Design of an Optimal Scheduling Control System for Smart Manufacturing Processes in Tobacco Industry

The whole process of tobacco production is composed of many components, in which their operation and administration are currently independent. It is required to deploy smart manufacturing workflow for the whole production process, in order to realize centralized effective global scheduling. This requires an advanced administration control platform that has strong abilities of multisource data integration and automatic decision support. To bridge such research gap, this paper designs an optimal scheduling control system for smart manufacturing processes of tobacco industry. First of all, this work discusses major characteristics of future-generation production control patterns in intelligent tobacco factories (ITF). Then, a five-layer architecture for optimal scheduling control of ITF is proposed, which contains Internet-of-Things layer, centralized control layer, model layer, platform layer and operation layer. In addition, a production scheduling optimization strategy is also developed for the proposed system to serve as the software algorithm that drives the running of whole smart manufacturing processes. Finally, this paper presents a comparative analysis of the proposed system’s transformation in a cigarette factory. Naturally, the effectiveness of the proposed production optimization scheduling strategy is verified through simulation.


I. INTRODUCTION
Intelligent manufacturing is the deep integration of the new generation of information technology represented by 5G, cloud computing, Internet of Things, big data and artificial intelligence with the field of manufacturing [1], [2]. This brings new mode and new ecological changes to the manufacturing industry [3], [4]. As the pillar of national economy and the core of industrial revolution, the manufacturing industry has received extensive attention from global industrial manufacturing powerhouses [5], [6]. The U.S. government pays more attention to the application of advanced information technology to penetrate the manufacturing field [7], [8]. It has introduced a series of strategies to revitalize The associate editor coordinating the review of this manuscript and approving it for publication was Huiyan Zhang . manufacturing development since 2009 [9]. Advanced software technology is also applied to promote the overall automation level, in order to ensure its leading position in the world manufacturing industry [10], [11]. After several years of research efforts, the investigation of intelligent manufacturing has achieved some preliminary achievements in these years [12], [13].
Intelligent factory acts as the application of advanced information technology such as Internet of Things, digital twin, big data, etc [14]. Besides, it also needs to be open up with the upstream and downstream raw and auxiliary materials supply chain, product sales chain, product service chain [15]. The goal is to achieve the interconnection of data and applications, and to further achieve ecological synergy of the whole value chain [16]. Generalized to tobacco industries, they need to switch into the mode of horizontal, vertical and end-to-end intelligent collaboration [17]. One is to realize the vertical information collaboration of management flow, production flow and logistics within the enterprise. Cigarette manufacturing includes both discrete and process-oriented manufacturing, which is relatively complex. The second is to connect with external value chain links such as material supply and product sales. This can achieve horizontal information collaboration from raw and auxiliary material supply. The third is to establish an end-to-end integrated collaboration system based on horizontal and vertical data collaboration, such as equipment procurement, installation, commissioning, operation, maintenance, etc.
Hochstein and Zhu proposed the overall architecture of the theoretical system of intelligent manufacturing with eight modules for the development process of intelligent manufacturing [18]. Jiang et al. proposed an industrial Internet system architecture 2.0 with Chinese characteristics for guiding the construction of industrial Internet platforms [19]. In practice, Chien and Lan proposed a software framework for intelligent manufacturing-supported platform in manufacturing enterprises [20]. Zhang et al. proposed a technical framework for an intelligent coal mine big data integration platform, in order to satisfy the needs of intelligent development in the coal mining industry. Specifically, data analysis techniques based on artificial intelligence are embedded into the designed platform [21]. Dolgui et al. proposed a five-layer architecture of Cyber-Physical System (CPS) for cigarette factories and gave some latent application cases. This research provides construction ideas for tobacco manufacturing enterprises to realize intelligent manufacturing [22].
In terms of current complete procedures of tobacco production manufacturing, digital support decision-driven centralized scheduling remains an urgent demand. Because the complete tobacco manufacturing process is composed of many components, including tobacco planting, tobacco picking, tobacco processing, cigarette packaging, product integration, logistics transportation, etc. Currently, all these components are administrated by different departments independently. In other words, their organization and operation are relatively distributed due to the lack of centralized information integration platform. Such conventional working mode is not beneficial to realizing effective management towards the complete production manufacturing process. Based on the realistic circumstances of production manufacturing process in tobacco industry, this paper explores applications of the new generation of intelligent production control method into tobacco industry [23], [24]. In detail, this work manages to design a centralized information integration platform for tobacco industry, so as to realize efficient optimal scheduling towards all the components of the whole production manufacturing process. The main goal is to improve intelligence level in smart manufacturing process in tobacco industries. As for technical road, this paper proposes a five-layer architecture for the new generation of intelligent production control technology. On this basis, functional architecture and key technologies are specifically designed for each level. This has important reference value for the construction of intelligent factories in the industry.

II. OVERVIEW OF INTELLIGENT CIGARETTE FACTORY A. SCENARIO AND ANALYSIS OF CIGARETTE FACTORY
At present, the manufacturing process of cigarette factories is distributed in various links such as cigaret primacy processing, roll jointing, packaging, energy, logistics, etc. The intelligent control system, as the core carrier of manufacturing, is shown in Figure 1. Upward, it interacts with the enterprise's production management execution system (MES). It is responsible for the management and control of production 33028 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.  equipment [25]. As the intelligent brain of each link in manufacturing, it can not only influence the manufacturing efficiency, but also closely relate to energy consumption. The higher the degree of intelligence of the system, the more obvious the effect of cost reduction and efficiency increase. The intelligent control system of cigarette factory mainly includes four core manufacturing systems: cigaret primacy processing centralized control, packet centralized control, energy centralized control, and logistics centralized control. Each centralized control system has a clear hierarchy and is operated independently distributed in a manufacturing area. Deficiencies are summarized in the following three aspects. 1) Less data sharing: The four manufacturing systems are developed, built, deployed and operated independently. And the development language, system architecture, and data interface adopted by the software suppliers of each system are different. Thus, there is less data interaction between the systems and serious information silos. This cannot completely meet requirement of cross-regional data collaboration construction [26].
2) Low degree of intelligence: After a long period of operation, although the system has accumulated a large amount of production and business data, but the data application is limited to simple report analysis. There still lacks data linkage analysis between systems. The problem discovery and processing mainly rely on after-the-fact discovery, manual intervention, or experience processing. For quality tracing and other production management applications, intelligence degree is also at a low level [4].
3) Scattered control process: The traditional production control process is a multi-level management mode divided into departments and regions. It is manifested in cost accounting, raw and auxiliary material supply, energy supply, safety production, quality inspection and other business aspects. The management functions are separated from each other and the process chain is long. Thus, it fails to form a closed-loop management system of production operation and leads to low synergy efficiency of production and manufacturing.

B. THE OBJECTIVE OF INTELLIGENT CIGARETTE FACTORY
The new generation of intelligent control is a large production organization mode that takes production chain, management chain and logistics chain as the core. The goal is to carry out the whole process and optimization from the global perspective of manufacturing management. Through the data integration technology, all manufacturing systems are unified to the intelligent integrated control platform. Thus an intelligent closed-loop system of supervision can be established to realize global collaboration. Hence, it further promotes the deep integration of information flow, value flow and logistics [4]. The new generation of intelligent production control gives a new interpretation to the production organizational structure model [27]. It focuses on solving the problems of poor equipment compatibility, scattered data scale, repeated functional investment, and low degree of data correlation. The main features are summarized as the following three aspects: 1) Unified scheduling and flat management: The integrated centralized control and management platform for cigaret affairs is established to transform the decentralized management into a unified mode. This realizes the flat management of organizational structure and all-round cascade control of VOLUME 11, 2023 production and manufacturing. Hence, the information sharing ability and business synergy efficiency can be enhanced.
2) Partition operation and centralized control: It keeps the bottom control system of production and manufacturing running independently. Unified data standards and communication protocols are applied. Under the ''mass production'' organization mode, heterogeneous data integration technology are employed to realize centralized real-time monitoring of equipment status. Through this, linkage control of manufacturing units can be integrated to promote digitalization, networking and intelligence of factories.
3) Comprehensive perception and deep excavation: Advanced sensing, data collection and network communication technologies are adopted to comprehensively perceive the enterprise's resources in the whole area. Such process is driven by digital models and based on big data, artificial intelligence and other advanced technologies. Thus, allround and multi-angle intelligent analysis of data assets can be conducted, and comprehensive information for one-stop scientific decision-making can be provided.

III. METHODOLOGY A. DESIGN OF FIVE-LAYER ARCHITECTURE FOR INTELLIGENT CIGARETTE FACTORY
Based on the actual production and operation of the tobacco industry, this paper proposes a five-layer architecture design that is consistent with the cigarette industry enterprises, as shown in Figure 2.

1) INTELLIGENT IoT LAYER
The intelligent IOT layer is composed of host devices, auxiliary devices and control system network. It adopts new generation information communication technologies to realize the full connection of man-machine-thing-loop [28]. Through full sensing of multi-source heterogeneous information sensing, it provides data support for inter-process collaborative manufacturing and edge control.

2) INTELLIGENT CENTRALIZED CONTROL LAYER
The intelligent centralized control layer consists of data acquisition system, equipment control system and edge computing system. The data acquisition system can use industrial intelligent gateway, compatible with data interfaces and communication protocols of different equipment suppliers. The edge computing system can be deployed to realize lightweight computing and real-time analysis of the manufacturing process. Their combination can reduce the impact of abnormalities through data interconnection among systems.

3) INTELLIGENT MODEL LAYER
The intelligent model layer digitally describes the entities, relationships, events and other elements of the whole manufacturing process. Its goal is to build a perfect environment for industrial mechanism model, digital twins and industrial data. The industrial mechanism can model the equipment  operation mechanism and provide services for equipment status monitoring. The digital twins includes packaging machine, auxiliary company equipment, cigaret primacy processing, etc. [28]. It provides the basis for the accurate operation of physical equipment in virtual space. The industrial data includes production model, quality model, process model, consumption model The industrial data can provide services for personnel management, process quality, knowledge mapping and other applications. Based on the unified model engine, it realizes the interconnection of control systems in cigaret primacy processing, bale winding, energy and logistics.

4) INTELLIGENT PLATFORM LAYER
Based on the data services provided by digital platforms such as data center and business center, the intelligent platform layer builds a highly scalable and reliable integrated platform for production management. The data center has the ability of unified data collection, aggregation, storage, calculation and service. The business center is an abstraction of the public business services of the enterprise and has the ability of public business reuse. The enterprise integration platform construction applies microservice architecture model according to the actual situation of factory operation. Meanwhile, based on information physical system (CPS) technology, the digital twin model is applied to realize the accurate mapping and real-time operation of physical equipment in virtual space, forming a flexible closed-loop control system with intelligent perception, intelligent analysis, intelligent decision making, accurate execution and intelligent feedback.

5) INTELLIGENT OPERATION LAYER
As the operation brain of the new generation of intelligent integrated control system, the intelligent operation layer is the operation center to realize the integration of intelligent manufacturing. Based on enterprise data assets, it fully utilizes industrial big data technology and artificial intelligence technology such as neural network and machine learning to realize deep mining and multi-dimensional analysis of data assets. Hence, it can provide services for applications such as production prediction, quality tracing, energy prediction, advanced scheduling, data statistics and analysis. Also, it can support scientific and intelligent decision-making of integrated enterprise operation [29], [30].

B. OPTIMIZATION OF CIGARETTE PRODUCTION SCHEDULING
Optimization of cigarette production scheduling is the key to improving the quality of tobacco production. Cigarette production scheduling is based on scientific management of the cigarette production process and pattern optimization to build a fine cigarette production system.
Scheduling strategy is the most fundamental condition to realize intelligent management control towards the whole production workflow. Due to this point, the formulation of scheduling algorithms is in demand here. It is necessary to judge the reasonableness of the scheduling instruction, and set the reasonableness of the scheduling instruction as p(ϵ). The p(ϵ) is calculated as: where g ϵ represents the number of remaining cigarette processes, L ϵ represents the equipment operation efficiency, φ represents the cigarette production, R c represents the tobacco refill, Q a represents the tobacco loss, and t i represents the production time. When 0 < p(ϵ) < φ, it means that the production scheduling order is reasonable.
If the production scheduling instruction is reasonable, the process of scheduling is optimized according to the internal asset structure level of the cigarette manufacturer, and the statistical feature analysis model is constructed and controlled by linear fitting, and the fitting equation of cigarette production scheduling is obtained as: where, ϵ i represents the secondary control term of the scheduling command; ϵ * i represents the cigarette production mapping transfer function; k represents the equipment operation attribution type energy; b represents the cigarette calibration index.
According to the fitting results, the hierarchical structure of efficient cigarette production control is analyzed, and the fuzzy comprehensive evaluation method is used to carry out the production scheduling MES system support design, and the MES system support model is expressed as follows: where, M is the cost overhead of cigarette production, C ij is the fuzzy matrix. According to the above analysis. The construction of the MES system support model is completed. The output calculation of cigarette production scheduling is carried out on the basis of this model to provide a feasible basis for optimizing the cigarette production scheduling method. We construct the mathematical model of efficient cigarette production scheduling under the MES system support model, combine the fuzzy information set of the mathematical model, assign the initial value of the decision function VOLUME 11, 2023 of cigarette production line control, and use the local optimization method to obtain the output value of the associated rule item to realize the output calculation of efficient cigarette production scheduling under the MES system support.
The scheduling of high-efficiency cigarette production is based on the process control and feature analysis of high-efficiency cigarette production, combined with decision making and fuzzy control methods, process effectiveness evaluation of high-efficiency cigarette production, combined with capital investment in cigarette production and the level of assembly line optimization. The mathematical model of efficient cigarette production scheduling is established as follows: where c ij represents the reasonable expectation of the production value level of cigarette production scheduling, and x ij is the output efficiency of the efficient cigarette production line. The fuzzy integrated decision making method is used for the adaptive optimization of the cigarette production scheduling process, and the fuzzy information set S of cigarette production scheduling is constructed and expressed as follows: where U represents the statistical features of cigarette production scheduling, A represents the attribute distribution set of cigarette production scheduling, and V is the set of cigarette production efficiency growth level functions: The attribute distribution of the cigarette production scheduling process is described by using the association rule optimization control method. The output value of the association rule term is obtained by using the local optimization method: where X denotes the feature data value of cigarette production scheduling and R denotes the attribute distribution data value  of cigarette production scheduling. When the X ⊆ U , R ⊆ A condition holds, the fuzzy control method is used to perform statistical analysis and fuzzy decision making for efficient cigarette production scheduling and optimize the adaptive level of cigarette production scheduling. In order to achieve the optimization of cigarette production scheduling, we first need to use the output value of cigarette production under the MES system support model. Then, a quantitative evaluation parameter model of cigarette production scheduling is built based on this model to filter the adaptive learning coefficient. Nextly, a fuzzy reference evaluation set is evaluated to judge whether the data supports the MES system. Finally, the test statistic of cigarette production scheduling is output, and the control process is analyzed combined with the test statistic. After that, the optimized cigarette production scheduling function is obtained. To this end, an efficient cigarette production scheduling method based on the MES system support is proposed. It uses the association rule scheduling method for efficient cigarette production adaptive scheduling of production output values and obtains adaptive learning coefficients. Its calculation process is as follows: 33032 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.
∂ n = |D n | /D n , D n ̸ = 0 0, D n = 0 (11) In the equation, ∂ n is the iteration coefficient of efficient cigarette production scheduling, B l(n) is the adaptive adjustment coefficient of efficient cigarette production scheduling, Wan) is the adaptive learning momentum factor, and D n represents the number of efficient cigarette production scheduling. To construct the quantitative evaluation parameter model of cigarette production scheduling, and the model expression is: where D represents the average adaptive learning coefficient of efficient cigarette production, and A represents the quantified weight value. The adaptive learning coefficients are screened in the quantitative evaluation parameter model, and the fuzzy information set is used to determine whether the data supports the MES system, and the judgment conditions are: where: From the above equation, when DW l i+1 min it means that the adaptive learning coefficient supports the MES system. The process control method is used to realize the fuzzy adaptive control of cigarette production scheduling. The test statistic of cigarette production scheduling is obtained: where, C i is a constant, ω i denotes the MES decision variable of efficient cigarette production scheduling. The F-test method is used. The standard error in the optimal cigarette production scheduling is obtained and is denoted as: where V i . The fuzzy scheduling function of cigarette production scheduling is obtained by using the full cost control method for the process adaptive control of cigarette production scheduling:  Combined with the MES system support model, the control process of the adaptation function for ω is analyzed as: where t ij denotes the model self-adaptation degree, n i the control function threshold.Using the control results,the optimization function of cigarette production scheduling of cigarette production scheduling is constructed,and its expression is: where k represents the function optimization coefficient.
In summary, to complete the optimization of cigarette production scheduling, simulation experiments are needed to test the feasibility and effectiveness of the method.

IV. EVALUATION AND NUMERICAL ANALYSIS A. BASIC SETTING
In order to verify the effectiveness of the proposed fivelayer system, we conducted a comparative analysis of the process index fluctuation rate, labor productivity, equipment downtime, and energy consumption of a cigarette factory after the completion of the transformation using this system. In particular, this work carries out a set of experimental simulations with use of the MATLAB platform. The whole production manufacturing process is simulated in MATLAB by setting simulative production situation with reasonable parameters and variables. There are totally four production lines in the manufacturing process. Among, two of them are the silk production lines. Their production capacity is 5000 kg/h and 4800 kg/h, respectively. The alcoholization time of cut tobacco in the storage cabinet should be more than 4 hours. There are 14 filter rod forming machines and  Figure 3 shows the fluctuation rate of the process index of different production lines in the cigarette factory before and after the modification of the system proposed in this paper. In this paper, we use 6σ to evaluate the fluctuation rate of the process index, and we can see from the figure that 6σ improves from 4.1% to 4.45%, which is an average improvement of about 5%, indicating that the process level of tobacco can be improved based on the system proposed in this paper. Figure 4 shows the energy consumption level per case for different production lines in the cigarette factory before and after the modification of the system proposed in this paper, from which we can see that the energy consumption per case decreases from 12.46 kgce/case to 11.09 kgce/case on average, with an average energy consumption reduction of 10% per case. Figures 5, 6, and 7 show the downtime rates of the filament making equipment and the Overall Equipment Effectiveness (OEE) of the roll jointing and packaging equipment for the different production lines in the cigarette factory before and after the system modification proposed in this paper, respectively. From the figures, we can see that the downtime rate of the filament making equipment decreases by about 8% on average, the operating efficiency of the roll jointing equipment increases by about 0.6%, and the operating efficiency of the packaging equipment increases by about 2.5%. In addition, we have verified and analyzed the key performance indicators using different scheduling strategies. Typical performance evaluation indicators include bottleneck equipment utilization, total workshop output, and on-time order delivery rate, etc. Figure 8 shows the total output of the workshop under different scheduling strategies, from which it can be seen that the total output of the workshop is improved by 7.2%, 8.2% and 8.9% respectively compared with FIFO, EDD and SPT strategies. Figure 9 shows the output of critical equipment under different scheduling strategies, from which it can be seen that the output of critical equipment is improved by 6.5%, 8.1%, and 8.2% respectively compared with FIFO, EDD, and SPT strategies. It shows that the scheduling optimization strategy proposed in this paper can well relieve the production pressure of the critical equipment and thus achieve the goal of improving the overall output. In addition, the comparison of critical equipment utilization with scheduling strategies is given in Table 1. The proposal is compared with three different schemes in terms of three indexes: equipment utilization rate, equipment switching rate and equipment idle rate. For the first two metrics, the proposal can show better evaluation performance compared with others. The on-time delivery rate of orders under different scheduling strategies is given in Figure 10. From the simulation results, we can see that the on-time order delivery rate of the proposed scheduling optimization strategy can reach 97.6%, which is the best among all scheduling optimization strategies.

C. DISCUSSION
This work designs a digital intelligent scheduling scheme for tobacco industry, in order to deal with current information dispersion problem in different procedures. The production process is simulated with use of MATLAB. The main goal of experiments is to explore the promotion gain brought by centralized information scheduling processes. It is compared with three different working strategies, under measurement of three different metrics. Major experimental results are shown from Figure 3 to Figure 10, which show that the proposed five-layer administration architecture can well improve scheduling efficiency of production administration. Thus, the proposal can be regarded as effective solutions for application of smart manufacturing in tobacco industry.
It is also noted that the designed framework is still far from realistic technical prevalence. First of all, industry production occurs in a relatively wide area. It is required to deploy sufficient fundamental conditions to realize the collection and transmission of massive business data. This costs much budget for relevant enterprises in current stage. Secondly, the fast integration of multisource data from different procedures is still a difficult issue in field of data science. For tobacco industry which contains many production procedures, considerable technical breakthrough is still required. From the proposed scheduling algorithm itself, many model parameters rely on manual setting, yet it cannot have the ability to dynamically adjust parameter setting according to real-time working conditions. These points are the future research direction of our team.

V. CONCLUSION
In this paper, the five-layer architecture of the new generation of intelligent production control for intelligent cigarette factory was proposed, and the functional architecture and key implementation technology of each layer were designed. Based on intelligent production control architecture, a production scheduling optimization strategy to optimize the scheduling of cigarette production process is proposed. Finally, this paper presents a comparative analysis of the proposed system's transformation in cigarette factory. In addition, the effectiveness of the proposed production optimization scheduling strategy is verified through simulation. The analysis and simulation show that the proposed architecture with production optimization scheduling strategy improve the scientific management of the cigarette production process, realize the fine management of cigarette production, and improve the self-adaptability and management level of cigarette production scheduling.
Although some outcomes can be observed from this research, there are still some points that need to be dealt in future works. From the perspective of hardware, both stability and security of the proposed five-layer structure still need to be further considered. These two points are also important characteristics to support a robust scheduling control platform. From the perspective of software, there still needs more research effort to promote self-adaptability of the intelligent scheduling control scheme. In all, the designed framework satisfies mainstream tendency of digital tobacco industry. Its reliability and feasibility are still required to be further tested in real-world engineering practice.
XIN LIU received the B.S. degree in management information system from Qingdao University, Shandong, in 1998. His current research interests include frontier information technology, information system architecture, and cigarette intelligent manufacturing. VOLUME 11, 2023