Keywords: virtual commissioning, emulation, performance evaluation, benchmark-
Current research and developments in next generation manufacturing control sys- tems, and specifically HolonicManufacturing Systems, recently emphasized the maturity of the underlying concepts and methods . In this context, next step is a dissemination of the concepts, primarily through a wide industrial acceptance of the related developments. However, those control architectures suffer from a lack of performance guarantee, as they are mainly based on emerging behaviour tech- niques, such as multi-agent systems or holonic paradigm, making the performance of the control system highly dependent on the context of execution of the experi- ment .
Keywords: Internet of Things, holonicmanufacturing approach, wireless sen-
sor networks, communicating material, data aggregation, energy efficiency
The Internet of Things (IoT) is widely used in manufacturing, monitoring applications and logistics, etc. The application of IoT makes the product more intelligent, its high flexibility brings new challenges to the Intelligent Manufacturing System (IMS): more flexible adaptability, higher computing capacity and control capabilities. In 2009, the Research Center for Automatic Control (CRAN) began to study the “communicating materials” concept, it is the material that can process, store data and communicate with the environment. With further study of Kubler  and Mekki , this concept has been applied to construction industry. The McBIM Project  (Material communicating with the BIM - Building Information Modelling) aims to design a “communicating con- crete”. Namely, it is the concrete equipped with embedded wireless micro-sensor net- work (WSN), which can measure the physical environment, store the information and exchange data with BIM platforms. Another objective is to demonstrate the usefulness of this new concept for different phases of a building's lifecycle, i.e. the manufacturing, construction and exploitation phases.
This solution is probably one of the most powerful that can be implemented with the constraints presented before. But, even if the implementation costs are relatively low, this solution is not so easy to implement and is thus made for large facilities. The purpose of this paper dealing with the Holonicmanufacturing systems, this constraint is very often respected as these systems are generally relatively large. A lot of technologies might be used to implement this observer, according to the objectives that were designed. The choice that was made here is to use discrete-event simulation. Indeed, a lot of simulation pieces of software meet the requirements of our study. First, it was widely used to model the behavior of such production systems. As a matter of fact, the model that was eventually made for the design of the facility can be used again in the production phase. This implies an interesting diminution of the investment time and costs. Then, the available means of communication are generally very well adapted to the communications inside such control architecture. Their graphical user interfaces is also very interesting, as it enables a clear vision on the behavior of the system for an operator. Finally, the state gathered on the observer is particularly well adapted to use to initialize online simulations in order to predict the future behavior of the system.
Abstract.Holonic Manufacturing Systems are a response solution for the emer-
gent need of flexible, reactive and productive manufacturing systems. This pa- per relies on PROSA, a classical holonic referencearchitecture, whichmakes use of a product specification, a process specification and a means to determine a resource’s productionabilitiesand capacity, but does not define a specific me- thod for representing such.This paper proposes an approach to define a prod- uct’s process specification model that integrates the principles and advantages of Service-oriented Architectures, Petri-Nets and Product Families. Then, are- definition of the basic holons is given to have a glimpse on a possible exploita- tion of this new approach, together with a short-term forecasting strategy, for the flexible orchestration of workflows. Finally, it is shown how the proposed- product’s process specification model enhances the HMS’s flexibility,reactivity and productivity giving rise to a Service-oriented HolonicManufacturing Sys- tem.
Since the beginning of the 21 st century, service oriented architectures developed,
dedicated to the interoperability of computer services in the companies. Along time, this architecture, mainly based on autonomy, negotiation and data distribu- tion, was transposed on the shop floor . In the field of systems involving a high level of information and control distribution, HolonicManufacturing Systems are more and more common in both academic and industrial worlds , and par- ticularly well adapted to the implementation of SOA. Many decisions must be tak- en by individual holons during a production. These decisions are based on the data that are retrieved on the HMS, but are generally insufficient for the operator to forecast the behavior of the system. In addition, the holons face the problem of myopia characteristic of HMS .
Abstract In the context of manufacturing in Industry 4.0, software systems be- come of prime importance. Efficient, adaptable and trusted software services are re- quired. Several approaches succeeded in creating a Service-oriented Holonic Man- ufacturing System that combines the advantages of Service-oriented Architectures and HolonicManufacturing Systems. However these systems until now suffer from many shortcomings, among which genericity, lack of proof of the functional be- haviour correctness, architecture modularity, etc. These systems are often manually implemented and become hardly adaptable and reconfigurable to different contexts (resources, workshop...). In this paper, we investigate a Model Driven Engineering approach to represent these systems in order to automate the generation of the ser- vices logic code from an abstract models and construct a new software chain that deals with all the shortcomings cited above.
Mass customization (MC) refers to a business strategy that combines two different business practices: mass production and craft production. The MC concept is relatively fresh in international business, first discussed by (Davis, 1987). Its development lagged behind because customer needs did not have effective means, i.e. technology, to be expressed and reached by product and service manufacturers. In the recent decade, changing economic and social environments gave the push for the demands of individualized products and services. Companies are now becoming more and more customer- centric. The major objective of MC is to improve the ability of companies to react faster to changing customers’ needs and to address the heterogeneity of demand more efficiently. Nonetheless, the MC concept requires new approaches due to the small volume/high variety order management and the maximization of the profitability of the firms (Blanc et al., 2008). (Molina et al., 2005) argue in this sense that the next generation manufacturing systems must therefore be able to provide increased levels of flexibility, re-configurability and intelligence to allow them to respond to product variety. These concerns are challenging the Intelligent Manufacturing Systems (IMS) community from two decades, through a worldwide industry-led research aiming at setting into practice agile Business to Manufacturing systems based on a networked heterarchy of autonomous units. This type of organization appears as more suitable to meet robustness to disturbances, adaptability to rapid changes and efficient use of available resources, which are still weak points of conventional manufacturing systems (Morel et al., 2003). The difficulty is to combine the ability and the capacity to
manufacturing systems. The combination of both of these principles appears to be a very attractive option as seen in works relating these two paradigms,[1–3]. However, there are no works been found describing how a process is composed, based on ser- vices, nor has been the composition of a service representing integrally a manufactur- ing process with an eye on its application on HMS. Moreover, flexibility, being the main attribute soled by these paradigms depends on the intrinsic flexibility found at all levels of the system. The way information is presented in the system will greatly define the flexibility limits at higher levels for finding new solutions; as in process planning and reconfiguration. For instance, if the information describing the system’s components and activities fails to express the underlying capacities and possibilities, the intrinsic flexibility present in the production floor will not be identified nor con- trol strategies will be exploited to its best. The objective of this paper is to propose a methodology for designing manufacturing process specifications based on Workshop services that welcomes product customization and is suitable for their application on distributed and product-driven systems. Conceptual models for Manufacturing Pro- cess and Workshop-services are proposed in this work designed to preserve the fractal characteristics of products and promote the reutilization of operations. Fractality of services (i.e. a same model of services from highest to lowest level services, com- pound or atomic) is especially efficient as it naturally fits to the nature of the re- quester. The products, resulting from the execution of services, therefore needs to be fractally modelled in order to take advantages of the structure of the corresponding services.
Big Data: One important aspect of the industry 4.0 is the massive usage of sensors, monitoring the production processes. Such
hyperconnected systems will generate rapidly a huge amount of unorganized data not always nor useful nor reliable. As underlined by (Babiceanu & Seker, 2016), manufacturing companies will progressively all be confronted to these data management issues grouped under the term Big Data. As stated before, Learning is a fundamental property of a holon, necessary to adapt the goals of the holon. Some attempts to couple Big Data techniques and HCAs have been proposed. As explained in previous sections, (Le Mortellec et al., 2013) defined a HCA aiming at limiting the data overload phenomenon (i.e an amount of data larger than the operator can handle) occurring with hyperconnected systems. However, far from being inappropriate, this new source of data can help to define new information such as machine health, traceability, production quality (Lee, Ardakani, Yang, & Bagheri, 2015; Lee, Kao, & Yang, 2014). Artificial Intelligence is then one key technologies (Jardim-Goncalves, Romero, & Grilo, 2017; László Monostori, 2014) to treat this sources of data. Moreover, new information of knowledge obtained via these techniques should then be integrated in the different elements of the HCAs. There have been attempts to apply Big Data techniques to manufacturing data (Lade, Ghosh, & Srinivasan, 2017; O. Morariu, Morariu, Borangiu, & Răileanu, 2018). However, some additional works are still required to integrate these techniques into HCAs, so that holons can autonomously learn from their environment. Data processing features should then be added to holonic architectures, to transform data into relevant information and/or knowledge for holons. One application of such techniques could be to help solving the myopia problem. Indeed, local decisional entities are subject to, at least, social and temporal myopia, i.e. suffer from a lack of information concerning their environment and consequences of their actions. This can lead to sub-optimal solutions or even deadlocks (Zambrano Rey et al., 2013). However, myopia is a needed characteristic of holonic architectures, ensuring a fast reaction. Big Data strategies can help to adjust this level of myopia depending on the information generated by analysing the shop-floor data. To keep the previous example related to potential fields, parameters of the potential field algorithm could be adjusted thanks to information coming from the manufacturing shop-floor.
Mass customization has become a main issue for many companies fighting on a world market. In order to define the customized product that fit each customer need, these companies use configuration software called configurators. Most of these configurators, mainly based on artificial intelligence techniques, are just interested in the product definition without addressing relevant manufacturing problems. The goal of this communication is to show that the same kinds of computer techniques can be used to define customized assembly operations and manufacturing routings.
In many areas illustrated in Fig. 1, workers operate often in very unergonomic postures. In such cases, they use multi- contact postures to relax stress, be in better equilibrium or cast their body posture (e.g. contact with knees, shoulders, back). Humanoids shall be endowed with similar multi-contact behaviors, which is a key technology in Comanoid. It allows transforming a humanoid robot into a reconfigurable multi- limb system that can adapt to narrow spaces, increase its equilibrium robustness and even its manipulation capabilities. 1) Multi-contact planning: A recent review in multi-contact technology  reveals that this problem is consensually ap- proached through a multi-level computation. The first level plans contacts around a sort of free-motion pre-planed “guide” that will exploit some properties under various hypothesis (there are many variants), but having as an output a set of contact sequences and associated transitions. In a second phase, the latter are the input for a simplified model (e.g. CoM) to generate consistent centroidal dynamics trajectory under balance criteria (presented later). In the last phase, this generated trajectory is the input to the whole body controller, which deals also with other task objectives and constraints. The problem in such multi-level computation is to make each phase likely feasible for the upcoming one, and make sure that if any phase turns out to be not feasible for the given input, the latter is quickly re-computed from the current state. Although impressive results have been obtained, this approach does not work well in practice. In the context of manufacturing, it is certainly not a good idea to plan contacts as if we do not know at all where and how they should be made. Assembly operations are quite repetitive in many aspects, only few variations are to be dealt with locally.
This special issue of Engineering contains ten papers—including two opinion papers and eight research papers—contributed by influential experts from China, the United States, the United Kingdom, Sweden, Japan, Singapore, and Australia. These papers focus on recent advances in a wide range of intelligent manufactur- ing fields, including artificial intelligence (AI) for intelligent manufacturing, design for intelligent manufacturing, human– cyber–physical system (HCPS) for intelligent manufacturing, the correlation and comparison of digital twins (DTs) and cyber– physical systems (CPSs), the design of context-aware smart products, intelligent machine tool (IMT), the online monitoring of laser welding status, anomaly diagnosis for machining processes, industrial big data analysis for manufacturing systems, and the upgrading pathways of intelligent manufacturing in China.
Figure 2: General manufacturing processes .
Most of the times primary processing isn’t enough to obtain finished mechanical parts because of process limitations. Some part features and their dimensional tolerances are impossible to obtain through casting. In these cases full functionality is attained once secondary manufacturing processes have been employed. In the present case study machining operations have been selected and additional models need to be created. To this purpose machining features need to be identified and a machining model needs to be created. This model contains information related to operation order and tool accessibility. Simulation is then performed to assess the part’s machinability and the validation of tool trajectories. Back loops between the designer experts’ activities are necessary to indicate the fact that the models obtained through simulation serve as controls for process optimisation. These models will guide the ranking of available process solutions according to significant technological and socio-economical parameters.
Integrated Design, Process Modeling, Process Simulation
It is well known that product development is constrained by the following demands from the part of the industrial actors: increase in production rate, cost and time to market reduction and part quality control. The product design department needs to rely on flexible and reactive solutions in order to efficiently respond to complex demands. These demands are most of the times translated into complex products and into an increase of part variants. Thus, the modeling of the product and of its manufacturing process becomes essential.
Fig. 15. Comparison of manufacturing indexes.
5. Conclusion and future work
This paper presents a new hybrid modular tool design methodology. Starting from the one-piece tool CAD model, global and local manufacturability indexes are calculated. In case of local indexes, which provide a well-detailed view of which areas of the tool are the most complex-to-manufacture, the manufacturability analysis is based on octree decomposition. This new approach allows focusing on the areas of the tool that are the most complex-to- manufacture because an accurate view of the manufacturing complexity distribution is obtained. Then, hybrid and modular points of view help designers to choose between a one-piece design and a hybrid modular one.
Figure 4 Schematic diagram showing the effect of particle size on critical velocity and particle velocity,
which defines the optimum particle size range for cold sprayability 
Cold Spray as an Additive Manufacturing Technology
The fact that cold spray is a method to consolidate metal powders automatically makes it under consideration as an additive manufacturing method. In fact, one of the main initial attractions of cold spray was to radically improve the buy-to-fly ratio of aerospace components. The commonly cited example is the creation of ‘bosses’ on components, which was probably conceived of because cold spray began ‘life’ as a coating technology. However, one of the main differences between cold spray and other thermal spray methods, as mentioned above, is the rapid deposition rate, which effectively transitioned cold spray from 2-D coatings into 3-D coatings, and from there, to additive manufacturing. Returning to the manufacturing of a part containing a boss, if a boss is created by conventionally (i.e. subtractively) machining an ‘oversized’ blank, it is clear that this leads to considerable material wastage, compared to ‘adding’ onto a blank of the correct dimensions. In aerospace, the metals tend to be relatively exotic, thus making the buy-to-fly ratio a very important economic factor and a positive aspect of cold spray.
4.3. Instrumentation and Control Perspectives
Designers of soft robots also face many other challenges that are yet to be properly addressed ( Rus and Tolley, 2015; Laschi et al., 2016 ). Some ambitious challenges lie in the instrumentation and control of such systems, and also in the design of embedded components adapted to the specificities of soft robotics. Multimaterial manufacturing makes it today possible to embed various elements in the soft matrix of the robot, in order to improve the integration of the functionalities. Modularity increases Onal and Rus (2012) , with soft robots including their own pressure sources ( Onal et al., 2011 ) and control valves ( Marchese et al., 2011 ). The inclusion of channels for liquid metal injection ( Park et al., 2012; Farrow and Correll, 2015 ) or the embedding of conductive hydrogels and electroactive fillers in flexible materials ( Larson et al., 2016 ), open perspectives in soft electronics ( Correll et al., 2014 ), a step further in the development of sensing systems for soft robots. To date, reliable sensors able to measure the state of a soft system only provide partial information on the state of the system, though more and more solutions have been proposed, using for example liquid metals ( Park et al., 2012; White et al., 2017 ) or microstructured metal on polymeric substrate ( Araromi et al., 2016; Atalay et al., 2017 ). The control of soft robots is also a very open challenge, and it has to be acknowledged that modeling soft device for realtime control is currently at a very early stage in spite of noticeable contributions ( Duriez, 2013; Coevoet et al., 2017 ).
increased incidence of email use within a region/city on manufactures’ performance. The effect on sales and sales per worker is of equal size, suggesting that such Internet spillovers are neutral in terms of employment level. However, these spillovers are subject to important threshold effects: the network of email users has to reach a critical size in order for the positive externalities of internet diffusion to be effective. Otherwise, our findings show that the burgeoning diffusion of internet technologies may only benefits first-movers and large and productive firms with sufficient absorptive capacity, at the expense of more fragile ecosystems, thereby concomitantly causing manufacturing output to decline. We indeed find evidence of negative spatial spillover effects on manufacturing firms’ sales and productivity from internet diffusion across industries on manufactures’ sales and productivity, where email incidence is be- low a threshold corresponding approximately to 50%. At the firm level, we also show that the sign of these spillover effects critically hinges on manufactures’ absorptive capacity. These results therefore lend credibility to the hypothesis that email diffusion may, in certain contexts, primarily benefit the highest performing firms, as already evidenced in other studies (Bustos, 2011; Paunov & Rollo, 2016; Rodrik, 2018).
KNOWN CYBERSECURITY RISKS POINT TO VULNERABILITIES IN
U.S. biopharmaceutical companies together spend nearly $160 billion each year on R&D, and their accumulated intellectual property (IP) is likely worth trillions of USD ( Research America., 2016 ). An advanced, persistent attack could allow corporate rivals to steal internal communications, IP related to the product or process, and facility monitoring data to gain a competitive advantage. A malware program called Dragonfly specifically targets cyber-physical systems used in pharmaceutical manufacturing equipment, stealing trade and manufacturing secrets as a form of corporate espionage ( Carman, 2014 ). Some have suggested that Dragonfly could also be used for physical sabotage in the future ( Symantec Security Response., 2014 ). Pharmaceutical companies hold patient data related to clinical trials and disease management in their corporate networks. Since the data is both highly sensitive personal information and regulated, breaches can both incur large fines and damage a firm’s reputation. Assessing emerging cybersecurity risks across the biopharmaceutical industry is especially important and timely as many companies work to establish digital strategies and data lakes that serve as repositories of data from across