Factories of the future: challenges and leading innovations in intelligent manufacturing

Abstract

This paper reviews some of the most recently reported research into challenges and leading innovations in intelligent manufacturing for the Factories of the Future (FoF). Such research can be categorised as addressing five broad topic areas: manufacturing systems frameworks, theories and models; the pervasiveness of Cyber-Physical Systems (CPSs); the critical role of semantic technologies and interoperability; the Virtual Organisation (VO) of manufacturing systems and the servitisation of manufacturing systems. The paper analyses conceptual, theoretical, empirical and technological contributions from several leading authors in domain area. This paper identifies a wide range of research topics from the elaboration of manufacturing systems frameworks to models, from sensors to CPSs, to the application of semantic technologies and interoperability architectures of the data and information generated by manufacturing agents, how VOs are shaping manufacturing environments and the increasing challenge of deploying manufacturing systems to support servitisation. The paper concludes elaborating final considerations on the challenges and leading innovations in intelligent manufacturing for the FoF research area.

Keywords:

• factories of the future
• industry 4.0
• manufacturing systems
• intelligent manufacturing
• cyber-physical systems
• virtual organisations
• servitisation

1. Introduction

The manufacturing industry needs to lead innovations to face the global competitive pressures in the advent of intelligent manufacturing across the broad range of manufacturing sectors. According to the European Commission’s (EC) vision for the way forward, a new model of production systems should address transformable, networked and learning factories, depending on several drivers such as high-performance, extreme customisation, environmental-friendliness, superior efficiency of resources, eminent human potential and significant knowledge creation (EC 2016). The EC coined this domain as Factories of the Future (FoF), and more recently others have coined this evolution in the manufacturing environment as Industry 4.0 – The Fourth Industrial (R)evolution (BCG 2015).

FoF embrace information and communication technology (ICT)-based production systems with high-quality manufacturing technologies and intelligent capabilities aimed to optimise their performance with a high degree of autonomy and adaptability for a balanced combination of high throughput with accurate production. Also, it embraces the implementation of the sustainable manufacturing concept, throughout methodologies and processes that have the capability of cost-efficiently shaping, handling and assembling products composed of complex and novel materials. The relevance of FoF is acknowledged by the funding agencies, as referred by Filos (2016) that following a positive assessment of the Public–Private Partnerships (PPPs) in the seventh Framework Programme, the EC decided to continue and even increase its stake in manufacturing Research, Development and Innovation (R + D + I) in the new HORIZON 2020 programme by continuing the FoF initiative with a funding envelope of €1.15 billion and also with a new initiative that addresses Sustainable Process Industry through Resource and Energy Efficiency (SPIRE) with a funding envelope of €900 million until the year 2020. As referred by Filos (2016), the main R + D + I priorities so far have been on resource and energy efficiency in manufacturing, including end-of-life issues, attractive work environments, mass-customisation and personalised manufacturing including embedded services and increased flexibility of the manufacturing capacity and enhanced process optimisation, modelling, simulation and ICT for Small and Medium-sized Enterprise (SME) manufacturing environments.

This paper presents the analysis of 16 research works in the domain of present and future of industrial manufacturing systems and production sites inspiring industry to produce more sustainable products using smart processes and smart factories, focusing on the challenges and leading innovations in intelligent manufacturing. The aim of this paper is to be a reference source for research on the state-of-the-art and a knowledge transfer medium to the industry on leading technological innovations towards the envisaged FoF.

The authors analyse novel contributions from researchers and practitioners who are exploring the identified major challenges and anticipated leading innovations of the FoF in a global perspective, putting focus on novel strategies, methods and tools in a scientific-based standpoint.

The paper analyses conceptual, theoretical, empirical and technological contributions from several leading authors in domain area. This paper considers five intertwined dimensions: manufacturing systems frameworks, theories and models; the pervasiveness of Cyber-Physical Systems (CPSs); the critical role of semantic technologies and interoperability; the Virtual Organisation (VO) of manufacturing systems; and the servitisation of manufacturing systems. These five dimensions cover much of the most relevant focus of current research in the ‘Factories of the Future: Challenges and Leading Innovations in Intelligent Manufacturing’.

2. Manufacturing systems frameworks, theories and models

Ostrom (2005) stresses the necessity of distinguishing between levels of specificity of academic work that are often confused. She attributes special importance to the clear consideration of frameworks, theories and models. She stresses that the development and use of a general framework helps to identify the elements and relationships among these elements that one needs to consider for analysis. Frameworks organise diagnostic and prescriptive research and inquiry. Hence, they provide the most general list of variables that should be used to analyse all types of manufacturing systems research and development. Manufacturing systems frameworks will provide a meta-theoretical language that is necessary when developing theories for manufacturing systems and that can be used to compare theories. They attempt to identify the universal elements that any theory relevant to manufacturing systems would need to include. The development and use of theories enables the analyst to specify which elements of the framework are particularly relevant for certain kinds of questions and to make general working assumptions about these elements (Ostrom 2005). Thus, manufacturing systems theories will focus on a framework and make specific assumptions that are necessary for an analyst to diagnose a phenomenon, explain its processes and predict outcomes. Several theories are usually compatible with any framework. Finally, models make precise assumptions about a limited set of parameters and variables (Ostrom 2005). Logic, mathematics, game theory, architectures, experimentation and simulation, and other means are used to explore the consequences of these assumptions systematically on a limited set of outcomes. Multiple models are compatible with most theories.

In this section, the contributions by Moghaddam and Nof (2016), Marcelino-Jesus et al. (2016) and Ilie-Zudor et al. (2016) are revised that attempt to organise the manufacturing systems body-of-knowledge by creating frameworks and models for the FoF(no theory has been put forward).

Moghaddam and Nof (2016) provide a framework of the Collaborative Factory of the Future (CFoF). This framework is applied to enable the abstraction of disparate e-Factories as collaborative networks of entities, which in turn can be engineered and augmented through the Collaborative Control Theory (CCT) operators in order to address FoF challenges of scalability and sustainability. The authors review a set of examples of ‘e-Factories’ and they abstracted a task-resource model as the basis for the Task Administration Protocol (TAP). Hence, they stress that all trends described by holons, IT/shop-floor subsystems, sensors, RFIDs, knowledge bases, humans, robots, manufacturing enterprises can be regarded as resources (physical and/or cyber) that accomplish a set of defined, desired and allocated tasks (designated as ‘e-Activities’). The objectives of such Complex Collaborative e-Systems will not be materialised without effective and dynamic identification of requirements, assignment, parallelisation and synchronisation of tasks, resources and information, and systematic detection and prevention of errors and conflicts. The outcome of such engineering processes will be a fault-tolerant system with fewer failures, capable of handling dynamic changes in its configuration, complexity and emergence with higher or lower scalability as needed. These functionalities, however, will require high levels of intelligence and autonomy of the processes, for example, design and redesign, negotiation, feedback, system improvement, reconfiguration and evolution.

Hence, Moghaddam and Nof (2016) state that what they designate as CFoF designs are challenged to provide promising solutions and address the emerging performance criteria (named as: ‘e-Criteria’), such as agility, dependability, integrality, resilience, robustness and scalability, and that these emerging e-Criteria combine with security, information assurance, collaboration incentives and multi-/cross-cultural factors and sustainability. The authors foresee future manufacturing systems as ‘systems of systems’, which will require effective interaction between dynamic and complex systems to realise the goals at system-wide and local levels. Service-oriented manufacturing (SOA-Mfg.) will be an emerging manufacturing design and control pattern that enables interaction and collaboration between different layers. As an example, ‘cloud manufacturing’ is an extension of the cloud computing concept in manufacturing and in line with SOA-Mfg., is an emerging manufacturing paradigm to support the notion of ‘manufacturing-as-a-service’. This will need to advance in order to address the premise of cloud manufacturing that is integration/distribution of distributed/integrated resources. According to the authors, in the context of the CFoF, such advancement will follow through optimal real-time, intelligent and autonomous execution of e-Activities and e-Transactions, conflict prevention and resolution, and emergence handling.

Marcelino-Jesus et al. (2016) present a framework targeted at entrepreneurs to help them in the evaluation and validation of new manufacturing technologies development. From the various phases of a project lifecycle, in which the proposed methodology is handled, it is the Technology Assessment (TA) phase that the research work of the authors focuses on. This phase intends to assure that the project developments reach the desired output accordingly to the identified requirements. It also seeks to validate the technological research results through an effective business analysis and by its end-users testing, to warrant an efficient exploitation and change management of such results appliance. Based on an evaluation of frameworks to analyse the most appropriate to be used in the project viability methodology, the authors chosen DECIDE, because of its generic but well-structured approach, and mainly due to its openness in aggregating in its process different mechanisms.

Marcelino-Jesus et al. (2016) developed a project pilot evaluation for textile SMEs that included four different mechanisms: (1) GQM – Goal-Question-Metric – approach to define evaluation metrics assuring the addressing of all the objectives; (2) the software quality evaluation methodology, which was defined by the authors based on standards such as ISO/IEC-25040, to technically evaluate the project results; (3) the qualitative evaluation mechanism for the business evaluation of the project outputs; and finally, (4) the performance indicators mechanism to evaluate the work conducted in the project assuring its completeness.

Grounded on two application scenarios, Ilie-Zudor et al. (2016) research work focused on developing a set of models that enable examining the relationship of decision levels, performance measures and modelling, and decision support approaches for manufacturing and logistics application fields for strategic, tactical and operational decisions. In the proposed Virtual Factory Framework (VFF), strategic to tactical decision support service offers ‘best practice’ solution templates to manufacturing challenges formalised in an adequate way. When few restrictions are made regarding the branch of industry or any other distinctive features, a high degree of abstraction is required to be able to cover a reasonably large variety of cases. In the case of the VFF, a set of generic and manufacturing-related features were selected as an organising principle of the underlying knowledge base, and the Intermediate Knowledge Representations (IKRs) case-based reasoning (CBR) service was deployed for retrieval and maintenance of the discrete cases. The corpus of knowledge relied on is, in VFF, a collection of best practice solutions that have already been validated and proved successful in industrial production.

Ilie-Zudor et al. (2016) approach stresses that best practices can be regarded as an abstraction of concrete cases – selecting a best practice is a high-level and usually strategic decision that must be followed by careful adaptation and elaboration of details matching the individual implementation cases. While the suggestion of a best practice can be machine-aided, the need for human attention at adaptation makes the entire procedure at best semi-automatic; in turn, the latter can go hand-in-hand with the revision of stored knowledge based on actual experience, which is an important step in the maintenance of a knowledge corpus for CBR.

Ilie-Zudor et al. (2016) also present the ADVANCE model focus on operational and tactical decision support in logistics networks, with two main objectives: (1) improving process transparency by an adequate layout of dataflows, processing functionalities and user interfaces that deliver the right data at the right time and in the right form to the given user; and (2) employing model building and prediction to ‘patch up’ missing information that would be needed for making the proper decisions. A dataflow framework was developed that allows the quick prototyping of dataflows and the assembly of customised data-processing blocks.

Ilie-Zudor et al. (2016) conclude advocating that while strategic decisions represent switching between complex packages of actions, with most of the model building and elaboration to be carried out after making the decision, operational decisions, on the other hand, determine specific and ‘atomic’ actions to be taken on a regular basis in a constant or slowly changing environment that has to be modelled beforehand. While VFF uses CBR to aid strategic decisions and gradually unfold a highly abstract corpus of semantic knowledge into lower levels of industrial production, ADVANCE builds-up decision structures in logistics that remain human-interpretable and can thus be constantly evaluated and gradually adapted to changing requirements, in addition to providing a practically usable ‘common ground’ of understanding and problem judgment among operating personnel. Practical evaluation of project outcomes has suggested their potential as starting points for the integration of manufacturing and logistics processes in enterprise networks.

3. The pervasiveness of cyber-physical systems

Internet of Things (IoT) has attracted attention of major players in industrial landscape and it is currently one of the most expected emerging technologies. The development of IoT platforms is driven by the need to facilitate machine-to-machine connectivity, which is emerging at unprecedented rate. Brynjolfsson and Mcafee (2014) predict that machine-to-machine connections will rise from 2 billion in 2012 to 12 billion in 2020. Hobbs, Manyika, and Woetzel (2015) value IoT market to 19 trillion USD.

CPS is similar to the IoT, sharing the same type of architecture, though CPSs present a higher combination and coordination between physical and computational elements. Hence, a CPS is a system composed of physical entities such as mechanisms controlled or monitored by computer-based algorithms. Common applications of CPSs typically fall under sensor-based communication-enabled autonomous systems. Evolving architecture and technological infrastructure of CPSs include the coupling of CPSs with cloud platforms. The coupled model implies both real machines that operate in a cloud platform and simulate the conditions with an integrated knowledge from both data-driven analytical algorithms as well as other available physical knowledge. CPSs are likely to become ‘pervasive’ in industrial settings, and leading innovations for intelligent manufacturing surely have a strong focus on the development of this scientific and engineering domain. This is reflected in the works of Ghimire et al. (2016) and Delgado-Gomes, Oliveira-Lima, and Martins (2016).

Ghimire et al. (2016) proposed the application of evolving technologies in the domain of IoT in project management, by capturing events on the shop-floor and determining the meaning of information about the perceived events. This can help decision-making in heterogeneous, highly dynamic environments, particularly by reducing the time for decision-making, and enables by establishing situational awareness on the top of existing manufacturing processes. The authors developed a Reference Architecture for Situational-Aware Applications through a layered representation of IoT-based situational awareness framework. The lowest layer represents the real world and associated smart objects and sensor networks. Those are the continuous producers of data and also the point where automated actions are performed when necessary. The information integration layer and the legacy system layer are responsible for modelling of the real-world representation into the computational model. The layer is also used for integration of data obtained from the real world into the computational model. The layer above that is for the implementation and integration of the situational awareness module. The module then processes, and interprets, data collected from lower layer to produce useful information for the applications to consume at the topmost layer. Implementations in each of the layers provide well-defined interfaces to allow seamless flow of data and actions. Ghimire et al. (2016) validate their reference architecture through the implementation into a scenario of project management in the construction industry, where project supervisors have a deeper insight of the situation of the project. The results demonstrated how CPSs could be deployed taking into consideration the complexity of a dynamic project site condition and multilayered communication, due to the involvement of multiple stakeholders.

The improvement of energy usage in manufacturing has been receiving large attention during the last years due to the increasing energy cost and environmental awareness and has become a relevant challenge for manufacturing environments. Delgado-Gomes, Oliveira-Lima, and Martins (2016) propose a CPS-based Infrastructure to collect and monitor energy data in real time for manufacturing and production systems, along with a Manufacturing Energy Management System (MEMS). The collected data improve energy consumption awareness and allow the MEMS to make further analysis and to identify where to take actions in the manufacturing process in order to reduce the energy consumption. The developed MEMS enables energy consumption monitoring with different granularities, depending on where the monitoring devices are placed, that is, per machine, per product process or for the factory as a whole. The authors validate their CPS-based Architecture and Infrastructure in a case study that demonstrated how the awareness system can use the energy data to identify action points to reduce the total energy consumption and also how the combined use of Device Profile for Web Services (DPWS) and IEC-61850 allows the integration and monitoring of energy-related devices, enabling the plug-and-play capability of new Intelligent Electronic Devices (IEDs) without the configuration process needed.

We can thus conclude through the work of Ghimire et al. (2016) and Delgado-Gomes, Oliveira-Lima, and Martins (2016) that CPSs Infrastructure can help decision-making processes, grounded on shop-floor generated data, for such disparate objectives such as project management or energy efficiency management.

4. The critical role of semantic technologies and interoperability

The relevance of ‘semantic technologies’ and ‘interoperability’ in the manufacturing environment is well documented and research works have been on going for over 20 years. There is an overwhelming body of research addressing how semantic technologies and interoperability may improve the manufacturing systems (e.g. Panetto et al. 2016; Agostinho et al. 2016). However, the works of Repta et al. (2016), Milicic et al. (2016), Mehrbod et al. (2016) and Nodehi et al. (2016) give us further the demonstration of the critical role of these technologies in disparate domains.

While most manufacturing processes are structured and well documented, there is another important category of processes, also referred to as being ‘flexible’ or ‘less structured’, because their execution is not strictly enforced by an external system. Repta et al. (2016) propose the design of a semantically enabled system capable of monitoring and analysing flexible or transient processes that typically occur in manufacturing environments, in which the human factor plays a very important role in the execution and planning of activities. In order to facilitate the development of such a system, the authors provide a framework for the management of processes that also involve physical activities, thus extending the scope of business process management into physical systems. The research work approach of Repta et al. (2016) considers a different approach, starting with the modelling phase, and instead of a dedicated language or notation, in order to capture the entire complexity of the considered domain, ontologies based on Description Logics (DL) were used. By employing this approach, the subsequent tasks involving processes can be cast as logic reasoning problems. The approach starts with the tasks of modelling events, activities and plans, and monitoring the unfolding of activities – the activity recognition task. In order to maintain the consistency of the defined plans, the authors considered a solution based on Semantic Web Rule Language (SWRL) for enabling a form of temporal reasoning in DL, the underlying logic formalism. The framework represents the integration with the physical layer going beyond the concepts of ‘activity’ and ‘resource’ and considers real-time events collected from the physical environment through sensors, thus enabling the detection of activities.

Milicic et al. (2016) provide a product-based semantic perspective on manufacturing systems, by developing a system for automated data analysis and data modelling. Hence, data from the Product Lifecycle Management (PLM) domain were successfully modelled mathematically by exploiting the advantages of having a semantic model of data, relaxed time constraints and by allowing suboptimal accuracy. The development of the system encompasses six steps: Data Understanding, Data Pre-processing Module, Correlation Detection, Modelling, Result Evaluation and Visualisation.

In the approach of Milicic et al. (2016), Data Understanding is the process where a human expert needs to understand the domain from which the data are collected, how they are collected and how they are co-dependent. This implies for a data miner collecting the data and an ontology expert understanding the domain to be modelled. Data Pre-processing Module uses the ontology as a base layer of a data mining system, and is crucial for automation of the pre-processing step. When the ontology covers an entire product lifecycle where no actor has access to all parts of the domain, some valuable dependences are left unnoticed, and thus the importance of Correlation Detection in this stage, with the focus on correlations between only two attributes, as even in this case a variety of cases need to be tackled. For Modelling, and since the system is automatic and self-initiated, it runs as a background engine that can be left running and storing results. As a consequence, a number of modelling attempts can be performed, and sequentially, one by one, variables from the data set are declared as targets and then an attempt to model the system is made using the remaining data. Many of these attempts will show no results, but if one or a few do it is achieved as gained value. Every data-modelling procedure starts with data splitting into training, validation and testing subsets. As is usual, training data are used for teaching the model, validation for parameter tuning and testing data are for final model estimation. Three algorithms were selected for each type of problem, classification and regression in such way that they cover a wide range of data set characteristics. Examples of covered sets are those with existing outliers, noisy data and high-dimensional data. Hence, Result Evaluation for the classification problems uses Matthews Correlation Coefficient (MCC) as it is known as the best summary of the confusion matrix. Thus, the models coming from all three classification algorithms are evaluated on test set data and the one with the highest MCC coefficient is chosen to be presented to the user. For regression problems, a mean square error is used since all algorithms and their models are evaluated on the same test data set. Finally, according to Milicic et al. (2016), the question of Visualisation becomes a non-trivial issue when it comes to presenting the results of un-asked questions and is further complicated when there is no information about the end-user’s level of expertise. The authors approach to ensure that the system acts appropriately in every situation is to always select the ‘safe choice’ as the default mode in order to safeguard non-experts in the field of data analysis.

Ferreira  et al. (2016) propose a novel framework for a plug-in component as a software adapter for an End-to-End manufacturing platform development, composed by Web services, mappings, portal, Enterprise Service Bus (ESB), and databases (Ontology DB and Integration DB). The End-to-End manufacturing concept embraces the philosophy of directly connecting buyers and suppliers, eliminating middle steps as much as possible and improving their business processes’ effectiveness. The proposed plug-in is responsible for establishing and managing the sustainability of seamless communications between each company legacy system and the business front end. An adapter is a crucial component, becoming an ‘active gate’ that applies service interoperability, knowledge alignment and model transformation techniques to facilitate an environment of SMEs to adopt the Dynamic Manufacturing Networks (DMN).

Furthermore, Ferreira et al. (2016) present the application of the scenario in the furniture industry. It demonstrates how it is possible for manufacturing companies to interoperate within a manufacturing network management system, such as the i_Platform developed for the IMAGINE project. The methodology identifies the steps that each enterprise willing to use the i_Platform needs to perform: (1) subscription of the company to the adapter portal; (2) define services to enable access to public (business) information; (3) publish that information in the i_Platform, which requires internal mechanisms to transform data among legacy and i_Platform formats; (4) announce a production requirement; (5) search for partners to support in the production; (6) select the most adequate DMN; (7) collaborative production and monitoring of execution. The methodology addresses many concerns related to the interoperability, such as common data structures, semantics and methods for data extraction and input. Due to its generic principles, which embody the overall and present research challenges, the framework can be adapted to other industrial scenarios or application domains.

The ‘collaboration context’ within networked enterprises is also the focus for Mehrbod et al. (2016), addressing the challenge of matching heterogeneous e-catalogues in Business-to-Business (B2B) marketplaces in manufacturing settings for electronic procurement (e-procurement). The business challenge is to match an e-catalogue from a buyer with product e-catalogues provided by the suppliers, and thus buyer companies find supplier partners in e-marketplaces. Since each business actor may use a different structure and classification for e-catalogues, it is not easy to match a product with the e-catalogue requested by another partner. The authors use Vector Space Model (VSM) and customise it to solve the matching problem of the e-catalogues using a combination of values, names and location of attributes of structured documents to find the syntactic correlation of e-catalogues. The authors also use ontologies to expand the matching mechanism with semantic relationships of data attributes. The proposed approach makes it possible to use the benefits of all available ontologies and schemas but not to be dependent on them.

In order to test the e-catalogue matching mechanism, Mehrbod et al. (2016) deployed a supplier finder service implemented in an e-procurement platform, where one of the major services that is provided by the platform to the users is the capability to search and find a company in the B2B social network. This service is called ‘Supplier Finder’ and provides company search and e-catalogue matching facilities to the users and helps them to find a company using its business information. The service can be used to find companies using their profile data or/and their product e-catalogues. In order to evaluate the matching performance, a set of product e-catalogues files in the following three general cases were inserted into the e-catalogue repository: (a) unstructured text such as PDF files which are common in online commerce; (b) structured or semi-structured e-catalogues which are unknown for the system such as enterprise-specific formats; and (c) structured standard documents that are known for the system such as cXML and UBL e-catalogues. The authors demonstrated through experimental results that the proposed approach improves the similarity ratio between similar e-catalogues compared with the basic approach of VSM. Moreover, the most important value proposition of the proposed approach is its simplicity and practicality for implementation and is not dependent on any assumption or underlying structure and is extendable to any new type of e-catalogue.

Finally, exploiting cloud services will become a vital area for ‘FoF’, particularly as the level of CPS exponentially increases the consumption of resources in the clouds. Indeed, the collaboration between different cloud vendors can provide better quality of service at a lower price. Nodehi et al. (2016) designed a framework to support inter-cloud interoperability in a heterogeneous computing resource cloud environment with the goal of dispatching the workload to the most effective clouds available at runtime. The methodology considers also a Genetic Algorithm (GA)-based job-scheduler, as a part of the interoperability framework offering workload migration with the best performance at the least cost, and the resource selection model is evaluated using Agent-Based Simulation (ABS) approach.

The authors consider three scenarios to run the ABS model: single-cloud-provider, multiple-cloud-provider without using proposed GA solution and multiple-cloud-provider using proposed GA solution. The simulation results conclude that the response time improves using multiple-cloud-provider without using proposed GA solution compared with single-cloud-provider and improves further in multiple-cloud-provider using proposed GA solution. Additionally, the simulation results imply that the overall profit for cloud subscriber increases using multiple-cloud-provider without using proposed GA solution and compared with single-cloud-provider and increases further using multiple-cloud-provider using proposed GA solution compared with multiple-cloud-provider without using proposed GA solution, thus the total cost for cloud subscriber considerably reduces using the Inter-Cloud Interoperability Framework (ICIF) with GA-based solution.

5. The virtual organisation of manufacturing systems

A VO has been defined as a Collaborative Networked Organisation (CNO) that comprises a temporary alliance among organisations that share their skills or core competencies and resources in order to better respond to business opportunities, where cooperation is supported by computer networks (Camarinha-Matos and Afsarmanesh 2005). VOs have some unique characteristics like geographic dispersion, temporary existence and being enabled by an ICT infrastructure. Manufacturing environments have been evolving over the years from traditional integrated companies to extended enterprises supported by global supply chains, and more recently to more open and diversified nature of VOs, partly enabled by the new configurations of ICT. Whereas in the past, research on VOs in the manufacturing environments focused much on developing architectures and ICT infrastructures (Rabelo 2008) to enable the traditional forms of manufacturing buyer–supplier relationships, our analysis uncover that VOs are now framed in a new evolving manufacturing systems and VOs are being designed to support other forms of cooperation. The works of Knoke, Missikoff, and Thoben (2016), Shamsuzzoha et al. (2016) and Gorecky, Khamisa, and Muraa (2016) present evidence of how VOs research and development is contributing to the evolution of manufacturing systems.

Knoke, Missikoff, and Thoben (2016) address the challenge of ‘collaborative open innovation’ in virtual manufacturing enterprises (VMEs). A knowledge-centric approach and a business innovation reference framework (BIRF) are proposed and characterised by flexible guidelines capable of providing useful direction without hindering freedom and creativity with tight engineering practice. The authors design a set of methods to support open innovation management in virtual enterprises where the innovation-centric components and their underlying methodology are: (a) the BIRF; (b) Virtual Innovation Factory (VIF); and (c) the Collaborative Innovation Capability Maturity (CICM).

The BIRF avoids the traditional idea of basing an innovation framework on a process-oriented approach. The authors argue that innovation needs freedom in choices and behaviours; therefore, the authors believe that a process-oriented approach risks tying creativity. Thus, the challenge has been to identify a knowledge-oriented approach, where innovation objectives are associated with knowledge objectives in a declarative, non-prescriptive way. VIF is a software platform that supports the innovation activities, and was conceived to support collective knowledge creation. The primary components are: the Shared Semantic Whiteboard; the Open Innovation Observatory; and the Production and Innovation Knowledge Repository. Finally, the basic paradigm for the CICM has been set by the Capability Maturity Model (CMM) and, accordingly, five levels of innovation capability maturity have been defined that scale from a ‘reactive’ and ‘poorly’ controlled to ‘continuously optimising’ processes. Knoke, Missikoff, and Thoben (2016) describe that the methodology have been experimented in two very different industrial settings: (a) a high-tech context (robotics and embedded control systems) and (b) low-tech context (wood and furniture). The experiments proved the positive impact of the flexible framework, and the integrated platform for shared knowledge collection and exchange.

Shamsuzzoha et al. (2016) define a collaborative business process monitoring system within Virtual Factory (VF) environment, tracking through visualisation over a user interface dashboard that features state-of-the-art business intelligence and provides data visualisation, user interfaces and means to support VF partners to execute collaborative processes. This data visualisation provides critical operational matrices required to manage VFs. Key reporting outputs such as Key Performance Indicators (KPIs) and day-to-day operational data can be used to monitor and empower partners’ processes that help to drive collaborative decisions. VF broker or VF partners also retain full flexibility to create, deploy and maintain their own dashboards using an easy-to-understand wizard-driven widget and an extensive array of data visualisation components such as gauges, charts, maps, etc.

Shamsuzzoha et al. (2016) defined a VF architectural framework with a set of components and their interfaces. The proposed framework consists of three layers: User Interface Layer, Process Management Layer and Data Management Layer. The User Interface Layer is considered as the top level which is interfaced directly with middle layer named as Process Management Layer. The bottom level consists of the Data Management Layer. The User Interface Layer is also considered as the ‘dashboard layer’, where VF partners directly are interfaced with the various collaborative processes. This ‘dashboard layer’ consists of several user interface components such as visualisation and configuration, message exchange, process designer, user role and information management. The Process Management Layer of the VF architectural framework is responsible for process-related activities and consists of components such as execution, monitoring, adaptation, forecasting and simulation, and optimisation. Data Management Layer can be defined as the activities related to data processing within the VF. As the other two layers, this layer also consists of several related components such as cloud-based data storage, data exchange, data search, data discovery and other database.

The VO of manufacturing environments will demand a continuous qualification of human worker about new and changing technology trends since the human is the most flexible entity in the production system. Gorecky, Khamisa, and Muraa (2016) address novel approaches for knowledge delivery and skill transfer and propose an advanced virtual training system. The authors developed the VISTRA system, a solution for the automatic generation of diverse training content with no mandatory manual authoring stages. The basic data for the training experience are derived from the engineering systems of the PLM. Every aspect of the VISTRA training is automatically created without the need for experts or instructional designers. All aspects of the training are driven by the system itself, from the templates for the creation of training plans to the instructional scaffolding ensuring the most effective transfer and retention of knowledge and development of skilled performance, clear communication of the results of training to the trainee, the possibility to train using games or to take tests, such that the overall experience is highly immersive and extremely simple to navigate.

To address an automated training content generation created by Gorecky, Khamisa, and Muraa (2016), the VISTRA training system incorporates an integration methodology based on a reference architecture for an interoperable information interface. The information interface merges heterogeneous enterprise data from planning processes in a unified information model that is provided as input to the virtual training application. The central component of the VISTRA system is the VISTRA Knowledge Platform (VKP) that is responsible for building the bridge between the existing digital factory data and the virtual training as client application. The second component of the system is called VISTRA Training Simulator (VTS). The VTS module represents a UNITY 3D-based virtual assembly simulation where the virtual training is performed in an interactive and playful manner. Training scenarios are loaded dynamically from the VKP considering available products, stations and assembly sequences, trainer guidelines and trainee profiles. The third component is the VISTRA Knowledge Sharing Centre (VKSC), which trainers execute for planning training sessions and reviewing training results. The VKSC module also allows checking, editing and manually extending imported enterprise and user data, when necessary. The authors have successfully validated the VISTRA virtual training systems prototype at two automotive end-user sites.

6. The servitisation of manufacturing systems

The process of enhancing value to products by adding services, often designated by ‘servitisation’, through the creation of bundles of products and services (Vandermerwe and Rada 1988; Thoben, Eschenbächer, and Jagdev 2001) has been an important trend in manufacturing environment and caused relevant impact on manufacturing systems by demanding new business models and new business relationships between end-customer, distributor, manufacturers and suppliers. This research area is addressed by Wiesner and Thoben (2016) and Angulo et al (2016) who provide new insights on how FoF should address the servitisation challenge.

Wiesner and Thoben (2016) address the requirements for models, methods and tools to support servitisation through collaboration of manufacturing enterprises and service providers. A novel approach to elicit these requirements is proposed and aims to fulfil them by establishing Manufacturing Service Ecosystems (MSEs), and developing models, methods and tools to realise products and services in VMEs. The approach covers the challenges of servitisation, namely, service innovation, ecosystem governance and support of the transition by guidelines, techniques and ICT tools in three different domains: (a) physical resource-related, (b) organisational/human-related and (c) IT-related. The distributed architecture of the system developed spans three levels: (a) single enterprise, (b) manufacturing ecosystem and (c) future Internet of Services.

Wiesner and Thoben (2016) refer that the level of the single enterprise (manufacturers, suppliers, customers, service providers and service consumers) includes knowledge models and ontologies, service descriptions of tangible and intangible assets, access services to private data vaults and legacy systems. For the level of the manufacturing service ecosystem (VFs and virtual enterprises, supply chains and global value networks), it includes the assets owned by the ecosystem itself, as well as the experiences gained by the VFs and virtual enterprises from time to time (created, operated and dissolved) in the ecosystem to respond to collaboration business opportunities. The level of the future Internet of Services (including Cloud Computing, Platform as a Service, Software as a Service Utility) is where basic platform services, utility services and value-added services will be provided globally to all the enterprises under a flexible and dynamic composition. Based on the requirements, analysis, supporting models, methods and tools could be specified for servitisation and collaboration, leading to the development of an Innovation Reference Framework, a Feedback Management Service and a Virtual Marketplace. For the MSE in general, an Innovation Ecosystem Platform (IEP) provides a platform to manage in terms of level of activity, interactions, roles and flow of information between its members. The need for support in the organisational, functional and governance dimensions led to a Methodology for Management and Governance of MSEs that supports the setup of a VME for a specific collaborative business opportunity. The Virtualisation Method for Tangible/Intangible Assets supports the composition of the right assets for a desired product–service combination.

Grounded on the servitisation research area, but more from a technological perspective on services, Angulo et al (2016) proposed a Service-Oriented Architecture (SOA) and ICT infrastructure to support eco-efficiency performance monitoring in manufacturing enterprises. The authors developed the Factory ECO-MATION architecture that integrates data from environmental sensors, production information and energy consumption records in order to allow their simultaneous monitoring, enabling the production planning and control parameters to be adjusted to keep all production goals at a reasonable scale while protecting the environment. The ICT infrastructure is based on an SOA that integrates production and environmental information from diverse data sources in order to support the eco-efficiency performance monitoring of manufacturing enterprises. Data for distributed applications are divided into two layers: (a) the Sensor Virtualisation layer and (b) Service Provisioning layer; and together, these layers constitute the virtual counterpart architecture.

Angulo et al (2016) describe the Sensor Virtualisation layer as representing the lower level of the virtual counterpart, gathering data from the sensors and delivering it to the database in the service provisioning layer. The Interaction sub-layer contains the set of functions and methods that enable and manage data delivery towards the service provisioning layer. The Coordination sub-layer manages communication with the physical sensors and requests access to the Communication layer. The Communication layer coordinates the dataflow from individual virtual sensor networks and stores the attributes of the whole communication structure according to the participating networks. The Service Provisioning layer represents the high level of the virtual counterpart. It stores sensors’ information in the database, and the information collected in this layer can be used by high-level applications. The Web Service based on RESTful sub-layer defines the SOA layer of the architecture.

7. Final considerations on the challenges and leading innovations for intelligent manufacturing systems for the factories of the future

FoF embrace ICT-based production systems with high-quality manufacturing technologies and intelligent capabilities aimed to optimise their performance with a high degree of autonomy and adaptability for a balanced combination of high throughput with accurate production, also embracing the implementation of the sustainable manufacturing concept. Through the analyses of the most recent advances in research into challenges and leading innovations in intelligent manufacturing for the FoF, the authors have uncovered five broad categories: manufacturing systems frameworks, theories and models; the pervasiveness of CPSs; the critical role of semantic technologies and interoperability; the VO of manufacturing systems; and the servitisation of manufacturing systems. Within each of the categories considered, authors see common themes, as well as diversity of focus within the theme.

In the first category, manufacturing systems frameworks, theories and models, two papers provide frameworks, one for the ‘CFoF’, enabling the abstraction of disparate digital factories as collaborative networks of entities, which in turn can be engineered and augmented through operators in order to address FoF challenges of scalability and sustainability, and the second framework, targeted at entrepreneurs to help them in the ‘evaluation and validation of new manufacturing technologies development’. In this category, a paper is presented with a set of models that enable examining the relationship of decision levels, performance measures and modelling, and decision support approaches for manufacturing and logistics application fields for strategic, tactical and operational decisions.

In the second category, the authors demonstrate how the pervasiveness of CPSs is enabling intelligent manufacturing. In one of the papers, it is discussed how evolving technologies in the domain of ‘IoT’ in project management, by capturing events on the shop-floor and determining the meaning of information about the perceived events can help decision-making in heterogeneous, highly dynamic environments. In another paper, the authors present how a ‘CPS-based infrastructure’ collects and monitors energy data in real time for manufacturing and production systems for improving energy consumption awareness and analysis to identify where to take actions in the manufacturing process in order to reduce the energy consumption.

The third category addresses the critical role of semantic technologies and interoperability in manufacturing intelligent environments. It proposed a design of a semantically enabled system for automated data analysis and data modelling, capable of monitoring and analysing flexible or transient processes that typically occur in manufacturing environments, in which the human factor plays a very important role in the execution and planning of activities. It also proposed a novel framework for a plug-in component as a software adapter for an End-to-End manufacturing platform development, composed by Web services, mappings, portal, ESB and databases. In another paper, VSM is customised to solve the matching problem of the e-catalogues using a combination of values, names and location of attributes of structured documents to find the syntactic correlation of e-catalogues, using ontologies to expand the matching mechanism with semantic relationships of data attributes. Another paper created a framework to support inter-cloud interoperability in a heterogeneous computing resource cloud environment with the goal of dispatching the workload to the most effective clouds available at runtime.

In the fourth category, the VO of manufacturing systems is addressed. One of the papers presents a promising solution for the challenging problem of collaborative open innovation regarding VOs in the manufacturing context, based on a knowledge-centric approach and by flexible guidelines capable of providing useful direction without hindering freedom and creativity with tight engineering practice. A novel approach is also presented for knowledge delivery and skill transfer through an advanced virtual training system, supporting the VOs of manufacturing environments as they demand a continuous qualification of human worker about new and changing technology trends.

Finally, the servitisation of manufacturing systems category starts by addressing the requirements for models, methods and tools to support servitisation through collaboration of manufacturing enterprises and service providers, covering the challenges of servitisation, mainly of service innovation and ecosystem governance. Also in this category, an SOA and ICT-Infrastructure to support eco-efficiency performance monitoring in manufacturing enterprises are proposed.

All of the categories of research are yielding research into challenges and leading innovations in intelligent manufacturing for the FoF, though there is a clear need for ongoing research. Nevertheless, future work needs to embrace the technology development that could support the implementation of these concepts, resulting in an integrated framework, theories and models able to support the implementation of leading innovations in intelligent manufacturing for the FoF.

Disclosure statement

No potential conflict of interest was reported by the authors.