Construction of intelligent integrated model framework for the workshop manufacturing system via digital twin

With the boosting development of the advanced manufacturing industry in the world, the original production pattern transformed from the traditional industries into the intelligence mode is completed with the least delay possible, which are still facing new challenges. The timeliness, stability, and reliability of them are significantly restricted due to the lack of real-time communication. Therefore, a model framework of intelligent workshop manufacturing system based on a digital twin is proposed in this paper, driving the deep information integration among the physical entity, data collection, and information decision-making. The traditional digital twin of conceptualization and fuzziness needs to be refined, optimized, and upgraded on the basis of the four-dimension collaborative model thinking. The model framework of a refined nine-layer intelligent digital twin is established. Firstly, the physical evaluation is refined into entity layer, auxiliary layer, and interface layer, scientific managing the physical resources and the instrument, and coordinating the overall system. Secondly, dividing the data evaluation into the data layer and the processing layer can greatly improve the flexible response-ability and ensure the synchronization of the real-time data. Finally, the system evaluation is subdivided into information layer, algorithm layer, scheduling layer, and functional layer, developing flexible manufacturing plan more reasonably, shortening the production cycle, and reducing logistics cost. Simultaneously, combining SLP and artificial bee colonies is applied to investigate the production system optimization of the textile workshop. The results indicate that the production efficiency of the optimized production system is increased by 34.46%.


Introduction
With the boosting development of the advanced manufacturing industry in the world, the original production pattern transformed from the traditional industry into the intelligence mode is finished with the least delay possible, which are still facing new challenges [1][2][3]. Because of the dynamics of customer orders, the uncertainty of resource acquisition, the complexity of the manufacturing process, and the dynamics of production chains and logistics information in the conventional manufacturing system became worse, which led to the inability to realize quickly the coordinated operation from the perspective of global optimization. Hence, building a well-functioning production system has become an urgent problem for the enterprise [4][5][6].
The prominent problem of the dynamic communication and deep integration of the information, entities, and data from the intelligent production workshop has become the focus of attention [7][8][9]. In view of these issues, scientific researchers put forward the intelligent modeling concept of digital twin which would establish the interactions of various production links [10][11][12][13][14][15]. The investigated results indicated that the digital twin would significantly optimize the layout, production lifecycle, and intelligent scheduling [16,17]. The uncertain factors from the production activities were managed and controlled comprehensively to thus accurately and availably guide the actual production [18,19]. Nevertheless, application of the digital twin to optimize the manufacturing systems had also some limitations: no real-time tracking and responding to the dynamic results of the whole physical entity, a large deviation between the local and the global optimized results, and failure to do the best optimization [20,21]. So far, few papers had been published on the implementation of optimization algorithms in the transforming and upgrading of the digital intelligent manufacturing workshop. In order to reinforce the interaction of the information system in the intelligent manufacturing workshop, it is necessary to combine the algorithm optimization with the advanced technologies composing of the intelligent digital network, the digital platform, and the application architecture.
Systematic layout design (SLP) is a set of methodical, gradual, and applicable methods optimizing the layout and operation unit of the manufacturing workshop and the planning and adjustment of the equipment. Researchers had combined SLP with a genetic algorithm and applied the workshop layout to successfully achieve the reduction of the logistics quantity and transportation cost [22][23][24]. But the results suggested that the SLP method was only limited to the local optimization of the physical entities, which failed to the interconnection of other production systems. This required that an intelligent algorithm is supplied to integrate and communicate the information of each layer [25]. A fresh global optimization intelligent algorithm that emerged in the layout optimization of the industrial workshop is an artificial bee colony (ABC) algorithm based on the swarm intelligence [26][27][28]. When the SLP method was combined with the intelligent algorithm to solve the multi-objective problems of the optimized workshop layout, the former had some problems with strong subjective intention and large error of the scheme design, whereas the initial conditions of the intelligent algorithm, such as the ABC algorithm, were randomly generated, easily leading to the local optimum, and the convergence speed became slower. Hence, the algorithm applied to the digital twin in the traditional manufacturing system still had the following limitations: the specific application model and the integration and management of the physical and information spatial data are lacking and the interaction between workshop physical and information space is an open-loop process [29].
In light of the issues above, the model framework of an intelligent workshop manufacturing system is proposed based on the intelligent digital twin in this paper. Compared with the traditional digital twin [30][31][32], this paper makes the four-dimensional collaborative model as the guiding principle. A more detailed and optimized construction is performed for each structure layer. More emphasis is placed on the dynamics of the data as well as the comparison and interaction between the virtual and the real-time to achieve the real-time, continuous and transient simulation and optimization. Then, the proposed method is applied to a textile manufacturing workshop.
The remaining part of this paper is described in detail below. In the second section, the fundamental thought of constructing digital twin workshop is introduced on the basis of a four-dimensional collaborative model. In the third section, the traditional digital twin framework is refined, optimized, and upgraded. The model framework of a refined nine-layer intelligent digital twin is established. In the fourth section, the framework of the proposed digital twin in workshop manufacturing system is conducted in the case study of a textile workshop, which is optimized for the workshop layout according to the requirements of the textile workshop. Finally, the conclusions and prospects are described.

Digital twin
Digital twin is defined as a technique of digital mapping and twin body model reconstruction in the virtual space conducting full factor elements, such as composition, function, performance, process, and state of the real space objects in the whole lifecycle [33]. Based on the entity object and virtual framework, digital twin can perform hyper-realistic mapping and transmit the twin data feedback into the real space, providing information reference and decision-making support for entity object and enhancing the coupling timeliness between physical objects and digital twin in a timely and accurate manner [34]. The information carrier of the data twin is the software system integrated with digital technologies such as big data and artificial intelligence to obtain data and information beyond the existing cognition and then predict the future development trend of the real space. At the same time, Virtual control of the entity object is realized through the fusion of the virtual reality. Thus, the entity object can implement the transmission and reception of digital twin virtual analysis information through the perception, control, and IoT. Finally, the entity object is optimized and improved.

Basic hierarchy structure of the digital twin
The digital twin falls into three basic parts: physical evaluation, data evaluation, and information evaluation. The physical evaluation represents the interrelation and connection among objective entities for the manufacturing workshop. It plays a role in the main link of the production processing and the connecting hinge of enterprise management. Workshop production activities are carried out by the physical layer. In addition, all aspects of the information data are provided to the upper level by the physical layer, such as material flow, equipment operation data, personnel management information, and production environment monitoring.
The data evaluation is primarily employed to perform acquisition, processing, analysis on the system data. Stored real-time data in the workshop system is widely used in fields of controlling workshop production activities and layout due to its diverse system model, high-speed transmission, multi-source heterogeneous information system.
With the help of the data processing and analysis system, the collection, arrangement, analysis, and conclusion of the data will be performed as the information evaluation being an important decision-making basis for the regulation of workshop production activities. It is a pivotal bridge connecting the physical evaluation and the information evaluation.
According to the production workshop requirements and work dynamic changes, the information evaluation is served as the information management platform of the workshop production activities to realize the comprehensive scheduling operation and management analysis of the model framework system.

The intelligent integrated design ideology of the digital twin
In this paper, as shown in Fig. 1, the integrated design ideology of intelligent manufacturing system mainly has four fundamental characteristics, as follows: In the course of the design and implementation, applying the analysis of the actual demand identifies the primary production targets for the workshop activities in the lifecycle of information systems. The predetermined production target is accomplished by allocating the specific task breakdown and efficient programming reasonably within a given duration. According to the optimized analysis of alternative structural design for the swarm intelligence algorithm, this paper focuses on verifying the correctness of the scheme on the feasible structure of the swarm intelligence algorithm by the experimental data. Consequently, a lifecycle thought provides a reliable guarantee for the specific implementation of the program and operation, and the maintenance guarantee of the daily production.
The design ideology of the physical entity is that the foundation of system design is analyzed from the physical object of the manufacturing workshop as the template, Fig. 1 The integrated design ideology of intelligent manufacturing system including all kinds of machining equipment, underground logistics system, storage equipment system, sub-system controlling system, and overall controlling system in the workshop. Subsequently, with the help of the new physics entity, the idea is successfully explained by the digital twin for analysis from the local to the whole, such as individual equipment, production space, sub-controlling system, and master controlling system.
The process elements of the production are composed of five factors: product, output, technological process, auxiliary department, and time. For example, the logistics demand and material flow of the raw materials, the similarity of operation nature, the continuity of process flow, noise, vibration, and other environmental factors.
Construction of the physical objects in the manufacturing workshop is conducted by virtual space including the 3D model of the workshop, twin data, process optimization, and task programming. Then, the two-way interaction containing information, resources, and data between entity space and virtual space is realized by the controlling system. Simultaneously, the fusion of multi-source data, for instance, the whole process, all elements, the whole system, and the whole discipline, is implemented by the controlling system in the production workshop.

Construction of an integrated model framework
The method of the integrated information system in the traditional production workshop is prevailingly employed to obtain valid data of production and logistics equipment. According to different production tasks, the information collection method for arranging the workshop of work area layout and production activities is mutually independent in the process of data processing, for instance, the implementation process of the manufacturing execution system. In spite of the execution of the system, the automation level has been elevated to a certain extent. Nevertheless, the respective work areas and information exchanges are carried out independently, resulting in the blockage of the information dynamic transmission, lacking the exchange of the information, and lowering the intelligent level of the management and control. To strengthen the interdependence, intelligence, and integration of the production information system, artificial intelligence algorithm, data dynamic acquisition, IoT, and other advanced digital technology are introduced into the production information system to resolve the above information system defects. After considering dimensions of the lifecycle, physical entities, process elements, and virtual space, the intelligent production system of the integrated workshop based on the digital twin is developed in this paper.

Physical evaluation
Construction of the physical evaluation mainly consists of workshop entity, physical modeling, auxiliary detection equipment, and workshop interface layer, as shown in Fig. 2.
The workshop entity is defined as the indispensable entity elements of the daily production and processing in the production workshop including interactions factors comprised of processing equipment, material transport system, operators, and production environment. Meanwhile, the entity elements in accordance with the entity size and function are transformed into the simulation environment, including the equipment model, raw material model, virtual environment, and virtual workers, which is constructed in a Software environment by combining three-dimensional entity model and mapping method. The equipment operational performance, the materials transport, the working conditions, and the production environment need to be assisted by an auxiliary monitoring device, consisting of the sensor, controller, PRID, and industrial camera configuration, to monitor the production requirements in real-time. The real-time data is stored and dynamically transmitted to the control system in the perceiving process. In such cases, the task commands from the interface layer are received and obtained via the controller and the controlling terminal interface. Subsequently, the information control system is fed back to the interface layer accessible regarding the decision information to realize the dynamic interaction of the perceptual controlling and the decision instructions and information.
The interface layer of the workshop, mainly composing of a local area network (LAN), Fieldbus, platform network, server, and client, can achieve physical synchronization of the information between virtual space and physical system. The communication interface network between the virtual space OPC client and the physical system OPC server is established by utilizing LAN [35,36], Fieldbus, and platform information network communication system under the functional definition of the data mapping. The interface information transmission between the virtual space and the physical system is finally available.

Data evaluation
Data evaluation is a technology that can manage, process, and analyze data during the interaction between the database management system and the data processing operation platform, as shown in Fig. 3. The database management system mostly comprises three aspects: logistics data, non-logistical data, and basic data. The logistics data focuses on managing the fresh material consumption, the division of operation area, the transportation path of AGV, and the operation as well as the maintenance of logistics equipment. Whereas, the non-logistical data is divided into the workshop workers, the cycle time of the processing equipment, the relevant data of the production environment, and the inspection of the processing process. Moreover, the basic data aims to cover all aspects dealing with the total area of the manufacturing workshop, the essential occupied area of the processing equipment, and the required area of the reserved safety channel. The whole data system is fundamentally covered by collecting the above-mentioned three broad categories of the data.
The data processing operation platform performed data activities are essentially relegated to the real-time data and the twin data system, such as cloud computing, big data, data storage, and deep learning. Nevertheless, the collection, storage, processing, and analysis of the intelligent production system can be realized by adopting cloud computing and big data processing technology.
The analysis of various industrial network protocols for digital workshop data sources such as production equipment, sensors, control equipment, and data acquisition system in the important production process of the digital manufacturing workshop need to collect data, which is uniform as OPC UA data access interface to ensure the real-time data transmission, security access, storage, management, etc. [35][36][37][38].
Then, the data processing operation platform is triggered by extracting effective information and realizing the association and transmission of the multi-source heterogeneous data. Ultimately, the big data storage and processing is  Figure 4 exhibits that the detailed structure diagram of the information evaluation consists of the system layer, functional layer, algorithm layer, and scheduling layer.

Information evaluation
The system layer is throughout responsible for the whole process of workshop production, including quality traceability, equipment perception, operation and maintenance guarantee, data analysis, and experience learning, which can promote the interconnectivity of each segment.
The functional layer is predominantly consisting of the order analysis, process design, structural analysis, operation and maintenance evaluation, health management, fault prediction, and life evaluation. For the production process in the whole lifecycle of the workshop production system, the service menu bar of the overall design is developed according to the requirements of the system layer divided into various service functions.
The algorithm layer can optimize the workshop requirements by combining artificial intelligence algorithms with the data layer collections under the adjustment of correction conditions and actual constraints. Then, the optimized results are outputted to the receiving layer through the algorithm layer.
The scheduling layer receives the final optimized results outputted by the algorithm layer, which governs the adjusting AGV scheduling, workshop layout optimization, processing equipment management, process optimization, and personnel management to enable a quick response to the dynamic workshop production process and task orders.

Case study
The textile workshop production is used as the case study in this paper. The integrated layout framework of the intelligent production system in the textile workshop is developed via the actual case analysis using the digital twin method. The integrated layout framework includes the whole process of task analysis, physical model, data collection, information analysis, and result optimization.

Construction of the physical evaluation
The spindles are the key object of investigation for the production process of the electronic glass fiber cloth in the intelligent textile workshop and are related to timely feeding the logistics links and operating state of the textile machine. One of the fresh materials is transported by accessing services of AGV to dominatingly provide for three stages: (1) intelligent logistics conveying equipment for weaving process and drawing fresh materials equipment from warehouse district through the AGV services; (2) fresh materials transported to automatic feeding and unpacking area, the manipulator taking out the spindles and putting a double-layer reflux displacement device for transmission according to the feeding transmission requirements; and (3) multiple AGV distributing in different textile regions to carry out the real-time detection of the textile machine and travel between the feeding point and various textile areas to ensure automatic delivery and recycling of the textile spindles.
Based on the process of building the physical evaluation above, 3D model technology is conducted in the light of the processing process above and physical analysis. Structure diagram of the physical layer in the textile workshop is shown in Fig. 5.
Among them, fresh materials are dominatingly executed by accessing services of the AGV system. The transmission routes for the spindle's transportation are shown in Fig. 6.
Dimensions of the textile workshop are 130 m in length 280 m in width. According to the rational analysis of the operation process, the designated operations areas for the textile workshop are divided and labeled in the process of the created physical layer. The designated operation areas  Table 1.

Construction of the data evaluation
The establishment of the data evaluation in the textile workshop aims to evaluate the relationship between logistics and non-logistics of the working areas. The starting point of the investigative engineering layout problems is defined as five basic elements including product P, output Q, process R, auxiliary department S and time T, and are analyzed comprehensively [8]. In this case, the comprehensive relation of each working area and the layout of the workshop is obtained and optimized, respectively. Statistics of the logistics factors comprises of the usage amounts of the spindles and the logistics quantity of other fresh materials. Meanwhile, the non-logistics factors data is collected, such as similarity of the working nature, continuity of the technological process, influence of noise, vibration, and textile random thread on the environmental factors.
Subsequently, essential information of the data evaluation is partially processed and analyzed by the SLP algorithm. Major factors affecting the logistics are logistics quantity, transmission distance, and fresh materials direction transportation between different working departments. SLP analytical methods commonly expressed by the relation diagram method are prevailingly divided into  Fresh material recovery area B 10 5 5  Textile area 1  50  38  6  Textile area 2  50  38  7  Textile area 3  50  38  8  Textile area 4  50  38  9 Fresh material temporary storage area 5 5 10 Automatic unpacking and loading device 5 1. 5 11 Empty container temporary storage area 10 5 12 Double-layer reflux displacement device 22 1.2  [13], and the meaning and relationship are shown in Table 2.
According to the actual production of a textile Co., Ltd, the logistics intensity is obtained by the following summary table, as shown in Table 3.
In accordance with the characteristics of production equipment from the company, reasons for the degree of mutual affinity among the working departments are formulated, as illustrated in Table 4. A relationship diagram of the non-logistics working departments on account of the results formulated in Table 4 is drawn, as indicated in Fig. 7.
Analyses are carried out by the effect of the warehousing logistical and non-logistical factors, the logistics intensity and the non-logistics close degree are classified into A = 4, E = 3, 1 = 2, 0 = 1, U = 0, and X = − 1. Meanwhile, comprehensive relationship value of the working departments is calculated on the basis of the quantifiable levels and the experts determining weights, which is correspondingly transformed into six grades of A, E, I, O, U, and X. Finally, the comprehensive interrelation diagram of the working departments is shown in Fig. 7. According to the real demands of the processing production, proportional values of the logistics and non-logistics are 0.7 and 0.3, respectively.

Construction of the information evaluation
The dynamic information feedback and dynamic information evaluation is obtained and uploaded owing to the logistics information and basic information collected by data evaluation respectively.
Taking workshop layout as an example, a multi-objective optimized function is established to lower the production cost, maximize the production efficiency, and minimize the area occupied by the working area. The formulas are given by: where C is the total costs of production and transportation in the textile workshop, e is the cost of transportation and auxiliary equipment, m is the cost of the processing materials, and h is the cost of the labor workers.
where E is the total efficiency of the production and transportation in the textile workshop, d is the production efficiency of the mechanical processing, and t is the efficiency of the spindle transport.
where S is the total area of the site area required for the production and transportation of the machinery in the textile workshop, l is the area occupation of the equipment, and d is a safe reserved distance between devices.
Therefore, multiple objective functions are converted into a single objective function.
where w 1 is the weight of total production efficiency, w 2 is the weight of total production cost, w 3 is the weight of total production area, and the sum of the three parameters is equal to 1: In order to unify dimensions, the normalized factors f 1 , f 2 , and f 3 are given by:  The ABC algorithm first initializes the defined parameters for the model ensembles. Close relationship between the logistics and the non-logistics of working departments determined using the SLP method limits the range of the initial population, so that the randomly generated population is closer to meeting the real workshop demands. Random generation of NP population is the sum of leading bees and scout bees. If the amount of honey sources is equal to half of the population quantity, superior and lower boundaries of the initial honey source position are corresponding to ub and lb, where ub is lower boundary of the parameter and lb is superior boundary of the parameter. The maximum number of times for the searched honey source is defined as a "limit," which means that a honey source fails to improve after passing the "limit" for several times of the experimentation, and then is abandoned by hired bees. The maximum number of the iterations for a honey source refers to "maxCycle," which is also the number of foraging cycles.

Optimization routine and parameter design
In this case study, the comprehensive affecting factors on transportation cost, production efficiency, and production area are considered. Hence, enactment weighting coefficients in the objective function are w1 = 0.20, w2 = 0.45, and w3 = 0.35, respectively. comprehensive influencing factors between logistics and non-logistics are confirmed by the SLP method, imposing restrictions on the initially generated population by the artificial honey colony algorithm.
Here, the initial population (NP) is 24, number of honey sources (FN) is 12, maximum searched iteration number of a honey source "limit" is 100, and maximum number of the iterations "maxCycle" is 500. In order to avoid accidental Fig. 7 Analysis of the comprehensive interrelation of each working department errors, random simulations of 15 times are conducted by using MATLAB to obtain the optimized solution. The final layout is shown in Fig. 8, GlobaMin = 0.75. In the digital twin-based textile workshop production system, the layout of the workshop is simulated and optimized according to the production factors. Considering the correlation of each process, logistics distance, occupation area, cost, and other aspects, the high-fidelity model is employed for simulation verification. Before and after optimizing, a comparison of the factors layout in the workshop production is shown in Fig. 9. According to the digital twin model framework proposed in this paper, the optimization layout of the intelligent logistics design is completed by making use of the logistics transportation equipment for the automatic textile spindle. It is designed by the physical layer, the data collection and basic operation of each link in the workshop conducted by the data layer, and application of the artificial intelligence algorithm in the information layer. After application, debugging and operation of the cooperative enterprise entity, it is concluded that the timely logistics rate has increased from 73 to 89%, an increased rate of 16%. The utilization rate of the textile equipment has also increased from 78 to 93%.
Auxiliary time of the changing spindles is reduced from 180 to 60 s by coordination of the AGV and cooperative robotic arm. The total production efficiency of a textile workshop is also increased from 55.19% to 89.65%. Meanwhile, the number of staff and workers in the workshop is also reduced from the original eight to three people due to the application of the automated logistics equipment. The effect of the optimized plan implementation indicates that a model framework of digital twin-based intelligent Fig. 8 Overall system structure diagram manufacturing system has a fine effect in the textile manufacturing workshop. After employing the model framework of a digital twin-based intelligent manufacturing system, the optimized results are shown in Table 5.

Conclusions and prospects
In this paper, the main objective of this investigation is the construction of a model framework of the intelligent manufacturing system based on the digital twin. Firstly, with help of the fundamental thought of the four-dimensional collaborative model, a nine-layer model framework of the intelligent digital twin is established by refining, optimizing, and upgrading the traditional digital twin. Secondly, an algorithm for the information layer is optimized and upgraded by combining with SLP and ABC algorithm to overcome its disadvantage. The proposed model framework of the digital twin is conducted in the case study of a textile workshop, which is optimized for the production system according to the requirements. Finally, the presented model framework is applied to a textile manufacturing workshop system. The experimental results manifest that the optimized auxiliary time and transport of the fresh materials are reduced by 66.6% and 58.6%, and the optimized timely logistics and equipment utilization are increased by 16.0% and 15.0%, respectively.
At present, the digital twin technology applied to the manufacturing workshop has not been widely popularized, that is a certain gap between digital twin technology and the described real-physical entity. Future work focuses on virtual debugging, production scheduling, energy consumption Fig. 9 Comparison of the factor layout in the workshop production before and after optimizing management, and other aspects related to the manufacturing workshop, and establishes an effective evaluation tool of the digital twin model. With the continuous integration and fast development of industrial big data, IoT, artificial intelligence, virtual reality, and other technologies, it is believed that digital twin modified technology will have a promising future in the field of intelligent manufacturing.