Smart Factory Reference Architecture Based on CPS Fractal

Smart Factory Reference Architecture Based on CPS Fractal

9th IFAC Conference on Manufacturing Modelling, Management and 9th IFAC Conference on Manufacturing Modelling, Management and Control 9th IFAC Confere...

566KB Sizes 0 Downloads 1 Views

9th IFAC Conference on Manufacturing Modelling, Management and 9th IFAC Conference on Manufacturing Modelling, Management and Control 9th IFAC Conference on Modelling, Management and Control online at www.sciencedirect.com 9th IFAC Conference on Manufacturing Manufacturing Modelling, Management and Berlin, Germany, August 28-30, 2019 Available 9th IFAC Conference on Manufacturing Modelling, Management and Control Berlin, Germany, August 28-30, 2019 Control Control Berlin, Berlin, Germany, Germany, August August 28-30, 28-30, 2019 2019 Berlin, Germany, August 28-30, 2019

ScienceDirect

IFAC PapersOnLine 52-13 (2019) 2776–2781

Smart Smart Factory Factory Reference Reference Architecture Architecture Based Based on on CPS CPS Fractal Fractal Smart Factory Reference Architecture Based on CPS Fractal Smart Factory Reference Architecture Based on CPS Wei Wu*. Jianfeng Lu, Jr.** Smart Factory Reference Architecture Based on CPS Fractal Fractal

Wei Wu*. Jianfeng Lu, Jr.** HaoJianfeng Zhang*** Wei Lu, HaoJianfeng Zhang*** Wei Wu*. Wu*. Lu, Jr.** Jr.**  Wei Wu*. Jianfeng Lu, Jr.** Hao  Hao Zhang*** Zhang*** Hao Zhang***  *CMIS Research Center, Tongji University , No. 4800 Cao’an Road Jiading District,  *CMIS Research Center, Tongji University , No. 4800 Cao’an Road Jiading District, Shanghai, ChinaUniversity (e-mail: [email protected] tongji.edu.cn). *CMIS Research Center, Tongji ,, No. 4800 Cao’an Road Jiading District, *CMIS Research Center, Tongji University No. 4800 Cao’an Shanghai, China (e-mail: [email protected] tongji.edu.cn). ** CMIS Research Center, Tongji University, No. 4800 Cao’anRoad RoadJiading JiadingDistrict, District, *CMIS Research Center, Tongji University , No. 4800 Cao’an Road Jiading District, Shanghai, China (e-mail: [email protected] tongji.edu.cn). Center, China Shanghai, (e-mail: [email protected] ** CMIS Research Tongji University, No. 4800tongji.edu.cn). Cao’an Road Jiading District, Shanghai, China (e-mail: [email protected]) Shanghai, China (e-mail: [email protected] ** Center, Tongji University, No. 4800tongji.edu.cn). Cao’an Road Jiading District, [email protected]) ** CMIS CMIS Research Research Center, University, No. Shanghai, China ***CMIS Research Center, Tongji Tongji (e-mail: University, No. 4800 4800 Cao’an Cao’an Road Road Jiading Jiading District, District, ** CMIS Research Center, Tongji University, No. 4800 Cao’an Road Shanghai, China (e-mail: [email protected]) ***CMIS Research Center, China Tongji (e-mail: University, No. 4800 Cao’an Road Jiading Jiading District, District, Shanghai, [email protected]) Shanghai, China (e-mail: [email protected]) Shanghai, China (e-mail: [email protected]) ***CMIS Center, No. ChinaUniversity, (e-mail: [email protected]) ***CMIS Research ResearchShanghai, Center, Tongji Tongji University, No. 4800 4800 Cao’an Cao’an Road Road Jiading Jiading District, District, ***CMIS ResearchShanghai, Center, Tongji University, No. 4800 Cao’an Road Jiading District, China (e-mail: [email protected]) Shanghai, China (e-mail: [email protected]) Shanghai, China (e-mail: [email protected]) Abstract: In the Industrial Internet or cloud manufacturing platforms, companies should actively respond Abstract: In the Industrial Internet or cloud manufacturing platforms, companies should actively respond to customized manufacturing requirements, which means that the intelligent utilization ofactively manufacturing Abstract: In the Industrial Internet or cloud manufacturing platforms, companies should respond to customized manufacturing requirements, which means that the intelligent utilization ofactively manufacturing Abstract: In the Industrialadjustment Internet or of cloud manufacturing platforms, companies should respond resources and adaptive production processes should be realized in smart factories. Abstract: In the Industrial Internet or cloud manufacturing platforms, companies should actively respond to customized customized manufacturing requirements, which means means that the theshould intelligent utilization ofsmart manufacturing to manufacturing requirements, which that intelligent utilization of manufacturing resources and adaptive adjustment of production processes be realized in factories. However, related research at home and abroad rarely pays attention to the support of factory model for to customized manufacturing requirements, which means that the intelligent utilization of manufacturing resources related and adaptive adaptive adjustment of abroad production processes should be support realized ofin infactory smart model factories. However, researchadjustment at home and rarelyprocesses pays attention to the for resources and of production be realized smart factories. flexible manufacturing processes. Inand this scenario, based onattention the should connotation of fractal theory, this study resources and adaptive adjustment of production processes should be realized in smart factories. However, related research at home abroad rarely pays to the support of factory model for flexible manufacturing processes. Inand thisabroad scenario, based onattention the connotation of fractal theory, model this study However, related research at home rarely pays to the support of factory for proposes arelated reference architecture, abstracting therarely functions and structures ofsupport manufacturing resources of However, research at home and abroad pays attention to the of factory model for flexible manufacturing processes. In this scenario, based on the connotation of fractal theory, this study proposesmanufacturing a reference architecture, abstracting the functions and structures of of manufacturing resources of flexible processes. In this scenario, based on the connotation fractal theory, this study different scales into resource fractals(RFs) with the nature of CPS (cyber physical system), and flexible processes.fractals(RFs) In this scenario, based onand the structures connotation of fractal theory, this study proposes aascales reference abstracting the functions of manufacturing resources of differentmanufacturing into architecture, resource with the nature of CPS (cyber physical system), and proposes reference architecture, abstracting the functions and structures of manufacturing resources of transforming the into manufacturing requirements into reconfigurable rules(cyber tomanufacturing organize fractals, which proposes a reference architecture, abstracting the functions and structures of resources of different scales resource fractals(RFs) with the nature of CPS physical system), and transforming the into manufacturing requirementswith intothe reconfigurable rules(cyber to organize fractals, which different scales resource fractals(RFs) nature of CPS physical system), and ultimately achieves customized production process. In addition, the effectiveness of the proposed smart different scales into resource fractals(RFs) with the nature of CPS (cyber physical system), and transforming the requirements into reconfigurable rules to organize fractals, which ultimately achieves customized production process. addition, the effectiveness of the proposed smart transforming the manufacturing manufacturing requirements intoIn reconfigurable rules organize fractals, factory modelling reference architecture is process. verified by a case study of the to processwhich of a transforming the manufacturing requirements intoIn reconfigurable rules tomanufacturing organize fractals, which ultimately achieves customized production addition, the effectiveness of the proposed factory modelling reference architecture is verified by a case study of the manufacturing process of a ultimately achieves customized production process. In addition, the effectiveness of the proposed smart smart cylinder product. ultimately achieves customized production process. In addition, the effectiveness of the proposed smart factory modelling modelling reference architecture architecture is is verified verified by by aa case case study study of of the the manufacturing manufacturing process process of of aa cylinder product. reference factory factory modelling reference architecture is verified by a case study of the manufacturing process of a cylinder product. © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. cylinder product. Keywords: intelligent manufacturing, smart factory, fractal theory, CPS, customized production, cylinder product. Keywords: intelligent manufacturing, smart factory, fractal theory, CPS, customized production, intelligent utilization resources. Keywords: intelligentof manufacturing, smart smart factory, factory, fractal fractal theory, theory, CPS, CPS, customized customized production, production, intelligent utilization of resources. Keywords: intelligent manufacturing, Keywords:utilization intelligentof resources. manufacturing, smart factory, fractal theory, CPS, customized production, intelligent intelligent utilization of resources.  intelligent utilization of resources.  activities, which will affect its own production status and  1. INTRODUCTION activities, which will affect its own production status and organizational form. Self-organization and self-adaptation are 1. INTRODUCTION  activities, which which willSelf-organization affect its its own own and production status and and activities, will affect production status organizational form. self-adaptation are 1. reflected in the production capacity allocation based on 1. INTRODUCTION INTRODUCTION activities, which will affect its own production status and In recent years, emerging information and communication reflected organizational form. Self-organization and self-adaptation are organizational form. Self-organization self-adaptation are in the capacity allocation based on 1. INTRODUCTION In recent years, emerging information and Data, communication manufacturing bigproduction data, the choice ofand production methods, organizational form. Self-organization and self-adaptation are technologies such as Internet of Things, Big and Cloud reflected in the production capacity allocation based on In recent emerging information and communication reflected thebigproduction capacity based on data, the choice of allocation production methods, In recent years, years, emerging information and Data, communication technologies such as Internet of Things, Big and Cloud manufacturing and so on.in Moreover, research has made a lot of progress in reflected in the production capacity allocation based on Computing have achieved significant development, manufacturing big data, the choice of production methods, In recent years, emerging information and communication technologies such as Internet of Things, Big Data, and Cloud manufacturing big data, the choice of production methods, and so on. Moreover, research has made a lot of progress in technologies such as Internet of Things, Big Data, and Cloud Computing have achieved significant development, the intelligence of manufacturing resources and the support manufacturing big data, the choice of production methods, promoting the and upgrading of the so research has made lot of progress in technologies such astransformation Internet of Things, Big Data, and Cloud Computing have achieved significant development, and so on. on. Moreover, Moreover, research has resources made aaservices, lotand of the progress in intelligence of manufacturing support promoting the transformation and upgrading of and Computing have achieved significant development, of decision-making and manufacturing and also and so on. Moreover, research has made a lot of progress in manufacturing companies towards intelligent manufacturing the intelligence of manufacturing resources and the support Computing have achieved significant development, promoting the the transformation and upgrading upgrading of of the intelligence of manufacturing resources and the support decision-making and manufacturing services, and also promoting transformation and of manufacturing companies towards intelligent manufacturing explores the advantages ofmanufacturing distributed, multi-agent models in the intelligence of manufacturing resources and the support (Tao, 2016; Lin, 2017). "Made in China 2025" strategy of decision-making and services, and also promoting the transformation and upgrading of manufacturing companies towards intelligent manufacturing of decision-making andofmanufacturing services, and2016; also explores the advantages distributed, multi-agent models in manufacturing companies towards intelligent manufacturing (Tao, 2016;that Lin, 2017). "Made in China 2025" strategy the optimization of manufacturing tasks (Taratukhin, of decision-making and manufacturing services, and also emphasised in order to meet uncertainties of the market explores the advantages of multi-agent models in manufacturing companies towards in intelligent 2025" manufacturing the optimization of manufacturing tasks (Taratukhin, 2016; (Tao, 2016; Lin, 2017). explores the Li advantages of distributed, distributed, multi-agent models in (Tao, 2016;that Lin, 2017). "Made in China China 2025" strategy emphasised order to"Made meet of thestrategy market Xie, 2017; 2017). But the organization of company is explores the advantages of distributed, multi-agent models in environment andin demand for uncertainties customized products, the Xie, the optimization optimization of manufacturing manufacturing tasks (Taratukhin, 2016; (Tao, 2016;that Lin, 2017). "Made in China 2025" strategy emphasised in order to meet uncertainties of the market the of tasks (Taratukhin, 2016; 2017; Li 2017). But the organization of company is emphasised that in order to meet uncertainties of the market environment and demand for customized products, the different. Information, data interaction and completed the optimization of manufacturing tasks (Taratukhin, 2016; manufacturing pattern of factory should be service-oriented Xie, 2017; 2017; Information, Li 2017). 2017). But Butdata the interaction organization and of company company is emphasised that in order to meet uncertainties products, of the market environment and demand for the Xie, Li the organization of is completed environment and demand for customized customized products, the different. manufacturing pattern of factory should manufacturing also various in different automation Xie, 2017; Information, Li tasks 2017).areBut the organization of company is rather than production-oriented. This be shiftservice-oriented has brought different. data interaction and completed environment and demand for customized products, the manufacturing manufacturing pattern of factory should be service-oriented different. Information, data interaction and completed tasks are also various in different automation manufacturing pattern to of the factory should be service-oriented rather thanchallenges production-oriented. This of shift has factories. brought level. Therefore, a new smart factory model isand needed, which different. Information, data interaction completed enormous modelling smart manufacturing tasks are also various in different automation manufacturing pattern of factory should be service-oriented rather than production-oriented. This shift has brought manufacturing tasks are also various indistributed different automation level. Therefore, a the new smart factoryofmodel is needed, which rather than production-oriented. This and shift has factories. brought enormous to the modelling of smart is convenient for management resources, manufacturing tasks are also various in different automation Within the challenges smart factory, digital world physical world is level. Therefore, aa the new smart factory model is needed, which rather than production-oriented. This and shift has factories. brought enormous challenges to the modelling of smart level. Therefore, new smart factory model is needed, which convenient for management of distributed resources, Within the smart factory, digital world physical world enormous challenges integrated, to the modelling of the smart factories. and secondly has anew loosely coupled architecture to support level. Therefore, a smart factory model is needed, which should be seamlessly realizing decentralized is convenient for the management of distributed resources, enormous challenges to the modelling of smart factories. Within be the seamlessly smart factory, factory, digital realizing world and andthephysical physical world and is convenient for the management of distributed resources, secondly has a loosely coupled architecture to support Within the smart digital world world should integrated, decentralized the flexible manufacturing. is convenient for the management of distributed resources, management of resources, combining the intelligent data and secondly has a loosely coupled architecture to support Within the smart factory, digital world and physical world should be be seamlessly seamlessly integrated, realizing the decentralized andflexible secondly has a loosely coupled architecture to support the manufacturing. should integrated, realizing decentralized management ofto resources, combining thethe intelligent secondly has a loosely coupled architecture to support fusion method process industrial big data, and drivingdata the and the should be seamlessly integrated, realizing the decentralized management of resources, combining the intelligent data the flexible manufacturing. Thisflexible paper manufacturing. draws on the basic features of CPS and fractal management of resources, combining the intelligent data fusion method to process industrial big data, and driving the the flexible manufacturing. independently optimized manufacturing processes based on This paper draws on the basic features of CPS and fractal management ofto resources, combining the intelligent data fusion method industrial big data, and the to form aonsmart factory model. RFs and withfractal CPS fusion method to process process industrial big data, and driving driving the independently optimized manufacturing processes basedXu, on theory This paper draws the basic basic features of CPS CPS customer demand data model (Helu, 2016; Prause, 2017; This paper draws onsmart the features of and fractal theory to form a factory model. RFs with CPS fusion method optimized to process manufacturing industrial big data, and driving the independently processes based on characteristics and self-similarity are functional and structural independently optimized manufacturing processes based on customer demand data model (Helu, 2016; Prause, 2017; Xu, This paper draws on the basic features of CPS and fractal theory to to form form smart factory factory model. RFs with CPS 2017). Externally, companies should2016; provide more flexible theory aaself-similarity smart RFs CPS are model. functional andwith structural independently optimized manufacturing processes based on characteristics customer demand data model Prause, 2017; Xu, abstractions of and manufacturing resources at different scales in customer demandmanufacturing data model (Helu, (Helu, 2016; Prause, 2017; Xu, 2017). Externally, companies should provide more flexible theory to form aself-similarity smart factory model. RFs with CPS characteristics and are functional and structural and diversified services to jointly build an characteristics and self-similarity are functional and structural abstractions of manufacturing resources at different scales in customer demand data model (Helu, 2016; Prause, 2017; Xu, 2017). Externally, companies should provide more flexible factory. Self-similarity is reflected in the fact that the agent 1) 2017). Externally, companies should provide more flexible and diversified manufacturing services to2018). jointly build an characteristics and self-similarity are functional and structural abstractions of manufacturing resources at different scales in open, fully interconnected platform (Aleš, In short, the abstractions ofpackage manufacturing resources at different scales in factory. Self-similarity is reflected fact thatbythe agent 1) 2017). Externally, companies should provide more flexible and diversified manufacturing services to jointly build an has the same structure 2) in is the organized the fractal open, fully interconnected platform (Aleš, 2018). In short, the and diversified manufacturing services to jointly build an abstractions of manufacturing resources at different scales in factory. Self-similarity is reflected in the fact that the agent 1) flow of information will break through the traditional factory. Self-similarity is reflected in the fact that the agent 1) has the same package structure 2) is organized by the fractal and diversified manufacturing services to jointly build an open, fully fully interconnected platform (Aleš, 2018). Intraditional short, the the factory. mapping method drivenstructure by the production relationship 3) can open, interconnected platform (Aleš, 2018). In short, flow of information will break through the Self-similarity is reflected in the fact that the agent 1) has the same package 2) is organized by the fractal automation hierarchy. The essential of the2018). smart factory has the same package structure 2) is organized by the service fractal method driven bymatching the production relationship 3) can open, interconnected short,will the mapping flow fully of information information willplatform break (Aleš, through the In traditional provide complete and manufacturing flow will break through the traditional automation hierarchy. The essential of the smart factory will has the same package structure 2) is organized by the fractal mapping method driven by the production relationship 3) can be theof utilization of manufacturing resources and provide mapping method driven by the production relationship 3) can complete and matching manufacturing service flow ofintelligent information willessential break through the traditional automation hierarchy. The of the smart factory will independently. The resulting fractal systemservice (CFS) automation hierarchy. The essential of the smart factoryWith will be the intelligent utilization of manufacturing manufacturing resources and mapping method driven bymatching thecompany production relationship 3) can provide complete and manufacturing the flexible adjustment of the process. independently. The resulting company fractal system (CFS) provide complete and matching manufacturing service automation hierarchy. The essential of the smart factory will be the intelligent utilization of manufacturing resources and provides various customized service activities, at the same be the intelligent utilization of manufacturing resources and the flexible adjustment of the manufacturing process. With provide complete and matching manufacturing service independently. The resulting company company fractal system system (CFS) this in intelligent mind, researchers believe that smart factories independently. The resulting fractal various service ofactivities, atcapabilities the (CFS) same be the utilization of manufacturing resourceshave and provides the flexible of the process. time, realizing thecustomized self-optimization resource the flexible adjustment of believe the manufacturing manufacturing process. With With this intypical mind,adjustment researchers that smart Openness factories have independently. The resulting company fractal system provides various customized service ofactivities, activities, atcapabilities the (CFS) same some features of complex systems. and provides various customized service at the same time, realizing the self-optimization resource the flexible adjustment of the manufacturing process. With this in mind, researchers that smart factories have this intypical mind,arefeatures researchers believe that smart factoriesservice have some ofinbelieve complex systems. Openness and provides variousthecustomized service ofactivities, atcapabilities the same time, realizing self-optimization resource randomness reflected the company's production time, realizing the self-optimization of resource capabilities this in mind, researchers believe that smart factories have some systems. Openness and some typical typicalarefeatures features ofincomplex complex systems. Openness and time, realizing the self-optimization of resource capabilities randomness reflectedof the company's production service some typicalarefeatures ofincomplex systems. Openness and randomness reflected the company's production service randomness are reflected in the company's production service Copyright © 2019, 2019 IFAC 2836Hosting by Elsevier Ltd. All rights reserved. 2405-8963 © IFAC (International Federation of Automatic Control) randomness are reflected in the company's production service Copyright © 2019 IFAC 2836 Copyright 2019 responsibility IFAC 2836Control. Peer review©under of International Federation of Automatic Copyright © 2019 2019 IFAC IFAC 2836 10.1016/j.ifacol.2019.11.628 Copyright © 2836 Copyright © 2019 IFAC 2836

2019 IFAC MIM Berlin, Germany, August 28-30, 2019

Wei Wu et al. / IFAC PapersOnLine 52-13 (2019) 2776–2781

and complex manufacturing processes and the service-driven flexible organization adjustment.

2777

and resources cannot meet the rapidly customized process. Therefore, this study introduces the fractal theory to make effective support for the model.

2. RELATED RESEARCH 2.2 Fractal Organization Theory 2.1 Cps Theory In order to enable companies and factories to achieve intelligent production process, researchers have proposed to introduce CPS to reflect the self-adaptation and openness of complex systems. CPS is an intelligent system that deeply embeds the sensing, computing and communication capabilities into the actual physical process and has real-time analysis, scientific decision-making and accurate execution (Fig. 1). Specifically, a large amount of implicit data contained in the physical domain is transformed into explicit data through in-depth perception, and then converted into valuable information in the cyber domain. The reasoning and learning engine integrates information, and then combines the actual manufacturing conditions and domain knowledge to generate intelligent decisions. The final precise execution applies the optimized decision into the physical domain, forming a closed-loop flow of data. In the field of intelligent manufacturing, CPS theory has been widely applied to various aspects of research. For example, Shin HJ (2018) et al. introduced support vector machine in CPS systems to predict exception in workshop, triggering structure reconfiguration to achieve flexible and agile manufacturing. Adamson (2016) et al. constructed a distributed model of robotic control based on manufacturing features. The model incorporates the adaptive capabilities of CPS. And the task drives the functional blocks that encapsulate the processing and assembly features of products to produce optimized management and control decisions of distributed devices. Zhang (2017) et al. proposed a workshop CPS that integrates smart machine agent (SMA), self-organization and selfadaptation modules. SMA senses data in real time and achieves optimal control based on rules or knowledge. The task-driven self-organization module is responsible for intelligently matching tasks with devices. The self-adaptation module recognizes the exception during the task and addresses it automatically. real-time analysis

explicit data

in-depth perception

information

reasoning & learning

cyber domain decision

physical domain

The concept of “fractal” was first proposed by Mandelbrot, a mathematician, in 1975 when he studied the length of the British coastline. Later, the researchers continued to analyse the nature of the “fractal”. Self-similarity is the core of fractals, that is, the details of the parts that are separated from the whole are not less than that of the whole. The nesting and recursive mechanism allow fractals to ensure both similarity and infinite hierarchical structure. In the manufacturing field, the increasingly dynamic, complex, and non-linear characteristics have caused problems in company modelling research. Professor Warnecke (1993) first proposed the concept of fractal factory, believing that there is a statistical self-similarity in the enterprise organization and the various levels of information flow. A fractal factory consists of independent, similar, and privileged dynamic fractal elements at different levels. Shin M (2009) et al. pointed out that the fractal organization can be function-driven or relation-driven divided by the different self-similarity pattern, when researching the plant's self-adaptation capabilities for the environment. The fractal factory is function-driven, in which the fractal structure division corresponds to the internal functional division of the plant, so fractals are easier to manage and maintain. But the relation-driven fractal organization pays more attention to the connection between different units of different fractal layers, and can flexibly adjust the structure combining with various event relationships while facing rapidly changes of external environment. Pirani (2016) et al. proposed a service-oriented fractal architecture of robotic control. The sensors, actuators, and electromechanical equipment in the workshop are virtualized into fractal service units, which directly reduce the manufacturing cost of shop by nesting the production bottleneck analysis algorithm and adjusting the fractal structure. Bider (2016) et al. proposed a fractal structure based on the type of business and asset relationship. The model categorizes the relationship between "business processes" and the "assets" and forms two different prototypes. The use of prototypes alternately to construct the fractal structure of the smart factory makes the relationship between the various components of the enterprise clearer. It can be seen that the relation-driven fractal organization structure will enable the smart factory to have a serviceoriented structural reconfiguration capability and facilitate a customized, agile production process, combining selfoptimizing ability of manufacturing RFs.

accurate execution

implicit data

3. SMART FACTORY REFERENCE ARCHITECTURE

Fig. 1. Basic features of CPS In summary, CPS can realize the optimal configuration of distributed resources and the flexible production process. However, the above studies all regard the whole smart factory as CPS. Hence, the manufacturing resources are not independent and autonomous, and the static matching of task

3.1 Cps Fractal Smart factories are made up of intelligent manufacturing units of different scales – equipment, workshops, factories. Combining the CPS model and the fractal idea, this study

2837

2019 IFAC MIM 2778 Berlin, Germany, August 28-30, 2019

Wei Wu et al. / IFAC PapersOnLine 52-13 (2019) 2776–2781

considers extracting the features of the manufacturing units of different scales in the cyber and physical domain and forming the unified functions and structures, and constructing the independent and adaptive CPS fractal (Fig. 2). Referring to the four important characteristics of typical CPS (in-depth perception, real-time analysis, scientific decision-making, accurate execution), the fractal contains five parts: perception, analysis, decision, execution, and data protocol and service interface. The perception unit is responsible for extracting explicit data from the resource entity and the external environment of the agent. Because fractals adopt the message-based lightweight communication method, the perception unit is responsible for receiving the message data generated when agents interact. The analysis unit is responsible for merging explicit data with different types and structures, and further extracting and mining manufacturing information, and then combining semantic features to form information clusters, which greatly improves the quality of the data. The decision unit combines the knowledge of the manufacturing domain to apply the learning and reasoning engine to the real-time manufacturing information, and make important decisions while discovering new knowledge and improving the timeliness and effectiveness of the decision. The execution unit is primarily responsible for the accurate decision execution and the precise control of physical entities to optimize the actual manufacturing process. In addition, the execution unit forwards and shares the optimized manufacturing information to other associated agents in the external environment, and accelerates the flow of information. Data protocols specify the way data is exchanged, and data security should also be considered to ensure data availability and real-time requirements. The service interface specifies the way the service is packaged, published, and invoked, by which the function of manufacturing resources can be encapsulated as a manufacturing service. Since RFs directly lives in the Industrial Internet, the Internet-based data communication scheme could be adopted. For example, the OPC-UA protocol uses Web services with XML or JSON as the data exchange format when transmitting data, and provides powerful data modelling and data security functions.

3.2 Operating Mechanism of RFs The CPS fractal provides a unified functional and structural description for manufacturing resources. However, RFs also need the data-driven operating mechanism to truly achieve optimal execution of tasks. The types and volumes of data generated by different scale manufacturing units in smart factory is different, but the data flow drives execution of task in the same way (Fig. 3), which includes collection, preprocessing, storage, analysis, modelling and application. RFs receive a wide range of data, which may come from the actual state feedbacks of production, the interactions of agents, the knowledge databases, and the manufacturing service activities. But all data is transmitted to the destination fractal through the data gateway and router of the Industrial Internet. Afterwards, the perception unit divides the received data into real-time field data and batched service data. For the 3V features of industrial big data (volume, variety, velocity), the perception unit performs cleaning, data repackaging and fast edge computing of massive raw data, which initially extracts key data, and then stores it to distributed databases. The analysis unit is responsible for extracting information automatically or according to service requirements. The data analysis process provides statistics of the manufacturing process data, as well as the state transitions or changes in key indicators in the manufacturing process. Data modelling provides different data models. The resource data model fuses the actual state of the physical manufacturing resource, the manufacturing function descriptions, and the visual simulation models. The customer data model stores the detailed manufacturing requirements. In addition, the production process, scheduling, operation and maintenance models describe the manufacturing mechanism and knowledge, which the agent can independently refer to and make optimized decision. The data application process represents decision-making and execution agent based on the acquired key manufacturing information. execution decision

extraction explicit analysis data

optimization reasoning info. decision learning

perception

analysis

decision

execution

data collection

data fusion

decision making

message routing

data management

data mining

knowledge discovery

decision execution

message management

semantic processing

knowledge management

accurate control

analysis

decision and execution demonstration of product/process

application

knowledge discovery demand analysis fault diagnosis performance prediction

modelling

analysis

management model material/customer data model scheduling model product/device/plant process model data model

statistics data reports visualization data fusion data mining

pre-processing &storage data packaging & unpackaging

perception

data protocol and service interface

data cleaning

edge computing

real-time collection of field data batch collection of service data

external environment

distributed storage

collection data gateway data router of IoT

sources environment

Fig. 2. Structure of CPS fractal

services cloud/supplychain/app..

production workshop/ device/...

database design/ process/...

Fig. 3. Data-driven operating mechanism

2838

fractals connection/...

2019 IFAC MIM Berlin, Germany, August 28-30, 2019

Wei Wu et al. / IFAC PapersOnLine 52-13 (2019) 2776–2781

3.3 Organization Mechanism of RFs A complete and complex manufacturing process typically requires manufacturing units to cooperate. Therefore, RFs also rely on the relation-driven fractal organizing method to form the final company fractal system (CFS). In the data modelling process, agent records the feature and status description of the managed manufacturing resources and the mapping relationship and location relationship with the associated agents in real time (Fig. 4). When participating in manufacturing service activities, the resource agent(host) first analyses the requirement data in the customer data model contained, and performs semantic matching and reasoning on the obtained production demands and production process knowledge to generate indivisible task sequences. Then, the agent retrieves the function lists of the adjacent hierarchical RFs according to the location relation. If a subtask in the task sequences can match the function of an agent (other), and that agent is unoccupied and its processing performance satisfies the task requirement, a fractal mapping described by the indivisible manufacturing task is established between the two RFs. At the same time, agent update the occupancy status, the task list, and insert the newly generated mapping relation into the relation table. By analogy, after the manufacturing tasks are continuously matched and refined to the indivisible manufacturing tasks, RFs in smart factory are organized into a CFS according to a series of fractal mappings generated. The completion of the original service order marks the disappearance of the original fractal mappings, and the new manufacturing service drives the CFS to organize kinds of manufacturing resources according to the new fractal mappings. Moreover, the continuous optimization of production process knowledge will make the task dismantling more consider the factors such as the ability of manufacturing resources. location relation mapping relation

URI feature description status description type

busy/idle

location

task list

function list

fault diagnosis

data model

performance prediction

...

...

self

URI1

URI2

realize the production mode of agile manufacturing, while the traditional enterprise CPS neither has the structural reconfiguration capability driven by the service demand nor guarantees that the different scales manufacturing resources can complete the manufacturing task independently. 3.4 Company Fractal System The implementation architecture of the CFS is shown in Fig. 5. Firstly, the physical manufacturing resources of each level in smart factory are encapsulated into intelligent fractals according to the structure of the CPS fractal. The encapsulation of the agent is a process of resource-oriented modelling, which reduces the coupling of functions, communication, and control of manufacturing entity. Each virtual manufacturing resource has the same structure but provides different granularities and types of manufacturing function. Fractal mappings can be generated based on order’s requirements in manufacturing cloud, which provides technical support for flexible production and adaptive adjustment of company organizations. CFS is executors of on-demand manufacturing processes while providing diverse service activities. Service activities are classified into internal services and external services according to the source of demand. The company management is the main consumer of internal services that establishes data models according to different analysis and application requirements understand and intervene in company operation and maintenance, or is compatible with traditional digital management systems such as ERP and MES. The external services of CFS are oriented to manufacturing needs in the Industrial Internet. The industrial cloud platform provides manufacturing service registration, requirements registration, and service-demand matching. Many companies use this platform to complete the customized production process. In addition, the various RFs in company have the ability to independently control, make decisions, and provide manufacturing services. Hence, they can participate in collaborative manufacturing tasks across enterprises or geographies.

...

manufacturing activities implement

relation of resources

A URI location relation mapping relation B URI location relation mapping relation ... ... ...

fractal details

URI1-1

URI1-2

2779

customization

service activities

...

fractal mappings

industrial cloud

supply

company fractal system fractal mapping virtual manufacturing resources encapsulation

collaborative company

manage

fractal

Fig. 4. Service-driven organization mechanism In summary, the service-driven organization mechanism of RFs in the customized manufacturing process ensures that the CFS maximizes the utilization of manufacturing capabilities and rapid organizational reconfiguration, while completing the optimal matching of resources. The data-driven operating mechanism of RFs ensures the optimal execution of the indivisible manufacturing tasks. Therefore, the CFS can

unified structure

structural abstraction

features of CPS

physical manufacturing resources

Fig. 5. CFS implementation architecture

2839

functional abstraction

self

2019 IFAC MIM 2780 Berlin, Germany, August 28-30, 2019

Wei Wu et al. / IFAC PapersOnLine 52-13 (2019) 2776–2781

4. CASE STUDY Considering an example from the industry field, hydraulic cylinders are the heart of hydraulic transmission systems and are commonly used in port cargo storage, shield equipment and in a variety of manufacturing scenarios. Hydraulic cylinders are available in a variety of types and can be highly customized. However, due to the wide variety of machining processes required for parts, it is not possible to process in a production line. A factory has mastered the complete machining process and product assembly process of the cylinder parts, but the layout of the equipment in the workshop is messy and the processing capability and equipment status of the equipment are different. Moreover, after adopting the enterprise digital management platform, the factory can only adopt the manual workshop scheduling method to organize production. Therefore, enterprises are currently faced with problems such as low equipment production efficiency, poor scheduling of workshops, and limited factory capacity. In order to achieve product customization and flexible production, companies need to establish a CPS fractal-based CFS in conjunction with the manufacturing processes of the cylinder product to verify the availability and effectiveness of the reference architecture proposed in this paper (Fig. 6). At first, according to the basic structure of the CPS fractal, the packaging of the machine tools, workshops and production plants of the parts is completed, and a number of machine tool agents, workshop agents and factory agents are obtained, which initially constitute the CFS. In the data modelling process, agents form lists of cylinder processing capabilities according to its corresponding manufacturing resources. The factory agent forms product data models based on the known cylinder production process and processing capability of machine tools, and finally registers various types of cylinder manufacturing services in the industrial cloud through the service interface. The platform sends the

results to CFS after completing the manufacturing servicerequirement match. After receiving the matching messages and service orders forwarded by several Industrial Internet routers, the shop fractal begins to form a service-driven cylinder production organization. After analysing the manufacturing requirements and cutting the machining tasks, the shop fractal retrieves the functions and status of machine devices. Then the shop fractal needs to find the optimal taskequipment binding scheme, which not only ensures that the utilization of each machine is not reduced but also meets the delivery period conditions. The optimized matching scheme includes rules between the shop fractal and the relevant machine fractals. According to this set of rules, the original sales order is transformed into multiple production orders and delivered to the corresponding machine to begin processing. The device fractal analyses the part processing requirements and material preparation. In the case of physical machine with no faults and excellent performance, device fractal may need to adjust the order in which the machining tasks are performed to ensure that all operations are completed within the deadline as much as possible, or to minimize the delay of the machining tasks. Finally, the device fractal informs the logistics fractal to transport the raw materials and starts processing after all preparations are ready. In summary, such workshop realizes the intelligent utilization of devices and the autonomous and adaptive processing flow. It shows that the smart factory reference architecture based on CPS fractal can effectively help company to face fast and customized manufacturing requirements. At the same time, compared with the traditional enterprise CPS, the case also proves that CFS has the characteristics of flexible reconfiguration of system architecture driven by manufacturing demand and intelligentization of production resources of different scales.

workshop fractal perception

receipt of order

manufacturing service-requirement matching in cloud platform

customer data model

sales order delivery date, list of products, ..

extraction of production needs decomposing task with knowledge

perception

receipt of production order

machining cylinder piston rod

machining demands of part

fixed type of piston rod

small oil cylinder and a number of supporting piston rod

extraction of material data

imported piston rod material, no need for surface plating

creating production order of parts including processes

processes confirmation

retrieving function list of device

machine condition confirmation

viewing fault diagnosis and performance analysis results

optimizing the execution sequence of tasks

ensuring utilization and minimizing expected delay time of all operations

logistics preparing

sending material logistics data to logistics device fractal

precise machining of device

loading NC program ,starting processing when materials are ready

analysis

decision

device fractal

analysis

retrieving the func. list the machining range, precision and and status of device waiting time of the CNC machines optimizing task-device matching

decision

balancing utilization and delivery time binding device for task

inputting decisions into writing mappings to relation table workshop fractal forming fractal structure execution

execution

sending production device fractals start processing and orders to device fractals real-time optimization

Fig. 6. Cylinder product manufacturing process based on CPS fractal

2840

2019 IFAC MIM Berlin, Germany, August 28-30, 2019

Wei Wu et al. / IFAC PapersOnLine 52-13 (2019) 2776–2781

5. CONCLUSIONS Considering that traditional manufacturing companies can't cope with the rapidly changing manufacturing needs, this paper proposes a smart factory reference architecture based on CPS fractal, with the goal of optimizing the utilization of manufacturing resources and adaptive production processes. Firstly, combining the features of CPS and the relation-driven fractal organization, the functions and structures of manufacturing units of different scales are abstracted into CPS fractal. RFs with CPS features adopt the operating mechanism driven by industrial data. Multiple Agents form CFS through the service-driven organizing mechanism, on the one hand to achieve customized manufacturing, and on the other hand to provide multiple manufacturing service activities. Finally, the effectiveness and availability of the proposed architecture was verified with a highly customized cylinder manufacturing process analysis. ACKNOWLEDGEMENT Some of research work in this paper is supported by the Major Program of National Natural Science Foundation of China (Grant No. 71690230,71690234) and National Key R&D Program of China, No. 2017YFE0100900. REFERENCES

2781

System in Industry 4.0. Wireless Communications and Mobile Computing, 1-13. Shin M, Mun J and Jung M. (2009). Self-evolution framework of manufacturing systems based on fractal organization. Computers & Industrial Engineering, 56(3), 1029-1039. Tao F, Li C, Liao TW and Other. (2016). BGM-BLA: A New Algorithm for Dynamic Migration of Virtual Machines in Cloud Computing. IEEE Transactions on Services Computing, 9(6), 910-925. Taratukhin VV and Yadgarova Y V. (2016). Industrial Internet Reference Architectures and Agent-Based Approach in Design and Manufacturing. Emerging Trends in Information Systems, 117-124. Xie C, Cai H, Xu L and Other. (2017). Linked Semantic Model for Information Resource Service towards Cloud Manufacturing. IEEE Transactions on Industrial Informatics, PP (99), 1-1. Xu G, Huang GQ, Fang J and Other. (2017). Cloud-based smart asset management for urban flood control. Enterprise Information Systems, 11(5), 719-737. Warnecke H J. (1993). The Fractal Company: A Revolution in Corporate Culture. Springer Berlin. Zhang Y, Qian C, Lv J and Other. (2017). Agent and cyberphysical system based self-organizing and self-adaptive intelligent shop-floor. IEEE Transactions on Industrial Informatics, 13(2), 737-747.

Adamson G, Wang L and Moore P. (2016). Feature-based control and information framework for adaptive and distributed manufacturing in cyber physical systems. Journal of Manufacturing Systems, 43(2), 305-315. Aleš Popovič, Ray Hackney, Rana Tassabehji, et al. The impact of big data analytics on firms' high value business performance[J]. Information Systems Frontiers, 2018, 20(2): 209-222. Bider I, Perjons E and Elias M. (2016). A fractal enterprise model and its application for business development. Software & Systems Modelling, 1-27. Helu M, Libes D, Lubell J and Other. (2016). Enabling Smart Manufacturing Technologies for Decision-making Support. ASME 2016 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference. V01BT02A035. Li JQ, Yu FR, Deng G and Other. (2017). Industrial Internet: A Survey on the Enabling Technologies, Applications, and Challenges. IEEE Communications Surveys & Tutorials, 19(3), 1504-1526. Lin J, Yu W, Zhang N and Other. (2017). A Survey on Internet of Things: Architecture, Enabling Technologies, Security and Privacy, and Applications. IEEE Internet of Things Journal, (99): 1-1. Pirani M, Bonci A and Longhi S. (2016). A scalable production efficiency tool for the robotic cloud in the fractal factory [C] Industrial Electronics Society, 68476852. Prause G and Atari S. (2017). On sustainable production networks for Industry 4.0. Entrepreneurship & Sustainability Issues, 4(4), 421-431. Shin HJ, Cho KW and Oh C H. (2018). SVM-Based Dynamic Reconfiguration CPS for Manufacturing 2841