Automation Pyramid As Constructor for a Complete Digital Twin, Case Study: a Didactic Manufacturing System

sensors

Article Automation Pyramid as Constructor for a Complete Digital Twin, Case Study: A Didactic Manufacturing System

Edwin Mauricio Martinez, Pedro Ponce * , Israel Macias and Arturo Molina

Tecnológico de Monterrey, Department of Mechatronics, School of Engineering and Sciences, Mexico City Campus, Calle del Puente 222, Tlalpan, México City 14380, Mexico; [email protected] (E.M.M.); [email protected] (I.M.); [email protected] (A.M.) * Correspondence: [email protected]

Abstract: Nowadays, the concept of Industry 4.0 aims to improve factories’ competitiveness. Usually, manufacturing production is guided by standards to segment and distribute its processes and implementations. However, industry 4.0 requires innovative proposals for disruptive technologies that engage the entire production process in factories, not just a partial improvement. One of these disruptive technologies is the Digital Twin (DT). This advanced virtual model runs in real-time and can predict, detect, and classify normal and abnormal operating conditions in factory processes. The Automation Pyramid (AP) is a conceptual element that enables the efficient distribution and connection of different actuators in enterprises, from the shop floor to the decision-making levels. When a DT is deployed into a manufacturing system, generally, the DT focuses on the low-level that is named field level, which includes the physical devices such as controllers, sensors, and so on. Thus, the partial automation based on the DT is accomplished, and the information between all manufacturing stages could be decremented. Hence, to achieve a complete improvement of the manufacturing system, all the automation pyramid levels must be included in the DT concept. An Citation: Martinez, E.M.; Ponce, P.; artificial intelligent management system could create an interconnection between them that can Macias, I.; Molina, A. Automation manage the information. As a result, this paper proposed a complete DT structure covering all Pyramid as Constructor for a Complete Digital Twin, Case Study: automation pyramid stages using Artificial Intelligence (AI) to model each stage of the AP based A Didactic Manufacturing System. on the Digital Twin concept. This work proposes a virtual model for each level of the traditional Sensors 2021, 21, 4656. https:// AP and the interactions among them to flow and control information efficiently. Therefore, the doi.org/10.3390/s21144656 proposed model is a valuable tool in improving all levels of an industrial process. In addition, It is presented a case study where the DT concept for modular workstations underpins the development Academic Editors: Sotiris Makris and of technologies within the framework of the Automation Pyramid model is implemented into a Nikolaos Papakostas didactic manufacturing system.

Received: 12 May 2021 Keywords: digital twin; automation pyramid; industry 4.0; innovative products; manufacturing Accepted: 2 July 2021 model; educational innovation; higher education Published: 7 July 2021

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in 1. Introduction published maps and institutional affil- iations. New smart-connected devices facilitate information flow to different levels of decision- makers. Consequently, the traditional factories are migrating their manufacturing processes to more innovative, optimized enterprises. Thus, the smart factory concept emerges as dramatically intensified manufacturing intelligence applications throughout the manufacturing and supply chain enterprise [1]. Smart factories are called to transform into Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. companies with more dynamic economic and supply chain services determined by busi- This article is an open access article ness and market demand. Some experts addressing the manufacturing changes attend to distributed under the terms and customer demands, the nature of products, the production economics, and the value chain conditions of the Creative Commons economics [2]. Attribution (CC BY) license (https:// Within the smart factory are emerging systems that can supplant or complement the creativecommons.org/licenses/by/ automated processes currently found in most industries worldwide; these are the cyber- 4.0/). physical systems (CPS). These new elements are the fundamental part of recent smart

Sensors 2021, 21, 4656. https://doi.org/10.3390/s21144656 https://www.mdpi.com/journal/sensors Sensors 2021, 21, x FOR PEER REVIEW 2 of 29

Sensors 2021, 21, 4656 2 of 29

physical systems (CPS). These new elements are the fundamental part of recent smart enterprises.enterprises. CPS CPS are collaborating, are collaborating, computational computational entities entities which which connect connect with the with real the phys- real iphysicalcal world world;; the CPS the get CPS online get online information information and process and process datadata using using Internet Internet services services [3]. As [3]. systemsAs systems comprising comprising various various technological technological elements elements,, CPSCPS requires requires knowledge knowledge of many of many dis- ciplinesdisciplines such such as assoftware software engineering, engineering, communica communicationtion protocols, protocols, control control theory, theory, design design strategies,strategies, and knowledgeknowledge ofof novel novel areas areas such such as as edge edge computing computing [4]. [4]. The The literature literature reports re- portsseveral several methodologies, methodologies, frameworks, frameworks and, implementationsand implementations of CPS. of CPS. For For instance, instance, in [5in], [5],a cyber-application a cyber-application framework framework was was proposed, proposed, advocating advocating several several patterns patterns for partiallyfor par- tiallyhandling handling ordered ordered knowledge knowledge to build to build pervasive pervasive applications. applications. TheThe CPS CPS is is comprised comprised of of assorted assorted technological technological elements elements with with different different purposes purposes accordingaccording to the to the task task assigned. assigned. In [6], In [the6], authors the authors classified classified the elements the elements of CPS of CPSfrom froma pro- a ductionproduction perspective: perspective: smart smart objects, objects, CPS CPS objects, objects, and and digital digital-twinning-twinning processes. processes. Then, Then, theythey proposed aa systematicsystematic framework framework for for the the CPS, CPS which, which includes includes two two primary primary states: states the: theinterconnection interconnection of three of three physical physical segments segments and and the communicationthe communication of real of real physical physical and andvirtual virtual devices. devices. RegardingRegarding the the third third element element above, above, the the Digital Digital Twin Twin (DT) is one of the fundamental partsparts of of CPS. CPS. In In recent recent years, years, DT DThas hasacquired acquired relevance relevance due to due its tousefulness its usefulness in the indus- in the try.industry. Several Several authors authors have also have discussed also discussed the definition the definition of this technological of this technological entity. The entity. first definitionThe first definition is attributed is attributed to [6,7], which to [6, 7defined], which the defined DT generally the DT generallyas nonphysical as nonphysical elements describedelements describedby digital model by digitals to represent models to physical represent elements physical’ behaviors elements’ in behaviors real-world in envi- real- ronmentsworld environments [8]. With the [8 experience]. With the and experience the effort and of theapplying effort ofthis applying conceptual this technology, conceptual sometechnology, authors some added authors characteristics added characteristics to the first definition. to the first In [9], definition. the authors In [ 9discussed], the authors that discussed that DT must comprise five segments: the physical, virtual, connection, data, and DT must comprise five segments: the physical, virtual, connection, data, and service seg- service segment. This model was initially proposed by [10], whose authors argued that the ment. This model was initially proposed by [10], whose authors argued that the five ele- five elements must have the same importance in developing any DT. The real-world device ments must have the same importance in developing any DT. The real-world device is the is the base for building the virtual segment; the virtual one allows to run the simulations, base for building the virtual segment; the virtual one allows to run the simulations, deci- decision-making, and control of the physical device. The core of DT is the data form since sion-making, and control of the physical device. The core of DT is the data form since it it can allow generating new knowledge (Figure1). Besides, DT can increment the complete can allow generating new knowledge (Figure 1). Besides, DT can increment the complete performance of manufacturing systems. performance of manufacturing systems.

Figure 1. Digital Twin General Scheme. Figure 1. Digital Twin General Scheme. Through innovative and effective tools such as DTs, developing new manufacturing processesThrough and innovative functions and that effective collaborate tools to such incorporate as DTs, developing the newly new emerging manufacturing CPS into processescompanies and in theirfunctions current that state collaborate is conceivable. to incorporate They can the adapt newly these emerging emerging CPS systems into com- in paniesprocesses in their that alreadycurrent havestate a is certain conceivabl degreee. ofThey automation. can adapt these emerging systems in processesHence, that the already integration have ofa certain standards degree to enable of automation. the communication between different hi- erarchicalHence, levels the integration of manufacturing of standards processes to enable is a standard the communication best practice inbetween the industry. different For hierarchicalinstance, the levels international of manufacturing standard ANSI/ISA processes 95is a usually standard facilitates best practice integrating in the enterprises’ industry. Forfunctions instance, and the the international control systems standard through ANSI/ISA models 95 and usually terminology facilitates that integrating determine enter- the prisesinformation’ functions exchanged and the among control different systems manufacturing through models operation and terminology levels [11 ].that Moreover, determine by theintegrating information new exchanged CPS into the among industrial different processes, manufacturing the standards operation for manufacturing levels [11]. Moreo- execu- ver,tion by processes integrating (such new as ISA-95)CPS into must the industrial be conﬁgured processes, to incorporate the standards new smartfor manufacturing technologies and communication protocols.

Sensors 2021, 21, x FOR PEER REVIEW 3 of 29

Sensors 2021, 21, 4656 3 of 29

execution processes (such as ISA-95) must be configured to incorporate new smart technologies and communication protocols. TheThe AutomationAutomation PyramidPyramid (AP) derives from from the the standard standard ANSI/ISA ANSI/ISA 95. 95. This This concep- con- ceptualtual representation representation suggests suggests the the levels levels defined defined in in the the standard ISA-95ISA-95 (Figure(Figure2 ).2). Some Some effortsefforts toto updateupdate thethe AP AP through through the the concept concept of of digital digital twin twin have have been been made. made. InIn [ 10[10],], a a shopshop floor floor virtualization virtualization case case was was proposed. proposed. This This work work explored explored the the digitalization digitalization of of the the four main shop-floor components: physical, virtual, service, and data shop-floor systems. four main shop-floor components: physical, virtual, service, and data shop-floor systems.

Figure 2. Automation Pyramid according to ISA 95 model. Figure 2. Automation Pyramid according to ISA 95 model.

OnOn the the other other hand, hand, data data management management in thein the manufacturing manufacturing process process must must be considered be consid- relevantered relevant in the in transition the transition towards towards process process virtualization. virtualization. AA data-drivendata-driven andand model-basedmodel-based conceptionconception ofof aa digitaldigital twintwin waswas presentedpresented inin [[1122].]. TheThe model, model, in in combination combination with with machine machine learning learning techniques, techniques, aimed aimed to to achieve achieve quicker quicker improvementimprovement ofof manufacturingmanufacturing systems.systems. DataData managementmanagement inin thethe newnew architectures architectures of of manufacturingmanufacturing processesprocesses isis crucialcrucial forfor achieving achieving a a smart smart factory’s factory’s transition. transition. Therefore,Therefore, somesome research research works works have have focused focused on on proposing proposing thethe mostmost appropriateappropriate modelsmodels andand algo-algo- rithmsrithms for for handling handling the the large large amount amount of of data data obtained obtained from from the the CPS. CPS. In In [ 13[13],], the the authors authors highlightedhighlighted datadata propertiesproperties suchsuch asas data data volume, volume, variety, variety, and and critical critical selection selection to to handle handle thethe information information with with certainty. certainty. Also, Also, different different works works concentrate concentrate their their efforts efforts on on applying applying intelligentintelligent algorithms algorithms to to streamline streamline the the process process to to handle handle large large amounts amounts of of data. data. Undoubt-Undoubt- edly,edly, artificial artificial intelligence intelligence (AI) (AI) algorithms algorithms are are the the resource resource most most suitable suitable to to manage manage data data accurately.accurately. In In [ 14[14],], the the machine machine learning learning (ML) (ML) applications applications in in the the industry industry are are extensively extensively discussed.discussed. TheThe authorsauthors presentedpresented aa broadbroad overviewoverview ofof thethe relevance relevance of of ML ML in in industrial industrial applications.applications. FromFrom a a manufacturing manufacturing perspective, perspective, some some authors authors have have digitalized digitalized AP AP decision-making decision-mak- levels.ing levels. For example, For example in [15, in], the[15] paper, the paper focused focused on the on practical the practical implementation implementation of a digital of a twindigital concept twin concept in a Manufacturing in a Manufacturing Execution Execution System System (MES). (MES). The proposed The proposed digital digital twin wastwin built was tobuilt monitor to monitor energy energy consumption consumption inside inside a laboratory a laboratory dedicated dedicated to Industry to Industry 4.0. Also,4.0. Also, some some authors authors have have proposed proposed modifying modifying the architecture the architecture of MES of fromMES afrom static a statestatic tostate a flexible, to a flexible, dynamic dynamic information information model. model. In [In16 [16],], the the proposal proposal focused focused on on the the case-case- basedbased approach approach to to information information modeling modeling to to generate generate new new manufacturing manufacturing systems systems that that can can integrateintegrate new new flexible flexible materials materials models. models. Thus, Thus, a a link link between between classical classical MES MES and and Enterprise Enterprise ResourceResource Management Management (ERM) (ERM) is is shown. shown However,. However, the the novel novel modeling modeling was was designed designed using us- theing Reference the Reference Architecture Architecture Models Models applied applied to Industry to Industry 4.0 (RAMI4.0). 4.0 (RAMI4.0). Currently,Currently, the the manufacturing manufacturing industryindustry tendstends toto incorporate incorporate smarter smarter elements elements into into moremore sophisticated sophisticated systems. systems. Thus, Thus, manufacturing manufacturing begins begins to have to have more more efficient, efficient, optimized opti- processesmized processes with better with communication better communication and connectivity, and connectivity, which provides which provides openness openness to real- time decision-making at any stage of the production chain. The interactions of technological elements and human capital have been increasing.

Sensors 2021, 21, 4656 4 of 29

Therefore, current manufacturing models must evolve with new proposals incorporating elements that provide decision-makers with a more precise overview that helps them determine the direction of a product in each production stage. Consequently, the relevance of improving manufacturing through virtualizing the processes that comprise it is evident. Unfortunately, the current proposals only consider virtualization in some of the levels that comprise the current AP. As noted in the transition to smarter manufacturing, the free and sufficient flow of information among its levels is crucial, so the virtualization of just one stage is unwise. Consequently, this work conceptualizes a new complete model of the automation pyramid based on the digital twin concept. In addition to being a valuable tool in the decision-making process, total virtualization can facilitate the flow of information among its different levels under a single platform. Also, this paper shows a case study of a DT applied to an educational environment. The characteristics of the proposed model are identified in the DT addressed. Augmented Reality (AR) and Virtual reality (VR) are discussed as part of DT development as emerging technologies.

2. Digital Twin General Concept The DT description was shown in 2003 at the University of Michigan by Grieves in a PLM presentation [10]. Over time, several authors added their perspective to define the digital twin (DT) concept [17,18]. In [19], the authors defined the DT as a king of ultra-high-fidelity simulation, including a health management system, a historical vehicle, and swift data. However, the definition of DT was described in different manners, so the DT definition was not unified. One of the most frequent and accepted definitions of DT was presented in 2012 by [20], where the DT was defined as an integrated, multi-physics, multi-scale, probabilistic simulation of a complex product using the most suitable physical models to emulate the life of its related twin. In [21], it was mentioned that the DT is composed of a physical product, a virtual product, and the connected data that links both products. Also, the authors defined some relevant characteristics of DT: (1) The physical and virtual worlds are combined and interchange information based on virtual models with high fidelity. (2) Interaction and convergence: (a) in physical space, (b) between historical and real- time data, and (c) between physical and virtual space. Related to (a), DT integrates full-flow, full-element, and full-service characteristics, then, the data generated in the physical-space phases can connect. Regarding (b), the DT data requires expert knowledge and accumulates data from deployed real-time systems. Finally, the physical and virtual spaces are smoothly connected, and they interact easily. (3) DT can update data in real-time so that virtual models can tolerate continuous up- grading by relating virtual space with physical space in parallel. In Figure1, the Digital Twin (DT) concept is illustrated. Some authors consider that this representation highlights the essential characteristics that a DT must possess [22]. However, some others have considered that a complete DT model should have essentially five elements in its structure: physical part, virtual part, connection, data, and service [10]. Some authors propose theoretical frameworks which consider the mentioned elements incorporating information processing layers [18]. By its definition, a DT could be a quintessential cyber-physical system when IoT is included as well as cloud computing, according to [22–24]. In [6], the authors proposed a systematic framework for cyber-physical production systems (CPPS). Its processing involves two situations: (1) the interconnection of three physical sections; (2) the mapping and collaboration of physical sections with virtual ones. The application in the manufacturing of DTs is a constant evolution; therefore, the proposal of DT application has different perspectives. In [25], the authors presented a variety of applications in manufacturing environments for the DT. The information, representing the authors’ perspective, is presented in four sections: DT-based production design, smart manufacturing in DT workshop/factory, product DT for usage monitoring, and DT as an enabler for smart MRO. Sensors 2021, 21, 4656 5 of 29

Some proposals focused their efforts on the theoretical foundation for the development of DTs. For example, in [9], the research for developing DTs comprises mainly four parts: DT modeling, simulation and VV&A (Verification, Validation, and Accreditation), data fusion, and interactive collaboration and service. As relevant information, the work presents an interesting review of DT industrial applications, specifically DT-related patents and DT applications by industrial leaders. On the other hand, most of the DTs applied in the industry are closely related to the levels below the automation pyramid (AP) [10]. In [17], the authors proposed a framework of DT-based smart production management and control; then, the authors addressed the main contribution of this work: the prediction services for a satellite assembly shop-floor, using digital twin and Big Data technologies. The case study demonstrated how to apply the proposed framework on a satellite assembly shop floor. Equally, in [21], they proposed a framework for product design, manufacturing, and service-driven by a DT. The authors concluded with discussions of problem solutions about data in the product lifecycle, a new method for digital-twin-driven product design, manufacturing, and service, and, finally, the detailed application methods and framework of digital-twin-driven product design manufacturing and service. Undoubtedly, the data plays a crucial role in the application of DTs. Therefore, theoretical proposals and applied knowledge of the management of data within DT are quite valuable. In [26], the authors addressed representative applications focusing on aligning with their proposed reference model. They presented a general data processing framework for constructing DT. It comprises data acquisition and cleansing, data storage, and time-sensitive data processing. The intrinsic relationship of data with the ERP and MES systems must be considered one of the main factors in the DT concept and its application. The research in [16] focuses on transforming the statistical architecture of MES to flexible and dynamic information models. The main contribution is the case-based presentation of a new approach for information modeling dedicated to new manufacturing systems. The shown models were deployed on the manufacturing system of smart electronic devices production, so they focused on several representative limitations in the previous generation of industrial computer systems. Indeed, the presence of DT will be more constant and have a greater impact on applications within the industry. A well-structured theoretical base, standards for the regulation of its application, and its correct implementation per se, are part of the work that employees of the industry and the academic community must seek and continue to improve. The correct evolution and application of technological elements in the industry will be essential for transforming industrial life. This work seeks to be an innovative and supportive proposal for the industrial community, relating a technological element developed in and for the industry to educate human resources to be realistically prepared for their future work.

3. The ISA 95 Standard and Automation Pyramid Concept This enterprise-control system integration was developed by the International Society of Automation (ISA) [27]. When a consistent and standard communication structure is required for integrating enterprise-control systems, the ANSI/ISA 95 could be implemented. ISA-95 also allows selecting the information transmitted and received between systems. Besides, the complete manufacturing system could be improved [3]. Thus, the ISA 95 standard is a guide to determine the requirements, select manufacturing execution systems (MES) suppliers, and lead the development of MES and databases [28]. As depicted in Figure2, the ISA 95 standard proposes the enterprise manufacturing operation in ﬁve levels. The zero level consists of the industrial process per se, the machinery, and the needed human resources. The subsequent level consists of the automation components: the interaction between the physical components and the more basic control elements such as PLCs and their Sensors 2021, 21, 4656 6 of 29

peripherals, sensors, and actuators in general. This level is mentioned as the industry hardware where the top-level systems’ interaction and the process occur. The equipment can be monitored through a human-machine interface (HMI) or supervisory control and data acquisition systems (SCADAs) at the monitoring and supervision level. Level three controls the manufacturing operations through manufacturing execution systems (MES), which control what processes should be executed and their order. At the top level, the financial, accounting, and marketing programs reside. The Enterprise Resources Planning systems (ERP) usually manage the inventory, billing, accounting, and logistics. This tool supports the accounting for company billings, expenses, and inventory. The ISA-95 standard is based on models through which the interfaces, business systems, and manufacturing control systems are defined. The standard address benefits such as [27]: • The time required for achieving the total production is reduced • Correct tools for integrating control systems in enterprises are included • A better and fasts manner of determining users’ needs • The cost of the automated manufacturing process is reduced • Optimizing supply chains • Reducing lifecycle. The operational models clarify the application functionality about how the information must be used within the different hierarchical levels. The standard determines which information must be exchanged within systems for sales, finance, logistics, maintenance systems, production, and quality areas. The UML models are the basis for the development of interfaces between ERP systems and MES. In the literature, the primary efforts focus on developing proposals for optimizing and applying conceptually and practically the DT concept to the elements located in the first three levels. However, it is relevant for this work to discuss the functions executed by the planning and management levels, MES and ERP systems, respectively.

3.1. Enterprise Resource Planning (ERP) Systems The enterprise resources must be incorporated into an information system efficiently to deploy Enterprise resource planning (ERP). This program coordinates the complete process and collects data analyzed and managed in different areas [29]. Also, in [29], the authors address some benefits of the ERP systems to Industry 4.0 (I4.0) implementations: • A real-time data is monitored and examined so atypical situations could be detected • A set of rules for business can be incorporated into the ERP systems to create an autonomous system • Portable devices are integrated into the ERP to set a communication channel between the manager and manufacturing machines • A well-organized online operation system could be set • The orders’ information of orders can be tracked, and the status is updated Some interesting research about ERP systems is presented in the recent literature. Some of these are about implementing ERP systems and integrating new technologies, such as cloud systems, into the traditional ERP systems [30]. While some authors conceptually integrate new emerging technologies, some proposals aim to implement cloud computing and ERP systems [31]. Compared to a single ERP implementation, the authors found that a hybrid solution is more suitable for interactions with the user. Their methodology undoubtedly opens the way to I4.0 technological integration. The ERP system allows transactions to be managed by the business in the company using a shared, robust database with standard procedures and data exchange between functional areas [32]. In a wide range of production companies, ERP systems support production and distribution. These are implemented to partially integrate and automate financial, resource management, manufacturing, and sales functions [33]. Sensors 2021, 21, x FOR PEER REVIEW 7 of 29

The ERP system allows transactions to be managed by the business in the company using a shared, robust database with standard procedures and data exchange between functional areas [32]. In a wide range of production companies, ERP systems support production and distribution. These are implemented to partially integrate and automate financial, resource management, manufacturing, and sales functions [33]. The personalization of the systems is a trend, especially among medium-sized companies, so ERP plans are offered as a service through providers. Thus, new ways of provid- Sensors 2021, 21, 4656 ing the software are investigated, mainly related to cloud system hosting [34]. Nowadays,7 of 29 accessibility to remote devices is a critical feature of personal and industry devices. Re- mote access to information verification for decision-making is crucial for optimizing man- ufacturingThe personalization processes. That of is the why systems ERP syst is aem trend, providers especially face among these medium-sizedchallenges by offering companies,mobile so- ERPcapable plans ERP are solutions offered as [3 a5]. service through providers. Thus, new ways of providing the softwareStandard are ISA investigated,-95 content allocates mainly relatedspecific to functions cloud system to the hostingvarious levels [34]. Nowadays,of business accessibilityenterprise management to remote devices (ERP) is systems. a critical Some feature of ofthese personal activities and are industry summarized devices. in Remote Figure access2. As it to shows, information information verification flows forbetween decision-making the functions is crucialthat cross for the optimizing enterprise/control manufac- turinginterface. processes. According That to the is why standard, ERP system a “function providers” is a group face these of tasks challenges classified by as offering having mobile-capablea common objective. ERP solutions These are [35 organized]. hierarchically and identified with a name and number.Standard The number ISA-95 identifies content allocatesthe data model specific hierarchy functions level. to the The various Figure 3 levels illustrates of busi- the nessboundary enterprise of the management enterprise/control (ERP) interface systems. [27 Some]. of these activities are summarized in FigureThe2. general As it shows, model informationrepresented in flows Figure between 3 depicts the an functions organizational that cross structure the enter- based prise/controlon functions. interface.Thus, this According work proposes to the standard,a Digital Twin a “function” structure is ausing group functions of tasksclassified as repre- assented having in Figure a common 3. As objective. a result, in These Table are 1 are organized presented hierarchically functions that and correspond identified to with those a nameaccomplished and number. in the TheERP number and MES identifies systems, the on datawhich model the DT hierarchy model is level. based, The as shown Figure 3in illustratesFigure 4. the boundary of the enterprise/control interface [27].

FigureFigure 3.3. Functional enterprise/control (A (A—Order—Order processing, processing, B B—Production—Production Scheduling Scheduling and and C C—— ProductionProduction control)control)

TableThe 1. Production general model Scheduling represented (PS) Details in Figure. 3 depicts an organizational structure based on functions. Thus, this work proposes a Digital Twin structure using functions as represented Production Scheduling (PS) in Figure3. As a result, in Table1 are presented functions that correspond to those accomplishedFunctions in the ERP and MES systems,The information on which the generated DT model or is modified based, as shownby (PS) in FigureDetermine4. the production schedule. The production schedules. Identify long-term raw materials The actual production versus the planned Table 1. Productionrequirements. Scheduling (PS) Details. production. Determine pack-out schedule for The production capacity and resource availability. end products. Production Scheduling (PS) The information generated or modiﬁed by Determine availableFunctions products Current order(PS) status. for sales. Determine the production schedule. The production schedules. Identify long-term raw materials The actual production versus the requirements. plannedproduction. Determine pack-out schedule for end The production capacity and resource products. availability. Determine available products for sales. Current order status.

Sensors 2021, 21, 4656 8 of 29 Sensors 2021, 21, x FOR PEER REVIEW 8 of 29

FigureFigure 4. Automation 4. Automation Pyramid: Pyramid: (a) Traditional(a) Traditional AP andAP and (b) Proposed(b) Proposed DT-based DT-based Automation Automation Pyramid. Pyra- mid. Order processing (OPR) Order processingThe (OPR) tasks assigned to this function are: The tasks• assignedCustomer to this order function handling, are: acceptance, and confirmation.  Customer• orderSales handling, are estimating. acceptance, and confirmation.  Sales are• estimatingWaiver and. reservation handling. •  Waiver and Grossreservation margin handling. reporting. •  Gross marginDetermining reporting. production orders.  DeterminingProduction production scheduling orders. (PS) Functions interact with functions in the production control system through a produc- Production schedulingtion schedule (PS) based on actual production information and production capacity information— Functionsthe interact functions with and functions information in the created production or adapted control by system the production through planning.a produc- The production scheduletion based scheduling on actual functions production are listed information in Table1 . and production capacity information—the functionsProduction and information control (PC) created or adapted by the production planning. The production schedulingThe fundamental functions functions are listed of productin Table control1. are shown below. • Raw materials that have to change into the final product. Production control• Updating (PC) of process plans. The fundamental• Plant functions engineering. of product control are shown below.  Raw materials• Raw that materials’ have to change requirements. into the final product.  Updating• ofGenerate process plans the reports. for performance and costs.  Plant engineering• Assessing. restrictions on capacity and quality.  Raw materials• Self-test’ requirements and diagnose. production and control equipment.  Generate• theCreate reports production for performance standards and and costs directions. for SOPs (Standard Operating Procedures),  Assessing restrictionsrecipes, and on equipmentcapacity and management quality. for defined processing equipment.  Self-test and diagnose production and control equipment. 3.2. Manufacturing Execution Systems (MES)  Create production standards and directions for SOPs (Standard Operating Proce- dures), recipes,Manufacturing and equipment execution management systems for (MES) defined were processing developed inequipment. the early 1990s as a specific application at the shop level. These systems mainly consist of production management and 3.2. Manufacturingcontrol Execution functions Systems such as (MES) scheduling, product quality management, control, production cost control, etc. This software aims to realize integrated production management and Manufacturing execution systems (MES) were developed in the early 1990s as a spe- control on the shop floor [6]. cific application at the shop level. These systems mainly consist of production manage- According to ANSI/ISA-95, MES can link the office ERP systems to the shop-floor ment and control functions such as scheduling, product quality management, control, equipment by employing operational manufacturing management (MOM) functions in the production cost control, etc. This software aims to realize integrated production manage- factories [36]. Some authors include online characteristics in the MES definition, adding ment and control on the shop floor [6]. the MES accumulation of methods and tools to accomplish production, where the online Accordingcharacteristics to ANSI/ISA are-95, used MES as acan connecting link the featureoffice ERP [37]. systems to the shop-floor equipment by empMESloying are theoperational essential manufacturing elements in the management fourth industrial (MOM) revolution functions (I4.0). in MES can the factoriescooperate [36]. Some between authors complex include tasksonline that characteristics associate several in the tools MES and definition, procedures add- to accomplish ing the MES accumulation of methods and tools to accomplish production, where the online characteristics are used as a connecting feature [37].

Sensors 2021, 21, 4656 9 of 29

the resulting collaboration. Owing to the cooperation of different IT systems inside and outside the enterprise, the MES must have open and flexible features. These include: (i) uninterrupted communication with the production level, (ii) conversion of information models online, (iii) transformation of data from the business model to data that can be used in the in-dustrial control systems and (iv) the conversion and processing of the data from the control systems into data that can be understood by different enterprise systems [16]. Nowadays, there are some improving areas in the implementation process of MES. The cost of commercial MES is high; besides, it needs an additional enterprise resource planning (ERP) system. Thus, it is a problem for small manufacturing enterprises (SMEs). On the other hand, the implementation of MES and ERP systems could need specialized hardware. Data analytics is not included for individual employees [38], so the current MES are implemented in large enterprises instead of customizable short-run products produced by SMEs [36]. I4.0 tends to customize the products, which seems not to be compatible with the ideology. Thus, problems identified in the current MES are generated by nonautomatic data entry, nonreal-time capability, interoperability problems, and data sharing among different elements of the MES [39]. Therefore, some proposals have addressed the characteristics of an MES to ensure compatibility with I4.0, including decentralization, mobility, connectivity, and integration on the cloud [40]. Despite the evolution of business and resource management systems, the connection between the process control devices and MES appears missing whenever an MES is applied. This gap reproduces that the two areas evolved unconnectedly and addressed different requirements regarding application context or data characteristics. Furthermore, there is evidence in the industry that different networking concepts also separate these two worlds. For example, LANs are fundamentals in offices, while the control level is the domain of field buses. Hence, the approach links two different network types and the interconnection of two approaches [41,42]. In the beginning, the purposes of MES were focused on finite operational scheduling, resource production management, and dispatching the production units. MES would also supply automated data collection and deliver detailed documents to the workstation. The MES manages uninterruptedly, up-to-the-minute, and on the spot [32].

4. Automation Pyramid and Digital Twin The proposed model elements are defined in detail. As presented in the introduction, several efforts have been made to introduce DT schemes in different levels of the automation pyramid (AP); however, there have been no proposals applying the DT concept to the whole ISA 95 structure. The main reason is that the MES and ERP systems are essentially digital. However, their gaps and improvement areas discussed in previous sections are enough to justify seeking new alternatives in management systems in the industrial environment. Thus, aiming to attend to the gaps found in the literature closely related to real applications in industry, this work seeks to overcome the exposed weaknesses through an innovative tool such as the DT concept. By making the entire AP model virtual through a simulation, the information flow through the stages should be automatically managed through an artificial intelligence algorithm. As shown, this flow would be bidirectional between the levels represented by three DT-based virtual models. Besides, with the entire pyramid virtual, the efficiency of the simulated process can be improved by running a massive quantity of simulations in a shorter time, which would be difficult to achieve with the one-stage pyramid virtualization. Then, the automation pyramid model based on DT is discussed below, and its general configuration is shown in Figure4. The virtualized AP model is made up of 3 DT models. Each DT is interconnected using the automation pyramid structure having and an AI-based management information Sensors 2021, 21, x FOR PEER REVIEW 10 of 29

Sensors 2021, 21, 4656 The virtualized AP model is made up of 3 DT models. Each DT is interconnected10 of 29 using the automation pyramid structure having and an AI-based management information system. The low level is based on math modeling functions; this DT integrates a system.virtual model The low that level accurately is based represents on math modeling the physical functions; structure. this DT integrates a virtual modelBesides, that accurately as an initial represents phase, the it is physical proposed structure. that DT-3 can execute the ERP and MES functionsBesides, presented as an initial at the phase, beginning it is proposed of this work that DT-3(OPR, can PS, execute and PSE, the OC, ERP OP). and Mean- MES functionswhile, the presented role currently at the assigned beginning to of DT this-2 workwill be (OPR, supervisory. PS, and PSE,This OC,functi OP).on Meanwhile,is one of the themain role characteristics currently assigned of the totraditional DT-2 will level be supervisory.-2 systems such This as function SCADA is and one HMI. of the For main DT- characteristics1, the mathematical of the traditionalmodeling function level-2 systems is assigned, such essentially as SCADA representing and HMI. For any DT-1, element the mathematicalat level 0-1 that modeling different function simulation is assigned, software essentially such as Matlab representing and LabVIEW any element performs. at level As 0-1a result, that different the complete simulation representation software such of the as DT Matlab-based and Automation LabVIEW performs.Pyramid could As a result, be de- thefined complete by DT-1, representation DT-2, and DT of-3. the DT-based Automation Pyramid could be defined by DT-1, DT-2, and DT-3. 4.1. AI-Based Management Information System (AI-MIS) 4.1. AI-BasedAt this time, Management artificial Information intelligence System has been (AI-MIS) demonstrated to be a convenient problem- solvingAt this tool time, for decision artificial-making. intelligence Besides has been, applying demonstrated this typeto ofbe mathematical a convenient algorithms problem- solvingfor handling tool for large decision-making. amounts of data Besides, opens applyingthe gap for this recent type machine of mathematical learning algorithmsalgorithms. for handlingThe amoun larget amountsof information of data gener opensated the in gap the for moder recentn manufacturing machine learning process algorithms. is chal- lengingThe to amount analyz ofe and information manage. generatedBesides, this in theinformation modern manufacturingcould be used for process predicting is chal- or lengingimproving to analyze the manufacturing and manage. process. Besides, Thus, this information understanding could information be used for efficiently predicting is or a improvingfundamental the step manufacturing in the automation process. industry Thus, understandingsince several sensors information integrated efficiently into IoT is can a fundamentalsend videos, step pictures, in the audios automation, and so industry on [14]. since Moreover, several the sensors generated integrated data intocan be IoT easily can senddeployed videos, and pictures, used. The audios, performance and so on of [ 14manufacturing]. Moreover, thesystems generated could data be incremented can be easily as deployedthe decision and-making used. The system performance performance of manufacturing is improved and systems the co couldoperative be incremented working. Thus, as thean decision-makingagile manufacturing system process performance can be implemented. is improved andWhen the agile cooperative manufacturing working. based Thus, on anmanagement agile manufacturing information process systems can be is implemented. implemented, When the agile partner manufacturing cooperation based is incre- on managementmented. The informationmanufacturing systems industry is implemented, is a dynamic the virtual partner enterprise cooperation that is can incremented. quickly re- Thespond manufacturing to market demands industry [22 is,43 a dynamic]. virtual enterprise that can quickly respond to marketThus, demands the interactions [22,43]. among the different DTs generate a large amount of various data.Thus, So the the exchange interactions of information, among the the different selection DTs of generatedata types a that large should amount be ofexchanged, various data.the weighting So the exchange of transition of information, times, and the the selection route ofto datathe DT types level that must should be bemanaged exchanged, effi- theciently weighting and safely. of transition For this times, reason, and the the proposed route to the model DT level must must have be an managed information efficiently man- andagement safely. system For this based reason, on machine the proposed learning, model given must the have nature an informationof identification, management classifica- systemtion, and based prediction on machine of these learning, AI algorithms. given the In nature Figure of 5 identification,, the proposed classification, model for the and AI- prediction of these AI algorithms. In Figure5, the proposed model for the AI-MIS is shown. MIS is shown.

FigureFigure 5. 5.Proposed Proposed AI-MIS. AI-MIS.

TheThe management management systemsystem modelmodel for for information information management management amongamong DTsDTs is is based based onon a a presented presented model model [ 14[14].]. The The data data set set preparation preparation would would involve involve extracting extracting the the data data fromfrom a a historical historical and and real-time real-time database, database, examining examining this this structure, structure, and and selecting selecting through through a sample and variable directions. The data pre-processing usually helps to improve the data quality by making data transformations. These shifts are related to ﬁtting the time axis data together to avoid inconsistencies in the essential variable. Also, the outliers Sensors 2021, 21, x FOR PEER REVIEW 11 of 29

Sensors 2021, 21, 4656 a sample and variable directions. The data pre-processing usually helps to improve11 of the 29 data quality by making data transformations. These shifts are related to fitting the time axis data together to avoid inconsistencies in the essential variable. Also, the outliers have haveto be to eliminated be eliminated from from the dataset. the dataset. In addition, In addition, some some missing missing relevant relevant data need data needto be toad- bedressed, addressed, and andfinally, ﬁnally, the scale the scale has to has be to adjusted be adjusted according according to the to process the process variables variables.. Once Oncethe training the training data set data is prepared, set is prepared, the ML the algorithm ML algorithm is selected is selectedaccording according to the data to char- the dataacteristics characteristics (different (different within DTs), within the DTs), complexity the complexity of the data of model, the data etc model,. The ML etc. algorithm The ML algorithmbetween DTs between would DTs be would different; be different; the selection the selection would be would based be on based different on different criteria criteria(Akaike (AkaikeInformation Information Criterion, Criterion, Bayesian Bayesian Information Information Criterion, Criterion, and the Hannan and the– Hannan–QuinnQuinn Criterion) Criterion)[44]. [44]. TheThe proposed proposed data data analytics analytics model model is is intended intended to to be be used used either either ofﬂine offline or or online online accordingaccording to to the the functions functions assigned assigned to to each each DT. DT.

4.2.4.2. Field Field and and Control Control Level Level Model Model MathematicalMathematical models models commonly commonly describe describe the the physical physical elements elements contained contained in the in ﬁeld the andfield control and levels. control These levels. are These usually are simulated usually in simulated dedicated in software dedicated such softwareas MATLAB such and as LabVIEW.MATLAB This and work LabVIEW precisely. This proposes work precisely the mathematical proposes simulationthe mathematical of actuators, simulation sensors, of andactuators, control systemssensors, toand be control developed systems in any to of be these developed platforms. in any To be of correctly these platforms. synchronized To be withcorrectly the rest synchronized of the levels, with these the simulations rest of the must levels, be developedthese simulations with specialized must be hardwaredeveloped thatwith supports specialized real-time hardware simulations. that supports Usually, real-time this simulations. kind of hardware Usually, is this based kind on of ﬁeld- hard- programableware is based gate on arrays field-programable (FPGAs). This gate characteristic arrays (FPGAs). enables This the proper characteristic communication enables ofthe allproper the systems communication integrating of theall the virtual systems pyramid. integrating Thus, the the virtual initial twopyramid. levels Thus, are proposed the initial totwo be virtuallylevels are modeledproposed in to the be dedicatedvirtually modeled software in mentioned the dedicated and software executed mentioned in a real-time and environment.executed in a Figurereal-time6 shows environment. the DT-1. Figure 6 shows the DT-1.

FigureFigure 6. 6.DT-1 DT-1 Virtual Virtual Model. Model.

TheThe virtual virtual representation representation of of the the ﬁrst first levels levels of of the the original original pyramid pyramid will will be be designed designed inin addition addition to to the the 3-D 3-D mathematical mathematical modeling. modeling. The The representation representation of of mechanical mechanical devices devices inin 3D 3D models models is is essential essential for for user user interaction. interaction. Through Through CAD CAD tools, tools, it it is is possible possible to to have have a a virtualvirtual representation representation of of the the physical physical models models of of any any machine. machine. Currently, Currently, CAD CAD modeling modeling softwaresoftware allows allows its its platforms platforms to to interact interact with with mathematical mathematical modeling modeling software software such such as as LabVIEWLabVIEW or or MATLAB. MATLAB. A A server server is requiredis required to interactto interact with with the virtualthe virtual model model to access to access the DTthe remotely. DT remotely. TheThe data data obtained obtained from from DT-1 DT-1 can can be be extracted extracted to to analyze analyze and and apply apply algorithms algorithms that that contributecontribute to to maintenance maintenance prediction, prediction, raw raw material material monitoring, monitoring, and and production production analysis. analysis. It canIt can be availablebe available for for any any ofthe of the rest rest of the of the virtual virtual models. models.

4.3. Supervisory Level The traditional pyramid automation considers SCADA and HIM systems to supervise the production levels. To simulate this characteristic, this work proposes to replace these

Sensors 2021, 21, x FOR PEER REVIEW 12 of 29

Sensors 2021, 21, 4656 4.3. Supervisory Level 12 of 29 The traditional pyramid automation considers SCADA and HIM systems to supervise the production levels. To simulate this characteristic, this work proposes to replace supervisorythese supervisory systems systems with virtual with virtual vision-based vision- monitoringbased monitoring systems. systems. Hence, Hence, the proposed the pro- functionposed function of supervision of supervision will be simulatedwill be simulated in a virtual in a environmentvirtual environment by the DT-2. by the DT-2. Currently,Currently, in in industry, industry, the thesupervision supervision of production of production through through image analysis image analysis is widely is studied.widely studied. Techniques Techniques for classifying for classifying and detecting and detecting failures failures based on based artiﬁcial on artificial intelligence intel- (AI)ligence are usually(AI) areemployed usually employed to ﬁnd anomalies to find anomal in theies production in the production quality. Hence,quality. the Hence, virtual the supervisoryvirtual supervisory level should level should use an use AI-algorithm an AI-algorithm to classify to classify previously previously captured captured images. im- Imagesages. Images with different with different characteristics characteristics would would be stored be stored in an in image an image bank bank according according to the to productionthe production process. process. Later, Later, these these can be can employed be employed to classify to classify those those processed processed in real-time, in real- capturedtime, captured by a high-precision by a high-precision camera. camera. Figure 7Figure shows 7 theshows DT-2 the virtual DT-2 model.virtual model.

FigureFigure 7. 7.DT-2 DT-2 Virtual Virtual Model. Model.

TheThe images images willwill bebe capturedcaptured by the camera, whose whose handling handling software software must must be be oper- op- eratingating in in re real-time.al-time. The The images, images, as asshown shown in inFigure Figure 7, 7are, are extracted extracted from from the the physical physical pro- productionduction process. process. In In the the image image processing processing stage, stage, the the images have aa preparationpreparation process process improvingimproving their their quality quality successfully successfully in in the the interpretation interpretation stage. stage. Frequently, Frequently, real real noisy noisy envi- en- ronmentsvironments can can affect affect the computer the computer vision vision algorithm algorithm because because the lighting the lighting conditions conditions could alsocould change also mechanicalchange mechanical vibrations vibrations in the camera’s in the camera mechanism,’s mechanism, etc. [45]. Therefore, etc. [45]. Therefore, to obtain anto accurateobtain an result accurate from result the selected from the AI algorithm,selected AI the algorithm, real images the mustreal images be pre-processed must be pre to - removeprocessed the imageto remove noise the completely. image noise For completely instance, ensuring. For instance, mechanical ensuring circumstances mechanical such cir- ascumstances proper mobility such as of proper the camera mobility mount of the and camera adequate mount distance and toadequate the capture distance point to can the facilitatecapture techniquespoint can facilitate such as atechniques frame-difference such as approach a frame-difference [46] to remove approach image [46 background.] to remove imageFor background. image representation, the pre-processed images are compared with those pre- selectedFor and image stored representation, in the image the bank. pre The-processed images storedimages should are compared be unique with images those with pre- non-defectiveselected and stored patterns. in the These image images bank. can The be images from astored speciﬁc should subsystem, be unique subarea images of with the product,non-defective or a product patterns. full-view These images to detect can wrong be from positions, a specific for subs instance.ystem, The subarea inspection of the interfaceproduct, mustor a beproduct a type full of- HMIview whereto detect the wrong AI results positions, can be for deployed. instance. The The AI inspection outputs generallyinterface willmust involve be a type the of vision-based HMI where inspection the AI results of a can determined be deployed. product The or AI product outputs component.generally will Also, involve if a faulty the product/componentvision-based inspection is detected, of a deter themined DT-model product must or deploy product a pre-short report with immediate actions to be executed by the operators. component. Also, if a faulty product/component is detected, the DT-model must deploy a 4.4.pre Planning-short report and Managementwith immediate Level actions to be executed by the operators.

As an initial proposal in this work, the planning and management levels are thought to be performed by AL-learning systems. These focus on the functions deployed by the ERP (order processing and production scheduling) and MES (the functions related to production control), as indicated in Figure3. Decision-making systems, such as MES and ERP, can be considered highly nonlinear systems as they involve a considerable variety of scenarios and combinations due to the number of entrances, conditions, functions, and outputs with Sensors 2021, 21, x FOR PEER REVIEW 13 of 29

4.4. Planning and Management Level As an initial proposal in this work, the planning and management levels are thought to be performed by AL-learning systems. These focus on the functions deployed by the ERP (order processing and production scheduling) and MES (the functions related to pro- Sensors 2021, 21, 4656 duction control), as indicated in Figure 3. Decision-making systems, such as MES13 of and 29 ERP, can be considered highly nonlinear systems as they involve a considerable variety of scenarios and combinations due to the number of entrances, conditions, functions, and outputs with which they operate. Usually, AI-learning algorithms such as Multilayer Neu- which they operate. Usually, AI-learning algorithms such as Multilayer Neural Networks ral Networks (MNNs) are used as alternatives to learn linear and nonlinear relationships (MNNs) are used as alternatives to learn linear and nonlinear relationships between input between input and output vectors. and output vectors. RegardingRegarding usingusing MNN,MNN, this this method method is is based based on on primary primary interconnected interconnected neurons neurons or or nodes.nodes. The The neuron neuron receives receives signals signals from from other other neurons; neurons the; the neurons’ neurons inputs’ inputs are are affected affected by associatedby associa weightsted weights so that so thethat output the output signal signal can be can calculated be calculated as the as output the output of an activationof an acti- function.vation function The input. The is input the sum is the of the sum neuron’s of the neuron inputs’ multiplieds inputs multipli by theed associated by the associ weights.ated Theweights interconnection. The interconnec of neuronstion of that neurons are grouped that are into grouped layers into allows layers MNN allows to approximate MNN to ap- nonlinearproximate relationships nonlinear relationships [47]. Thus, [47 the]. Thus, initial the proposal initial proposal is to use is to MNN use MNN to execute to execute the functionsthe functions and and tasks tasks already already mentioned mentioned by by the the ERP ERP and and MES MES systems. systems. Figure Figure8 shows8 shows thethe approachapproach consideringconsidering the the MNN MNN and and the the functions functions mentioned mentioned as aspart part of the of the DT- DT-33 vir- virtualtual model. model. Figure Figure 8 integrates8 integrates a multilayer a multilayer neural neural network network in which in which the inputs the inputs are avail- are availableable and andPCA PCA Data; Data; those those inputs inputs are arerepresented represented by only by only one one input input neuron, neuron, but but they they are areexpanded expanded to a to cluster a cluster of neurons of neurons according according to the to the required required inputs. inputs. On Onthe theother other hand, hand, the theoutputs outputs are areproduction production orders, orders, finished ﬁnished goods goods waiver, waiver, and and PCA PCA Data; Data; those those outputs outputs also alsohave have to be to expended be expended to accomplish to accomplish the the required required outputs. outputs. The The MNN MNN only only represents, represents, in ingeneral general terms, terms, the the input input-output-output relationship, relationship, but but it itdoes does not not represent represent the the total total number number of ofneurons neurons required required in in the the input input and and output output layer. layer.

FigureFigure 8. 8.DT-3 DT-3 Virtual Virtual Model. Model.

TheThe proposalproposal consistsconsists ofof threethree MNNs MNNs that that represent represent the the three three functions functions considered considered forfor this this initial initial proposal. proposal. For For practical practical graphical graphical representation representation purposes, purposes, only only the the first first MNNMNN illustratesillustrates aa generalgeneral configurationconfiguration ofof thethe MNNMNN body.body. It It is is observed observed that that the the input input andand outputoutput layerslayers contemplate the the tasks tasks performed performed by by the the order order processing processing function. function. In In this case, the OPR function has two input tasks and three output tasks. It is worth mentioning that referring to the name of a task involves all the possible values obtained from that specific task. Remember that an artificial neuronal network tries to model the

behavior of a biological neural network. The network’s hidden layers will have a certain number of neurons, each associated with an activation function. In MNN, the body of each neuron represents a linear adder of external stimuli followed by a nonlinear activation function whose task is to use the sum of the stimuli to determine the output activity of the neuron [48]. Sensors 2021, 21, 4656 14 of 29

Some of the inputs and outputs of each network are labeled as “data.” Regarding the PC network, the inputs and outputs with this label are shown for illustrative simplicity, since these tasks consist of two inputs and two outputs, i.e., the value of “QA Data” as input is two sets of values: “Standards and Customers” and “QA results,” according to the functional control model on which this DT is based. In the first OPR network, the PCA Data is the information shared between the PCA and OPR functions as determined in the standard. Accordingly, the input PCA Data value does not correspond to similar values to the output PCA data information. This relationship applies to the rest of the outputs associated with the PC network labeled “Data”. On the other hand, some of the tasks in each of the OPR, PS, and PC networks have an arrow after or before the label with a particular color. Those with the same color indicate that the set of values derived from this task as output will be the same set of input values that some other of the two networks will receive. For instance, the PS network output data set for availability will be the same data set that the OPR network will receive as input. The graphic union between each network indicates that they are part of the same DT. The networks will have to be executed as a single system whose internal subsystems interact simultaneously, yielding partial and final results in real-time. Consider the environment where the activities of the input layer are executed. The expected results are at the exit of the proposed networks. The main objective of applying the MNNs is to have an intelligent virtual system that can learn automatically; thus, the Backpropagation algorithm is proposed for MNN training. Backpropagation networks work under supervised learning, so there must be a set of training instructions that describe each output and its expected output value. To “train” the neural network, one must set a data set containing input signals connected with corresponding targets. In each iteration of the training process, the values of the weights are adjusted using new data from the dataset defined for training. Weight modifications are calculated using the backpropagation error algorithm for supervised training. Each step of the training begins by forcing the inputs out of the training set. Then, it is possible to determine the output values of the signals of each neuron in each layer of the network [48]. The DT-3 virtual model has as its initial main objective to emulate the decision-making carried out in typical ERP and MES systems, focusing (for now) on covering those discussed in the development of this work. Therefore, the DT-3 model can serve as an interface in which the complete detailed information of each task contemplated in real-time can be obtained. The DT-3 can be a valuable tool in simulating scenarios with real-world conditions and real-time responses or programmed responses to be displayed in a certain period. The DT-3 virtual model, with its proposed functions, can monitor production, provide information to control product batches, and label them for identification, information on material waste, machine downtime, and status-of-the-process data. This means the activities related to production control are particularly contemplated for the PC function and all those considered in the ISA-95 standard for the OPR and PS functions.

5. Case Study: Production System for Educational Purposes Using a Digital Twin Topology In the same way that the traditional pyramid can represent all the company levels, the DT model proposed in this work can be applied to a company as a whole, a manufacturing plant, a designated plant area, or a particular process (Figure9). On the other hand, this case study is not considered a complete AI-MIS model since it is an academic application that illustrates the performance of the proposed structure. However, a classiﬁcation model to detect defective based on a simple neuron is included to show the proposed framework using IA. Sensors 2021, 21, x FOR PEER REVIEW 15 of 29

this case study is not considered a complete AI-MIS model since it is an academic appli- Sensors 2021, 21, 4656cation that illustrates the performance of the proposed structure. However, a classification 15 of 29 model to detect defective based on a simple neuron is included to show the proposed framework using IA.

Figure 9. DT-baseFigure AP model 9. DT-base for complete AP model company for complete application company. application.

In the case studyIn the shown case study below, shown the application below, the applicationof the model of theis part model of the is part develop- of the development ment of DTs forof DTsworkstations for workstations applied applied in the teaching in the teaching of undergraduate of undergraduate students. students. In situations ofIn social situations isolation of social (COVID isolation 19), such (COVID as today, 19), such having as today, access having to remote access to remote workspaces becomesworkspaces crucial becomes to day crucial-to-day to day-to-dayactivities. The activities. DT concep The DTt offers concept the offers oppor- the opportunity tunity to run tosystems run systems remotely remotely without withoutpersonal personal presence presence in the workplace in the workplace or classroom or classroom to manipulate the equipment. This case study applies the DT concept to remotely manipulate to manipulate the equipment. This case study applies the DT concept to remotely manip- industrial equipment used as a learning tool for undergraduate students at Tecnologico de ulate industrial equipment used as a learning tool for undergraduate students at Tecno- Monterrey in Mexico City. logico de Monterrey in Mexico City. For the teleoperated manufacturing system, some projects had been developed pro- For the teleoperated manufacturing system, some projects had been developed viding technological elements. In the current DT-based AP model, the Modular Production providing technological elements. In the current DT-based AP model, the Modular Pro- System (MPS) has elements that can be classified into the DT-1 and DT-2 models. Thus, an duction System (MPS) has elements that can be classified into the DT-1 and DT-2 models. overview of this case study is presented first, followed by a detailed description of each Thus, an overview of this case study is presented first, followed by a detailed description element in the proposed DT-based AP model framework. of each element in the proposed DT-based AP model framework. 5.1. Description 5.1. Description The MPS belongs to the remote laboratory network of Tecnologico de Monterrey. The The MPSMPS belongs objective to the isremote to develop laboratory technological network distanceof Tecnologico education de Monterrey. platforms The that facilitate the MPS objectivetransmission is to develop of technological information and distance knowledge education and developplatforms skills, that abilities, facilitate and the competencies transmission obtainedof information through and experimentation knowledge and and develop practice skills, in real abilities environments., and competencies obtained throughThe experimentation MPS has four workstations and practice (Figure in real 10 environments.) that together represent a product line that Sensors 2021, 21, x FOR PEERThe REVIEW MPScan has be four configured, workstations monitored, (Figure programmed, 10) that together and represent remotely manipulateda product line through that a platform16 of 29

can be configured,that manages monitored, access programmed and system, integrity.and remotely manipulated through a platform that manages access and system integrity.

FigureFigure10. 10.Physical Physical Workstations.Workstations.

This access system is a web application stored on a TEC System server, allowing users’ access through a section system; only the user who has reserved a time slot can use the MPS. The students access the virtual laboratory for uploading and configuring the physical system (Figure 11).

Figure 11. Complete Virtual Model of MPS.

It has a three-dimensional computational model in which students can run experi- ments avoiding damages to the physical equipment; they can optimize their user time because the three-dimensional model can be accessed simultaneously by a limited number of users. Once the instructor validates that the students’ settings and programs do not have errors, the instructor generates an access key on an internet page so that students reserve a time slot in the remote lab to test their programs and settings (Figure 12).

Figure 12. View of one module regarding the Virtual Laboratory Platform.

Sensors 2021, 21Sensors, x FOR 2021 PEER, 21 REVIEW, x FOR PEER REVIEW 16 of 29 16 of 29

Sensors 2021, 21, 4656 16 of 29

Figure 10. PhysicalFigure Workstations10. Physical Workstations. . This access system is a web application stored on a TEC System server, allowing users’ This access Thissystem access is a systemweb application is a web applicationstored on a storedTEC System on a TEC server, System allowing server, us- allowing us- accessers’ access through through a section a section system; system; only only the user the user who who has reservedhas reserved a time a time slot canslot usecan theuse ers’ access MPS.through a section system; only the user who has reserved a time slot can use the MPS. the MPS. TheThe studentsstudents access access the the virtualvirtual laboratorylaboratory for for uploadinguploading and and conﬁguring configuring the the physical physical The studentssystem access (Figure the 11 virtual). laboratory for uploading and configuring the physical system (Figuresystem 11). (Figure 11).

Figure 11. CompleteFigureFigure 11. 11. VirtualComplete Complete Model VirtualVirtual of MPS. Model Model of of MPS. MPS.

It has a threeItIt has-hasdimensional aa three-dimensionalthree-dimensional computational computationalcomputational model in which modelmodel students inin whichwhich can studentsstudents run experi- cancan runrun experi-experi- ments avoidingmentsment sdamages avoidingavoiding to damagesdamages the physical toto thethe equipment; physicalphysical equipment; equipment;they can optimize theythey cancan the optimizeoptimizeir user theirtimetheir user usertime time because thebecausebecause three-dimensional the the three-dimensional three-dimensional model can modelbe model accessed can can be besimultaneously accessed accessed simultaneously simultaneously by a limited by by number a a limited limited number number of users. Onceofof users. users. the instructor Once Once the the instructor validates instructorvalidates that validates the thatstudents that the the students’’ settings students settings and’ settings programs and programsand do programs not do not do have not have errors,errors,have the instructorerrors, the instructor the generates instructor generates an generates access an access key an key onaccess onan an internetkey internet on anpage pageinternet so so that that page students students so that reserve students a reserve a timetimereserve slot slot ina in timethe the remote slot remote in labthe lab toremote to test test their theirlab toprograms programs test their and andprograms settings settings and (Figure (Figure settings 12 12).). (Figure 12).

Figure 12. View of one module regarding the Virtual Laboratory Platform. Figure 12.Figure View 12. of Viewone module of one moduleregarding regarding the Virtual the VirtualLaboratory Laboratory Platform Platform..

To access the remote laboratory, students have to install an application speciﬁcally

designed for MPS that allows them to work with the equipment remotely. They can also have visual information in real-time using video cameras (Figure 13). Sensors 2021, 21, x FOR PEER REVIEW 17 of 29

Sensors 2021, 21, 4656 To access the remote laboratory, students have to install an application specifically17 of 29 designed for MPS that allows them to work with the equipment remotely. They can also have visual information in real-time using video cameras (Figure 13).

FigureFigure 13. 13.Application Application toto accessaccess thethe RemoteRemote Laboratory.Laboratory.

5.2.5.2. TheThe MPSMPS inin thethe ContextContext ofof thethe DT-BasedDT-Based AP AP Model Model DT-1:DT-1: CADCAD ModelsModels andand Server—TheServer—The modularmodular trainingtraining systemsystem consistsconsists ofof fourfour recon-recon- ﬁgurablefigurable manufacturing manufacturing stations stations which which reproduce reproduce an an automatic automatic and and modular modular production production system.system. WorkstationWorkstation 1—The 1—The MPS MPS initiates initiates this this workstation workstation for for the the production production and and circulation circulation ofof materials.materials. First,First, itit receivesreceives rawraw materialsmaterials (plastic(plastic boxesboxes withwith twotwo circularcircular spacesspacesand and emptyempty metalmetal cylinderscylinders insideinside thethe boxes)boxes)to to deliver deliver the the end-product end-product boxes boxes already already armed armed withwith two two metal metal cylinders cylinders (“armed (“armed with” with refers” refers to the to cylinders the cylinders in the boxes).in the boxes). Then, it Then, moves it themoves product the product already already armed towardarmed toward the end the of end a conveyor. of a conveyor. Finally, Finally, a manipulating a manipulating arm removesarm removes the assembled the assembled product produc fromt thefrom input the input of the of next the workstation next workstation (Figure (Figure 14). 14).

FigureFigure 14. 14.Virtual Virtual Model Model of of Workstation Workstation 1. 1.

WorkstationWorkstation 2—This 2—This station station process process begins begins when when a a loaded loaded box box (chips) (chips) is is transported transported toto the the conveyor conveyor belt belt by by the the robot robot manipulator manipulator of of Workstation Workstation 1. 1. This This is is an an automatic automatic unit unit forfor sortingsorting andand temporary temporary storage.storage. AlongAlong thethe conveyor,conveyor, path path three three sensors sensors are are arranged arranged thatthat identiﬁesidentifies thethe colorcolor ofof the the box. box. TheThe sensorssensors classifyclassify thethe cargocargo boxbox toto storestore it it in in one one of of thethe three three vertical vertical storages storages arranged arranged along along the the conveyor conveyor belt belt (Figure (Figure 15 15).). The SCADA system counts the parts and sends signals to the workstation controller (PLC) to dispatch a certain number of boxes from the three storages to a second output conveyor belt that moves boxes to the table.

Sensors 2021, 21, x FOR PEER REVIEW 18 of 29

Sensors 2021, 21, 4656 18 of 29 Sensors 2021, 21, x FOR PEER REVIEW 18 of 29

Figure 15. Virtual Model of Workstation 2.

The SCADA system counts the parts and sends signals to the workstation controller Figure(PLC)Figure 15. to15. Virtual dispatchVirtual ModelModel a certain ofof WorkstationWorkstation number 2. of2. boxes from the three storages to a second output conveyor belt that moves boxes to the table. WorkstationWorkstationThe SCADA 3—This3system—This counts workstation the parts is is the the and arranging arranging sends signals and and packing to packing the workstation station. station. It receives It controller receives the theclassified(PLC) classiﬁed to dispatchproduct product ofa certain Worksta of Workstation numbertion 2, ofthen 2, boxes then organizes from organizes the it on three itthe on storagespallets the pallets to to facilitate a to second facilitate its output trans- its transportationportationconveyor tobelt the that to last the moves workstation last workstation boxes toof thethe of table.platform. the platform. Once Oncethe boxes the boxesare organized are organized and armed and armedon theWorkstation onpallets, the pallets, they 3— areThis they placed workstation are placedon a chain onis the a that chain arranging transports that transportsand them packing to them Workstation station. to Workstation It receives 4 (Figure the 4 (Figure16classified). 16 ).product of Workstation 2, then organizes it on the pallets to facilitate its trans- portation to the last workstation of the platform. Once the boxes are organized and armed on the pallets, they are placed on a chain that transports them to Workstation 4 (Figure 16).

FigureFigure 16.16. VirtualVirtual ModelModel ofof WorkstationWorkstation 3. 3.

WorkstationWorkstation 4—This 4—This storage storage and and delivery delivery station station is is responsible responsible for for storing storing the the palletspallets delivereddeliveredFigure 16. byVirtualby WorkstationWorkstation Model of Workstation 33 thatthat containcontain 3. thethe ﬁnishedfinished productproduct inin twotwo verticalvertical storagesstorages forfor subsequentsubsequent release.release. TheThe stationstation consistsconsists ofof twotwo verticalvertical warehouses,warehouses, twotwo turntables,turntables, and and a a manipulatormanipulatorWorkstation armarm with with4—This threethree storage degreesdegrees and ofof delivery freedomfreedom station (DOF) (DOF) is (Figure ( Figureresponsible 17 17).). for storing the pallets deliveredTheThe Server—TheServer by Workstation—The systemsystem 3 that hashas contain anan SQLSQL the server,server, finished whichwhich product allowsallows in different differenttwo vertical operationsoperations storages andand for improvesimprovessubsequent thethe release. encodingencoding The andand station orientationorientation consists of ofof objects. objects.two vertical AA singlesingle warehouses, operationoperation two isis equivalent equivalent turntables, to to and one one a ororma moremorenipulator programsprograms arm with inin aa low-levelthreelow-level degreeslanguage. language. of freedom TheThe serverserver (DOF) hashas (Figure anan ASMLabASMLab 17). platform platformdesigned designed toto givegiveThe studentsstudents Server— easyeasyThe useusesystem ofof aa has remoteremote an SQL lab. lab. server, TheThe platformplatform which allowsallows different monitoringmonitoring operations ofof thethe MPSMPS and throughthroughimproves camerascameras the encoding andand microphonesmicrophones and orientation to to supervise supervise of objects. the the A process. process. single operation is equivalent to one or more programs in a low-level language. The server has an ASMLab platform designed to give students easy use of a remote lab. The platform allows monitoring of the MPS through cameras and microphones to supervise the process.

Sensors 2021, 21, 4656 19 of 29 Sensors 2021, 21, x FOR PEER REVIEW 19 of 29

FigureFigure17. 17.Virtual Virtual ModelModel ofof WorkstationWorkstation 4. 4.

5.3.5.3. DT-2:DT-2: Real-TimeReal-TimeManipulation Manipulation Interface,Interface,Monitoring Monitoring System, System, and and Augmented Augmented and and Virtual Virtual RealityReality Currently,Currently, thethe teleoperatedteleoperated systemsystem hashas aa SCADASCADA systemsystem thatthat performsperforms thethe typicaltypical monitoringmonitoring taskstasks ofof thisthis system.system. Additionally,Additionally, camerascameras havehave beenbeen placedplaced toto monitormonitor thethe entireentire MPSMPS process.process. ForFor now,now, thethe monitoringmonitoring processprocess ofof thethe camerascameras isis performedperformed brieﬂy,briefly, withoutwithout usingusing anyany AIAI algorithmalgorithm toto detectdetect failuresfailures inin thethe productsproducts oror thethe productionproduction line.line. Before reaching this point, it was found that a priority should be a user interface allowing Before reaching this point, it was found that a priority should be a user interface allowing real-time communication of the SCADA system with the MPS. Thus, the solution for real-time communication of the SCADA system with the MPS. Thus, the solution for com- communication between SCADA and MPS is discussed below. munication between SCADA and MPS is discussed below. 5.4. OPC Integration with LabVIEW 5.4. OPC Integration with LabVIEW To achieve communication between LabVIEW and the PLCs in this project, it was necessaryTo achieve to use ancommunication OPC server. The between OPC server LabVIEW allows and declaring the PLCs the sharedin this variablesproject, it used was innecessary the PLC programsto use an OPC and the server. interface The forOPC the server SCADA. allows declaring the shared variables usedShared in the PLC variables programs consist and of the Boolean interface memories, for the byteSCADA. memories, counters, inputs, and outputs.Shared The variables four workstations consist of wereBoolean created memories, in the OPCbyte memories, panel to perform counters, as clientsinputs, that and willoutputs. communicate The four withworkstations the OPC were server. created Table in2 illustratesthe OPC panel some to of perform the shared as clients variables that betweenwill communicate the LabVIEW with interface the OPC and server. the OPC Table server 2 illustrates for Workstation some of 1.the shared variables between the LabVIEW interface and the OPC server for Workstation 1. Table 2. Variables in OPC server and LabVIEW. Table 2. Variables in OPC server and LabVIEW. Name in LabVIEW Name in OPC Server PLC Direction Description EnvironmentName in OPC Name in LabVIEW PLC Description Server Environment DirectionThrough this variable, it is possible to monitor Q1_1 Extended Piston Q1.1 Throughthe this piston variable, position it is possible to Q1_1 Extended Piston Q1.1 M23_0 Enter Button M23.0 Through thismonitor variable, the the piston production position is started Through this variable, it is possible to monitor I1_1 Box sensor I1.1 Through this variable, the production is M23_0 Enter Button M23.0 the sensor state, which detects the boxes started Through this variable, it is possible to Similarly, the variables were declared into the four workstations using LabVIEW. Thus, I1_1 Box sensor I1.1 monitor the sensor state, which detects it was possible to integrate any program, libraries, controls, and modiﬁed indicators into the boxes one single location.

Similarly, the variables were declared into the four workstations using LabVIEW. Thus, it was possible to integrate any program, libraries, controls, and modified indicators into one single location.

Sensors 2021, 21, 4656 20 of 29 Sensors 2021, 21, x FOR PEER REVIEW 20 of 29

5.5. Blocks in the5.5. Interface Blocks in LabVIEW the Interface in LabVIEW Figure 18 showsFigure the complete 18 shows interface the complete of the interface SCADA. of A thedetailed SCADA. description A detailed of each description of each block is below.block is below.

Figure 18. MPS Interface in LabVIEWFigure 18.. MPS Interface in LabVIEW.

Emergency BlockEmergency—In this Block—Inblock, the thisuser block, can stop the the user SCADA can stop system the SCADA process system if an process if an emergency or emergencyerror occurs, or stop error a occurs,specific stopblock, a specificor completely block, orstop completely the program stop from the program from running. running. Selection BlockSelection—In this Block—In block, the this user block, selects the user how selects the system how the work systems manually works manually (push) (push) or automaticallyor automatically (pull). (pull).The production The production of the oftemporary the temporary and assembled and assembled storage storage depends depends only ononly the on stock the in stock the warehouse. in the warehouse. Also, it is Also, possible it is to possible select if tothe select user needs if the user needs a a production basedproduction on the basedtype of on card the or type color. of card or color. CommunicationCommunication Block—A communication Block—A communication alert panel is implemented alert panel is between implemented the between the four modules, fourplus modules,the status plus of critical the status sensors of critical in the sensorsproduction in the process. production The sensor process.’s The sensor’s status for aluminumstatus forcylinders aluminum and the cylinders sensor andfor plastic the sensor cases for are plastic shown cases on the are right shown side on the right side of the box. Theof user the will box. be The able user to identify will be able the toavailability identify the of material availability to produce. of material If there to produce. If there is no availabilityis noof material, availability the of user material, must feed the user module must 1 feedwith module the material 1 with. the material. On the lower leftOn side the lowerof the leftblock, side two of indicators the block, show two indicatorsthe status showof Workstation the status 2. of Workstation The user can detect2. The in userreal-time can detectif the module in real-time is busy if theperforming module isany busy task, performing such as classi- any task, such as fying a new cardclassifying and identifying a new card the andstatus identifying of availability the status of the of carousels. availability In of addition, the carousels. to In addition, identify and correctto identify physical and anomalies, correct physical the upper anomalies, right thepart upper is the right communication part is the communication alerts alerts for module 4. Forfor moduleexample, 4. t Forhe user example, can observe the user when can observea platform when has a been platform received has beenfor received for storage or a differentstorage condition. or a different condition. Also, the communicationAlso, the communication indicators for indicatorsmodule 3 forare moduledisplayed 3 are. For displayed. example, For it can example, it can be seen whether 5 or 10 pieces have been received with aluminum or 5 or 10 pieces classified be seen whether 5 or 10 pieces have been received with aluminum or 5 or 10 pieces classi- by color. Likewise, the user detects errors in the connection between modules 2 and 3 and fied by color. Likewise, the user detects errors in the connection between modules 2 and order errors. The system only allows combinations of pieces whose add up are five or ten; 3 and order errors. The system only allows combinations of pieces whose add up are five any other combination is not processed. or ten; any other combination is not processed. Also, the user can visualize the availability of module 3 and observe its present status: Also, the user can visualize the availability of module 3 and observe its present status: available, stowing tokens, or transporting the tokens to module 4’s reception position. available, stowing tokens, or transporting the tokens to module 4’s reception position. Workstation Block 1—Here, the users can select the number of tokens they want to Workstation Block 1—Here, the users can select the number of tokens they want to manufacture with one or two inserted aluminum. This option is only available if the manufacture with one or two inserted aluminum. This option is only available if the man- manufacturing is done by type. Additionally, the program hides the keyboard when the ufacturing is done by type. Additionally, the program hides the keyboard when the pro- production process runs to avoid overwriting values. duction process runs to avoid overwriting values.

Sensors 2021, Sensors21, x FOR PEER21 REVIEW 21 of 29 2021, , 4656 21 of 29

Workstation Block 2A—Station block 2A allows the user to view the status of the Workstation Block 2A—Station block 2A allows the user to view the status of the three three module carousels. This block indicates the number of pieces in each carousel to ver- module carousels. This block indicates the number of pieces in each carousel to verify the ify the availability of pieces and empty or full carousels. availability of pieces and empty or full carousels. Also, two indicatorsAlso, two report indicators whether report carousels whether 2 carousels and 3 are 2 in and the 3 process are in the of processstoring of storing parts by type.parts The by indicators type. The easily indicators identify easily errors identify in the errors communication in the communication between the between field the ﬁeld devices anddevices the SCADA and the system. SCADA system. WorkstationWorkstation Block 2B— BlockThis all 2B—Thisows the allowsuser to theselect user the to number select the of numbertokens to of request tokens to request module 2, withmodule the 2,restriction with the restrictionthat the sum that of thethe sumcombination of the combination is 5 or 10. Both is 5 or types 10. Bothand types and colors showcolors the two show controllers the two. controllers. The selection The process selection will process begin when will begin the user when selects the user the selects the number of numberchips the ofy chipswant theyto accommodate want to accommodate on a pallet, on followed a pallet, by followed pressing by the pressing Enter the Enter button. button. WorkstationWorkstation Block 4—In Block station 4—In block station 4, the block user 4,can the visualize user can the visualize availability the availability of the of the 20 warehouse20 warehousespaces. If the spaces. indicator If the is indicator red, then is that red, place then is that unavailable place is unavailable;; otherwise, otherwise,the the place is accessibleplace is. accessible. Modular ProductionModular Production System Animation System— Animation—TheThe workstations workstations were simulated were in simulated Solid- in Solid- Works to beWorks integrated to be into integrated the human into-machine the human-machine interface designed interface in LabV designedIEW. This in LabVIEW. an- This imation allowsanimation the operator allows the to observe operator the to observe functions the offunctions the stations of the virtually. stations The virtually. ad- The ad- vantage of vantagehaving this of having simulation this simulationis to simulate is to the simulate process the to processdetect improvements to detect improvements or or failures. failures. The animationThe animationis controlled is controlled by LabVIEW by LabVIEW and CAD and files CAD of ﬁlesthe ofworkst the workstationsations inte- integrated grated intointo the LabVIEW the LabVIEW project. project. The Theworkstation workstation varia variablesbles (inputs, (inputs, outputs, outputs, sensors, sensors, etc.) etc.) found found in thein animation the animation LabVIEW LabVIEW project project are arelinked linked to the to theproject project’s’s shared shared variables variables that that control control the theinterface, interface, making making the theMPS MPS and and the theanimation animation run run synchronously. synchronously. Taking WorkstationTaking Workstation 3 as an example, 3 as an it example, was simulated it was simulatedthe conveyor the belt conveyor movement belt movement through a throughsubproject a subprojectcarried out carried in LabV outIEW in. LabVIEW.This action This was action repeated was for repeated all moving for all moving componentscomponents such as boxes, such platforms, as boxes, platforms,and chains. and Figure chains. 19 depicts Figure 19the depicts indicators the indicatorsand and controls usedcontrols to simulate used to the simulate movements the movements of the station. of the station.

Figure 19. Workstation 3 AnimatioFigure 19.n. Workstation 3 Animation.

Similarly, FigureSimilarly, 20 displays Figure 20 the displays indicators the and indicators controllers and controllersfor the variables for the of variables Work- of Work- station 4. station 4.

Sensors 2021, 21, 4656 22 of 29 Sensors 2021, 21, x FOR PEER REVIEW 22 of 29 Sensors 2021, 21, x FOR PEER REVIEW 22 of 29

Figure 20. Workstation 4 Animation. Figure 20. WoFigurerkstation 20. 4Workstation Animation. 4 Animation. Monitoring System—Regarding the monitoring system, security cameras allow the MonitoringMonitoring System—RegardingSystem—Regarding thethe monitoringmonitoring system,system, securitysecurity camerascameras allowallow thethe physical system to be monitored in real-time, mainly to prevent collision between compo- physicalphysical systemsystem toto bebe monitoredmonitored inin real-time,real-time, mainlymainly toto preventprevent collisioncollision betweenbetween compo-components. The cameras are embedded in the ceiling of the laboratory. The cameras have an IP nents.nents. TheThe camerascameras areare embeddedembedded inin thethe ceilingceiling ofof thethe laboratory.laboratory.The Thecameras cameras have have an anIP IP address that allows remote monitoring through the web (Figure 21). addressaddress thatthat allowsallows remoteremote monitoringmonitoring throughthrough the the web web (Figure (Figure 21 21).).

Figure 21. MPS view of monitoring camera. FigureFigure 21.21. MPSMPS viewview ofof monitoringmonitoring camera.camera. The communication of all workstations could be done via ethernet through the TheThe communication communication of ofall workstationsall workstations could could be done be via done ethernet via ethernet through through the SCADA the SCADA system to ensure continuous production, control, and supervision. Now, the systemSCADA to system ensure continuousto ensure continuous production, production control, and, control supervision., and supervis Now, theion monitoring. Now, the monitoring system is based on the LabVIEW environment. It is using an OPC system that systemmonitoring is based system on is the based LabVIEW on the environment.LabVIEW environment. It is using It an is using OPC systeman OPC thatsystem allows that allows incorporating new variables that are controlled. incorporatingallows incorporating new variables new variables that are that controlled. are controlled. In the MPS, the user currently has total control from the SCADA system to each sta- InIn thethe MPS, MPS, the the user user currently currently has has total total control control from from the the SCADA SCADA system system to each to each station, station, and the total manipulation of production must be from the graphic screen. andtion the, and total the manipulationtotal manipulation of production of production must bemust from be thefrom graphic the graphic screen. screen. Augmented Reality (AR)—The augmented reality (AR) application was made for AugmentedAugmented Reality Reality (AR)—The (AR)—The augmented augmented reality reality (AR) (AR) application application was was made made for for Workstations 3 and 4. The objective was to generate an operation manual for these work- WorkstationsWorkstations 3 3 and and 4. 4. TheThe objectiveobjective waswas toto generategenerate anan operationoperation manualmanual forfor thesethese work-workstations through AR. The AR application was made with the Vuforia software, part of the stationsstations throughthrough AR.AR. TheThe ARAR applicationapplication waswas mademade withwith thethe VuforiaVuforia software, software, part part of of the the PTC package, where it can be designed from two views, 2D and 3D. PTCPTC package,package, wherewhere itit cancan bebe designeddesigned fromfrom two two views, views, 2D 2D and and 3D. 3D.

Sensors 2021, 21, 4656 23 of 29 Sensors 2021, 21, x FOR PEER REVIEW 23 of 29 Sensors 2021, 21, x FOR PEER REVIEW 23 of 29

TheThe AR application application was was observed observed thro throughugh a Thingmark a Thingmark that that works works as a as QR a QRcode. code. The The VuforiaThe AR Viewapplication application was observed made it thro possibleugh a to Thingmark scan and viewthat works the jobs as a hosted QR code. in that The Vuforia View application made it possible to scan and view the jobs hosted in that envi- environment.Vuforia View Figureapplication 22 shows made the it Thingmarkpossible to codes.scan and view the jobs hosted in that environment. Figure 22 shows the Thingmark codes. ronment. Figure 22 shows the Thingmark codes.

Figure 22. Thing mark codes in the AR application. FigureFigure 22.22. ThingThing markmark codescodes inin thethe ARAR application.application. Also, the application permits interacting with the 3D model as a single piece or with Also,Also, the application application permits permits interacting interacting with with the the 3D 3D model model as a as single a single piece piece or with or each element comprising the workstation. Each button has an assigned event in the appli- witheach eachelement element comprising comprising the workstation. the workstation. Each Eachbutton button has an has assigned an assigned event eventin the inappli- the applicationcation when when pressed pressed (see Figure (see Figure 23). 23). cation when pressed (see Figure 23).

Figure 23. 3D model interaction through Vuforia software. FigureFigure 23.23. 3D3D modelmodel interactioninteraction throughthrough VuforiaVuforia software. software. Similarly, the workstations can be visualized using Augmented Reality glasses and Similarly,Similarly, thethe workstationsworkstations cancan bebe visualizedvisualized usingusing AugmentedAugmented RealityReality glassesglasses andand the designed APP, see Figure 24 (AR for Workstations 3). thethe designeddesigned APP,APP, see see Figure Figure 24 24 (AR (AR for for Workstations Workstations 3). 3). The buttons on the panel allow selecting an element from the table, which is indicated TheThe buttonsbuttons onon thethe panelpanel allowallow selectingselecting anan elementelement fromfrom thethe table,table, whichwhich isis indicatedindicated in a virtual model; a pop-up window will be displayed with relevant information on the inin aa virtualvirtual model;model; aa pop-uppop-up windowwindow willwill bebe displayeddisplayed withwith relevantrelevant informationinformation onon thethe selected element. The selection of the PLC is shown in Figure 24. One must scan the Thing- selectedselected element. The The selection selection ofof the the PLC PLC is shown is shown in Figure in Figure 24. One 24. must One scan must th scane Thing- the mark in Figure 25 with the Vuforia View application to access the augmented reality Thingmarkmark in Figure in Figure 25 with 25 with the the Vuforia Vuforia View View application application to to access access the the augmented reality reality model. model.model.

Sensors 2021, 21, 4656 24 of 29 Sensors 2021, 21, x FOR PEER REVIEW 24 of 29 Sensors Sensors 2021 2021, ,21 21, ,x x FOR FOR PEER PEER REVIEW REVIEW 2424 of of 29 29

Figure 24. AR of Workstation 3. FigureFigureFigure 24. 24.24. AR AR of ofof Workstation WorkstationWorkstation 3. 33. .

Figure 25. Thingmark for Workstation 3. FigureFigureFigure 25. 25.25. Thingmark Thingmark for forfor Workstation WorkstationWorkstation 3. 33. . A second application allows accessing the AR model from Workstation 2. This is AA second second application application allows allows accessing accessing the the AR AR modelmodel from from Workstation Workstation 2. 2. This This is is shownA in second Figure application 26. Access to allows this mo accessingdel is done the by AR scanning model fromthe Thingmark, Workstation as 2.shown This in is shownshown inin FigureFigure 2626.. AccessAccess toto thisthis modelmodel isis donedone byby scanningscanning thethe Thingmark,Thingmark, asas shownshown inin Figureshown 27in. Figure 26. Access to this model is done by scanning the Thingmark, as shown in FigureFigureFigure 27 2727...

Figure 26. AR model of Workstation 2. Figure 26. AR model of Workstation 2. FigureFigure 26. 26.AR AR model model of of Workstation Workstation 2. 2.

Sensors 2021, 21, 4656 25 of 29 Sensors 2021, 21,Sensors x FOR PEER2021, 21REVIEW, x FOR PEER REVIEW 25 of 29 25 of 29

Figure 27. ThingmarkFigureFigure 27.27. for ThingmarkThingmark Workstation forfor 2. WorkstationWorkstation 2.2.

Virtual RealityVirtualVirtual (VR) Reality—RealityThrough (VR)—Through(VR) the— Throughvirtual reality thethe virtualvirtual application, realityreality application,oneapplication, can interact oneone canandcan interactinteract andand monitor Workstationsmonitormonitor WorkstationsWorkstations 3 and 4, Three 3 3 and andSiemens 4, 4, Three Three products Siemens Siemens were products products used were to were make used used this to make toapplica- make this this application: application: ProcessProcess tion:Simulate, Process Simulate, OPC Simulate, Scout, OPC andScout, OPC TIA Scout, and Portal. TIA and Portal.With TIA thePortal. With latter the With’s latter’shelp, the thelatter help, PLC’s the help, of PLCWork- the of PLC Workstation of Work- station 3 was3station programmed, was programmed, 3 was programmed, whose whose mainmain taskwhose is task palletizing.main is palletizing. task is palletizing. The animationTheThe of animation animationthe 3D model ofof thethe of 3DWorkstation3D modelmodel ofof Workstation3 replicates its 33 replicatescorrespondingreplicates itsits correspondingcorresponding physical physicalphysical model (DT). modelmodelThe interaction (DT).(DT). TheThe between interactioninteraction both betweenbetween models bothbothis carried modelsmodels out isis through carriedcarried outaout client throughthrough and aana clientclient andand anan OPC server, OPCwhichOPC server,server, obtains whichwhich valuable obtainsobtains information valuablevaluable informationandinformation data from andand the datadata PLC. fromfrom The thethe information PLC.PLC. TheThe informationinformation of available devicesofof availableavailable is giving devicesdevices by the isis givinggiving OPC, byanbyd thethe the OPC,OPC OPC, and anclientd thethe links OPCOPC the clientclient server linkslinks and thethe access serverserveres and and accessesaccesses the offered data.thethe offered offered data.data. Process SimulateProcessProcess virtually SimulateSimulate validates virtuallyvirtually and validatesvalidates simulates andand the simulatessimulates movements thethe of movementsmovements each subsystem ofof eacheach subsystemsubsystem on the workstation.onon thethe workstation. workstation.The interaction TheThe between interactioninteraction this between betweensoftware thisthis and softwaresoftware the variables andand the thedeclared variablesvariables for declareddeclared forfor the PLC has theathe simil PLCPLCar has hasbehavior aa similarsimil ofar the behaviorbehavior virtual ofof and thethe physical virtualvirtual and andsystems physicalphysical in real systemssystems-time. inThein real-time.real OPC-time. TheThe OPCOPC communicationcommunicationcommunication collaborates collaboratesincollaborates a continuous inin aa flow continuouscontinuous of information ﬂowflow ofof informationwithoutinformation interruptions. withoutwithout interruptions.interruptions. The virtual identificationTheThe virtualvirtual identiﬁcationidentification of the output ofof data thethe under outputoutput certain datadata underunder stimuli certaincertain to the stimulistimuli system toto’s theinputthe system’ssystem by ’s inputinput byby actuators, switches,actuators,actuato andrs, switches,switches, sensors andallowsand sensorssensors monitoring allowsallows the monitoringmonitoring system in thethe real systemsystem-time. inin real-time.real-time. On the otherOnOn hand, thethe otherthroughother hand,hand, the throughthroughProcess Simulate thethe ProcessProcess platform, SimulateSimulate one platform,platform, can see onetheone DT cancan of seesee thethe DTDT ofof Workstation 3. With this DT, one can manipulate and monitor Workstation 3 remotely Workstation Workstation3. With this DT,3. With one thiscan DT,manipulate one can andmanipulate monitor andWorkstation monitor Workstation3 remotely 3 remotely (Figure 28). (Figure 28). (Figure 28).

FigureFigure 28. Process 28.FigureProcess Simulate 28. SimulateProcess platform Simulate platform with platformthe with virtual the with virtual model the model ofvirtual DT ofW model DTorkstation Workstation of DT 3 .W orkstation 3. 3.

Sensors 2021, 21, x FOR PEER REVIEW 26 of 29 Sensors 2021, 21, 4656 26 of 29

5.6. DT-3: Neural Network-Based Algorithm for Quality Detection in Boxes Stored in 5.6. DT-3: Neural Network-Based Algorithm for Quality Detection in Boxes Stored in WorkstationWorkstation 22 DuringDuring day-to-dayday-to-day workwork atat WorkstationWorkstation 2,2, itit isis observedobserved thatthat boxesboxes sometimessometimes ex-ex- ceededceeded thethe specifiedspecified heightheight andand weight.weight. Therefore,Therefore, itit isis proposedproposed anan intelligentintelligent systemsystem toto detectdetect thosethose productsproducts thatthat dodo notnot complycomply withwith thethe specifications.specifications. Currently,Currently, workwork isis performedperformed usingusing anan AIAI algorithmalgorithm basedbased onon neuralneural networksnetworks toto classifyclassify the produ productct stored stored in in Workstation Workstation 2. 2. The The product product is classified is classified as defective as defective or non or- non-defectivedefective according according to the to weight the weight and height and height data of data the ofbox, the obtained box, obtained through through the sensors the sensorsattached attached to the band to the conveyor band conveyor and station and chassis, station chassis, respectively. respectively. ThisThis algorithmalgorithm collaboratescollaborates directlydirectly toto guaranteeguarantee productproduct qualityquality (quality(quality assurance)assurance) relatedrelated toto productionproduction control.control. FigureFigure 2929 showsshows thethe generalgeneral classificationclassification algorithmalgorithm model.model.

FigureFigure 29.29. ClassiﬁcationClassification modelmodel toto detectdetect defectivedefective products.products.

TheThe mathematicalmathematical representationrepresentationof ofthe the Perceptron Perceptron is: is:

Y =Y Wx= Wx + b+ b (1)(1) where Y represents the output of the classification model, W represents the set of coef- where Y represents the output of the classiﬁcation model, W represents the set of coefﬁcients ficients or weights of the model, and b is the bias to trigger the Perceptron’s output. The or weights of the model, and b is the bias to trigger the Perceptron’s output. The use of use of a single perceptron model was chosen due to the data distribution that, in this a singleuse of perceptron a single perceptron model was model chosen was due chosen to the due data to distributionthe data distribution that, in thisthat, case, in this is linearlycase, separable,is linearly asseparable, shown in as Figure shown 30 in. The Figure blue 30. points The representblue points a non-defectiverepresent a non piece,-de- andfective the red piece, ones and are defectivethe red ones production are defective pieces. production This datawas pieces. taken This from data Workstation was taken numberfrom 2Workstati throughon the number SCADA 2 system.through the SCADA system.

FigureFigure 30.30. DataData distributiondistribution ofof piecespieces withwith goodgood qualityquality andand badbad quality,quality, accordingaccording toto theirtheir weightweight andand height.height.

ForFor trainingtraining thethe PerceptronPerceptron model,model, itit waswas setset thethe numbernumber ofof epochsepochs toto 10001000 andand thethe learninglearning raterate toto 1.1. TheThe aboveabove datadata waswas divideddivided intointo threethree segments: segments: training,training, validation,validation, andand datadata testing, to to avoid avoid overfitting. overﬁtting. The The obtained obtained perceptron perceptron model model generated generated from from Fig- Figureure 30 30data data can can be beseen seen next next (Equation (Equation (2) (2)):): Y = [0.90958486, −0.816145]x − 1 (2)

Sensors 2021, 21, 4656 27 of 29

If the output value Y is equal to or greater than 0, the new piece x is classiﬁed as a defective piece; conversely, if the output value of y is less than 0, the value is classiﬁed as a non-defective piece. The activation function chosen for this purpose was a single unit step. Due to the simplicity of the model and the data, it could be embedded this model into a microcontroller with low computational requirements, thereby enhancing the inspection process of the pieces for the operator and improving the system’s manufacturing process.

6. Conclusions This paper proposes a reference manufacturing AP model based on the DT concept. The main objective of proposing a new model is the continuous evolution of manufacturing systems incorporating new structures and smart technologies for more flexibility in enterprise operations. Thus, the entire manufacturing system could increment its robustness. Besides, a fast response between the pyramid levels could be achieved. Emerging technological structures such as DT, manufacturing systems can evolve into dynamic systems that offer assistive solutions for decision-making at strategic points in the production chain. In addition, through incorporating artificial intelligence algorithms, new manufacturing systems can autonomously improve their performance securely and supervised. Regarding the case study, undergraduate and graduate students can access learning resources that they do not have available offline by applying new technologies. Those technologies can be run online. Although each application corresponds to a DT of the proposed model, and despite the larger-scale scope that the DT-based AP model may have, the work presented permeates the relevance of the proposed model and the importance of filling new technological gaps. For future work, it is essential to complement each application in the case study to achieve the proposed model in its entirety. In this way, the proposed theoretical model can be reliably evaluated. Sporadically, the development of DT-3 should be thoroughly worked on, which will complement the rest of the work carried out, corresponding to DT-1 and DT-2. In this way, users can have a complete manufacturing system that can be a radically valuable tool for training with a highly technological, avant-garde manufacturing system.

Author Contributions: Conceptualization, E.M.M., P.P., I.M. and A.M.; methodology, E.M.M., P.P. and I.M.; software, E.M.M. and I.M.; validation, E.M.M., P.P., I.M. and A.M.; formal analysis, E.M.M., P.P. and I.M.; investigation, E.M.M., P.P., I.M. and A.M.; resources, E.M.M., P.P., I.M. and A.M.; data curation, E.M.M. and I.M.; writing—original draft preparation, E.M.M., P.P. and I.M.; writing— review and editing, E.M.M., P.P., I.M. and A.M.; visualization E.M.M., P.P., I.M. and A.M.; supervision, E.M.M., P.P., I.M. and A.M.; project administration, E.M.M., P.P. and I.M.; funding acquisition, E.M.M., P.P. and I.M. All authors have read and agreed to the published version of the manuscript. Funding: The authors would like to acknowledge the financial support of Writing Lab, TecLabs, Tecnologico de Monterrey, Mexico, in the production of this work. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: Not applicable. Acknowledgments: The authors would like to acknowledge the technical support of Writing Lab, TecLabs, Tecnologico de Monterrey, Mexico, in the production of this work as well as the Department of Mechatronics. Conflicts of Interest: The authors declare no conflict of interest.

References 1. Davis, J.; Edgar, T.; Porter, J.; Bernaden, J.; Sarli, M. Smart manufacturing, manufacturing intelligence and demand-dynamic performance. Comput. Chem. Eng. 2012, 47, 145–156. [CrossRef] 2. Seely, J.; Kulasooriya, D.; Gifﬁ, C.; Chen, M. El Futuro de la Manufactura; Deloitte University Press: Westlake, TX, USA, 2015. Sensors 2021, 21, 4656 28 of 29

3. Monostori, L.; Kádár, B.; Bauernhansl, T.; Kondoh, S.; Kumara, S.; Reinhart, G.; Sauer, O.; Schuh, G.; Sihn, W.; Ueda, K. Cyber-physical systems in manufacturing. CIRP Ann. 2016, 65, 621–641. [CrossRef] 4. Shi, W.; Cao, J.; Zhang, Q.; Li, Y.; Xu, L. Edge Computing: Vision and Challenges. IEEE Internet Things J. 2016, 3, 637–646. [CrossRef] 5. Lee, J.; Bagheri, B.; Kao, H.-A. Cyber–Physical Systems architecture for Industry 4.0-based manufacturing systems. Manuf. Lett. 2015, 3, 18–23. [CrossRef] 6. Ding, K.; Chan, F.T.; Zhang, X.; Zhou, G.; Zhang, F. Defining a Digital Twin-based Cyber-Physical Production System for autonomous manufacturing in smart shop floors. Int. J. Prod. Res. 2019, 57, 6315–6334. [CrossRef] 7. Grieves, M. Digital Twin: Manufacturing Excellence Through Virtual Factory Replication. International Journal of Production Re- search. 2014. Available online: https://www.researchgate.net/publication/275211047_Digital_Twin_Manufacturing_Excellence_ through_Virtual_Factory_Replication (accessed on 1 July 2021). 8. Qi, Q.; Tao, F. Digital Twin and Big Data Towards Smart Manufacturing and Industry 4.0: 360 Degree Comparison. IEEE Access 2018, 6, 3585–3593. [CrossRef] 9. Tao, F.; Zhang, H.; Liu, A.; Nee, A.Y.C. Digital Twin in Industry: State-of-the-Art. IEEE Trans. Ind. Inform. 2019, 15, 2405–2415. [CrossRef] 10. Tao, F.; Zhang, M. Digital Twin Shop-Floor: A New Shop-Floor Paradigm Towards Smart Manufacturing. IEEE Access 2017, 5, 20418–20427. [CrossRef] 11. ANSI/ISA-95. Enterprise-Control System Integration-Part 5: Business To Manufacturing Transactions; The International Society of Automation: Research Triangle Park, NC, USA, 2018. 12. Jaensch, F.; Csiszar, A.; Scheifele, C.; Verl, A. Digital Twins of Manufacturing Systems as a Base for Machine Learning. In Proceedings of the 2018 25th International Conference on Mechatronics and Machine Vision in Practice (M2VIP), Stuttgart, Germany, 20–22 November 2018; pp. 1–6. [CrossRef] 13. Raptis, T.P.; Passarella, A.; Conti, M. Data Management in Industry 4.0: State of the Art and Open Challenges. IEEE Access 2019, 7, 97052–97093. [CrossRef] 14. Ge, Z.; Song, Z.; Ding, S.X.; Huang, B. Data Mining and Analytics in the Process Industry: The Role of Machine Learning. IEEE Access 2017, 5, 20590–20616. [CrossRef] 15. Cimino, C.; Negri, E.; Fumagalli, L. Review of digital twin applications in manufacturing. Comput. Ind. 2019, 113, 103130. [CrossRef] 16. Cupek, R.; Drewniak, M.; Ziebinski, A.; Fojcik, M. “Digital Twins” for Highly Customized Electronic Devices—Case Study on a Rework Operation. IEEE Access 2019, 7, 164127–164143. [CrossRef] 17. Zhuang, C.; Liu, J.; Xiong, H. Digital twin-based smart production management and control framework for the complex product assembly shop-floor. Int. J. Adv. Manuf. Technol. 2018, 96, 1149–1163. [CrossRef] 18. Zheng, Y.; Yang, S.; Cheng, H. An application framework of digital twin and its case study. J. Ambient. Intell. Humaniz. Comput. 2018, 10, 1141–1153. [CrossRef] 19. Reifsnider, K.; Majumdar, P. Multiphysics Stimulated Simulation Digital Twin Methods for Fleet Management. In Proceedings of the 54th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Boston, MA, USA, 8–11 April 2013. [CrossRef] 20. Glaessgen, E.; Stargel, D. The Digital Twin Paradigm for Future NASA and U.S. Air Force Vehicles. In Proceedings of the 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, Honolulu, HI, USA, 23–26 April 2012. [CrossRef] 21. Tao, F.; Cheng, J.; Qi, Q.; Zhang, M.; Zhang, H.; Sui, F. Digital twin-driven product design, manufacturing and service with big data. Int. J. Adv. Manuf. Technol. 2018, 94, 3563–3576. [CrossRef] 22. Alam, K.M.; El Saddik, A. C2PS: A Digital Twin Architecture Reference Model for the Cloud-Based Cyber-Physical Systems. IEEE Access 2017, 5, 2050–2062. [CrossRef] 23. Liu, C.; Jiang, P.; Jiang, W. Web-based digital twin modeling and remote control of cyber-physical production systems. Robot. Comput. Manuf. 2020, 64, 101956. [CrossRef] 24. Zhang, H.; Zhang, G.; Yan, Q. Digital twin-driven cyber-physical production system towards smart shop-floor. J. Ambient. Intell. Humaniz. Comput. 2019, 10, 4439–4453. [CrossRef] 25. Wang, P.; Luo, M. A digital twin-based big data virtual and real fusion learning reference framework supported by industrial internet towards smart manufacturing. J. Manuf. Syst. 2021, 58, 16–32. [CrossRef] 26. Lu, Y.; Liu, C.; Wang, K.I.-K.; Huang, H.; Xu, X. Digital Twin-driven smart manufacturing: Connotation, reference model, applications and research issues. Robot. Comput. Manuf. 2020, 61, 101837. [CrossRef] 27. ISA Society. Enterprise-Control System Integration. Models, and Terminology; The International Society of Automation: Research Triangle Park, NC, USA, 2000; Available online: https://www.isa.org/standards-and-publications/isa-standards/isa-standards- committees/ (accessed on 1 July 2021). 28. SIEMENS. ISA 95 Framework and Layers. 2019. Available online: https://www.plm.automation.siemens.com/global/es/our- story/glossary/isa-95-framework-and-layers/53244 (accessed on 1 June 2021). 29. Oztemel, E.; Gursev, S. Literature review of Industry 4.0 and related technologies. J. Intell. Manuf. 2020, 31, 127–182. [CrossRef] Sensors 2021, 21, 4656 29 of 29

30. Elmonem, M.A.A.; Nasr, E.S.; Geith, M.H. Benefits and challenges of cloud ERP systems—A systematic literature review. Futur. Comput. Inform. J. 2016, 1, 1–9. [CrossRef] 31. Johansson, B.; Alajbegovic, A.; Alexopoulo, V.; Desalermos, A. Cloud ERP Adoption Opportunities and Concerns: The Role of Organizational Size. In Proceedings of the 2015 48th Hawaii International Conference on System Sciences, Kauai, HI, USA, 5–8 January 2015; pp. 4211–4219. [CrossRef] 32. Aloini, D.; Dulmin, R.; Mininno, V. Risk assessment in ERP projects. Inf. Syst. 2012, 37, 183–199. [CrossRef] 33. Mourtzis, D.; Papakostas, N.; Mavrikios, D.; Makris, S.; Alexopoulos, K. The role of simulation in digital manufacturing: Applications and outlook. Int. J. Comput. Integr. Manuf. 2013, 28, 3–24. [CrossRef] 34. Mourtzis, D.; Doukas, M.; Bernidaki, D. Simulation in Manufacturing: Review and Challenges. Procedia CIRP 2014, 25, 213–229. [CrossRef] 35. Su, C.J. Effective Mobile Assets Management System Using RFID and ERP Technology. In Proceedings of the 2009 WRI International Conference on Communications and Mobile Computing, Kunming, China, 6–8 January 2009; Volume 3, pp. 147–151. [CrossRef] 36. Iarovyi, S.; Mohammed, W.; Lobov, A.; Ferrer, B.R.; Lastra, J.L.M. Cyber–Physical Systems for Open-Knowledge-Driven Manufacturing Execution Systems. Proc. IEEE 2016, 104, 1142–1154. [CrossRef] 37. McClellan, M. Applying Manufacturing Execution Systems; Routledge: London, UK, 1997. 38. Coronado, P.D.U.; Lynn, R.; Louhichi, W.; Parto, M.; Wescoat, E.; Kurfess, T. Part data integration in the Shop Floor Digital Twin: Mobile and cloud technologies to enable a manufacturing execution system. J. Manuf. Syst. 2018, 48, 25–33. [CrossRef] 39. Jeon, B.W.; Um, J.; Yoon, S.C.; Suk-Hwan, S. An architecture design for smart manufacturing execution system. Comput. Des. Appl. 2016, 14, 472–485. [CrossRef] 40. Almada-Lobo, F. The Industry 4.0 revolution and the future of Manufacturing Execution Systems (MES). J. Innov. Manag. 2016, 3, 16–21. [CrossRef] 41. Sauter, T. The continuing evolution of integration in manufacturing automation. IEEE Ind. Electron. Mag. 2007, 1, 10–19. [CrossRef] 42. MESA International. MES Functionalities & MRP to MES Data Flow Possibilities; MESA International: Chandler, AZ, USA, 1997. 43. Song, L.; Nagi, R. Design and implementation of a virtual information system for agile manufacturing. IIE Trans. 1997, 29, 839–857. [CrossRef] 44. Mitchell, T.M. Machine Learning; McGraw-Hill: New York, NY, USA, 1997. 45. Martinez, P.; Ahmad, R.; Al-Hussein, M. A vision-based system for pre-inspection of steel frame manufacturing. Autom. Constr. 2019, 97, 151–163. [CrossRef] 46. Guo, J.; Wang, J.; Bai, R.; Zhang, Y.; Li, Y. A New Moving Object Detection Method Based on Frame-difference and Background Subtraction. IOP Conf. Ser. Mater. Sci. Eng. 2017, 242, 012115. [CrossRef] 47. Gardner, M.; Dorling, S. Artificial neural networks (the multilayer perceptron)—A review of applications in the atmospheric sciences. Atmos. Environ. 1998, 32, 2627–2636. [CrossRef] 48. Cruz, P.P. Inteligencia Artificial. Con Aplicaciones a la Ingeniería; Alfaomega Grupo Editor: Mexico City, Mexico, 2010; p. 378.