiting effect of gap-graded asphalt surpasses that of continuous-graded asphalt. This may be attributed to the deeper texture depth and greater surface roughness of gap-graded pavements, which result in a larger contact area with ice and, consequently, slightly higher salt release compared to continuous-graded pavements. Since the amount of LCSMF incorporation has a significant impact on interface adhesion force, this study has prioritized the optimization of ice-inhibiting performance as one of its main objectives.

Analysis of test results on road performance

Salt storage asphalt mixture is typically employed as the top layer of roads designed for ice and snow control. During its utilization, asphalt pavements are subjected to vehicle loads and various environmental factors. To ensure that the road can provide long-term stable and reliable service for vehicles, asphalt pavements must possess adequate high-temperature stability, low-temperature crack resistance, water stability, fatigue resistance, and aging resistance. This chapter will focus on studying the high-temperature stability, low-temperature crack resistance and water damage resistance of two gradations of salt-storing asphalt mixtures, namely AC-13 and SMA-13. Comprehensively evaluate the road performance of asphalt mixture for replacing mineral powder under different volume substitution rates of self-made fillers, and conduct a comparative analysis with the technical requirement values of road performance of asphalt mixture.

Fig. 14

Road performance test results.

Figure 14a presents the results of the rutting test, where it is visually observable that as the proportion of LCSMF replacing mineral powder increases in equal volume, the high-temperature stability of both continuous-graded and gap-graded asphalt mixtures exhibits a declining trend, albeit remaining well above the specified values. Figure 14b displays the results of the beam bending test, which indicates the flexural tensile performance of asphalt mixtures at low temperatures and serves as a reference standard for low-temperature stability. From the data presented, it can be discerned that with the continuous increase in LCSMF replacement rate, the low-temperature crack resistance of both grades declines overall, with a more pronounced effect on continuous-graded asphalt. Considering the results from Fig. 14a and b in conjunction with the specified values, and given that the increasing LCSMF incorporation has a deteriorating effect on high-temperature stability yet remains significantly above the specified values, this study prioritizes the enhancement of low-temperature stability as one of its key optimization objectives.

The water stability of salt-stored asphalt mixture was verified by immersion Marshall test and freeze-thaw splitting test, and the residual stability and freeze-thaw splitting residual strength ratio were characterized. According to the analysis in Fig. 14c and (d), it is found that with the increase of LCSMF substitution rate, the overall water stability performance shows a downward trend. However, the residual stability of the discontinuous graded mixture shows a decreasing trend and then increases. Because this situation is complicated, the study decides to take residual stability as a representative of water stability and become one of the optimization objectives.

The optimal dosage of LCSMF is studied through a comprehensive algorithm

Basic principles of NSGA-II and ideal point synthesis algorithm

• 1. Basic Principles of NSGA-II algorithm.

  As a kind of road material, the salt-storage asphalt mixture is desired to show the performance of ice inhibition and snow melting while ensuring strong road performance. From the above experimental results analysis, it can be observed that with the increase of the salt storage filler replacement ratio of mineral powder, the road performance of ice and snow-inhibiting asphalt mixture conflicts with the ice-inhibiting effect. The mechanical properties (water stability and low-temperature crack resistance) tend to decrease with increased salt storage filler content. In contrast, the ice-inhibiting effect increases significantly with the salt storage filler replacement rate.

  NSGA-II is often used for multi-objective optimization problems, where these objective functions often conflict. It is widely used because of its excellent performance in computational efficiency and set quality³⁶. The core steps of NSGA-II are Non-dominated Sorting and Crowding Distance calculation. The purpose of non-dominated ranking is to divide the individuals in the population into multiple non-dominated levels, and the implementation steps are as follows: for each p in the population, the set of the individuals S_p and the number of dominated n_p are calculated. Put all individuals with n_p=0 into the first layer of non-dominant frontier F₁. For each p in F₁, traversing the set of individuals Sp that it dominates, n_q-1 of the individual that will be dominated. If n_q=0, then put q into the next layer front F₂. The process is repeated until all individuals are assigned to a specific layer front. The crowding calculation aims to measure the distribution density between individuals in the same non-dominant frontier to maintain population diversity.

  Fig. 15

  

  Principle of non-dominated ordering and congestion calculation.

  The implementation steps are as follows: the individuals in each non-dominant frontier F_i are sorted according to the value of each objective function. For each objective function, the crowding degree of each individual is calculated, and the crowding degree of boundary individuals is set to infinity to ensure that the boundary solution is preserved. The principle is shown in Fig. 15.

  Domination relation: Solution x dominates solution y if and only if

  $$\forall j \in \{ 1,2 \cdot \cdot \cdot ,m\} ,f_{j} (x) \le f_{j} (y)AND\exists j \in \{ 1,2 \cdot \cdot \cdot ,m\} ,f_{j} (x) \le f_{j} (y)$$

  (2)

  Crowding degree formula:

  $$p = \sum\limits_{i = 1}^{m} {\frac{{f_{j} (p + 1) - f_{j} (p - 1)}}{{f_{j}^{\max } - f_{j}^{\min } }}}$$

  (3)

  Where f_j(p+1) and f_j(p-1) are adjacent objective function values, $f_{j}^{{\hbox{max} }}$ and $f_{j}^{{\hbox{min} }}$ are the maximum and minimum values of this objective function.

• 2. Basic principle of ideal point method.

  The ideal point method is a multi-objective optimization method mainly used to find the optimal solution between multiple objective functions. The basic principle is to measure the quality of the solution by defining an Ideal Point and Reference Point and guiding the search process to optimize in the direction of the ideal point. The optimal value f_iis optimized separately for each objective function f_i, and the ideal point (f₁*,f₂*,…, f_n*) is obtained. Secondly, the reference point can be an ideal point or an expected value defined by the designer, but the ideal point is usually set as a reference point. Then, for each solution x, the distance between its objective function and the reference point is calculated, and the nearest distance between minD (x) and the reference point is solved. The output point is the solution needed in this study.

  Distance formula:

  $$D(x) = \sqrt {[f_{1} (x) - f_{1}^{*} (x)]^{2} + [f_{2} (x) - f_{2}^{*} (x)]^{2} + ... + [f_{n} (x) - f_{n}^{*} (x)]^{2} }^{{}}$$

  (4)

  Where ( f₁(x),f₂(x),… ,f_n(x) ) is the objective function, ( f₁^*,f₂^*,… f_n^* ) is the ideal point.

Comprehensive solution process of NSGA-II and ideal point method

The independent variable ranges from 0 to 100. During function fitting, it was observed that although the road performance test results decreased within this substitution rate range, all values remained above the specification limits. Therefore, apart from constraining miny₁, maxy₂, and maxy₃, no additional constraints were imposed. The NSGA‑II algorithm was applied within the defined domain, with an initial randomly generated population size of N = 100, number of iterations I = 50, crossover probability of 0.8, and mutation probability of 0.1 to output the Pareto front (The population size and iteration setting I = 50 were based on established practices in similar material optimization studies³⁷. Preliminary parameter sensitivity analysis also confirmed that this configuration effectively converges to a stable Pareto front within reasonable computational cost). The solution set obtained by the NSGA‑II algorithm represents a group of Pareto-optimal solutions rather than a single solution. Therefore, this study adopts the “Ideal Point Method” to identify a balanced solution from the Pareto front. In this method, the multi-objective ideal point Z^* is defined as the point that minimizes the sum of Euclidean distances to all single-objective ideal points Z_i. (The Euclidean distance was chosen as the evaluation criterion for the ideal point method because it intuitively reflects the geometric proximity of a solution in the multi-dimensional objective space.) The Euclidean distances between all points on the Pareto front and Z^* were calculated, and the point P_min closest to Z^* was selected as the optimal solution. The detailed procedure follows the flowchart shown in Fig. 16.

Fig. 16

Flowchart of comprehensive solution of NSGA-II and ideal point method.

Analysis of optimization results

This paper encompasses three objectives: minimizing the ultimate tensile failure force at the interface (objective y₁), maximizing the maximum flexural tensile strain (objective y₂), and maximizing the residual stability (objective y₃). The objective functions are determined based on the fitted data from multiple sets of experiments mentioned above, with the substitution rate of salt-storing filler serving as the independent variable for the multi-objective optimization. The objective functions are outlined in Table 4.

Table 4 Objective function.

Figure 17a and b present the Pareto fronts obtained after 100 iterations of iterative optimization for the two grading types. This solution set comprises a total of 100 optimal solutions. In the figures, it can be observed that the Pareto front distribution exhibits no obvious discontinuity, appearing smooth and uniform with good convergence. Specifically, for the AC-13 grading type within the Pareto front solution set, there is an overall trend of increasing ice-suppression performance, low-temperature crack resistance, and water stability as the substitution rate of salt-storing filler increases.

Fig. 17

Scatter distribution of Pareto front of AC-13 and SMA-13 grading; (a) AC-13 (b) SMA-13.

However, for the SMA-13 grading type, within the range of the Pareto front solution set, while the ice-suppression performance significantly increases with the substitution rate of salt-storing filler, the low-temperature crack resistance deteriorates significantly. This is attributed to the salt-storing filler reducing the adhesion and cohesion between asphalt and aggregates, thereby severely impacting the low-temperature flexural tensile strain. Compared to continuous grading types, the low-temperature crack resistance of gap-graded asphalt mixtures is more sensitive to the substitution rate of mineral powder. As for water stability, it initially decreases with increasing substitution rate due to the potential disruption of asphalt-aggregate adhesion by the salt compounds. However, an appropriate increase in the salt-storing filler can enhance water stability by improving interfacial adhesion and increasing the use of anti-stripping agents.

The Pareto frontier obtained using NSGA-Ⅱ is not a solution but a set containing 100 optimal solutions. This study is committed to finding an optimal advantage in the solution set. It uses the algorithm to solve the “ideal point” according to the process described above and outputs the point Pmin nearest to the Pareto frontier solution set. At this point, Pmin is the final output result(the results are shown in the appendix 8).

In Fig. 18a and b, the green dots represent Pmin. From the figures, it is intuitive to observe that the optimal substitution rate of salt-storing filler for continuous grading is higher than that for gap grading. The primary reason for this lies in the different internal structural forms of the two types of mixtures. Gap grading is characterized by an abundance of coarse aggregates, mineral powder, and asphalt, with a lesser amount of fine aggregates. The superior performance of this mixture largely depends on the mastic formed by mineral powder and asphalt, which fills the voids between coarse aggregates and provides additional cohesive force and deformation resistance. When a large amount of salt-storing filler replaces mineral powder, it alters the composition and properties of the mastic, thereby affecting the cohesive force and deformation resistance of the mixture, and significantly impacting its low-temperature crack resistance. Additionally, the incorporation of salt-storing filler can disturb the internal mixture structure, increasing the porosity of the mixture and further altering its water stability. Therefore, salt-storing filler cannot fully replace mineral powder in gap grading.

Fig. 18

AC−13 and SMA−13 Pareto frontier combined ideal point; (a) AC−13 (b) SMA−13.

Conclusion

This study addresses the issue of strong chloride corrosion in conventional salt-storage fillers by developing a low-chloride alternative. Through carrier selection, optimization of ice-melting components, and surface modification, the optimal composition ratio was determined using simplex centroid design. The NSGA-II algorithm combined with the ideal point method was applied to identify the optimal replacement ratio of the prepared filler in asphalt mixtures. The results demonstrate effective maintenance of ice-melting performance while significantly reducing corrosion risks to infrastructure and the environment.

• 1. Optimization Results of LCSMF Materials:

  Bentonite was identified as the optimal carrier material, and stearic acid as the optimal hydrophobic modifying material, with the best modification conditions being a temperature of 30 °C, a time of 30 min, and a stearic acid doping amount of 6%. Through the simplex centroid design test, the optimal ratio of snow-melting components was determined to be CaCl₂: CH₃COOK: Mg(CH₃COO)₂ = 0.39:0.2:0.41.

• 2. Performance Evaluation Results:

  The salt-storage filler exhibited excellent snow-melting performance, with a complete ice-road adhesion failure interface and almost no ice residue during testing. The peak adhesion force significantly decreased as the substitution rate of salt-storage increased. As the substitution rate increased, the overall road performance showed a downward trend, with water stability and low-temperature crack resistance being more severely degraded by the salt-storage filler. However, the test data were still above the specified requirements, indicating good performance.

• 3. Multi-objective Optimization Results:

  Taking snow-melting performance, low-temperature crack resistance, and water stability as key reference indicators, the study generated the Pareto front using the NSGA-II algorithm and selected the optimal solution through the ideal point method to determine the optimal doping amount: 99.9% of mineral powder could be replaced by salt-storage filler in AC grading, and 86.7% in SMA grading.

However, this study was conducted at the laboratory scale, and the selected materials and processes have not yet been validated under actual road conditions. The optimization process focused only on AC and SMA mix types, and the applicability of the conclusions to other gradation systems requires further verification. Furthermore, the long-term performance of the filler and its durability under the combined effects of real traffic and environmental conditions necessitate ongoing monitoring and evaluation.

Potential environmental risks

The newly developed low-chloride salt-storage filler demonstrates significant environmental benefits by effectively reducing chloride-induced corrosion of infrastructure and minimizing hazards to vegetation. However, potential ecological impacts require further attention: first, the long-term accumulation effects of components such as acetate in soil-water systems and their ecological consequences need further clarification; second, the release behavior of the filler over extended service periods and its potential risk to groundwater necessitate validation through field monitoring. It is recommended that subsequent research focus on ecotoxicity testing and a full life-cycle environmental impact assessment to ensure environmental safety prior to large-scale application.