Research on the Location-Routing Optimization of International Freight Trains Considering the Implementation of Blockchain

Abstract

The purpose of this study is to solve the problem of low load factor and profit margin in the point-to-point transportation of international freight trains through the assembly transportation organization mode. A bi-objective location-routing optimization model is constructed to optimize problems, such as the location of the assembly center, route of freight assembly, frequency of international freight trains, and number of formations. The objectives are to minimize the total comprehensive cost and maximize the average satisfaction of the shippers. Considering the impact of blockchain technology, the proportion of customs clearance time reduction after blockchain implementation, the proportion of customs clearance fee reduction after blockchain implementation, and the cost of blockchain technology are introduced into the model. The case study is based on railroad transportation data for 2022. In this case, 43 stations in the Indo-China Peninsula are selected as origin stations, and two Chinese stations are designated terminal stations. An improved NSGA-II algorithm (ANSGAII-OD) is proposed to resolve the location-routing optimization model. This algorithm is based on opposition-based learning and its dominant strength. The case study indicates that assembly transportation is advantageous compared with direct transportation. Moreover, the comprehensive cost is reduced by 19.77%. Furthermore, blockchain technology can effectively reduce costs and improve transportation efficiency. After the implementation of blockchain technology, the comprehensive cost is reduced by 8.10%, whereas the average satisfaction of shippers is increased by 10.35%.

1. Introduction

Railways play a pivotal role in international freight transportation and support global supply chain operations in many industries [1]. In November 2006, the heads of railways in China, Germany, and Russia signed a memorandum proposing running international trains between China and Europe [2]. In March 2011, China Railway Express (CRE) was successfully run from Chongqing, China, to Duisburg, Germany. In September of the same year, the railway transport sectors of China, Russia, Kazakhstan, Germany, and the local government of Chongqing, China, signed a memorandum regarding the Chongqing–Sinkiang–Europe International Railway, which proposed the establishment of a unique platform enterprise [3]. In recent years, the scale of international freight trains has been growing steadily and rapidly [4]. It has extensively promoted the liberalization and facilitation of international trade and has been highly appreciated by Asian and European countries [5]. In addition, COVID-19 has highlighted its impact on international transportation, making it attractive to policymakers. This is because the volume of cargo grew rapidly during the COVID-19 pandemic, whereas maritime and air transportation became severely stagnant [6]. In 2015, Pakistan and the Chinese government signed the China–Pakistan Memorandum of Understanding (MOU), which was proposed to establish rail connections [7]. In December 2021, all sectors of the China–Laos Railway opened [8]. The value of freight trains in international trade is becoming increasingly important [9]. However, problems with international freight trains, such as poor coordination among participants, a single transportation organization mode, and significant time spent consolidating cargo, are prominent. This reduces competitiveness and is not conducive to the sustainable development of international liner shipping [10].

Information sharing is critical for coordination among participants, and information asymmetry can reduce the performance of supply chains that rely on international freight trains [11]. Consequently, many countries are actively promoting railway informatization. Rail operators in the United States have created the America Smarter Railroad to improve information interoperability within rail systems [12]. European countries have jointly developed InteGRail (Intelligent Integration of Railway Systems) [13]. Rail Route 2050, published in 2012, proposes the removal of barriers to connectivity by 2050 [14]. Rail operators in Japan have built a cyber-rail system that realizes an efficient connection between railways and other modes of transportation through its powerful information function [15]. The China State Railway Group Co., Ltd. issued the 14th Five-Year Plan for Railway Network Security and Informatization in 2022, which proposes optimizing and integrating the railway business information system and promoting the building of an information integration platform [16]. Blockchain systems possess technical characteristics such as integrity, stability, and traceability. Therefore, it can reduce the resource consumption of international freight trains, improve the efficiency of customs clearance, and reduce the fees and time for customs clearance [17]. Recently, it has attracted strong interest from relevant organizations, and some CRE operating companies have been actively exploring it [18].

Previously, scholars tended to improve the efficiency of international freight trains by optimizing transportation networks [19]. Some scholars selected hubs through evaluation, and factors such as economic and social development, geographic location, and current status of transportation facilities were considered [20]. Other scholars determined the location of the hub based on system optimization, where the least fee and shortest time are common model objectives [21]. Meanwhile, some studies have further analyzed transportation routes [22]. Additional factors, such as environmental protection [23] and system reliability [24], have also been considered for the various needs of the transportation industry. Regarding the implementation of blockchain technology in the transportation industry, scholars have achieved a series of results on the factors influencing its implementation [25], the construction of blockchain systems [26], and the impact of blockchain on supply chain performance [27]. Previous studies have not sufficiently analyzed the impact of implementing blockchain technology on optimizing transportation networks and did not provide effective theoretical support for applying blockchain technology to international liner shipping. It is worth noting that international freight trains must undergo customs clearance at the border station. The clearance process will consume much time and many resources. Furthermore, different countries have different rail agreements [28], and several different institutions are created to handle data integrity, consultation, border delay problems, and others related to customs clearance processes. By integrating the current customs clearance system with the blockchain network and applying the combined system in an IoT environment, the level of trust between participants in international freight trains will increase dramatically. While data security is ensured, the customs clearance process of cross-border railway freight transport will be simplified, the customs clearance efficiency of international trains will be improved, and the customs clearance costs will be reduced. But now, quantitative analysis related to these is scarce and is not sufficiently integrated with the optimization of transport systems.

In summary, this study makes three contributions. First, a bi-objective location-routing optimization model was constructed to study the locations of assembly centers and routes of international freight trains. Second, considering the impact of blockchain technology on customs clearance, the proportional reduction of fees for cross-border clearance after blockchain implementation, the proportional decrease in customs clearance time after blockchain implementation, and the cost of blockchain technology were introduced into the optimization model. Finally, a sensitivity analysis of the three parameters is conducted to study the relationship between each parameter and the optimization objective.

The main contents of this paper are as follows. Literature related to the location-routing optimization of international freight trains and the implementation of blockchain technology in the transportation industry are reviewed in Section 2. A detailed description of the problem investigated in this study and symbols in the location-routing model are discussed in Section 3. The location-routing optimization model for international freight trains and its solution algorithm is presented in Section 4 and Section 5, respectively. In Section 6, the optimization case of international freight trains and the sensitivity of the three parameters of the blockchain are analyzed. Finally, this study and future research are summarized in Section 7.

2. Literature Review

This section reviews the relevant literature and consists of two main sections. First, relevant research on location-routing optimization of international freight trains is reviewed. Next, research on implementing blockchain technology in the transportation industry is reviewed.

2.1. Location-Routing Optimization of International Freight Trains

Research on train hub locations is relatively mature. Some researchers have attempted to evaluate hub locations. Zhao et al. ranked nodes based on their centrality in the transportation network and determined the location of hubs for CRE [29]. Zhang et al. proposed five connectivity indicators to comprehensively evaluate the nodes and determine the locations of international rail logistics hubs under the ‘Belt and Road’ initiative [30]. Muravev et al. employed a hybrid multi-criteria decision-making model to identify the optimal location of an international logistics center for CRE. The decision-making model was based on 17 indicators selected from economic and social development, infrastructure construction, and service capacity of the transportation industry [20]. Other researchers have studied the location of train hubs based on system optimization. To optimize the location of the assembly center for CRE, Tang et al. constructed a bi-objective optimization model that considered the minimum transportation cost and the minimum transportation time [31]. Wei et al. optimized the location of hubs for CRE. Constraints such as transportation demand and operating capacity of the hub were considered in the optimization model [22]. Cheng et al. sought to identify the optimal location of assembly centers for CRE with the objective of minimizing both transport costs and carbon emissions [23]. Lu et al. constructed a two-echelon network comprising railways and roads and determined the location of an intermodal transport hub for the CRE. In the optimization model, uncertainties in transport time and capacity were considered [32]. Some researchers have analyzed freight transportation routes in addition to hub locations. Tang et al. allocated transportation demand points to assembly centers while determining their locations [31]. Wei et al. performed a joint optimization of freight transportation routes and train operation plans [31]. Fattahi et al. considered transportation routes for conventional and hazardous cargo in their location optimization model [33]. Fazayeli et al. constructed a multimodal transportation location-routing optimization model based on time windows and fuzzy demand [34].

2.2. Implementation of Blockchain Technology in the Transportation Industry

Blockchain technology has the advantages of decentralization, openness, and security, which ensure the authenticity and timeliness of data transmission. Consequently, it can build a cornerstone of trust among multiple subjects and promote the free interaction of information [35]. This has led to the widespread use of blockchain in sectors such as finance [36] and healthcare [37]. Many researchers have studied the applications of blockchain technology in the transportation industry. Zhang et al. analyzed the value of blockchain technology in the transportation industry [38]. Orji et al. analyzed the main factors affecting blockchain’s implementation in the transportation industry [25]. Regarding the specific application of blockchain technology, there is a series of technical frameworks and models for different scenarios, such as international transportation [39], port operations and logistics management [39], and urban transportation [40]. At the same time, several previous studies have confirmed the benefits of blockchain for transportation systems. Wamba et al. conducted a comparative analysis of supply chain performance before and after blockchain application [41]. Lohmer et al. showed that blockchain technology can improve supply chain resilience and control loss costs [27]. In addition, a few studies have explored the changes in transportation systems resulting from implementing blockchain technology. Chen et al. considered the enhancement of information interaction by blockchain. After applying blockchain, they analyzed the changes in tariff competition and the market structure of shipping logistics [42]. Hong et al. analyzed the changes in the distribution of container traffic in the transportation network of a CRE after applying blockchain technology. Their study considered the impact of blockchain technology on customs clearance time and costs [18].

2.3. Literature Summary

Many studies have been conducted on the locations of hubs for international freight trains. Some studies have explored cargo transportation routes. However, in previous studies, the point-to-point mode of freight consolidation led to a waste of transportation resources and increased transportation costs. In addition, research on implementing blockchain technology in the transportation industry focuses on the constraints of applying blockchain and constructing a blockchain system architecture. Few studies have been conducted on the optimization of transport networks after the implementation of blockchain. Therefore, a location-routing optimization model is constructed to jointly optimize the cargo consolidation route and location of the assembly center. In this model, international linear trains are used to collect cargo at multiple origin stations. Three parameters–the proportion of reduction in customs clearance costs after the implementation of blockchain, the proportion of reduction in customs clearance time after the implementation of blockchain, and the cost of blockchain technology are introduced into the model to analyze the impact of blockchain’s implementation on the transportation network of international freight trains.

3. Problem and Parameter Description

Considering the low transportation demand and long waiting times at some stations in international liner transportation, the consolidation mode is adopted for cargo transportation. The goods are transported to the consolidation center by train and then to the final station after consolidation. As the international trade goods of some countries are small or scattered, it is more favorable to go through cross-country consolidation. The implementation of blockchain technology significantly impacts the transportation time and costs of railway sections. As shown in Figure 1, shippers, consignees, international freight forwarders, financial institutions, railway operating sectors, customs, inspection agencies, and other relevant sectors are all linked by blockchain platforms. After the implementation of blockchain technology, data sharing and cross-verification among chain members and automatic capture and comparison of the required information can be realized. This simplifies the customs clearance process. As a result, the locations of assembly centers and cargo consolidation routes also change.

As shown in Figure 2a, before the implementation of blockchain, goods are assembled at two assembly centers $k1$, $k2$ through two routes $k1$→$\mathrm{i}1$→$\mathrm{i}2$→$k1$, $k2$→$\mathrm{i}3$→$\mathrm{i}4$→$k2$, respectively, and then transported to the final destination by the international freight train. After the implementation of blockchain technology, the time and fees for cross-border clearance change. Thus, the transportation routes for goods change accordingly. As shown in Figure 2b, the assembly center $k2$ is not used. Simultaneously, shorter customs clearance times and lower customs clearance costs facilitate cross-border transportation. As a result, the assembly transportation route becomes $k1$→$\mathrm{i}1$→$\mathrm{i}2$→$\mathrm{i}3$→$\mathrm{i}4$→$k1$.

Based on the above considerations, the location-routing problem of international freight trains is studied as shown in Figure 3.

For the convenience of this study, the following assumptions are made:

(1)

All train formations are at the same level.

(2)

The goods are transported by train to the assembly center or the destination station.

(3)

Goods are picked up when they arrive at their destination with no waiting times or storage fees.

The main symbols and their meanings are listed in

Table 1. Model symbols and variables.

Typology	Symbol	Meaning	Typology	Symbol	Meaning
Sets	$I$	Set of origin station	Parameters	$C_c$	Fee of cross-border clearance, CNY/TEU
	$K$	Set of candidate points of assembly centers		$r_{bc}$	Proportion reduction in the fee of cross-border clearance after the implementation of blockchain technology
	$D$	Set of the final destination station		$C_m$	Fee of train formation, CNY/TEU
	$S_i$	Set of originating stations in the assembly transportation route of the origin station $i$$, i\in I$		$c_b$	Fee of blockchain technology, CNY/TEU
Variables	$q_{id}$	Cargo volume from origin $i$ to destination $d$$, \mathrm{TEU}, i\in I,d\in D$		$C_k$	Construction fee of assembly center $k$$, \mathrm{CNY}, k\in K$
	$l_{ij}$	Transport miles from $i$ to $j$$, \mathrm{km}, i,j\in I\cup K\cup D$		$B_{\max }$	Maximum number of formations allowed, TEU
	$t_{wi}$	Congested waiting times at origin $i$$, \mathrm{day}, i\in I$		$p$	Profit generated by transporting a unit of goods over a unit distance, CNY/(TEU·km)
	$t_{wk}^1$	Congested waiting times at assembly center $k$$, \mathrm{day}, k\in K$		$c_{il}$	Loading fee at origin $i$$, \mathrm{CNY}/\mathrm{TEU}, i\in I$
	$t_{wk}^2$	Waiting time for transportation at assembly center $k$$, \mathrm{day}, k\in K$		$c_{du}$	Unloading fee at destination $d$$, \mathrm{CNY}/\mathrm{TEU}, d\in D$
	$n_{ij}^t$	Number of track changes between stations $i$ and $j$$, i,j\in I\cup K\cup D$		$c_{kt}$	Transit fee at assembly center $k$$, \mathrm{CNY}/\mathrm{TEU}, k\in K$
	$n_{ij}^c$	Number of cross-border clearance between stations $i$ and $j$$, i,j\in I\cup K\cup D$		$T_t$	Time of track change, day/TEU
	$f_{kd}$	Frequency of the international freight trains from assembly center $k$ to destination $d$$, \mathrm{train}/d, k\in K$$, d\in D$		$T_c$	Time of cross-border clearance, day/TEU
	$B$	Number of formations, TEU		$r_{bt}$	Proportion reduction in the time of cross-border clearance after the implementation of blockchain technology
	$v_{ij}$	Traveling speed between station $i$ and $j$$, \mathrm{km}/d, i,j\in I\cup K\cup D$		$T_m$	Time of train formation, day/TEU
	$t_{id}^{\max }$	Maximum allowed transport time from origin $i$ o destination $d$$, \mathrm{day}, i\in I,d\in D$		$t_{il}$	Loading time at origin $i$$, \mathrm{CNY}/\mathrm{TEU}, i\in I$
	$t_{id}^d$	Time of direct transportation from origin $i$ to destination $d$$, \mathrm{day}, i\in I,d\in D$		$t_{du}$	Unloading time at destination $d$$, \mathrm{day}/\mathrm{TEU}, d\in D$
	$t_{id}$	Time of assembly transportation from origin $i$ o destination $d$$, \mathrm{day}, i\in I,d\in D$		$t_{kt}$	Transit time at assembly center $k$$, \mathrm{day}/\mathrm{TEU}, k\in K$
Parameters	${\theta }_1$	The weighting factors for fee		$H_i$	Handling capacity of origin $i$$, \mathrm{TEU}/\mathrm{day}, i\in I$
	${\theta }_2$	The weighting factor for the cost of time value		$H_d$	Handling capacity of destination $\mathrm{d}$$, \mathrm{TEU}/\mathrm{day}, d\in D$
	$\beta$	Average time value of goods, CNY/(day·TEU)		$H_k$	Handling capacity of assembly center $k$$, \mathrm{TEU}/\mathrm{day}, k\in K$
	$c_{g1}$	Fixed cost of railway transportation, CNY/TEU	Decision variables	$x_{ij}$	The value is 1 if the station $i$ is linked with $j$$, \mathrm{and} 0 \mathrm{otherwise}, i,j\in I\cup$$K\cup D$
	$c_{g2}$	The base price of railway transportation, CNY/(TEU·km)		$X_{ik}$	The value is 1 if the origin $i$ as allocated to the assembly center $k$$, \mathrm{and} 0 \mathrm{otherwise}, i\in I,k\in K$
	$c_{si}$	Price of storage at origin $i$, CNY/(TEU·day)		$x_b$	The value is 1 if blockchain technology has been implemented, 0 otherwise
	$c_{sk}$	Price of storage at assembly center $k$, CNY/(TEU·day)		$X_k$	The value is 1 if the candidate point $k$$\mathrm{is} \mathrm{selected} \mathrm{as} \mathrm{an} \mathrm{assembly} \mathrm{center}, 0 \mathrm{otherwise}, k\in K$
	$C_t$	Fee of track change, CNY/TEU		$x_{id}^{j_1{j}_2}$	The value is 1 if the transportation route from origin $i$ to destination $d$$\mathrm{passes} \mathrm{through} \mathrm{the} \mathrm{section} \left(j_1,j_2\right)$$, 0 \mathrm{otherwise}, i\in I,d\in D,$$j_1,j_2\in I\cup K\cup D$

.

4. Model Building

The bi-objective location-routing optimization model is constructed as shown in Equations (1)–(34). The two objectives of the model are to minimize the comprehensive cost and maximize average shipper satisfaction. The comprehensive cost consists of fees and time values. The average satisfaction of the shippers is determined by the maximum allowed and direct transportation times.

$$\min Z_1={\theta }_1(f_1+f_2+f_3+f_4+f_5+f_6)+{\theta }_2\beta (f_7+f_8+f_9+f_{10})$$

(1)

$$\max Z_2=f_{11}$$

(2)

$$f_1={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{\displaystyle \sum _{j_1\in I\cup K\cup D}{\displaystyle \sum _{j_2\in I\cup K\cup D}{q}_{id}(c_{g_1}+c_{g_2}{l}_{j_1{j}_2}}})x_{id}^{j_1{j}_2}}}$$

(3)

$$f_2={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}\left[t_{wi}{c}_{si}+{\displaystyle \sum _{k\in K}{c}_{sk}(t_{wk}^1+t_{wk}^2)X_{ik}}\right]}}$$

(4)

$$f_3={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{\displaystyle \sum _{j_1\in I\cup K\cup D}{\displaystyle \sum _{j_2\in I\cup K\cup D}{q}_{id}[C_t{n}_{j_1{j}_2}^t+C_c(1-x_b{r}_{bc})n_{j_1{j}_2}^c+C_m+x_b{c}_b]x_{id}^{j_1{j}_2}}}}}$$

(5)

$$f_4={\displaystyle \sum _{k\in K}{C}_k{X}_k}$$

(6)

$$f_5={\displaystyle \sum _{k\in K}{\displaystyle \sum _{d\in D}{f}_{kd}(B_{\max }-B)p{l}_{kd}}}{X}_k$$

(7)

$$f_6={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}(c_{il}+c_{du}+{\displaystyle \sum _{k\in K}{\mathrm{c}}_{kt}{X}_{ik}})}}$$

(8)

$$f_7={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{\displaystyle \sum _{j_1\in I\cup K\cup D}{\displaystyle \sum _{j_2\in I\cup K\cup D}{q}_{id}{l}_{j_1{j}_2}/v_{j_1{j}_2}{x}_{id}^{j_1{j}_2}}}}}$$

(9)

$$f_8={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}\left[t_{wi}+{\displaystyle \sum _{k\in K}(t_{wk}^1+t_{wk}^2)X_{ik}}\right]}}$$

(10)

$$f_9={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{\displaystyle \sum _{j_1\in I\cup K\cup D}{\displaystyle \sum _{j_2\in I\cup K\cup D}{q}_{id}[T_t{n}_{j_1{j}_2}^t+T_c(1-x_b{r}_{bt})n_{j_1{j}_2}^c+T_m]x_{id}^{j_1{j}_2}}}}}$$

(11)

$$f_{10}={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}(t_{il}+t_{du}+{\displaystyle \sum _{k\in K}{t}_{kt}{z}_{ik}})}}$$

(12)

$$f_{11}={\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}\max (\frac{t_{id}^{\max }-t_{id}}{t_{id}^{\max }-t_{id}^d},0)}}{q}_{id}/{\displaystyle \sum _{i\in I_o}{\displaystyle \sum _{d\in D}{q}_{id}}}$$

(13)

$$t_{id}={\displaystyle \sum _{j_1\in I\cup K\cup D}{\displaystyle \sum _{j_2\in I\cup K\cup D}\frac{l_{j_1{j}_2}}{v_{j_1{j}_2}}{x}_{id}^{j_1{j}_2}}}+t_{wi}+{\displaystyle \sum _{k\in K}(t_{wk}^1+t_{wk}^2)X_{ik}},\forall i\in I,d\in D$$

(14)

$$t_{wi}=⌊{\displaystyle \sum _{d\in D}{q}_{id}}/H_i⌋[{\displaystyle \sum _{d\in D}{q}_{id}}-(⌊{\displaystyle \sum _{d\in D}{q}_{id}}/H_i⌋-3)H_i]t_{il}/2{\displaystyle \sum _{d\in D}{q}_{id}},\forall i\in I$$

(15)

$$\begin{array}{l}{t}_{wk}^1=(⌊({\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}{X}_{ik}}}-H_k)/B⌋+1)({\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}{X}_{ik}}}-H_k-⌊({\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}{X}_{ik}}}-H_k)/B⌋\cdot {\displaystyle \frac{B}{2}})\\ \hspace{1em}\hspace{1em}\hspace{1em}\cdot \max \{0,{\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}{X}_{ik}}}-H_k\}/\max \{1,{\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}{X}_{ik}}}\cdot ({\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in D}{q}_{id}{X}_{ik}}}-H_k)\cdot {\displaystyle \sum _{d\in D}{f}_{kd}}\},\forall k\in K\end{array}$$

(16)

$$\begin{array}{cc}{t}_{wk}^2=\hfill & (B{f}_{kd}-{\displaystyle \sum _{i\in I}{q}_{id}{X}_{ik}})[(B{f}_{kd}-{\displaystyle \sum _{i\in I}{q}_{id}{X}_{ik}})(3-2B)+B(2B-1)]\cdot \min \{0,B{f}_{kd}-{\displaystyle \sum _{i\in I}{q}_{id}{X}_{ik}}\}\hfill \\ \hfill & /-\max \{1,2B{f}_{kd}{\displaystyle \sum _{i\in I}{q}_{id}{X}_{ik}}|B{f}_{kd}-{\displaystyle \sum _{i\in I}{q}_{id}{X}_{ik}}|\},\forall k\in K,d\in D\hfill \end{array}$$

(17)

$$\mathrm{s}.t.{\displaystyle \sum _{k\in \mathrm{K}}{\mathrm{X}}_{ik}}=1,\forall i\in \mathrm{I}$$

(18)

$${\displaystyle \sum _{i\in I}{\displaystyle \sum _{d\in \mathrm{D}}{q}_{id}{X}_{ik}\le H_k}}{X}_k,\forall k\in K$$

(19)

$$X_{ik}\le X_k,\forall i\in I,k\in K$$

(20)

$$x_{id}^{j_1{j}_2}\le x_{j_1{j}_2},\forall j_1,j_2\in I\cup K\cup D$$

(21)

$${\displaystyle \sum _{i\in I}{x}_{ij}=1},\forall j\in I$$

(22)

$${\displaystyle \sum _{i\in I}{\displaystyle \sum _{j\in I}{x}_{ij}}}-{\displaystyle \sum _{i\in I}{\displaystyle \sum _{j\in I}{x}_{\mathrm{ji}}}}=0$$

(23)

$$x_{ik}\le X_{ik},\forall i\in I,k\in K$$

(24)

$$x_{ki}\le X_{ik},\forall i\in I,k\in K$$

(25)

$$x_{ij}+X_{ik}+{\displaystyle \sum _{m\in K,m\ne k}{X}_{jm}}\le 2,\forall i\in I,j\in I,k\in K$$

(26)

$${\displaystyle \sum _{j\in S_i}{\displaystyle \sum _{d\in D}{q}_{jd}}}\le B,\forall i\in I$$

(27)

$${\displaystyle \sum _{i\in I}{q}_{id}{X}_{ik}}\le B{f}_{kd},\forall k\in K,d\in D$$

(28)

$$B\in (0,B_{\max }]$$

(29)

$$x_{id}^{j_1{j}_2}\in \left\{0,1\right\},\forall i\in I,d\in D,j_1,j_2\in I\cup K\cup D$$

(30)

$$x_{ij}\in \{0,1\},\forall i,j\in I\cup K\cup D$$

(31)

$$X_{ik}\in \{0,1\},\forall i\in I,k\in K$$

(32)

$$X_k\in \left\{0,1\right\},\forall k\in K$$

(33)

$$x_b\in \left\{0,1\right\}$$

(34)

Equation (1) is the minimization of comprehensive cost, which is the weighted sum of the fee and time value of goods. Fees include transportation, storage during the waiting process, constructing assembly centers, loss of formation downsizing, cargo loading and unloading, cross-border clearance, track change, and train formation. Time includes transportation, waiting, cargo loading and unloading, cross-border clearance, track change, and train formation. Equation (2) is the maximization of the average shipper satisfaction. The average satisfaction of shippers is obtained using Equation (13). In addition, the average satisfaction of shippers is 0 when the assembly transportation time is more than the maximum time allowed by the shipper. It is worth noting that the direct transportation time in Equation (13) does not include the waiting time at the origin station; thus $t_{id}^d\le t_{id}$, $\forall i\in I,d\in D$. Equation (3) is the transportation fee. Equation (4) is the storage fee during the waiting process at origin stations and assembly centers. Equation (5) is the cross-border clearance, track change, and train formation fees. When blockchain technology is not implemented, $x_b=0$. After blockchain technology implementation, $x_b=1$, and the cross-border clearance fee was reduced $r_{bc}$. Simultaneously, the cost of applying blockchain technology $c_b$ must be considered. Equation (6) is the construction cost of the assembly center. Equation (7) is the loss of formation downsizing. Equation (8) is the cargo loading and unloading fees. Equation (9) is the transportation time. Equation (10) is the waiting time at the origin station and the assembly center. Equation (11) is the time of cross-border clearance, track change, and train formation. After the implementation of blockchain technology, $x_b=1$, and the time for cross-border clearance will reduce $r_{bt}$. Equation (12) is the time of cargo loading and unloading. Equation (13) is the average satisfaction of shippers. Equation (14) is the assembly transportation time. Equation (15) is the congested waiting times at the origin station. Equation (16) is the congested waiting times at the assembly center. Equation (17) is the waiting time for transportation at the assembly center. Equation (18) is the uniqueness constraint for the assembly center. Equation (19) is the capacity constraint of the assembly center. Equation (20) shows that the origin station can be allocated to the assembly center only when it is selected. Equation (21) shows that cargo can only be transported between two stations when there is connectivity. Equations (22) and (23) show that the origin station can only be visited once. Equations (24) and (25) show that routes can only be linked if the origin station is allocated to the assembly center. Equation (26) shows that route connectivity is only possible between two origin stations if allocated to the same assembly center. Equation (27) is the capacity constraint of the international freight train. Equation (28) is the sending capacity constraint of the assembly center. Equation (29) is the constraint on the value of the number of formations. Equations. (30)–(34) are 0–1 value constraints.

5. Solving Algorithm

The multi-objective optimization algorithm NSGA II is used to solve the model. To improve the performance of the algorithm, opposition-based learning and dominance strength are introduced. An improved congestion calculation method and adaptive elite retention strategy replace the same steps in the algorithm. As a result of these improvements, an adaptive NSGA II algorithm (ANSGA II-OD) based on opposition-based learning and dominance strength is proposed, the flow is shown in Figure 4.

5.1. Opposition-Based Learning Mechanism

To reduce the probability of local convergence of the algorithm, an opposition-based learning mechanism is introduced. Opposite population information is used to increase the search space and improve the global search capability [43]. If $y_i=(y_{i1},y_{i2},\dots ,y_{in})$ is a feasible solution, then there is an opposite solution, as shown in Equation (35).

$${\tilde{y}}_{ij}=a_j+b_j-y_{ij},j=1,2,\dots ,n$$

(35)

where $a_j$ and $b_j$ are the upper and lower bounds of the values of the variable $j$, respectively. $n$ represents the dimensions of the solution vector.

A dynamic mechanism that considers the effect of opposite solutions on the speed of convergence of the algorithm is introduced. The jumping rate and probability of opposition-based learning were obtained by analyzing the situation of the population and individuals. It can then be determined whether an individual must undergo opposition-based learning. The latter population distribution is more stable and reasonable. Therefore, the learning probability should be reduced. In addition, the opposite solution should be better for the individual at the back of the non-dominated sorting. Thus, the probability of learning should be high. Based on the above analysis, the jumping rate and probability of opposition-based learning are given by Equations (36)–(38).

$$J{r}_e=\gamma {\displaystyle \frac{e_{\max }-e+1}{e_{\max }}}Jr$$

(36)

$$ol{p}_i=[ol{p}_{\max }-{\displaystyle \frac{(ol{p}_{\max }-ol{p}_{\min })e}{e_{\max }}}]{\displaystyle \frac{l\left(y_i\right)}{ml(P_e)}}$$

(37)

$$ol{p}_e=rand(ol{p}_{\min },ol{p}_{\max })$$

(38)

where $e$ denotes the current number of iterations, $J{r}_e$ denotes the jumping rate, $\gamma$ denotes a random number, and $\gamma >1$. $e_{\max }$ denotes the maximum number of iterations. $Jr$ denotes a smaller constant, and $Jr\in (0,0.4)$. $ol{p}_i$ denotes the probability of the opposition-based learning. $ol{p}_{\max }$ and $ol{p}_{\min }$ denote the upper and lower bounds of $ol{p}_i$. $l\left(y_i\right)$ denotes the number of the dominant layer of the individual $y_i$. $ml(P_e)$ denotes the average number of dominant layers of all individuals in the current population. $ol{p}_e$ denotes a random constant, as shown in Equation (38), which determines whether opposition-based learning is necessary.

5.2. Calculation of Dominance Intensity

An indicator for judging the priority of individuals, namely the dominance intensity ${\xi }_{y_i}$, is introduced. The lower the dominance intensity of the individual in the non-dominated set, the better the outcome for most optimization objectives of the individual. Thus, they are more likely to enter the next generation. The dominant intensity is expressed by Equation (39).

$${\xi }_{y_i}={\displaystyle \sum _{a\in M}\frac{f_a^{\max }-f_a(y_i)}{f_a^{\max }-f_a^{\min }}}$$

(39)

where the individual $y_i$ belongs to the non-dominated set. $M$ denotes the number of optimization objectives. $f_a(y_i)$ denotes the value of the $a\mathrm{t}\mathrm{h}$ optimization objective of an individual $y_i$. $f_a^{\max }$ and $f_a^{\min }$ denote the upper and lower values of the $a\mathrm{t}\mathrm{h}$ optimization objective, respectively.

5.3. Improved Calculation of Crowding Degree

The original method for calculating the degree of crowding focuses only on the distance between neighboring individuals. Consequently, individuals with significant differences in optimization objectives are less likely to be inherited. This is not conducive to ensuring the distribution of the solution set. As shown in Equation (40), this study adopts the crowding degree calculation method considering variance [44].

$${\eta }_{y_i}={\displaystyle \frac{{\lambda }_{y_i}}{{\left[\frac{1}{M}{\displaystyle \sum _{a\in M}{(|f_a^{y_{i+1}}-f_a^{y_{i-1}}|-\frac{{\lambda }_{y_i}}{M})}^2}\right]}^{-1}+1}}$$

(40)

where ${\lambda }_{y_i}={\displaystyle \sum _{a\in M}|f_a^{y_{i+1}}-f_a^{y_{i-1}}|}$. $f_a^{y_{i+1}}$ and $f_a^{y_{i-1}}$ denote the values of the $a\mathrm{t}\mathrm{h}$ optimization objective for individuals $y_{i+1}$ and $y_{i-1}$, respectively.

Each individual has three attributes: non-dominance rank, dominance intensity, and crowding degree. These three attributes are sequentially compared. This can result in a better distribution of better individuals at the Pareto boundary.

5.4. Adaptive Elite Retention Strategy

The original elite retention strategy has a fixed retention size, which is not conducive to convergence. In the early stages of evolution, the scale of elite individual retention should be controlled to search for random individuals. Later in the evolution, as the size of the non-dominated set increases, the number of retained elite individuals increases. This allows the algorithm to converge to the Pareto boundary quickly. The $e\mathrm{t}\mathrm{h}$ generation elite individual retention size ${\phi }_e$ is calculated using Equations (41) and (42).

$${\phi }_e=N\cdot {\alpha }_e$$

(41)

$${\alpha }_{e+1}={\alpha }_e\left[1+\ln \left(\rho +1\right)\right]$$

(42)

where ${\alpha }_e$ and ${\alpha }_{e+1}$ are the factors that influence the retention size of elite individuals in generations $e$ and $e+1$, respectively. To maintain the modest size of the elite individual, let ${\alpha }_e\in \left[0.2,0.8\right]$. $\rho$ denotes the ratio of the size of the non-dominated set to the population size $N$.

6. Case Analysis

6.1. Data Sources

The 43 Indo–China Peninsula stations are selected as the origin stations, and Kunming and Nanning in China are selected as the destination stations. The topological network of international freight trains and the serial numbers of the stations are shown in Figure 5. Through field research at the railway operation department and a global freight forwarding company, the cargo volume of the origin stations, transport miles, transport speed, and maximum allowed transport time are determined.

6.2. Parameter Settings

Considering the economic fundamentals, locational conditions, transportation demand, and other factors, 12 cities, including Hanoi, Ho Chi Minh, Phnom Penh, Savannakhet Town, Vientiane, Bangkok, Chun Pung, Kota Baru, Kuala Lumpur, Singapore, Mandalay, and Yangon, are selected as candidate points for the assembly center. Meanwhile, through field research and by referring to the previous literature [5,45], the main parameters are determined, as listed in

Table 2. Values of the main parameters.

Parameters	Value	Parameters	Value
${\theta }_1$	0.55	$B_{\max }$	55 TEU/train
${\theta }_2$	0.45	$p$	1.5 CNY/(TEU·km)
$\beta$	1000 CNY/(TEU·day)	$c_{il}$$, c_{kt}$	1500 CNY/TEU
$c_{g1}$	500 CNY/TEU	$c_{du}$	3000 CNY/TEU
$c_{g2}$	3.28 CNY/(TEU·km)	$T_t$	1 day/TEU
$c_{si}$$, c_{sk}$	Singapore, 200 CNY/(TEU·day) Stations in Malaysia, Thailand, and Vietnam, 150 CNY/(TEU·day) Stations in Cambodia, Laos, and Myanmar, 100 CNY/(TEU·day)	$T_c$	1 day/TEU
$c_{sd}$	150 CNY/(TEU·day)	$T_m$	0.5 day/TEU
$C_t$	1250 CNY/TEU	$t_{il}$$, t_{kt}$$, t_{du}$	0.5 day/TEU
$C_c$	1200 CNY/TEU	$H_i$	100 TEU/day
$C_m$	80 CNY/TEU	$H_k$	1000 TEU/day
$C_k$	4,000,000 CNY

.

6.3. Comparison of Algorithm Performance

The process of solving the optimization scheme using the ANSGAII-OD is shown in Algorithm 1. After inputting the topological network and related parameters of the international freight train, the location-routing optimization scheme of assembly centers is output.

Algorithm 1. The implementation process of the algorithm

Input values of the main parameters and topological network of transportation network

1. Set the population size and dimension, and initialize the population according to the constraints (18)–(34),

2. Set the maximum number of iterations $e_{\max }$, let the current number of iterations $e=1$,

3. Non-dominant ranking of individuals in the initial population,

4. Calculating the jumping rate $J{r}_e$ according to Equation (36), and determine whether $J{r}_e<Jr$,

4.1 If $J{r}_e\ge Jr$,

4.1.1 Calculating the dominant intensity ${\zeta }_{y_i}$ according to Equation (39),

4.1.2 Calculating the crowding degree ${\eta }_{y_i}$ according to Equation (40),

4.1.3 Comparing the non-dominance rank, dominance intensity, and crowding degree sequentially, and sorting the individuals,

4.1.4 Calculating the $e\mathrm{t}\mathrm{h}$ generation elite individual retention size ${\phi }_e$ according to Equations (41) and (42),

4.1.5 Merging populations, let $e=e+1$, turn into 5,

4.2 Else if $J{r}_e<Jr$,

4.2.1 According to the non-dominant ordering, the population was divided into the non-dominant set $Q_1$ and other individual set $Q_2$,

4.2.2 Traverse through all individuals in set $Q_2$, and calculating the probability of opposition-based learning according to Equations (37) and (38),

4.2.3 Determine whether $ol{p}_i>ol{p}_e$,

4.2.4 If $ol{p}_i>ol{p}_e$, calculating the opposite solution according to Equation (35) and constraints (18)–(34), and add it into set $Q_3$,

4.2.5 Merging the set $Q_1$ and $Q_3$,

4.2.6 Non-dominant ordering of individuals in the merged set, then turn into 4.1.1.

5 Determine whether $e=e_{\max }$,

5.1 If $e<e_{\max }$, turn into 3,

5.2 Else if, end.

Output optimized solution

To compare the performances of the algorithms, the ANSGAII-OD and NSGAII algorithms are used to solve the location-routing optimization model of this paper. The convergence speed, number of Pareto solutions, spacing, HRS, and PR of the two algorithms are compared. Programming was performed in Python 3.8.7 on a computer with an Intel i7-9700K CPU@3.60 GHz, 16 G RAM, and a 64-bit Windows 10 operating system. The convergence of the two algorithms is shown in Figure 6, and the other indicators are listed in

Table 3. Comparison of the performance of ANSGA II-OD algorithm with NSGA II algorithm.

Algorithms	Running Time(s)	Number of Pareto Solutions	Spacing	HRS	PR	IGD
ANSGA II-OD	45.34	20	55.34	5.78	0.5	3.42 × 10−3
NSGA II	136.76	12	108.59	13.49	0.37	8.93 × 10−2

.

Figure 6 and Table 3 show that the ANSGA II-OD algorithm has a larger PR and better time complexity. This algorithm converges faster, and more efficiently generating Pareto solutions. Meanwhile, the Spacing and HRS of the ANSGAII-OD algorithm are smaller. The obtained solution set is more uniformly distributed and effective in solving such problems. In addition, the IGD of the ANSGAII-OD algorithm is also smaller than NSGA II, which means the solution obtained by ANSGAII-OD is closer to the real Pareto boundary.

6.4. Results

6.4.1. Optimization Results When Blockchain Technology Is Not Applied

The results of location-routing optimization when blockchain technology was not applied are shown in Figure 7 and

Table 4. Location-routing optimization result.

Assembly Center	Destination Station	Transportation Route	Frequency (Train/d)
3	1	3→7→6→5→4→3→1	1
	2	3→7→6→5→4→3→2	2
8	1	8→10→9→8→1	1
	2	8→10→9→8→2	1
14	1	14→16→15→14→1 14→11→12→13→14→1	2
	2	14→16→15→14→2 14→11→12→13→14→2	1
21	1	21→24→22→23→21→1 21→17→18→19→20→21→1 21→32→30→28→27→26→25→21→1 21→35→34→33→31→29→21→1	2
	2	21→24→22→23→21→2 21→17→18→19→20→21→2 21→32→30→28→27→26→25→21→2 21→35→34→33→31→29→21→2	2
39	1	39→36→37→39→1 39→42→41→45→44→43→40→39→1	1
	2	39→36→37→39→2 39→42→41→45→44→43→40→39→2	1

. Five candidate points were selected as assembly centers: Hanoi, Phnom Penh, Vientiane, Bangkok, and Mandalay. The number of formations corresponded to 53 TEU/train. The frequencies of international freight trains from assembly centers to destination stations are listed in Table 4. The two optimization targets, comprehensive cost, and average shipper satisfaction were 2.84 billion CNY and 0.73, respectively.

The comprehensive cost under the direct transportation mode was 3.54 billion CNY, which can be reduced by 19.77% using the assembly transportation mode. Concurrently, over 70% of international freight trains were less than 40 TEU/train in the direct transportation mode. This results in a waste of transportation resources. If the number of formations is set at 50 TEU/train, the frequencies of more than 75% of international freight trains will be less than one train/week. This results in longer waiting times. Therefore, the assembly transportation mode can solve this problem.

6.4.2. Optimization Result After Applying Blockchain Technology

Referring to previous studies, combined with field research, we set the proportion of reduction of customs clearance fees after the implementation of blockchain to $r_{bc}=0.3$, the proportion of reduction of customs clearance time to $r_{bt}=0.4$, and the cost of blockchain technology as $c_b=600$ CNY/TEU [46,47]. The location routing optimization results after applying the blockchain are presented in Figure 8 and

Table 5. Location-routing optimization result.

Assembly Center	Destination Station	Transportation Route	Frequency (Train/Week)
3	1	3→7→6→5→4→3→1 3→11→10→8→3→1 3→19→20→9→3→1	2
	2	3→7→6→5→4→3→2 3→11→10→8→3→2 3→19→20→9→3→2	2
14	1	14→12→13→14→1 14→16→15→14→1 14→18→17→14→1	2
	2	14→12→13→14→2 14→16→15→14→2 14→18→17→14→2	2
21	1	21→27→26→25→21→1 21→37→36→39→38→21→1 21→24→22→23→21→1 21→42→41→40→45→43→44→21→1	2
	2	21→27→26→25→21→2 21→37→36→39→38→21→2 21→24→22→23→21→1 21→42→41→40→45→43→44→21→2	2
32	1	32→31→29→28→30→32→1 32→35→34→33→32→1	1
	2	32→31→29→28→30→32→2 32→35→34→33→32→2	1

. Four stations—Hanoi, Vientiane, Bangkok, and Kuala Lumpur were selected as the assembly centers. The number of formations corresponded to 51 TEU/train. The frequencies of the international freight trains are listed in Table 5.

To analyze the effectiveness of blockchain, the comprehensive cost and average satisfaction of shippers before and after the implementation of blockchain technology were compared, as shown in

Table 6. Comparison of optimization objectives before and after the implementation of blockchain.

	Scenarios	Before the Implementation of Blockchain	After the Implementationm of Blockchain
Optimization Objectives
Comprehensive cost/Billion CNY		2.84	2.61
Average satisfaction of shippers		0.73	0.81

. When blockchain technology was applied, the comprehensive cost decreased by 8.10%, and the average satisfaction of shippers increased by 10.35%. Table 6 shows that blockchain was significant in reducing costs and improving service quality.

As we all know, some emerging technologies, such as the Internet of Things (IoT), can assist interested participants to efficiently locate, monitor, and share scarce resources to improve their productivity. Digital technologies can support operators of international freight trains to develop new plans to diversify their business and increase the efficiency of their production [48]. Moreover, they can facilitate the automation of operational processes, such as automated tracking and tracing of goods and automated processing and delivery of documents. Terminals at border stations can assist in the authentication of incoming containers and trucks through RFID tags. Detailed data about incoming trains can be effectively utilized by the managers to speed up the process of customs clearance and optimize the plan for the use of yard space and track-changing equipment [49]. The IoT network can ensure the real-time sharing of information on transportation between owners, carriers, railway operators, and other stakeholders to achieve the advancement of some operations as well as carry out related business. However, IoT and cloud computing are both centrally governed technologies and fall short in terms of security and privacy of the data. As a contrast, blockchain is a decentralized and distributed technology that can provide stakeholders with an operational environment, which is trusted, auditable, secure, and transparent.

Table 7. Blockchain’s features and its role in the transportation of international freight train.

Feature	Role in the Transportation of International Freight Train
Transparency	Data and transactions are stored in a decentralized way. All nodes in the blockchain record the same information. The transparency of freight transportation ensures trust between participants.
Availability	The distributed and shared nature of the blockchain ensures that authorized users always have access to the data and transactions stored. This availability can help managers improve the efficiency of decision-making during the transportation of international freight trains.
Integrity and authenticity	Asymmetric key encryption and timestamps ensure the integrity and authenticity of on-chain data and transactions. It in turn strengthens the mutual trust of the participants.
Audit and Data Provenance	Data and transactions in the blockchain system cannot be tampered with. This advantage can assist stakeholders in assessing the goods and their transportation to make decisions quickly.
Authorization	Blockchain assures the authorization of international freight train participants. Every participant has its own unique identifier, and only authorized nodes are allowed to carry out transactions. This ensures the security and privacy of data.

shows the advantages of the blockchain platform. Some operators in the transportation industry are actively promoting the exploration of blockchain. As a leader of shipping companies, Maersk’s blockchain platform, namely TRADELENS, officially launched commercial operations in December 2018 and publishes 2 million events on the platform every day. We can see that blockchain-based transportation systems can support the secure exchange of data information between participants for efficient and collaborative decision-making. For instance, blockchain can support the digitization of documents to help stakeholders access data and transactions online, rather than obtain information through the physical exchange of shipping documents [50]. Thus, participating organizations can access the logistics data in real time and make timely decisions. Therefore, the time and cost consumption during transportation will be reduced. Westergaard-Kabelmann’s case study has shown that the implementation of blockchain at the largest ports of India, namely Nhava Sheva, has reduced the processing time for import documents by 32 h [51]. While the comprehensive cost is reduced, the average time of receipt will be earlier, so that the average satisfaction will increase. As can be seen, applying blockchain technology is conducive to the sustainable development of international freight trains and the prosperity of global trade.

6.5. Sensitivity Analysis

To further study the impact of blockchain technology on the comprehensive cost and service quality, we set the proportion of reduction in customs clearance fees after applying blockchain technology $r_{bc}\in$ [0.2, 0.6], the proportion of reduction in customs clearance time $r_{bt}\in$ [0.2, 0.6] and the cost of blockchain technology $c_b\in$ [200, 1000] CNY. The changes in the comprehensive cost and average satisfaction of shippers were recorded, as shown in Figure 9 and Figure 10, respectively.

As shown in Figure 9a,b, the comprehensive cost is negatively related to $r_{bc}$. With the deep integration of blockchain technology and railway systems, information interoperability among the participants of the international liner was improved. Therefore, the proportion of reductions in fees for cross-border clearance after the implementation of blockchain technology $r_{bc}$ will increase. Thus, the fee for cross-border clearance was further reduced. Simultaneously, fees for cross-border transportation are reduced. As a result, some international freight trains select a cross-border route with shorter mileage and faster speed instead of an internal route. This further reduces overall costs. Figure 9c shows that the comprehensive cost is negatively correlated with $r_{bt}$. The operating sector and other relevant organizations should actively promote the implementation of blockchain technology. The advantages of blockchain in data security sharing have been fully utilized. In addition, the customs clearance time was reduced. Thus, the efficiency of cross-border transportation is significantly enhanced. This will further reduce comprehensive costs. In addition, as shown in Figure 9d, the comprehensive cost increased with increasing $c_b$. If the hardware and software costs for blockchain technology increase, the comprehensive cost also increases. Therefore, when applying blockchain technology, it is worth controlling its cost.

As shown in Figure 10a,b, the average satisfaction of shippers first decreased with $r_{bc}$ and stabilized when $r_{bc}$ ≥ 0.35. When $r_{bc}\in$ [0.2, 0.35), as the increase of the value of $r_{bc}$, more trains select cross-border transportation routes owing to the reduction of the cross-border transportation fee. The time of assembly transportation tends to increase. As a result, the average satisfaction of shippers decreases. When $r_{bc}\in$ [0.35, 0.60], the change in routes of international freight trains is finished. Therefore, the average satisfaction of shippers tends to stabilize even if $r_{bc}$ increases gradually. From Figure 10c, we can see that the average satisfaction of shippers is positively correlated with $r_{bt}$. The larger the $r_{bt}$, the more prominent the role of blockchain technology in shortening the customs clearance time of the international freight train. Furthermore, the time for assembly transportation will be shorter. As a result, the average satisfaction of shippers subsequently increases. $c_b$ has a small effect on the average shipper satisfaction. No clear pattern of change is observed in this optimization objective as $c_b$ increases. This is because the increase in the cost of blockchain technology mainly changes the transportation fee and has no significant effect on the timeliness of the international freight train.

7. Conclusions

A bi-objective location-routing optimization model was constructed to optimize the location of the assembly centers of international freight trains. The two optimization objectives are to minimize the comprehensive cost and maximize the average satisfaction of the shippers. Three parameters, namely, the proportion of reduction in customs clearance fees after the implementation of blockchain technology, the proportion of reduction in customs clearance time after the implementation of blockchain technology, and the cost of blockchain technology, were introduced into the model. The transport network of international freight trains between 43 stations in the Indo–China Peninsula and two stations in Southern China, Kunming, and Nanning, was examined as a case study. A scheme for the assembly transportation of international freight trains before and after the implementation of blockchain technology was developed. The effects of the three parameters on the two optimization objectives were also analyzed.

The case study showed that the assembly transportation mode could effectively reduce the comprehensive cost and improve the utilization of transport resources. Moreover, blockchain technology can significantly reduce comprehensive costs and improve shippers’ average satisfaction. In addition, sensitivity analysis shows that after the implementation of blockchain technology, the comprehensive cost decreases with an increase in $r_{bc}$ and $r_{bt}$, but increases with $c_b$. The average satisfaction of shippers first decreased with the increase of $r_{bc}$ and stabilized when $r_{bc}\ge 0.35$. An increase in $r_{bt}$ enhanced the shippers’ average satisfaction. In contrast, $c_b$ exhibited no significant effect on the average satisfaction of shippers. The application of blockchain technology is significant in the operating sectors of international freight trains. Thus, the integration of blockchain technology with existing transport systems should be promoted. The advantages of blockchain technology in facilitating information interconnection must be fully exploited. This effectively reduces costs and improves efficiency.

The limitations of this study should be addressed in future research. In the optimization model, only the impact of blockchain technology implementation on cross-border customs clearance was considered. However, insufficient attention has been given to the effect of blockchain technology on other parts of the supply chain. These factors include the advantages of blockchain technology in securing supply chain transactions and international settlements. Generally, each part of the supply chain interacts with the others. Customers are more concerned with the total cost and time of the supply chain. Future research should focus on the impact of blockchain technology on the entire international supply chain.