Flexible manufacturing workshop mechanical processing and product scheduling algorithm based on MES

Biao Ma; Yanfang Zhang

doi:10.1515/nleng-2024-0077

Artikel Open Access

Flexible manufacturing workshop mechanical processing and product scheduling algorithm based on MES

Biao Ma und Yanfang Zhang

Veröffentlicht/Copyright: 24. April 2025

Veröffentlicht von

Veröffentlichen auch Sie bei De Gruyter Brill

Manuskript einreichen Informationen für Autor*innen

Aus der Zeitschrift Nonlinear Engineering Band 14 Heft 1

Abstract

In modern manufacturing workshops, there are path conflicts and collisions between automated guided vehicles (AGVs), which affect the efficient scheduling efficiency of the workshop. To address this issue, a manufacturing execution system is introduced to optimize resource management and scheduling. This system combines the shortest path algorithm with a hybrid genetic algorithm to develop collision-free driving paths for AGVs. The results showed that after 20 iterations, the adaptability of workpiece transportation remained stable at 100, with an average of 9 workpieces per day for AGV and a completion time of 2,731 min. In addition, there were no vehicle conflicts between 750 and 1,000 min. These findings indicate that the proposed system improves the scheduling flexibility and efficiency of the manufacturing workshop, while minimizing collision risks to the greatest extent possible. Future research can explore extending the system to larger scheduling requirements and consider complex scheduling constraints comprehensively.

Keywords: MES; flexible manufacturing; path optimization; AGV; product scheduling

1 Introduction

Information technologies such as 5G, big data, and artificial intelligence have been integrated and applied in the intelligent manufacturing industry. Countries such as the United States, Germany, China, and Japan have successively proposed national-level strategies for the development of intelligent manufacturing [1]. The background for proposing advanced manufacturing technology development strategies varies among countries. However, a key common goal is to achieve interconnection, interoperability, and intelligence between the physical and information worlds in intelligent workshops [2]. In manufacturing systems, traditional single variety and large-scale manufacturing methods have been unable to meet the personalized needs of current customers. Optimizing production resource scheduling is a major challenge for manufacturing enterprises to achieve high-quality development. This is particularly reflected in the intelligent manufacturing practices of discrete manufacturing enterprises [3]. Manufacturing execution system (MES) is mainly responsible for workshop production management and scheduling. MES can track and record the production conversion from raw materials to finished products [4]. MES production monitoring focuses on material transportation and quality indicators between production and process. It is based on real-time data in production and utilizes MES configuration technology. The production progress, process quality, and material consumption of production areas such as production workshops, power and energy workshops, auxiliary material warehouses, and finished product warehouses can be monitored in real time. Flexible production first appeared in 1967. At that time, a British company Molins developed a new system consisting of six modular multi-process numerical control machines [5]. The core value of flexible manufacturing lies in the ability to rely on automated equipment to ensure production efficiency while also allowing production lines to have scalability. Flexible manufacturing can make timely adjustments to products based on market changes and customer needs. At present, the automation of processing equipment in manufacturing enterprises is relatively high. However, according to the actual statistical data on the workshop site, in 95% of cases, the workpiece waits for processing, materials, or loading and unloading during operation. About 5% of the time is in a processing state. The production management level of the workshop still needs to be improved urgently. Therefore, the research aims to improve workshop production efficiency by optimizing resource scheduling through the execution manufacturing system.

The MES operating system plays a crucial management role throughout the entire manufacturing workshop. Wang et al. [6] found that extreme weather can lead to serious emergencies. Machine learning technology has been widely applied to solve various power system problems. After reviewing the MES elasticity quantification method, it was found that this technology had great research value in preventing emergencies. In recent years, MES and related systems underwent development in terms of complexity and efficiency [6]. Nurdiyanto and Kindiasari [7] found that in the execution of manufacturing systems, the enterprise resource planning part operated in conjunction with various collaborative systems. The results indicated that the collaboration between the manufacturing system and various collaborative systems could greatly improve work efficiency [7]. The sentiment analysis was to detect viewpoints or polarities in textual data. Bernal and Limson [8] found that negative emotions could be detected through deep learning sentiment analysis techniques to identify abnormal activity in operating system logs. The research mainly used gated recurrent unit networks to identify emotions. The results showed that this method could detect abnormal events in operating system logs [8]. Sun et al. [9] proposed an optimized coordination method for energy scheduling and voyage scheduling in the case of a hybrid AC/DC transmission multi-energy ship MES connected to renewable energy grid. The results indicated the effectiveness of this method in coordinating multi-energy scheduling and voyage arrangements, minimizing operational costs/risks, and resisting various uncertainties [9]. Hebbi and Mamatha [10] found that unsupervised machine learning techniques, namely, K-means, had hierarchical clustering with run count features. A handwritten Kannada character dataset was established based on the system. Isolated Kannada vowels, consonants, modifiers, and Otaksha were identified, constructing a handwritten dataset of 85 characters. The results showed that the accuracy of the unsupervised method reached 80%, which greatly improved the efficiency of character collection [10].

The path problem in flexible manufacturing workshops also determines the efficiency of workshop operation. Martinez et al. [11] proposed a digital twin demonstrator to support the supervision activities of operators in the context of flexible manufacturing and robotics technology, addressing the large-scale production shifting toward mass customization and personalized restructuring faced by manufacturing companies. The results indicated that this supervision technology could monitor the self-adaptation of manufacturing systems to production and environmental changes in real time [11]. Liu et al. [12] found that integrating automated guided vehicle (AGV) resources into production scheduling has become a research hotspot. They established a mathematical model for optimizing dual resource scheduling of machine tools and designed a dual-layer input encoding method. The results indicated that the model can optimize the integrated scheduling of resources for AGVs [12]. Zhang et al. [13] found the complexity of manufacturing systems based on digital twins and the incompleteness of related models. Therefore, a five-dimensional fusion model of digital twin virtual entities for robot intelligent manufacturing systems supporting automatic reconstruction was proposed. The results indicated that the model improved the operational efficiency of such systems [13]. Zhang et al. [13] found that dynamic disturbances often occurred in real time in modern discrete flexible manufacturing systems. Therefore, a real-time scheduling method based on hierarchical multi-agent deep reinforcement learning was proposed. The results indicated that the proposed method had better performance compared to similar methods [13]. Luo et al. [14] found that the most important factors affecting AGV energy consumption were their cruising speed and travel distance. Therefore, an optimization method was proposed to minimize the number of performance indicators of a real AGVE system designed by Volvo Cars in Gothenburg, Sweden. The results showed that the maximum completion time, delay, and delay values of this method were better than those obtained from existing traffic controllers [14].

In summary, domestic and foreign researchers have conducted various studies on the flexible manufacturing of MES. However, few scholars have optimized the path and product scheduling of flexible manufacturing workshops through MES. A conflict-free path-based AGV movement method for manufacturing workshops is proposed to address this issue. Discrete flexible assembly job-shop (DFAJ) is proposed to achieve a better scheduling solution. The research content mainly includes five parts. The first part reviews the current status and related research on mechanical processing and product scheduling methods in flexible manufacturing workshops. The second part elaborates on the proposed method. The third part conducts experimental analysis on the method. The fourth part discusses the research results. The fifth part summarizes the research and proposes future prospects.

The innovation of the research lies in breaking the limitations of traditional assembly workshop scheduling problems that only consider time and sequential arrangements, and combining path planning to achieve collaborative optimization of job scheduling and material distribution. Second, in multi-AGV systems, dynamic priority strategies are used to solve path conflict problems, improving the flexibility and response speed of the system. The third is to use hybrid genetic algorithm for production scheduling, which optimizes scheduling schemes through specific crossover and mutation probabilities to improve scheduling efficiency and quality.

2 Methods and materials

The AGV driving path of MES flexible manufacturing workshop is optimized. The shortest path algorithm, Dijkstra, is proposed. The architecture design of MES is introduced. The essential difference between bidirectional two-way lanes and bidirectional one-way lanes is explained in the setting of path nodes. Second, the convenience of non-uninstalled desktop AGV is introduced. Finally, an algorithm flow for discrete flexible manufacturing systems is proposed. The introduction and explanation sequence of this chapter are shown in Figure 1.

Figure 1

Introduction and explanation sequence flow chart of this chapter.

2.1 Optimization strategy for mechanical processing in flexible manufacturing workshop based on MES

MES can provide enterprises manufacturing data, production scheduling, inventory, quality management modules, etc. The architecture hierarchy of MES visualization system from bottom to top is data perception, data fusion, data persistence, data reconstruction, and business and application layers [15,16]. Figure 2 shows the MES architecture.

Figure 2

Design diagram of the MES framework.

In MES flexible manufacturing systems, AGV belongs to the transportation equipment resources in material transportation systems, which are widely used in many manufacturing fields [17]. The AGV system improves efficiency through automated material handling, while the MES system is responsible for monitoring and managing the production process. The homework-oriented enterprise factory combines the AGV system with the MES system. The MES system assigns tasks to AGVs based on production plans and priorities, while the AGV scheduling system optimizes allocation based on the urgency of tasks and the current status of AGVs, improving overall operational efficiency. When AGV encounters abnormalities during task execution, it can promptly provide feedback to the MES system, which adjusts the production plan accordingly to reduce production interruptions. Therefore, the research first needs to model the layout of the job-shop under the MES system, and select appropriate algorithms to find a conflict free and shortest path for each AGV in the manufacturing system. Common algorithms for solving path optimization include Dijkstra, A* algorithm, and time window. The Dijkstra algorithm is used to find the shortest path from a single source point to all other vertices in a weighted graph. Its working principle is to set the estimated shortest path of all vertices to infinity, except for the starting vertex, whose value is set to 0. Next, the vertex with the smallest distance is selected from the previously unvisited vertices and marked it as visited. The path length from the current vertex to each adjacent point of the vertex is calculated. If the length is less than the known shortest path length of the adjacent point, the shortest path length of the adjacent point is updated. The aforementioned calculation steps are repeated until all path nodes have been accessed. The Dijkstra algorithm has high accuracy and easy implementation. The time window algorithm is mainly used to solve path planning problems with time constraints, such as each customer having a specific time window within which they must be accessed. The goal of the algorithm is to find a path that allows all customers to be accessed while satisfying the constraints of the time window. Other goals can also be considered, such as minimizing the total travel time or distance.

Therefore, the study combines time window and Dijkstra’s algorithm to investigate the path conflict problem in path planning. The Dijkstra is represented by Eq. (1):

(1) G 0 → k + 1 < G 0 → k + G k → k + 1 ,

where the calculation process starts with G k and traverses all nodes in set G , taking the node G k + 1 with the shortest distance from G k . G k + 1 is removed from set G and added to set S . The weights of S 0 in the set G are updated to reach the remaining nodes. Whether G k + 1 is S 0 is determined. If it is, the algorithm needs to be terminated to obtain the shortest path information and minimum weight. In the time window method, path L is often divided into two parts: the idle time window set F L and the reserved time window set R L . The commonly used mathematical model for time windows is represented by Eq. (2):

(2) F L = { f k = [ a k , b k ] } R L = { r k = [ c k , d k ] } ,

where f k represents the idle time period of path L . k represents the k th spatial time period, starting from a k and ending at b k . The path types of the workshop can be divided into bidirectional one-way lanes, bidirectional two-way lanes, and two-way single lane based on the direction of travel and the number of lanes. In the bidirectional one-way topology map, M represents the production equipment number. Numbers 4, 8, 6, and 3 represent general nodes. Number 1 and 9 are the raw material warehouse and finished product warehouse, respectively. Figure 3 is a topological map of the workshop layout.

Figure 3

Topological map of workshop layout.

In the manufacturing system of multi-AGV, vehicles should find suboptimal paths while avoiding collisions and reaching their destination with minimal time and cost losses, which should be addressed. When Dijkstra’s algorithm is applied to complex scenarios, such as dynamic environments and multi-objective optimization problems, the complexity may increase. Therefore, the combination of natural selection and search algorithms is used for optimization. The study proposes to combine time window algorithm with genetic algorithm to form a hybrid genetic algorithm. The combination of the two can enable AGV to effectively search for the optimal or nearest path solution while satisfying time window constraints. The mathematical model of mixed genetics aims to minimize the total travel time of all AGVs. The delay cost is shown in Eq. (3):

(3) min Z = min ∑ v = 1 W Q V ,

where Q V represents the nodes that the vehicle v must pass through to complete. v ∈ { 1 , 2 , … , w } , and s ∈ { 1 , 2 , … , p } . W represents the total travel distance of the car. This study first focuses on the mathematical modeling of possible path conflicts. Based on the aforementioned calculation formula, the study utilizes the shortest path algorithm and hybrid genetic algorithm. Combining these two algorithms can fully utilize the advantages of a single algorithm. Dijkstra can quickly determine the global optimal path of the initial solution, while hybrid genetic algorithm searches and optimizes within local regions through its crossover and mutation operations, thus avoiding getting stuck in local optimal solutions. When there are too many scheduling tasks, the hybrid genetic algorithm can effectively handle complex constraints, such as time windows and resource limitations, thereby finding better scheduling solutions and avoiding the characteristic of only seeking fast and not refined shortest paths. This study combines two algorithms, which increase computational complexity. However, when solving complex optimization problems, it usually provides better performance and higher-quality solutions. The research first requires mathematical modeling of possible path conflicts in path planning. The occupancy rate of the vehicle at this node is represented by Eq. (4):

(4) ∑ v = 1 w Z v s t ≤ 1 , ∀ s ∈ { 1 , 2 , ⋯ , p } .

In Eq. (4), when Z v s t is equal to 1, k v occupies a node in time t . p represents the adjustable parameters. On the contrary, multiple cars do not reach a certain node simultaneously. However, a car closest to this node should pass through that node first. To address such situations, a topology map is used to plan paths and determine the vehicle position. The device is represented as c 1 , c 2 , … c n . The node is represented as s 1 , s 2 , … s m . The warehouse is represented as w . The vehicle is represented as k 1 , k 2 , k 3 . If equipment c 1 and c 2 are on nodes s 1 and s 11 , the workshop layout is reasonable, and the current position of the car is confirmed. Warehouse w is located at node s 5 . The initial positions of the three small vehicles are set at s 3 , s 8 , and s 9 . Figure 4 shows the warehouse topology map and the distribution of vehicles and nodes.

Figure 4

Warehouse topology map and distribution map of vehicles and nodes.

2.2 MES flexible manufacturing product scheduling optimization model and method

The warehouse layout and the distribution of AGV operation nodes are preliminarily understood, and the shortest path algorithm and hybrid genetic algorithm are combined to optimize the path. While the path of the flexible manufacturing workshop is optimized, the processing equipment of its work vehicles is highly automated. Each device can also complete multiple different processes for different workpieces. Traditional AGVs play an important role in industrial automation and logistics systems, mainly used for automated handling and transportation of goods. Due to their small size, they can add more AGVs to achieve modular expansion. However, traditional AGVs are only suitable for command tasks that travel along predetermined paths and may not be suitable for non-repetitive or variable tasks. Therefore, to meet the demand for customized products, some large enterprises that produce high-end cars, large engines, and precision valve bodies begin to try non-unloading assembly forms [18]. Non-unloading AGV is an automated material handling equipment that can automatically transport goods from one place to another without the need for manual unloading. Non-unloading AGV transports differentiated assembly materials and workpieces together to the assembly island. General assembly materials are stored on the assembly island. This can effectively avoid problems such as missing or incorrect installation during the assembly process [19]. Compared with traditional AGVs, non-unloading AGVs do not require excessive deployment of AGVs on the production and transportation lines, resulting in more significant path optimization effects. Figure 5 shows the traditional AGV and non-unloading worktable AGV.

Figure 5

Comparison between traditional AGV and non-unloading worktable AGV: (a) traditional VGV mode and (b) non-unloading worktable AGV.

Conflict-free path planning brings uncertainty to the transportation time of multiple AGVs. Therefore, it is necessary to achieve integrated scheduling of processing equipment and AGV under conflict-free path planning. Assuming there are m processing equipment and w AGVs with the same transportation capacity in the flexible machining workshop, it is necessary to complete processing tasks for n different workpieces. The used non-unloading AGC with consistent transportation capacity is for the convenience of research and calculation, and to ensure that the algorithm meets the corresponding scheduling rules to the greatest extent possible. Based on this, a set of mathematical models are established as constraints for the algorithm, with the objective function of minimizing the maximum completion time. The mathematical model is shown in Eq. (5):

(5) C = min ( C max ) = min ( max i = 1 n ( C i ) ) ,

where C represents the completion time of the task. The optimal number of AGVs is equal to the minimum number of AGVs in the workshop that have the longest completion time and shortest working time, represented by Eq. (6):

(6) W opt = min ( v ) × min ( C max ) ,

where v is the average speed of the trolley operation. To make every operation processed by only one device, the cache of each device is constrained to 1. Its constraint is represented by Eq. (7):

(7) ∑ k = 1 m a ijk = 1 , ∀ i ∈ { 1 , 2 , … , n } , ∀ j ∈ { 1 , 2 , … , p } ,

where a i j k is the constraint interval. If a i j k is equal to 1, the processing equipment is on the specified k . Meanwhile, to ensure that only one AGV is responsible for the transportation task of each process, the continuous operation is represented by Eq. (8):

(8) ∑ v = 1 w x i j v = 1 , ∀ i ∈ { 1 , 2 , … , n } , ∀ j ∈ { 1 , 2 , … , p } ,

where x i j v is the average range of small car operations. When x i j v is equal to 1, the small car k v is responsible for the transportation task of process o i j . The empty time of the current transporting task is represented by Eq. (9):

(9) S ′ T ij v ′ ≥ C T i ( j − 1 ) v + ∑ k = 1 m α i ( j − 1 ) k ⋅ t i ( j − 1 ) k ∀ i ∈ { 1 , 2 , … , n } , ∀ j ∈ { 1 , 2 , … , u i } , ∀ v , v ′ ∈ { 1 , 2 , … , w } , ∀ k ∈ { 1 , 2 , … , w } ,

where S ′ T i j v ′ represents the start time of empty transportation in transportation task T i j v ′ , and C T i ( j − 1 ) v represents the end time of empty transportation in transportation task T i j v ′ . The start and end time when the car is operating under load is represented by Eq. (10):

(10) C T ij v ′ ≥ S T ijv + ∑ k = 1 m ∑ k ′ = 1 m α i ( j − 1 ) k ⋅ α ijk ⋅ t k k ′ ∀ i ∈ { 1 , 2 , … , n } , ∀ j ∈ { 1 , 2 , … , u i } , ∀ k , k ′ ∈ { 1 , 2 , … , m } , ∀ v ∈ { 1 , 2 , … , w } ,

where C T i j v ′ represents the end time of load transportation in transportation task T i j v , and S T i j v represents the start time of load transportation in transportation task T i j v . Eq. (10) indicates that the end time of the load travel for the AGV’s current transport task is greater than or equal to the sum of the remaining idle time and the required transport time between the two processing devices. The occupancy at each node at the same time is represented by Eq. (11):

(11) ∑ v = 1 w z vst ≤ 1 , ∀ s ∈ { 1 , 2 , … , p } , ∀ t ∈ C ,

where z v s t represents the node occupied by the car. When z v s t is equal to 1, the car k v occupies node N s at time t . In production, waiting time is always ineffective, which reduces the utilization rate of small cars and affects the final completion time of the entire production. The invalid waiting time is represented by Eq. (12):

(12) IWT = ∑ i , p = 1 n ∑ j , n = 1 n { ( C i ( j + 1 ) k − C T ijv ) + ( C pqk ⋅ β pqijk − C T ijv ) } ,

where IWT represents the invalid waiting time. In the flexible machining workshop, the current available quantity of AGVs w is 4. When the completing time is minimized, there is usually an upper limit to the maximum number of AGVs. The concept of DFAJ is proposed and combined with the assembly island concept. The assembly island is composed of multiple identical devices, which can simultaneously perform assembly tasks for multiple products, thereby improving the production efficiency of the assembly system. The schematic diagram of assembly island and assembly process in DAFJ is shown in Figure 6.

Figure 6

Assembly island and assembly process diagram in DFAJ.

This study proposes DFAJ and combines the concept of assembly islands. The assembly island is composed of multiple identical devices, which can perform assembly tasks for multiple products simultaneously, improving the production efficiency of the assembly system. Figure 6 shows the assembly process of the workpiece process, AGV, and assembly island. The devices inside the red box indicate that they have been occupied. RFID is responsible for identifying the workpiece process during AGV transportation. When there is a bottleneck or waiting situation during the workpiece assembly process, the staff can choose to assemble on other assembly islands. In this way, without increasing equipment, assembly efficiency can be effectively improved. The layout of DFAJ is similar to that of a discrete flexible machining workshop, which can be represented using a topology map, as shown in Figure 7.

Figure 7

Workshop map of DFAJ.

DFAJ generates schedulable process set data based on the assembly process sequence of the product, including pre-production and post-production processes. Based on the tight preceding process, it is first necessary to randomly select a process from the set of non-tight preceding processes and then delete the selected process from the list. If the immediately preceding process becomes empty after deletion, the process is no longer in the immediately preceding process. Based on the tight post-process, a process is randomly selected from the set of non-tight post-processes. The previous deletion work is repeated until all process searches are completed. According to the objective function of DFAJ, the AGV is optimized in the system by minimizing the completion time. The assembly process is represented by Eq. (13):

(13) Q = min ( c max ) = min ( max k = 1,2 , … M C k ) ,

where Q represents the maximum completion time for the assembly of all products. The optimal delivery capacity for assembly is represented by Eq. (14):

(14) V min = ∑ k = 1 M ( s t jkv − c ′ t jkv ) T max ,

where V min represents the optimal AGV quantity for the assembly process to meet the latest delivery date; s t j k v represents the start time of the car v load responsible for task j k ; and c ′ t j k v represents the end time of the j k load responsible for the task. The quantity of equipment workpieces is represented by Eq. (15):

(15) ∑ k = 1 M x ikt ≤ P i , ∀ i ∈ ( 1 , 2 , … , S ) , ∀ t ∈ Q ,

where x i k t indicates that the capacity of the assembly island with parallel equipment will not be overloaded, ensuring that only one workpiece is assembled for each parallel equipment. When x i k t is equal to 1, the workpiece j k is assembled at the assembly island at time t . Otherwise, it is 0. From the above, it can be concluded that in DAFJ workshop scheduling and vehicle route planning, it is necessary to rely on MES system to coordinate AGV work, ensure timely delivery of workpieces to designated locations, and coordinate with workshop equipment to avoid mechanical equipment idle or production delays. When scheduling MES systems, the shortest path algorithm is required to calculate the shortest path from the starting point to the endpoint and determine the conflict-free path for AGV. In addition, a hybrid genetic algorithm is used to gradually optimize the accuracy of path selection.

3 Results

The proposed conflict-free path based on Dijkstra is more efficient and nimbler for flexible manufacturing workshops. In this regard, a workshop mechanical processing task scheduling diagram is drawn. The scheduling tasks are divided into five groups for transportation. The scheduling situation of these three products is compared. Subsequently, this study validates the scheduling situation of discrete flexible manufacturing systems at different time periods and conducts comparative analysis.

3.1 Analysis of workshop mechanical processing scheduling with conflict-free paths

A workshop machining task scheduling diagram is constructed to verify that Dijkstra based on time windows can better optimize the path conflict problem of multi-task and multi-AGV. The study adopts a Windows 10 with a 64 bit operating system. The editing language is C++. The experimental processor and graphics card uses Intel^® Xeon^® Platinum 8124M, with a memory of 64CB. Due to the combination of hybrid genetics and Dijkstra’s algorithm, the computational complexity is increased. In actual task scheduling, there may be path planning confusion caused by demand delays. Therefore, to enable the proposed method to respond faster and have higher scalability when dealing with more complex manufacturing system requirements, the experiment requires first mastering the basic transportation tasks in the workshop, so that the algorithm can schedule the tasks. Figure 8 shows the scheduling of workshop processing tasks.

Figure 8

Workshop machining task scheduling diagram.

Figure 8 shows the scheduling diagram of machining tasks in the workshop. The scheduling task mainly consists of AGV loading task i and unloading task j, with a total of five sets of transportation tasks and three processing tasks. (i ₁, j ₁), (i ₂, j ₂), (i ₃, j ₃), (i ₄, j ₄), and (i ₅, j ₅) are transportation tasks. (i ₁, j ₂), (i ₂, j ₃), and (i ₄, j ₅) are processing tasks of production equipment. When processing workpieces, the car was waiting next to the equipment. Therefore, k ₁ waited for the workpiece processing to be completed at node 7 and returned to node 6. Then, the position of node 7 was provided to the small car k ₂. The aforementioned scheduling results indicate that the proposed algorithm can schedule and control AGVs based on workshop tasks, demonstrating high stability and system scalability. This study compares the results based on the research scheme of Japanese scholar Miyamoto. Figure 9 shows the Miyamoto path planning scheme and the proposed conflict-free path planning scheme.

Figure 9

Miyamoto path planning scheme and the proposed conflict-free path planning scheme: (a) Miyamoto path planning scheme and (b) conflict-free path planning diagram.

Figure 9(a) shows the Miyamoto path planning scheme. When car k ₁ made space for k ₂ at node 7, it could not accurately determine whether k ₁ was waiting at node 6 or on the path between nodes 6 and 7. In Figure 9(b), after unloading at node 7, k ₁ could wait on the path between nodes 6 and 7 until the workpiece was processed. K ₂ unloaded the workpiece from j ₄ and took the workpiece to node 11 for processing. This saved the ineffective waiting time at node 7 and advanced the processing time of node 11 for j ₂. As a result, the proposed path planning scheme resulted in a more optimal path planning scheme. Due to the need to consider the AGV required for transporting workpieces, this study comprehensively compares the scheduling problem using two-stage method and hybrid genetic algorithm. Figure 10 shows the applicability analysis of these two methods.

Figure 10

Two-stage method and hybrid genetic algorithm for comprehensive comparative scheduling. (a) The search process of two-stage method. (b) Hybrid genetic algorithm search process.

Figure 10(a) shows the two-stage method, and Figure 10(b) shows the hybrid genetic algorithm. According to Figure 10(a), the optimal fitness of workpiece transportation showed a significant decrease before 30 iterations, indicating that the algorithm could quickly find the optimal solution in the initial stage. However, the fluctuation was large, indicating that the stability of the solution was not high. The average fitness decreased before 60 iterations and tended to plateau after 60 iterations. Due to the fact that the scheduling results of flexible workshops become constraints for AGV scheduling problems, the shortcomings of the two-stage method are also evident. It can only obtain local optimal solutions for integrated scheduling problems. The optimal fitness of workpiece transportation in Figure 10(b) showed a decreasing trend before 20 iterations and stabilized at around 100 after 20 iterations. The average fitness decreased from 100 to around 80 before 6 iterations and then tended to plateau. This indicates that the hybrid genetic algorithm quickly converged to a stable state, and the stability of the solution was high. Overall, the hybrid genetic algorithm outperforms the two-stage method in comprehensive scheduling. The aforementioned data indicate that finding a better scheduling plan can reduce the waiting time and transportation time of workpieces in the workshop, thereby improving overall production efficiency. At the same time, the optimized scheduling scheme can ensure that the AGV’s travel path in the workshop is more reasonable, avoid unnecessary path intersections and conflicts, and reduce production delays and safety accidents caused by conflicts.

3.2 Analysis of MES path planning strategy

The established DFAJ consists of i independent assembly islands. Each island has assembly P _i identical assembly equipment. There are j different types of products in the assembly production planning task. The demand for each type of product is D _j. The number of AGVs in the system is V. This experiment uses three types of products. A schedulable process set data table based on tight preceding and tight following processes is generated according to the assembly process sequence of the products in Table 1.

Table 1

Data table for the schedulable process set of tight preceding and tight following processes

	Product A		Product B		Product C
Working procedure	Pre-production process	Post-production process	Pre-production process	Post-production process	Pre-production process	Post-production process
1	—	3	—	3	—	4
2	—	5	—	4, 5, 6	—	3
3	1	4, 5	1	7	2	5, 6
4	3	12	2	7	1	5
5	2	8	2	9	3, 4	9
6	—	9	2	8	3	9
7	4	12	3, 4	10, 12	6	8
8	5	10	6	10	7	10, 12
9	6	11	5, 7	11, 12	5, 6	10
10	7, 8	13	8	13	8, 9	14
11	9	13	9	13	7	13
12	7	14	9	14	8	13
13	10, 11	14	10, 11	14	11, 12	16
14	12, 13	15	12, 13	—	10	15
15	14	—	—	—	14	16
16	—	—	—	—	13, 15	—

Products A, B, and C in Table 1 underwent 16 processes, which were divided into pre-production process and post-production process. The total demand for products B and C was 45, with a delivery cycle of 1 week and five working days per week. Therefore, it is most suitable to assemble an average of 9 workpieces per day. To observe the scheduling situation of DFAJ with delivery time as a constraint more clearly, a scheduling Gantt chart is drawn in Figure 11.

Figure 11

Scheduling Gantt chart with delivery time as constraint.

In Figure 11, in DFAJ constrained by delivery time, only three AGVs were used to meet the order delivery time constraint. The 17th process of AGV3 means that AGV unloads the currently assembled workpiece to the finished product warehouse. The 18th process of AGV3 indicates that AGV returns to the blank warehouse to load the next workpiece. The transportation process between the 17th and 18th processes represents the AGV returning from the finished product warehouse to the blank warehouse. The vertical axis represents the AGV number, workpiece type, and production sequence, respectively. For example, 35 indicates that the production sequence of product C workpieces is the fifth. The final scheduling results indicated that the three AGVs in the workshop could meet production requirements. The completion time was 2,731 min, which was less than the specified delivery time of 2,880 min.

Two workpieces are randomly selected in the study and further validated based on this cycle process. Figure 12 shows the scheduling Gantt chart of AGV and processing equipment for workpieces 1 and 2. In Figure 12, the pink square represents the processing equipment, and the green square represents the workpiece, clearly displaying the scheduling results of each part and AGV. Figure 12(a) shows the path for transporting workpiece 1 using an assembly island. There were fewer path conflicts between processing equipment and workpieces. At 0 min, the AGV departed every other time unit to minimize the completion time of task production and avoid path conflicts. There was no conflict between 750 and 1,000 min. After 1,000 min, the loading and unloading of AGV workpieces and equipment production showed a stable state. Figure 12(b) shows the transportation path of workpiece 2 without the assembly islands. The conflicting nodes observed in the path area for workpiece 2 were significantly higher than that for workpiece 1. Workpiece 2 did not experience any overtaking or node conflicts in its path after 1,200 min. Therefore, the proposed DFAJ, combined with the assembly island model, can more accurately predict the production process and AGV path planning. This method facilitates the workshop to adjust the production plan in a timely manner and deliver on schedule.

Figure 12

Gantt chart for scheduling AGV and processing equipment for workpieces 1 and 2: (a) Gantt chart of the AGV and machine scheduling problem of job 1 and (b) Gantt chart of the AGV and machine scheduling problem of job 2.

4 Discussion

This study focused on the two major processes of mechanical processing and product assembly in flexible manufacturing workshops. The integrated optimization scheduling problem of production resources for AGV cars representing processing or assembly equipment and logistics equipment was analyzed to achieve the optimal allocation of production resources in manufacturing enterprises, improve production efficiency, shorten delivery time, and reduce production costs. The experiment showed that the optimal fitness for workpiece transportation occurred before 20 iterations. After 20 iterations, the fitness stabilized at around 100. This result is similar to the optimal value of AGV transportation path obtained by Vlachos et al., but the overall value is better [20]. The experiment compared the conveying time of AGV based on different scheduling tasks. The result showed that it was 149 min ahead of the delivery time. This result is even faster than the experimental results of Margherita and Braccini, indicating that the designed optimization scheme has higher delivery efficiency [21]. The final experiment was conducted based on real scheduling tasks. The optimal scheduling effect was achieved, without AGV collisions during the operation process. Choi et al. also conducted experiments on scheduling optimization of AGV vehicles. However, compared with the proposed method, the scheduling effect did not reach the best [22]. Overall, although the proposed method can achieve intelligent factory logistics automation and production transportation optimization, the MES system can experience sudden failures, such as power outages, equipment damage, and data overload, which seriously affect the time of workshop operations. Therefore, it is necessary to regularly optimize the indexing and data cleaning of the MES system, reproduce the technical basis, and cooperate with monitoring mechanisms to regularly evaluate and optimize the operating system. Manufacturers can choose suitable AGV scheduling systems and MES control systems based on production needs and operating environments in actual production, ensuring the stability of instruction transmission. More importantly, it is necessary to provide training for operators to ensure that they are proficient in operating the integrated system. Moreover, a monitoring mechanism can be established during the implementation phase to promptly identify and resolve issues.

5 Conclusion

In the manufacturing system, workshop equipment is constantly upgraded, and the manufacturing automation is greatly improved. Manufacturing enterprises are facing issues such as optimizing scheduling and configuration of production resources. Therefore, the study focuses on the analysis and research of the mechanical processing and workshop product assembly scheduling in flexible manufacturing workshops. The study introduced an executive manufacturing system to manage and schedule workshop production, while combining shortest path algorithm and hybrid genetic algorithm for conflict-free planning of the driving path of automatic guided transport vehicles in the workshop. The results showed that the optimal fitness for workpiece transportation demonstrated a decreasing trend before 20 iterations and remained stable at around 100 after 20 iterations. This indicated that the proposed method had strong stability and could adapt well to AGV scheduling. The final scheduling results indicated that three AGVs could meet production requirements and the completion time was 2,731 min, which was less than the delivery time of 2,880 min specified. This indicates that the proposed method could complete scheduling tasks on time under different scheduling requirements, with high processing efficiency. The study recorded the path of transporting workpieces using assembly islands, with no conflicts occurring between time 750 and 1,000. After time 1,000, AGV presented a stable state for the loading and unloading of workpieces and equipment production. The DAFJ proposed in the study combined with the assembly island model can more accurately predict the production process and AGV path planning. In summary, it can be proven that the proposed method can quickly adjust the layout of production lines and transportation processes, reduce the capacity reduction caused by human errors, and adapt to changes in the industrial market more quickly. In practical applications, it can be applied to manufacturing, electronics, automotive manufacturing, food industry, etc., where production must strictly follow safety regulations and standards. The optimization scheduling of the system can manage logistics during the handling process, while ensuring compliance with industry standards and reducing the possibility of human errors. However, the diversity of industries also tests the practicality of this method. Future research may consider developing more complex scheduling algorithms that support simultaneous scheduling of multiple processes and products to meet more complex production needs. In addition, it is also possible to explore how to incorporate more practical constraints (such as worker rest time and equipment maintenance time) into scheduling algorithms to further improve the rationality and efficiency of scheduling.

Funding information: This work was supported by the Youth Project of Science and Technology Research Program of Chongqing Education Commission of China, Design and Implementation of MES System for the Production Line of The Product as Wheel Hub (No. KJQN202203902); and the Youth Project of Science and Technology Research Program of Chongqing Education Commission of China, Design and Virtual Debugging of CNC Machining System Based on Industrial Robots (No. KON202303910).
Author contributions: Biao Ma: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing-original draft, writing-review and editing. Yanfang Zhang: data curation, methodology, formal analysis, resources, software, validation, visualization, writing-review and editing. All authors have accepted responsibility for the entire content of this manuscript and approved its submission.
Conflict of interest: The authors state no conflict of interest.
Data availability statement: The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

References

[1] Xu Y, Sahnoun M, Ben Abdelaziz F, Baudry D. A simulated multi-objective model for flexible job shop transportation scheduling. Ann Oper Res. 2022;311(2):899–920.10.1007/s10479-020-03600-0Suche in Google Scholar

[2] Lu N, Zhou W, Dou ZW. Can intelligent manufacturing empower manufacturing? –an empirical study considering ambidextrous capabilities. Ind Manage Data Syst. 2023;123(1):188–203.10.1108/IMDS-11-2021-0718Suche in Google Scholar

[3] Guo Y, Zhang W, Qin Q, Chen K, Wei Y. Intelligent manufacturing management system based on data mining in artificial intelligence energy-saving resources. Soft Comput. 2023;27(7):4061–76.10.1007/s00500-021-06593-5Suche in Google Scholar

[4] Zanchettin AM. Robust scheduling and dispatching rules for high-mix collaborative manufacturing systems. Flex Serv Manuf J. 2022;34(2):293–316.10.1007/s10696-021-09406-xSuche in Google Scholar

[5] Ani EC, Olu-lawal KA, Olajiga OK, Montero DJ, Adeleke AK. Intelligent monitoring systems in manufacturing: current state and future perspectives. Eng Sci Technol J. 2024;5(3):750–9.10.51594/estj.v5i3.870Suche in Google Scholar

[6] Wang J, Pinson P, Chatzivasileiadis S, Panteli M, Strbac G, Terzija V. On machine learning-based techniques for future sustainable and resilient energy systems. IEEE Trans Sustain Energy. 2022;14(2):1230–43.10.1109/TSTE.2022.3194728Suche in Google Scholar

[7] Nurdiyanto H, Kindiasari A. Critical role of manufacturing execution systems in digital transformation of manufacturing industry. J Electr Syst. 2024;20(7s):2432–6.10.52783/jes.4038Suche in Google Scholar

[8] Bernal EA, Limson V. Assessing manufacturing execution system critical factors and readiness of selected semiconductor companies in Laguna: basis for the development of management interventions. Ignatian Int J Multidiscip Res. 2024;2(5):258–71.Suche in Google Scholar

[9] Sun M, Cai Z, Zhao N. Design of intelligent manufacturing system based on digital twin for smart shop floors. Int J Comput Integr Manuf. 2023;36(4):542–66.10.1080/0951192X.2022.2128212Suche in Google Scholar

[10] Hebbi C, Mamatha H. Comprehensive dataset building and recognition of isolated handwritten Kannada characters using machine learning models. Artif Intell Appl. 2023;1(3):179–90.10.47852/bonviewAIA3202624Suche in Google Scholar

[11] Martinez S, Mariño A, Sanchez S, Montes AM, Triana JM, Barbieri G, et al. A digital twin demonstrator to enable flexible manufacturing with robotics: a process supervision case study. Prod Manuf Res. 2021;9(1):140–56.10.1080/21693277.2021.1964405Suche in Google Scholar

[12] Liu Q, Wang NJ, Li J, Ma T, Li F, Gao Z. Research on flexible job shop scheduling optimization based on segmented AGV. CMES Comput Model Eng Sci. 2023;134(3):2073–91.10.32604/cmes.2022.021433Suche in Google Scholar

[13] Zhang C, Xu W, Liu J, Zhang C, Xu W, Liu J, et al. Digital twin-enabled reconfigurable modeling for smart manufacturing systems. Int J Comput Integr Manuf. 2021;34(7–8):709–33.10.1080/0951192X.2019.1699256Suche in Google Scholar

[14] Luo S, Zhang L, Fan Y. Real-time scheduling for dynamic partial-no-wait multiobjective flexible job shop by deep reinforcement learning. IEEE Trans Autom Sci Eng. 2021;19(4):3020–38.10.1109/TASE.2021.3104716Suche in Google Scholar

[15] Kolasani S. Revolutionizing manufacturing, making it more efficient, flexible, and intelligent with Industry 4.0 innovations. Int J Sustain Dev AI ML IoT. 2024;3(1):1–17.Suche in Google Scholar

[16] Leng J, Zhu X, Huang Z, Xu K, Liu Z, Liu Q, et al. ManuChain II: Blockchained smart contract system as the digital twin of decentralized autonomous manufacturing toward resilience in industry 5.0. IEEE Trans Syst Man Cybern Syst. 2023;53(8):4715–28.10.1109/TSMC.2023.3257172Suche in Google Scholar

[17] Yu D, Hu X, Liang K, Ying J. A parallel algorithm for multi-AGV systems. J Ambient Intell Humaniz Comput. 2022;13(4):2309–23.10.1007/s12652-021-02987-3Suche in Google Scholar

[18] Boysen N, Schneider M. Picker routing in AGV-assisted order picking systems. INFORMS J Comput. 2022;34(1):440–62.10.1287/ijoc.2021.1060Suche in Google Scholar

[19] Singh A, Madaan G, Kumar H, Kumar A. Smart manufacturing systems: a futuristics roadmap towards application of industry 4.0 technologies. Int J Comput Integr Manuf. 2023;36(3):411–28.10.1080/0951192X.2022.2090607Suche in Google Scholar

[20] Vlachos I, Pascazzi RM, Ntotis M, Spanaki K, Despoudi S, Repoussis P. Smart and flexible manufacturing systems using Autonomous Guided Vehicles (AGVs) and the Internet of Things (IoT). Int J Prod Res. 2024;62(15):5574–95.10.1080/00207543.2022.2136282Suche in Google Scholar

[21] Margherita EG, Braccini AM. Industry 4.0 technologies in flexible manufacturing for sustainable organizational value: reflections from a multiple case study of Italian manufacturers. Inf Syst Front. 2023;25(3):995–1016.10.1007/s10796-020-10047-ySuche in Google Scholar

[22] Choi JY, Jeon JH, Lyu JH, Park J, Kim GY, Chey SY, et al. Current applications and development of composite manufacturing processes for future mobility. Int J Precis Eng Manuf Green Technol. 2023;10(1):269–91.10.1007/s40684-022-00483-3Suche in Google Scholar

Received: 2024-08-14

Revised: 2024-11-28

Accepted: 2024-12-20

Published Online: 2025-04-24

This work is licensed under the Creative Commons Attribution 4.0 International License.

Artikel in diesem Heft

https://doi.org/10.1515/nleng-2024-0077

Schlagwörter für diesen Artikel

MES; flexible manufacturing; path optimization; AGV; product scheduling

Creative Commons

BY 4.0