Introduction

Manufacturing and logistics have reciprocal attachments with each other. Advanced decision and intelligent technologies bring them transformational changes (Chien et al. 2012). Since 1980s, Just-in-time (JIT) manufacturing created by Toyota, has been viewed as a magic weapon to reduce waste and cost for manufacturers. However, it does not rule out a buffer stock operation in order to offset the disaccord between the level production and order fluctuation. The supplies of parts are leveled by a method of order split. Therefore, the delivery of auto parts from suppliers to production lines generally is required to be in a way of small-batch and high-frequency. The Milk-run mode which originates from the dairy industry practice where one tanker collects milk every day from several dairy farmers for delivery to a milk processing firm, is presently a very popular inbound logistics solution for auto manufacturers. It actually meets the demands of production activities and low inventory. However, through a deep investigation on the components of total cost, it can be found that the milk-run mode inevitably leads to a high ratio of the transport cost. Low inventory and high-cost transportation constitute the contradiction of JIT production in auto manufacturers, the balance between them has become the primary concern of the auto parts logistics managers.

Milk-run mode is a modern logistics mode with the characteristics of small batch, multiple frequency, and time limitations. However, this mode only works well under the condition that the auto-part supplier is close to the main engine plant. As for the farther supplier are concerned, they need to find a warehouse near the main engine plant to ensure a full supply, where a high inventory may be kept. Generally, many auto manufacturers will outsource the Milk-run logistics service to the third-party logistics (TPL) companies in actual operation. Regrettably, the TPL companies only implement the collection plan of auto manufacturers without a deep integration of logistics resource and business optimization. Next, most of the researches on Milk-run mode start from a view of route optimization and scheduling, which usually consider the co-existence of multiple constraints such as time window, vehicle loading and parking lot (Wang and Jiang 2009;Nemoto et al. 2010;Kaneko and Nojiri 2008; Anne and Brois 2018; Luo and Liu 2020; Vidal et al. 2020). However, fewer interests are shown on the integration on the procedures of collection and delivery. Therefore, logistics efficiency cannot be substantially improved with only partial optimization in its strategy or technology and with less breakthrough and innovation on its operation mode.

The main feature of Drop and Pull transport is the separability of tractors and trailers. Through the mutual coordination between tractors and trailers, resources can be effectively saved and transportation can be completed efficiently (Semet and Taillard 1993; Li 2004). It is known that as early as in the 1960s, the Drop and Pull transport mode have been widely used in Europe and the United States. The practice in food logistics (Caramia and Guerriero 2010), ocean logistics (Hsu and Hsieh 2007; Song and Dong 2012), steel logistics (Tang and Li 2009) and other logistics industries have made great benefits. The Drop and Pull transport mode has been very popular in developed countries, and the researches on which are mainly focused on vehicle scheduling (Wei et al. 2011; Emde and Michel 2017; Ladier and Alpan 2018) and path optimization (Canrong et al. 2014; Anurag and Pei-Chann 2014; Xiao et al. 2017). Although started late in China, it has developed quickly in both theories and practices with the encouraging policies in recent years (He et al. 2012; Li et al. 2014).

The purpose of this study is to break through the limitations of the existing inbound logistics mode, and establish a new inbound logistics mode that integrates Milk-run collection and Drop and Pull delivery. It is attempted to suggest an innovative mode with the characteristics of low-carbon and high efficiency, thus promote the innovation of logistics management in auto manufacturing enterprises.

Establishment of an integrated inbound logistics mode combining milk-run collection with drop and pull delivery and LNG vehicle

At present, in Europe, the United States and other developed countries, Drop and Pull transport has become the mainstream mode of transport, in which semi-trailer transport volume accounts for more than 70% of the total transport volume. The Chinese government has launched projects on Drop and Pull transport since 2011. With the beneficial practices in these years, a Drop and Pull transportation network has been built. It was proven that Drop and Pull transport can reduce fuel consumption by 20–30%, and greatly reduce the energy consumption and pollution emissions of the transport industry. Therefore, Drop and Pull transport is fully in conformity with the basic national policy of building a conservation-oriented society in China.

With the adjustment of China’s energy strategy, the application of LNG technology has been listed in the national key projects. Compared with gasoline and diesel vehicles, exhaust emissions from LNG vehicles are reduced by 85%, in which CO2 emission is reduced by 26%, CO emission 97% and NOX emission 35%, and no sulfur compounds, lead, benzene and other carcinogens are generated. Therefore, LNG has the characteristics of safe, environmental friendliness, economy and energy. A survey of car carrier in China disclosed the differences between diesel vehicle and LNG vehicle on fuel consumption, gas consumption and cost difference, as were summarized in Table 1. Although, the price of LNG truck is nearly 100 thousand Yuan higher than that of diesel truck, the difference may be offset by the reduction of fuel cost in one to two years. Additionally, a large amount of diesel truck will be withdrawn from the market with the imminence of implementation on National emission standard for motor vehicle pollutants in the sixth stage. Therefore, LNG vehicle will get a great market opportunity in logistics industry.

Table 1 Comparison of diesel and LNG vehicle
Full size table

Through technological innovation and mode innovation, we can bring many beneficial changes such as greenhouse gas emissions reduction, better economic benefits and environmental protection, thus achieve the goal of low-carbon economy. The Drop and Pull transport is suitable for the medium range and high frequency transportation operations, Milk-run collection mode is suitable for small batch, intensive transport, and LNG vehicles are economical and environment beneficial. If we build a new low-carbon and efficient inbound logistics mode integrating these advantages together, it would be full of expectation. With these considerations, the integrated inbound logistics (IIL) mode including milk-run collection, Drop with Pull delivery and LNG vehicles, is therefore proposed. The flow chart of IIL mode is shown in Fig. 1.

Fig. 1
figure 1

Illustration of IIL mode

Full size image

For the convenience of management, the auto manufacturer usually outsources logistics to a third part logistics (TPL) provider. So those TPL providers who are expert at integrated logistics would have the opportunity to run the IIL mode. In the actual implementation of IIL mode, the main engine manufacturer, auto parts supplier and the TPL provider sign a tripartite cooperation agreement. In addition, the business activities within/outside of the main engine factory are run through a supply chain management information system, which generally includes many applications such as ERP, logistics management, supplier management and customer management. The main engine factory makes the production plans which are shared with auto parts suppliers and TPL provider. Following by, The TPL provider makes the integration logistics solution through intelligent scheduling methods and takes charge of the collection and delivery of auto parts. The distribution center may be held exclusively by the auto manufacturer or joint held by the auto manufacturer and TPL provider. The milk-run collection is taken between the distribution center and auto parts suppliers. The drop and pull delivery is run between the distribution center and the main engine factory.

Compared with traditional milk-run mode, the IIL mode has three important features. First, it proposes a new concept of drop and pull delivery; Second, it is an integrated mode in which milk-run collection with drop and pull delivery are considered in the same systematic frame; Third, LNG vehicle are put into use. It may be deduced that the IIL mode is able to reduce empty vehicle transportation, energy consumption and exhaust emissions. It is essentially a high-clean, low-carbon and economical mode of production.

Mathematical model of the IIL mode

Problem description

The implementation process of IIL mode can be divided into two sub-processes. A sub-process is Milk-run collection in which the LNG trucks collect auto part from suppliers along different closed-loop routs. Generally, the routs are carefully arranged by the TPL provider according to production plans and vehicle conditions. The truck starts to run from distribution center for the collection. If the demand for auto parts from a supplier is up to the full truck load, the full loaded vehicle may return to distribution center, or even directly to main engine factory. If it does not exceed the limit, the vehicle will continue to collect auto parts at next supplier along the route. When it is full loaded, the truck returns to distribution center, and the auto parts are unloaded. A part of them may be temporarily stored, some may be reloaded on a trailer through cross-dock way. In order to ensure the stability of production, the truck must return at the regulated time. Next, the truck may begin next collection loop with empty unit load devices. Due to the complexity of demand, many trucks are arranged to collect auto parts along different routes. For the convenience of routing planning, suppliers will be classified into many groups according to geographical position and demand.

The other sub-process is drop and pull delivery. The auto parts will be loaded in advance on a trailer according to the production plan. So the milk-run collection of auto parts must be finished before the trailer’s loading time. Then a tractor towing a full loaded trailer departs from distribution center to main engine factory. After the trailer is detached, the tractor will return to distribution center, towing another trailer loaded with empty unit load devices. Both the sub-processes are linked cohesively under considerations of quantity, batch and time window.

In the IIL mode, the Milk-run collection and the drop and pull delivery are linked up together. A systematic planning on IIL mode actually is a complex optimization problem which involves vehicle rout problem with time window (VRPTW) and scheduling problem. In order to make the IIL mode feasible in practice, we build a mathematical model with intelligent optimization algorithm. The IIL model in this study takes into account the actual operation conditions including circular collection, time window and Drop and Pull delivery. The optimization objective is to minimize the total cost.

Assumptions

Considering the actual processes of inbound logistics, several assumptions are made as follows:

  1. (1)

    Only one distribution center is set up near the engine factory, and its location is known.

  2. (2)

    An auto parts supplier is only served by a vehicle one time; A supplier can only be allocated into a collection path. Every auto parts supplier has a time window for collection. If it doesn’t arrive within the prescribed time window, a vehicle will be considered as service failure.

  3. (3)

    Only a unique main engine factory exists, and its location is known.

  4. (4)

    The LNG trucks with medium size are used in the Milk-run collection sub-process. Whereas, large sized LNG trucks with tractor and trailer are used in the delivery sub-process. It is stipulated that the total weight of auto parts on each route should not exceed the truck’s load capacity. The volume limits of trucks are not taken into account.

  5. (5)

    The distance between arbitrary two places within suppliers, distribution center and main engine factory is expressed by the linear distance, regardless of the actual route situation.

  6. 6)

    Since either the distribution center or the main engine factory usually locates in the suburbs, large freight vehicles may not be limited by the urban road transportation regulations; Moreover, LNG vehicles have more environmental advantages than diesel trucks. Thus, the drop and pull delivery with LNG vehicles is feasible in practice.

  7. (7)

    The demands of auto parts are shared between main engine factory, distribution center and suppliers through a logistics management information system.

  8. (8)

    There are necessary service facilities provided for LNG vehicles such as parking lot, filling station and maintenance shop.

Model formulations

The following notations are defined as follows

Indices and parameters

\( i \) :

Index of collection points, \( i = 0,1,2, \ldots ,n \)

\( j \) :

Index of collection points, \( j = 0,1,2, \ldots ,n \)

\( k \) :

Index of vehicles, \( k = 0,1,2, \ldots ,K \)

\( V_{H} \) :

Index of main engine factory, \( V_{H} = 0 \)

\( V_{C} \) :

Index of distribution center, \( V_{C} = 1 \)

\( V_{S} \) :

Index of suppliers, \( V_{S} = 2,3,4, \ldots ,n \)

\( E_{1} \) :

The total mileage in the first stage

\( E_{2} \) :

The total mileage in the second stage

\( R_{1} \) :

The number of LNG trucks used at the first stage

\( R_{2} \) :

The number of LNG tractors used at the second stage

\( q_{i} \) :

The weight of auto parts comes from the supplier, \( i \in V_{S} \)

\( Q \) :

The main engine factory’s demand

\( g_{1} \) :

The LNG truck’s loading capacity

\( g_{2} \) :

The LNG tractor’s loading capacity

\( M \) :

A large constant

\( a_{i} \) :

The earliest service time at point \( i \)

\( b_{i} \) :

The latest service time at point \( i \)

\( f_{k2} \) :

The delivering frequency in the second stage

\( d_{ij} \) :

The distance between point \( i \) and point \( j \)

Variables

  • \( x_{ijk} = \left\{ {\begin{array}{*{20}l} 1 \hfill & {If\;vehicle\;k\;runs\;from\;point\;i\;to\;point\;j} \hfill \\ 0 \hfill & {Otherwise} \hfill \\ \end{array} } \right. \)

  • \( y_{ik} = \left\{ {\begin{array}{*{20}l} 1 \hfill & {If\;vehicle\;k\;serves\;point\;i} \hfill \\ 0 \hfill & {Otherwise} \hfill \\ \end{array} } \right. \)

\( t_{arr.ik} \) :

The vehicle’s arrival time at point \( i \)

\( t_{dep.ik} \) :

The vehicle’s leaving time at point \( i \)

\( t_{ij} \) :

The vehicle’s driving time from point \( i \) to point \( j \)

\( W_{r} \) :

The time that a tractor stays at the distribution center

Create an undirected graph \( G = \left( {V,E} \right) \), the \( V = \left\{ {V_{H} ,V_{C} ,V_{S} } \right\} \), \( V_{H} = \left\{ 0 \right\} \) represents main engine factory, \( V_{C} = \left\{ 1 \right\} \) represents distribution center, \( V_{S} = \left\{ {2,3,4, \ldots ,n} \right\} \) represents suppliers; \( E = E_{1} \cup E_{2} \) represents the routes.

Define that the Milk-run collection occurs at the first stage and the Drop and Pull delivery continues at the second stage. Set vehicle set \( \left( {R_{1} ,R_{2} } \right) \), \( R_{1} \) are LNG trucks using at the first stage, \( R_{2} \) are LNG tractors using at the second stage.

In terms of the above notations, the IIL model can be formulated as follows:

$$ {\text{min }}Z = \mathop \sum \limits_{i,j \in V} \mathop \sum \limits_{k \in E} x_{ijk} d_{ij} $$
(1)
$$ \mathop \sum \limits_{{i \in V_{S} }} x_{1ik} = 1,\forall k \in R_{1} ,i \ne 0 $$
(2)
$$ \mathop \sum \limits_{{i \in V_{S} }} x_{i1k} = 1,\forall k \in R_{1} ,i \ne 0 $$
(3)
$$ \mathop \sum \limits_{{k \in R_{1} }} y_{ik} = 1,i \in V_{s} $$
(4)
$$ \mathop \sum \limits_{{i \in V_{C} {\bigcup }V_{S} }} x_{ijk} = y_{jk} ,\forall k \in R_{1} , j \in V_{S} $$
(5)
$$ \mathop \sum \limits_{{j \in V_{C} {\bigcup }V_{S} }} x_{jik} = y_{ik} ,\forall k \in R_{1} , i \in V_{S} $$
(6)
$$ \mathop \sum \limits_{{i \in V_{C} {\bigcup }V_{S} }} \mathop \sum \limits_{{k \in R_{1} }} x_{ijk} \ge 1 ,\forall i \in V_{S} $$
(7)
$$ \mathop \sum \limits_{{i \in V_{C} {\bigcup }V_{S} }} x_{ijk} = \mathop \sum \limits_{{j \in V_{C} {\bigcup }V_{S} }} x_{jik} ,\forall k \in R_{1} , i,j \in V_{S} $$
(8)
$$ x_{01k} = 1,\forall k \in R_{2} $$
(9)
$$ x_{10k} = 1,\forall k \in R_{2} $$
(10)
$$ y_{0k} = 1,\forall k \in R_{2} $$
(11)
$$ \mathop \sum \limits_{{i \in V_{S} }} q_{i} y_{ik} \le g_{1} , \forall k \in R_{1} $$
(12)
$$ \mathop \sum \limits_{{i \in V_{S} }} q_{i} \ge Q $$
(13)
$$ f_{k2} g_{2} \ge Q $$
(14)
$$ a_{i} \le t_{arr.ik} \le b_{i} $$
(15)
$$ 0 \le t_{dep.ik} - t_{arr.ik} \le b_{i} - a_{i} $$
(16)
$$ t_{{dep.ik_{1} }} + t_{ij} - M\left( {1 - x_{{ijk_{1} }} } \right) \le t_{{arr.jk_{1} }} , \left\{ {i,j \ne 0} \right\} , \forall k_{1} \in R_{1} $$
(17)
$$ t_{{arr.1k_{1} }} \le t_{{arr.1k_{2} }} $$
(18)
$$ t_{{arr.1k_{2} }} + W_{r} + t_{01} \le t_{{arr.0k_{2} }} $$
(19)

Objective function (1) minimizes the total mileage which vehicles run at all the stages. Constraints (2) and (3) determine the starting point and ending point of Milk-run network. Constraint (4)–(6) ensure that each supplier can only be serve by one vehicle; Constraint (7) ensures that all suppliers are traversed; Constraint (8) indicates that a vehicle arrives at a supplier and will leave; Constraint (9)–(10) determine the starting point and ending point of Drop and Pull delivery network; Constraint (11) ensures that only one route link the main engine factory and the distribution center.

Constraint (12) indicates that the total amount of auto parts which a truck collects on a single path must not exceed its load capacity; Constraint (13) regulates the main engine factory’s demand can be met by its suppliers; Constraint (14) shows the load capacity constraint of the tractor.

Constraint (15)–(16) indicate that the arrival time should be within the receiving service period and the operation time should not exceed the receiving time period, respectively; Constraint (17) ensures the sequence that a truck collects auto parts; Constraint (18) ensures that the time that truck arrive at the distribution center is not later than the tractor’s arrival time; Constraint (19) ensures that the tractor’s arrival time at the distribution center plus the time of waiting, processing and delivery is no longer than the arrival time at the main engine factory.

Case study

Business information of AJ logistics company

AJ, founded in 2000, is a famous auto logistics company. It owns about 17,000 employees, more than 50 distribution centers, 4.44 million square meters warehouse, and a widespread distribution network covering 562 cities in China.

It mainly provides logistics services for automobile manufacturers and parts manufacturers. Its business includes three major segments: auto parts logistics, vehicle logistics and port logistics. Its customers mainly include Shanghai Volkswagen, Shanghai GM, FAW Toyota, SGMW, GTMC, BYD and most of main engine factories in China.

In recent years, the Milk-run mode which has been widely used in inbound logistics, has brought significant changes in supply chain management for auto manufacturing enterprises. However, AJ found that even enacting Milk-run mode, the logistics cost is relatively high, where the transportation costs still accounts for a rate of 44%. How to reduce costs scientifically and effectively is yet to be solved.

Solution under IIL mode

Since the auto parts suppliers are numerous and scatter geographically, it had better classify them into several areas though clustering analysis method. The main indices of classification are geographic location, transport mileage and order requirements. By this way, four areas are divided into from \( {\text{D}}_{1} \) to \( {\text{D}}_{4} \), respectively, as are shown in Table 2.

Table 2 Classification of areas (units: kilometer)
Full size table

For each of the four areas, a group of vehicles are arranged to take the Milk-run collection, respectively. The arrangement can be transformed into vehicle routing optimization problem with time window (VRPTW). Therefore, the IIL mode is actually a combination of VRPTW and drop and pull delivery. Since the computation time grows exponentially as the problem’s size increases, the IIL model was solved by the genetic algorithm under the programming environment of Matlab. The following parameters are adopted: The population size is 100, the number of iterations 1000, the crossover probability (Pc) 0.75, the mutation probability (Pm) 0.1, vehicle’s speed 30 km/h, truck’s load capacity 5t, and trailer’s load capacity 30 t.

Take the arrangement in area \( {\text{D}}_{2} \) for example, the program was run to get the solution of IIL model. In area \( {\text{D}}_{2} \), there are forty-four suppliers, a distribution center and a main engine factory. The auto parts list and package dimensions are shown in Table 3. The iterative process and optimization results are shown in Figs. 2 and 3.

Table 3 Auto parts type and package size
Full size table
Fig. 2
figure 2

The iterative process with genetic algorithm method

Full size image
Fig. 3
figure 3

Optimization Milk-run routes

Full size image

After 1000 iterations, 9 optimal vehicle travel routes were obtained. Since the coordinates used in the algorithm are longitude and latitude, the route length is calculated according to the actual mileage length, the optimized route summary table is shown in Table 4.

Table 4 Summary of Optimized Milk-run Routes
Full size table

Result discussion

  1. (1)

    Analysis of mileage and load rate

From Table 5, the total mileage in original mode is 1297.4 km, but it reduced to 710.5 km in IIL mode. The saving rate is as high as 43%. In addition, the average load rate in original mode is 87.4%, and it rises to 98% in IIL mode. Therefore, the optimized model can greatly improve transportation efficiency and reduce the transportation costs.

Table 5 Milk-run routes before optimization
Full size table
  1. (2)

    Analysis of Energy consumption

According to the total mileage and data in Table 1, the energy consumption cost of vehicles may be calculated in the currency of Yuan, as shown in Fig. 4. The energy consumption costs under Point-to-Point, Milk-run and IIL mode are 2908, 1502 and 1071, respectively. Under IIL mode, the energy consumption cost reduces by 28% than Milk-run mode, and 63% than Point-to-Point mode. It can be seen that IIL mode has more cost advantage than the other two modes. More importantly, CO2 emission under IIL mode was reduced by almost 37% than Milk-run mode.

Fig. 4
figure 4

Comparison of energy consumption

Full size image
  1. (3)

    Analysis of management improvement

From the perspective of suppliers, suppliers can arrange production activities flexibly and further reduce the inventory level through the IIL mode. Similarly, from the perspective of logistics companies, the Milk-run collection which is run along some preset routes and time window greatly reduces the difficulty of coordinating with the suppliers. Meanwhile, time window can reduce the waiting time of transportation vehicles at suppliers. For the main engine factory, due to the integrated inbound logistics service through information sharing and optimization technology, a quick response of order may be done. Next, the main engine factory may make a more reasonable cost control through identification of purchase cost, logistics cost and package cost. More importantly, the innovative IIL mode is contributable for the total manufacturing process from production to logistics cleaner and more efficient.

Conclusions

Based on the analysis of existing problems of auto parts inbound logistics, an integrated inbound logistics (IIL) mode, with an intelligent scheduling of Milk-run collection, drop and pull delivery and LNG vehicle, was established. It was formulated as a two-stage mathematical optimization model. The feasible solution was found by means of genetic algorithm.

Through the comparative analysis of mileage, energy consumption and logistics management, it is revealed that it may be a constructive strategy for auto manufacturers to enact IIL mode in order to keep logistics operation under a state of more efficiency, less cost and carbon. It may be contributable to achieve a total clean and high efficient manufacturing process from production to logistics. Its value may be extended to the innovation of logistics and production management in other large manufacturing industry.

To be noted, the problem described in this study was a simplified scene where only a distribution center and an engine factory were considered. Some limits in application are ignored. For example, the distribution of LNG filling station is lack of balance especially in the south of China and the maintenance is not very convenient, although LNG vehicle’s endurance mileage is presently up to 1000 km with the technology innovation. With the enforcement of policy and boost of LNG facility, these limits will be overcome. Future studies may be focused on more complicated scenes, such as large scaled inbound logistics operations with multiple distribution centers and multiple factories, traffic rules and weather conditions.