D2.1 – Dynamic Worker Model

D2.1 – Dynamic worker model

D2.1 –Dynamic worker model

Empowering and Participatory Adap- tation of Factory Automation to Fit for Workers

Abstract Smart factories are characterized by increasing automation and increasing customization. In these dynamic environments flexible and adaptive work organization is crucial, both for productivity and work satisfaction.

The Factory2Fit project will support this development by developing adaptation solutions with which people with different skills, capabilities and preferences can be engaged, motivated and productive members of the work community in manufacturing industries.

This deliverable describes the development of a work-related abstract profile of an employee, called Worker Model, and the development of a more dynamic worker model, Adaptive Worker Model. These models include worker characteristics and their relations with task-related, machine-related and contextual characteristics, respectively. Methodologies, standards and theories for user modelling development are also introduced in this deliverable. Alongside, this document provides a further analysis of the wearable measurement equipment that can be used for worker monitoring and the online worker monitoring and off-line measurements acquired from these devices. These measurements will be used for the definition of worker characteristics. Furthermore, an analysis of data storage mechanism that will be developed is provided.

Keywords: Used Model standards and methodologies, Worker Model, Adaptive Worker Model, worker characteristics, task characteristics, machine characteristics, contextual characteristics, Online monitoring, Off-line measures, Data storage mechanism.

D2.1 Dissemination Level: PU (Public) Deliverable Type: R

The research leading to these results has received funding from Horizon 2020, the European Un- Page 1 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Authoring and review process information

EDITOR DATE Anastasios Drosou / CERTH-ITI 17.03.2017

CONTRIBUTORS DATE Michael Bojko, Ralph Riedel, Sebastian Mach, Franziska Schmalfuß & Matthias Beggiato / TUC 14.03.2017 Anita Honka & Juhani Heilala/ VTT 14.03.2017 Rakesh Mehta, Cemalettin Ozturk/ UTRC-I 16.03.2017 Dimitris Zarpalas and Nikos Zioulis/ CERTH-ITI 16.03.2017

REVIEWED BY DATE Päivi Heikkilä / VTT 23.03.2017

APPROVED BY DATE Thomas Walter / Continental 30.3.2017

Version History Date Version Author Description

27.01.2017 0.1 Anastasios Drosou / CERTH-ITI ToC finalized Section assignments distributed to all involved 10.02.2017 0.2 Anastasios Drosou / CERTH-ITI partners

11.03.2017 0.3 Anastasios Drosou / CERTH-ITI Input from TUC (CEP) received 14.03.2017 0.4 Anastasios Drosou / CERTH-ITI Input from TUC (FPL) received

14.03.2017 0.5 Anastasios Drosou / CERTH-ITI Input from VTT received First draft sent circulated for early review. 14.03.2017 0.6 Anastasios Drosou / CERTH-ITI Input from UTRC-I & final contributions form VTT expected. 15.03.2017 0.7 Anastasios Drosou / CERTH-ITI Final contributions from VTT received.

16.03.2017 0.8 Anastasios Drosou / CERTH-ITI Minor inconsistencies spotted. 16.03.2017 0.9 Anastasios Drosou / CERTH-ITI Final contributions from UTRC-I received.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 2 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

17.03.2017 0.91 Anastasios Drosou / CERTH-ITI Final version for internal review.

23.03.2017 0.92 Päivi Heikkilä / VTT Review complete and suggested edits provided.

24.03.2017 0.93 Anastasios Drosou / CERTH-ITI Final version for approval.

31.3.2017 1.0 Eija Kaasinen / VTT Completed list of abbreviations.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 3 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Table of Contents

1 Introduction ...... 8 1.1 Purpose of the document ...... 8 1.2 Intended readership ...... 8 1.3 Relationship with other Factory2Fit deliverables ...... 9 1.4 Acronyms and abbreviations ...... 9 2 Common (Virtual) Worker Models Glossary...... 11 3 Relevant Standards & Methodologies ...... 13 3.1 Introduction to Ontology based models ...... 13 3.2 Standards Related to User/Worker Modelling ...... 17 3.3 User Interface Description Languages (UIDLs) ...... 20 3.3.1 UIDLs Overview ...... 20 3.3.2 UIDLs Comparison ...... 24 3.4 Physical Modelling ...... 26 3.5 User Behaviour Simulation ...... 33 4 Theories of work satisfaction and performance ...... 38 4.1 Definition of job satisfaction ...... 38 4.2 Factors influencing job satisfaction...... 38 4.3 Models of work performance ...... 41 5 Factory2Fit Worker Modelling Methodology ...... 44 5.1 Factory2Fit Worker modelling methodology outline ...... 44 5.2 Work-related Profile...... 46 5.2.1 Strong characteristics of the Worker Model ...... 50 5.2.2 Soft characteristics of the Worker Model ...... 51 6 Adaptive Worker Model...... 54 6.1 Definition of the Adaptive Worker Model ...... 54 6.2 Task characteristics ...... 56 6.3 Machine characteristics...... 59 6.4 Contextual characteristics ...... 60 7 Online Worker monitoring and off-line measurements definition ...... 62 7.1 Smart Sensor Network ...... 62 7.2 Online Worker monitoring ...... 64

The research leading to these results has received funding from Horizon 2020, the European Un- Page 4 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

7.3 Off-line measurements ...... 64 7.4 Definition and Classification of Physical, Sensorial and Cognitive abilities ...... 65 7.5 Calculation of Overall Labour Efficiency/Effectiveness (OLE) and Overall Equipment Efficiency (OEE) ...... 66 7.6 (Persistence) Data storage mechanism ...... 71 8 Conclusions ...... 73 9 References ...... 74

Table of Figures

Figure 1: Relations among deliverables in Factory2Fit project ...... 9 Figure 2: Upper concepts [37]...... 13 Figure 3: Entities’ class hierarchy [33] ...... 14 Figure 4: Operations’ class hierarchy [33] ...... 14 Figure 5: Ressources’ class hierarchy [33] ...... 15 Figure 6: Section of the function ontology [34] ...... 16 Figure 7: Section of the device classes ontology [34]...... 17 Figure 8: ISA-95 information on personnel ...... 18 Figure 9: ISA-95 information on capability of resources ...... 18 Figure 10: Relationship between planning modelling and simulating in manufacturing systems...... 26 Figure 11: Framework for object models in manufacturing systems [49]...... 27 Figure 12: Data flow scheme in the virtual factory [47]...... 28 Figure 13: Schematic class overview of a possible reference model [52]...... 28 Figure 14: Taxonomy of visualization techniques from the information-related point of view [57] ... 31 Figure 15: Classification of simulation models [55] ...... 31 Figure 16: Information behaviour model [73] ...... 33 Figure 17: Original TAM [63] ...... 33 Figure 18: Framework for human performance modelling [60] ...... 37 Figure 19: Job Characteristics Model [84] ...... 39 Figure 20: Theoretical framework for human performance [85] ...... 42 Figure 21: High performance cycle [91]...... 43 Figure 22: Factory2Fit Worker Modelling methodology ...... 45 Figure 23: Characteristic analysis into its components ...... 46 Figure 24: Worker Model characteristic's categories ...... 47 Figure 25: Worker characteristic analyses...... 49 Figure 26: Worker Model division in Strong characteristics Worker Model and Soft characteristics Worker Model ...... 50 Figure 27: Adaptive Worker Model ...... 54 Figure 28: Adaptive Worker Model characteristic’s categories ...... 55 Figure 29: Mind map with all the worker characteristics for the Adaptive Worker Model ...... 56 Figure 30: Task characteristics ...... 57 Figure 31: Microsoft Band 2 device ...... 63

The research leading to these results has received funding from Horizon 2020, the European Un- Page 5 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 32: Samsung Gear S3 device...... 63 Figure 33: FitBit devices (a) Fitbit Surge,(b) Fitbit Charge HR and (c) Fitbit Charge2 ...... 63 Figure 34: Multiple dimensions- abilities a worker has (physical, cognitive, sensorial)...... 66 Figure 35: OEE calculation ...... 68 Figure 36: Data storage framework...... 71

List of Tables

Table 1: List of Abbreviations ...... 9 Table 2: Properties Comparison of UIDLs [19]...... 25 Table 3: Reference activity model [50] ...... 29 Table 4: Classification of modelling tools (similar to VDI 4499 [56])...... 29 Table 5: Overview of typical software applications for modelling in (industrial) engineering ...... 30 Table 6: Resolution levels [48] ...... 31 Table 7: Relations between application purpose and targeted group [55] ...... 32 Table 8: Characteristics of modelling methods for humans [53] ...... 32 Table 9: Comparison between adaptive and adaptable systems [65] ...... 35 Table 10: Categories of worker characteristics and characteristics’ variables ...... 47 Table 11: Characteristics included in Strong characteristics Worker Model ...... 50 Table 12: Characteristics included in Soft characteristics Worker Model ...... 51 Table 13: Characteristics in Current work context category...... 55 Table 14: Task characteristics and the defining variables ...... 57 Table 15: Machine characteristics ...... 59 Table 16: Important contextual characteristics and their measures ...... 60 Table 17: Examples of Physical, Cognitive and Sensorial abilities of a worker ...... 66 Table 18: OEE Six big losses ...... 68 Table 19: OEE factors used in measuring human workforce effectiveness ...... 69

The research leading to these results has received funding from Horizon 2020, the European Un- Page 6 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Executive Summary

This deliverable provides evidence of advances towards the achievement of project objectives in a wider scope of the development of Factory2Fit worker models. The main aim of this document is to define an extended and dynamic worker model, called Adaptive Worker Model, considering the worker characteristics and other working area related characteristics, i.e. task-related, machine-related and contextual characteristics. An overview of standards and methodologies related to user modelling is also presented in this document. Alongside, this deliverable introduces the equipment that will be used for online worker monitoring and off-line measurements needed to define worker characteristics. The mechanisms that will be used for the development of a (persistence) data storage mechanism for worker models storage are also analysed. A detailed description of the above-mentioned process and the constituting elements is given in the following chapters: Chapter 1: is an introduction of the current deliverable and provides links to different activities and deliverables of the Factory2Fit project. Chapter 2: provides a common (Virtual) Worker Models Glossary. This chapter provides the exact definitions of the terminology used in this document. Chapter 3: presents a detailed analysis of existing standards, methodologies and models related to user modelling. Each standard is followed by a brief analysis of the most popular User Interface De- scription Languages (UIDLs) and a comparative analysis between them. In this chapter there is also an overview of the physical modelling methodologies and several used approaches / solutions in ergonomics and factory planning that will be used in Factory2Fit worker modelling development. Chapter 4: provides an analysis of the existing theories for job satisfaction and performance. Chapter 5: analyses in detail the methodology used in the development of Factory2Fit Worker Model. Worker Model is a work-related abstract profile of employee needed for the development of a dynamic Adaptive Worker Model that includes the worker characteristics and also characteristics in the working area related with the worker. Worker Model segmentation into Strong characteristics of the worker model and in Soft characteristics of the worker model is also provided in this Chapter. Chapter 6: provides a detailed analysis of the Adaptive Worker Model. Adaptive Worker Model is an ontology-based highly adaptive model that combines worker characteristics, task-related, machine- related and contextual characteristics. This chapter provides a brief analysis of all these categories of characteristics. Chapter 7: presents an overview of the devices used in Smart Sensor Network and the online monitoring and off-line worker measurements used to define each worker characteristic presented in Chap- ter 5. Chapter 8: describes a data storage mechanism that will be used for Factory2Fit project. This data storage mechanism will be developed to enable access to worker profile data and store data for ana- lysing worker’s progress in following WPs.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 7 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

1 Introduction

1.1 Purpose of the document This document (D2.1) reports the research and activities in Task 2.1, Development of an integrated dynamic worker model. The purpose of this deliverable is to describe the content and dynamics of the models developed in Task 2.1, the Worker Model and the Adaptive Worker Model. Also, this deliverable introduces in detail all the worker and workplace - related characteristics that compound those models, i.e. task, machine and contextual characteristics. The basic type of Worker Model is an abstract profile of employees that includes all the worker characteristics needed for the proper definition of the worker. In contrast, Adaptive Worker Model is a dynamic update of Worker Model, which includes both worker characteristics and working area – related characteristics. Furthermore, this document provides an analysis of the existing standards, models, techniques and methodologies, in order to develop a more structured worker model. The models developed in Task 2.1 aim to define the worker his/her interrelation with the working area, sufficiently. This project follows a human centered approach, which is used also in the models development. These models are the core of the project as they include the characteristics, variables and measures that define the worker both as a human being and his/her work in a unit. The different measures acquired form each worker included in these models, will be used in further WPs in order to be evaluated and used in the adaptation of the working environment according to the worker’s needs. Proper construction and development of models is the key to developing a proper design and execution of all processes that exist in Factory2Fit project. This report also provides a brief analysis of the hardware used for Smart Sensor Network and the online monitoring and off-line measurements, which define the worker characteristics. This deliverable contains also the data storage mechanism that will be developed for Factory2Fit project. Each research and development partner has been expected to contribute to this deliverable with respect to the areas of its competence and expertise. After the deliverable D2.1 is submitted on M6, the worker model development and update shall continue throughout the project and results will be directly introduced into research.

1.2 Intended readership D2.1 is a public document (PU) and therefore is intended for the European Commission, the Fac- tory2Fit Project Officer, the members of the Factory2Fit consortium, members of other H2020-funded projects as well as a broad range of different target audiences from factory workers, employers, industrial networks and associations, to other related national and EU-funded projects, Commission staff, media, and the wider public. Even though the Worker model has been defined based on the needs of the Factory2Fit project, the results can be utilised by other projects and organizations that plan to utilise similar technologies.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 8 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

1.3 Relationship with other Factory2Fit deliverables Deliverable D2.1 is a direct outcome of task T2.1 on “Development of an integrated, dynamic worker model”, whereby worker, task, machine and environmental characteristics provided, originate from the D1.2 related with “Industrial requirements”. D1.2 includes the results of T1.1 “Industrial requirements” and T1.2 on “Enabling technologies, platforms and tools” which are related to technical, business and user requirements. D2.1 has also dependencies with other tasks in the WP1 (either done or ongoing) and specifically with T1.5 on “Adaptation architecture” where the corresponding adaptation for Factory2Fit project is defined. In WP2, D2.1 is related with all WP2 deliverables and tasks. Worker Model definition forms the basis for the next tasks, which will use this model to develop next Fac- tory2Fit adaptation features and modules (Figure 1).

The worker model defined in D2.1 will be an input in WP3. Worker Model will be used as input in T3.1 on “Adapting Levels of Automation” for the development of an ability ranking system, in order to match each worker to the proper level of automation. Figure 1 illustrates more elaborately the dependencies between the current task/deliverable and its related deliverables/WPs.

Figure 1: Relations among deliverables in Factory2Fit project

1.4 Acronyms and abbreviations

Table 1: List of Abbreviations Abbreviation Description AR Augmented Reality D Deliverable

DBMS Database Management System DISL Dialogue and Interface Specification Language

EAB External Advisory Board

The research leading to these results has received funding from Horizon 2020, the European Un- Page 9 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

EFFRA European Factories of the Future Research Association EHS Environment, Health and Safety

EU European Union GIML Generalized Interface Markup Language

ISML Interface Specification Meta-Language ISO The International Organization for Standardization JCM Job Characteristics Model

KOM Kick-Off-Meeting OAM Open Assembly Model

OEE Overall Equipment Efficiency OLE Overall Labour Efficiency/Effectiveness RIML Renderer-Independent Markup Language

SQL Structured Query Language SunML Simple Unified Natural Markup Language

T Task TAM Technology Acceptance Model

TPM Total Productive Maintenance UI User Interface UIDLs User Interface Description Languages

UIML User Interface Markup Language UsiXML User Interface eXtensible Markup Language

VW Virtual Worker WSXL Web Service eXperience Language

WP Work Package XICL eXtensible User Interface Markup Language XIML eXtensible Interface Markup Language

UIDLs User Interface Description Languages

The research leading to these results has received funding from Horizon 2020, the European Un- Page 10 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

2 Common (Virtual) Worker Models Glossary

Worker Model - A Worker Model is a work-related abstract model that refers to a high level description of several worker characteristics. Those characteristics are required to describe the worker of a unit and are broken down according to the characteristics category. This model will be stored in ontologies and will include also multiple characteristic’s related parameters like characteristic’s description, metrics, etc.

Adaptive Worker Model - Adaptive Worker Model refers not only to worker characteristics composing the Worker Model and other working area related characteristics. These elements refer to machine, task and contextual characteristics with which the worker interacts daily. Adaptive Worker Model is a dynamic and highly adaptable model.

Worker Profile – Worker Profile refers to a set of characteristics that a worker has. This model can be comprised of one or more Adaptive Worker Models, because it includes many of the worker characteristics included in this. The difference between each Worker Profile refers to the different and specific characteristics a single worker has.

Virtual worker (VW) - The virtual worker is a representation of a worker based on a Worker Profile, intended for simulation purposes. It includes components, which are able to interact with other virtual entities e.g. virtual products or software applications. VW may represent the human body as e.g. a kinematic system, a series of links connected by rotational degrees of freedom (DOF) that collectively represent musculoskeletal joints such as the wrist, elbow, vertebra, or shoulder. The basic skeleton of the model is described usually in terms of kinematics. In this sense, a human body is essentially a series of links connected by kinematic revolute joints. Each DOF corresponds to one kinematic revolute joint, and these revolute joints can be combined to model various musculoskeletal joints. A VW also includes low level control loops for muscles (basal ganglia muscle loops) and higher levels of control in a hierarchical fashion: action (motion planning, cognition etc.).

Environmental Model - An Environmental Model is a formal machine-readable set of characteristics used to describe the use environment. It includes all required contextual characteristics besides the Worker Model, the interaction model, the device model, the product and related worker tasks.

Machine Model - A formal machine-readable representation of the features and capabilities of one or several physical components involved in worker interaction. It is important to carefully discriminate between worker and machine model as they are two different kinds of models. Too often they are conflated together, with machine properties sprinkled into worker profiles and vice versa. The machine model expresses capabilities of the machine. A given machine can be used by many different users and a given user could use different devices and in different ways. By carefully separating the different functionalities of machine modelling and worker modelling in design scenarios it will be eas- ier to enumerate the attributes for each model and from them develop the matching function and attributes of the adaptation process.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 11 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Interaction Model - The interaction model is a machine readable representation of the interaction behaviour of an application. The interaction model is maintained UI-agnostic, which means it is independent of the concrete format of user interface output- and input data. Interaction model often refers to an abstract user interface model, like for example UIML, UI Socket, XForms, etc. It should be noted that the Interaction model can be used for adaptation of Human Machine Interfaces (HMI) and for simulating the use of an application /product with a virtual user.

Context Model – Context model is a machine-readable representation of information that can be used to characterize the situation of an entity. An entity is a person, a place, a device, or a product that is considered relevant to the interaction between a worker and an application, including the worker and applications themselves. [111]

Simulation - The process that enables the interaction of the virtual worker with the application model within an artificial environment. The simulation can be online or off-line. Online simulation can be performed autonomously or manually, where the operator can interact with the environment from a 1st or 3rd person perspective. Accessibility assessment and evaluation can be performed automatically or subjectively by the operator.

Adaptive Worker Interfaces - Worker interfaces that adapt their appearance and/or interaction behaviour to an individual worker according to a Worker Profile. In contrast to adaptable worker interfaces, which are modified by a deliberate and conscious choice of a worker, adaptive worker interfaces automatically initiate and perform changes according to an updated worker profile. Changes in the worker profile can be done manually by the worker or can be inferred automatically by the system in a machine learning process during the interaction between the system and an individual worker.

Worker Interface Design Pattern - Approved worker interface solution to a recurring design problem. User Interface Design has a formalized description. For the use in Adaptive Worker Interfaces, design patterns have a representation in form of reusable software components which can be put together to complete worker interfaces during run-time.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 12 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

3 Relevant Standards & Methodologies

This chapter analyses the standards related to user modelling along with popular methodologies and languages for user interface description and physical modelling. Alongside, section 3.1 introduces the ontology based models and specifically analyses the ontologies used in manufacturing and for smart manufacturing. For each individual element an overview is initially given followed by a brief analysis of its relevance to user modelling.

3.1 Introduction to Ontology based models

Ontology is a formal explicit description of concepts in a domain of discourse (classes). Properties of each concept describe various features and attributes of the concept (slots, roles). Ontology together with a set of individual instances of classes constitutes a knowledge base. [40]

Ontologies or ontology-based models have already been widely used in manufacturing systems engineering. Kiritsis [32] provides a quite comprehensive overview on recent developments.

Among the applications in recent research are the following:

• Knowledge sharing and knowledge reuse. [36] [44] • Factory systems design, factory planning. [19] [25] [45] • Assembly systems design. [28] • Design of reconfigurable [23], modular [27] [43] or flexible production systems. [31] • Design for sustainability. [30] [42] • Industry 4.0, cyber-physical systems design. [26] [29] • Production automation and sensors application. [37] [41] • Multi-agent systems for production control. [38] • Lifecycle management. [32] • Project management, semantic interoperability. [35] • Service modelling. [20]

Of special importance are so called upper concepts:

Figure 2: Upper concepts [37].

The generalization of a product is a part; a component or a sub-assembly can also be a part. Parts are usually manipulated by manufacturing processes or combined by assembly processes (Figure 2).

There are also so called upper ontologies, which consist of very general terms that are common across all applications and a multitude of domains. One example is MASON that uses a function-behaviour-

The research leading to these results has received funding from Horizon 2020, the European Un- Page 13 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model structure (FBS) approach [33]. This approach consist of three head concepts, which are entities (all common helper concepts, to specify the product like geometric entities, raw material, cost entities), operations (cover all processes linked to manufacturing like manufacturing operations, logistic operations, human operations, launching operations), and resources (stand for all manufacturing resources like geographic resources (plants, workshops), tools, and human resources). Examples are given below (Figure 3, Figure 4 and Figure 5). The FBS function-behaviour-structure approach ensures the transformation of functional representations F to a design description or structure S through the behaviours B.

Figure 3: Entities’ class hierarchy [33]

Figure 4: Operations’ class hierarchy [33]

The research leading to these results has received funding from Horizon 2020, the European Un- Page 14 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 5: Ressources’ class hierarchy [33]

Other upper ontologies or foundation ontologies [24] [39] are for instance SUMO, which has been developed by the IEEE Standard Upper Ontology Working Group, or DOLCE, which is more complex than SUMO and contains categories of natural language and of human common sense.

A similar approach is the OAM (open assembly model) which consist at the top level of product ontology, production system ontology, and process ontology [19]. Similar are the proposals of Zhang and Agyapong-Kodua [45] or Cutting-Decelle et al. [27].

A recent approach is the development of capability ontologies [20]. The idea behind is that a capability model characterizes a manufacturing facility and its constituting elements including devices, machine, cells, operators, and processes with respect to the range of applicability, speed, cost, quality, and associated constraints and uncertainties. This is then used for supply chain configuration [20] and supplier selection [21].

The different dimensions of manufacturing capability include:

• Technological capabilities such as the resolution, accuracy, feed, speed, power, and automation level of the manufacturing equipment. • Operational capabilities such as production capacity, throughput time, cost per unit, etc. • Geometric capabilities such as shape producible, dimensions, wall thickness, work envelope, etc. • Quality capabilities such as defect rate, surface finish, and tolerances. • Relational capabilities that refer to interfaces with other systems and processes both hardware and software. • Stochastic capabilities such as reliability, variations, etc.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 15 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Loskyll and Schlick [34] pursue a similar concept. They use an ontology-based approach for the abstract description of field device functionalities. This is then used for the dynamic orchestration of automated processes based on service-oriented architectures. Sections of the function and device classes ontology are shown in the following figures (Figure 6, Figure 7).

Figure 6: Section of the function ontology [34]

The research leading to these results has received funding from Horizon 2020, the European Un- Page 16 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 7: Section of the device classes ontology [34]

3.2 Standards Related to User/Worker Modelling

The structure of the data models for personnel and capability information will be presented in section 5.2. The development of these models needs to be a machine readable standardized data format. One candidate for this is ISA-95 (IEC 62264), which is an international standard for enterprise-control system integration. The data model approach presented in the ISA-95 could be useful for personnel information as well as for capability information of resources (Figure 8, Figure 9).

The research leading to these results has received funding from Horizon 2020, the European Un- Page 17 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 8: ISA-95 information on personnel

Figure 9: ISA-95 information on capability of resources

The data can be used for scheduling of the tasks for capable operators.

ISA-95 is based on the hierarchical structuring of ISA-88 of the physical assets of the industrial company. ISA-88 is a standard addressing batch process control, and ISA-95 was developed to be applied in all industries, and in all sorts of processes, like batch processes, continuous and repetitive processes. More information on relationship of ISA-88 and ISA-95 are available at the “Technical: ISA-88 and ISA- 95” article [109].

The research leading to these results has received funding from Horizon 2020, the European Un- Page 18 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Adaptation and Industry 4.0 standardisation

This section is related to Physical modelling issues.

There are a lot of initiatives on standardisation, with a lot of overlaps. There are also many initiatives ongoing that establish overviews of standards in the field of the digitalisation of manufacturing (see for example German Standardization Roadmap on Industry 4.0 [100].)

The aim of the roadmap is to present the current state of standardization relevant to Industry 4.0. The roadmap not only describes the current status and gives an overview of all relevant standards and specifications, but also gives concrete recommendations for action and outlines standardization needs in the various areas of Industry 4.0. Being a "living" document, the Roadmap will be published and regularly revised by DIN and DKE to reflect the current situation as needed.

The roadmap has following recommendation on human related issues, Human beings in Industry 4.0:

x Further develop standards and specifications for people-friendly work design in Industry 4.0. x Technology design – Adaptive design of work systems in Industry 4.0. x Concepts for a functional division of work between human beings and machines. x Design of the interaction between human beings and technical systems.

There is no standard list presented for work adaptation, just recommendation to use existing standards, for instance on machine safety or ergonomics and identify need for new standardisation.

ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies. The list below includes references to some standards on ergonomics that appear to be relevant, without being exhaustive [101].

ISO/TC 159 - Ergonomics subcommittees:

x ISO/TC 159/SC 1 General ergonomics principles. x ISO/TC 159/SC 3 Anthropometry and biomechanics. x ISO/TC 159/SC 4 Ergonomics of human-system interaction. x ISO/TC 159/SC 5 Ergonomics of the physical environment.

Few examples on workplace design issues on ISO/TC 159/SC 3 13.180 [102]:

x ISO 6385:2016 Ergonomics principles in the design of work systems. x ISO 11226:2006 Ergonomics -- Evaluation of static working postures. x ISO 11228-1:2003 Ergonomics -- Manual handling -- Part 1: Lifting and carrying. x ISO 11228-2:2007 Ergonomics -- Manual handling -- Part 2: Pushing and pulling. x ISO 11228-3:2007 Ergonomics -- Manual handling -- Part 3: Handling of low loads at high frequency. x ISO/TR 12295:2014 Ergonomics -- Application document for International Standards on manual handling (ISO 11228-1, ISO 11228-2 and ISO 11228-3) and evaluation of static working postures (ISO 11226).

The research leading to these results has received funding from Horizon 2020, the European Un- Page 19 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

x ISO 14738:2005 Safety of machinery -- Anthropometric requirements for the design of work- stations at machinery. x ISO/TS 20646:2014 Ergonomics guidelines for the optimization of musculoskeletal workload.

Those standards set limits to human material handling capabilities, working posture, lifting, carrying, pushing and pulling. Same standardisation group have set evaluation methods for human mental load as well. There are other standards on protective clothing and equipment e.g. ISO/TC 94 Personal safety -- Protective clothing and equipment.

Workplace and task design guidelines

EHS, Environment, Health and Safety, laws and directives define the use of protective equipment, work ergonomic and so on. There are plenty of useful guides available from various organisations e.g.

x Ergonomics and human factors at work [103]. x Ergonomic guidelines for Manual material handling [104].

3.3 User Interface Description Languages (UIDLs) 3.3.1 UIDLs Overview A user interface description language (UIDL) consists of a specification language that describes various aspects of a user interface under development. In this section we present an overview of UIDLs that have been considered for different reasons: they are available for testing; they have been used in some development cases; they are widely used in general.

UIDLs are very relevant to User Modelling as they provide a formal way of describing various aspects of User Interfaces and useful means for adaptation of User Interfaces according to user needs/preferences.

Dialog and Interface Specification Language (DISL) Dialog and Interface Specification Language (DISL) [1] is a user interface markup language (UIML) sub- set that extends the language in order to enable generic and modality independent dialog descriptions. Modifications to UIML mainly concern the description of generic widgets and improvements to the behavioural aspects. Generic widgets are introduced in order to separate the presentation from the structure and behaviour, i.e., mainly to separate user- and device-specific properties and modalities from a modality independent presentation. The use of generic widget attribute enables assigning each widget to a particular type of functionality it ensures (e.g., command, variable field, text field, etc.). Further, a DISL rendering engine can use this information to create interface components appro- priated to the interaction modality (i.e., graphical, vocal) in which the widget will operate. The global DISL structure consists of an optional head element for Meta information and a collection of templates and interfaces from which one interface is considered to be active at one time. Interfaces are used to describe the dialog structure, style, and behaviour, whereas templates only describe structure and style in order to be reusable by other dialog components.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 20 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

3.3.2 Generalized Interface Markup Language (GIML) The Generalized Interface Markup Language (GIML) is used for the generalized Interface Toolkit (GITK) [2]. GIML is used in this context as an interface descriptor. Following the OMG principles of separation of concerns GIML splits functionality and presentation. While the functionality is preserved in GIML the UI is derived from XSL files, which come from user and system profiles. This information is merged with the functional descriptions by using XSLT to form a final interface description. The profile data could come directly from a file system or from a remote profile server.

3.3.3 Interface Specification Meta-Language (ISML) Interface Specification Meta-Language (ISML) [3] was developed with the purpose that metaphors (shared concepts between the user and the computer) are made explicit in design. ISML decouples that metaphor model from any particular implementation, and express mappings between the concepts shared between the user and the system. It provides a framework that supports mappings between both user-oriented models (such a task descriptions) and software architecture concerns (in- teractor definitions). The ISML framework composites these concepts within five layers (devices, components, meta-objects, metaphor, interactors), using a variety of mappings to link them together.

3.3.4 Renderer-Independent Markup Language (RIML) Renderer-Independent Markup Language (RIML) [4] is a markup language based on W3C standards that allows document authoring in a device independent fashion. RIML is based on standards such as XHMTL 2.0 and XFORMS. Special row and column structures are used in RIML to specify content adaptation. Their semantics is enhanced to cover pagination and layout directives in case pagination needs to be done. Due to the use of XForms, RIML is device independent and can be mapped into a XHTML specification according to the target device. RIML semantics is enhanced to cover pagination and layout directives in case pagination needs to be done, in this sense it was possible to specify how to display a sequence of elements of the UI.

3.3.5 Software Engineering for Embedded Systems using a Component-Oriented Approach (See- scoaXML) Software Engineering for Embedded Systems using a Component-Oriented Approach (SeescoaXML) [5] consists of a suite of models and a mechanism to automatically produce different final UIs at runtime for different computing platforms, possibly equipped with different input/output devices offering various modalities (e.g. a joystick). This system is context-sensitive as it is expressed first in a modality- independent way, and then connected to a specialization for each specific platform. The context sen- sitivity of the UI is here focusing on computing platforms variations. An abstract UI is maintained that contains specifications for the different rendering mechanisms (presentation aspects) and their related behaviour (dialog aspects). These specifications are written in an XMLcompliant UIDL that is then transformed into platform specific specifications using XSLT transformations. These specifications are then connected to a high-level description of input/output devices. The entry point of this forward engineering approach is therefore located at the level of Abstract UIs.

3.3.6 Simple Unified Natural Markup Language (SunML)

The research leading to these results has received funding from Horizon 2020, the European Un- Page 21 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Simple Unified Natural Markup Language (SunML) [6] is an XML language to specify concrete user interfaces that can be mapped to different devices (PC, PDA, voice). The innovation of this language is the capacity to specify components dynamically. In SunML it is also possible to encapsulate the style and the content of each widget independent of the others. Two different files are used for that purpose. Another interesting feature offered in SunML is widget composition. Some operators have been defined for that purpose: union (semantically common widgets), intersection, subtraction, substitu- tion, inclusion. Widgets Merging Language (WML) is the extension used for that purpose. SunML presents a reduced set of elements that seems to be not enough, but the composition of widgets is used to specify more complex widgets.

3.3.7 TeresaXML TeresaXML [7] is a UIDL for producing multiple final UIs for multiple computing platforms at design time. They suggest starting with the task model of the system, then identifying the abstract UI specifications in terms of its static structure (the presentation model) and dynamic behaviour (the dialog model): such abstract specifications are exploited to drive the implementation. This time, the transla- tion from one context of use to another is operated at the highest level: task and concepts. This allows maximal flexibility, to later support multiple variations of the task depending on constraints imposed by the context of use. Here again, the context of use is limited to computing platforms only. The whole process is defined for design time and not for runtime. For instance, there is no embarked model that will be used during the execution of the interactive system, contrarily to the SEESCOA approach [5]. At the AUI level, the tool provides designers with some assistance in refining the specifications for the different computing platforms considered. The AUI is described in terms of interactors that are in turn transformed into Concrete Interaction Objects (CIOs) once a specific target has been selected.

3.3.8 MariaXML MariaXML [17] is the successor of TeresaXML and is a new UI specification language that t can be used for the abstract and concrete user interface definition. MariaXML takes into account new technical requirements in order to support dynamic behaviours, events, rich internet applications and multi- target user interfaces, in particular those based on web services. In this way, it is possible to have a UI specified in MariaXML attached to a web service.

3.3.9 User Interface Markup Language (UIML) User Interface Markup Language (UIML) [8] is an XML-compliant language that provides a device-independent method to describe a UI, and a modality-independent method to specify a UI. UIML allows describing the appearance, the interaction and the connection of the UI with the application logic. The following concepts underlie UIML: 1. UIML is a meta-language: UIML defines a small set of tags (e.g., used to describe a part of a UI) that are modality-independent, target platform-independent (e.g., PC, phone) and target language- independent (e.g., Java, VoiceXML). The specification of a UI is done through a toolkit vocabulary that specifies a set of classes of parts and properties of the classes. Different groups of people can define different vocabularies: one group might define a vocabulary whose classes have a 1-to-1 cor- respondence to UI widgets in a particular language (e.g., Java Swing API), whereas another group might define a vocabulary whose classes match abstractions used by a UI designer.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 22 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

2. UIML separates the concerns of an interactive application in such a way that moving one program from one platform to another might require only small or no changes at all. Because UIML is based on XML, it is easy to write transformations that take the language from one abstract representation to a more concrete representation. 3. UIML separates the elements of a UI and identifies: (a) which parts are composing the UI and the presentation style, (b) the content of each part (e.g., text, sounds, images) and binding of content to external resources, (c) the behaviour of parts expressed as a set of rules with conditions and actions and (d) the definition of the vocabulary of part classes. 4. UIML groups logically the UI in a tree of UI parts that changes over the lifetime of the interface. During the lifetime of a UI the initial tree of parts may dynamically change shape by adding or de- leting parts. UIML provides elements to describe the initial tree structure and to dynamically modify the structure. UIML also allows UI parts and part-trees to be packaged in templates: these templates may then be reused in various interface designs.

3.3.10 USer Interface eXtensible Markup Language (UsiXML) USer Interface eXtensible Markup Language (UsiXML) [9] is structured according to different levels of abstraction defined by the Cameleon reference framework [58]. The framework represents a reference for classifying UIs supporting a target platform and a context of use, and enables to structure the development life cycle into four levels of abstraction: task and concepts, abstract UI (AUI), concrete UI (CUI) and final UI (FUI). Thus, the Task and Concepts level is computational-independent, the AUI level is modality-independent (In the cockpit it can be several physical, Vocal, GUI, Tactile) and the CUI level is toolkit independent. UsiXML relies on a transformational approach that progressively moves among levels to the FUI. The transformational methodology of UsiXML allows the modification of the development sub-steps, thus ensuring various alternatives for the existing sub-steps to be explored and/or expanded with new sub-steps. UsiXML has a unique underlying abstract formalism represented under the form of a graph-based syntax.

3.3.11 Web Service eXperience Language (WSXL) Web Service eXperience Language (WSXL) [10] [11] is designed to represent data, presentation and control. WSXL relies on XML based standards such as XPath, XML Events, DOM, XForms and XLink as well as Web Services standards such as SOAP, WSDL and WSFL. Also, WSXL includes an extensible Adaptation Description Language where explicit locations of adaptation points, the permissible operations on adaptation points (e.g. insert, delete, modify), and the constraints on the contents of adaptation (e.g. via an XML Schema) can be specified. The Adaptation Description Language can be used during a post-processing step where the output of a WSXL component can be adapted independently without invoking the component. Furthermore, a WSXL collection provides an execution and management environment for WSXL components. It calls the lifecycle operations on WSXL components it in- stantiates, and implements a set of interfaces and a processing model for use by WSXL components and objects external to the collection. An object implementing the WSXL collection interface needs not be a WSXL component. The developer can create new and more abstract UI components. Finally, WSXL can be used for interactive Web Services applications, involving human participants.

3.3.12 eXtensible user-Interface Markup Language (XICL)

The research leading to these results has received funding from Horizon 2020, the European Un- Page 23 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

The eXtensible user-Interface Markup Language (XICL) [12] is an easy way to develop User Interface Components to Browser-based software. New UI components are created from HTML components and other XICL components. The XICL description is translated into DHTML code. An XICL documents is composed by a UI description composed by HTML or XICL elements and several components (Struc- ture, Properties, Events and Methods). XICL is a language for the UI development whose structure and behaviour is an abstract level. It also promotes reuse and extensibility of user interface components.

3.3.13 eXtensible Interface Markup Language (XIML) The eXtensible Interface Markup Language (XIML) [13] [14], is a language developed by Redwhale Software, derived from XML. It enables a framework for the definition and interrelation of interaction data items [15]. XIML can provide a standard mechanism for applications and tools to interchange interaction data and to interoperate within integrated user-interface engineering processes, from design, to operation, to evaluation. The XIML language is mainly composed of four types of components: models, elements, attributes, and relations between the elements. The presentation model is composed of several embedded elements, which correspond to the widgets of the UI, and attributes of these elements representing their characteristics (colour, size…). The relations at the presentation level are mainly the links between labels and the widgets that these labels describe. XIML supports design, operation, organization, and evaluation functions; it is able to relate the abstract and concrete data elements of an interface; and it enables knowledge-based systems to exploit the captured data. XIML is also able to store the models developed in MIMIC [16]. MIMIC is a meta-language that structures and organizes interface models. It divides the interface into model components: user-task, presentation, domain, dialog, user, and design models. The design model contains all the mappings between elements belonging to the other models. The XIML is thus the updated XML version of this previous language.

3.3.2 UIDLs Comparison A comparative review of some selected user interface description languages is produced in order to understand their scopes and their differences. Table 2 compares the properties of the different UIDLs according to the following criteria: x Models: This criterion gives the aspects of the UI that can be specified in the description of the UIs. o The task model is a description of the task to be accomplished by the worker, o the domain model is a description of the objects the worker manipulates, accesses or visualizes through the UIs, o the presentation model contains the static representation of the UI and o the dialog model holds the conversational aspect of the UI. x Tools: Some of the languages are supported by authoring tools and rendering engines. x Supported languages: Specifies the programming languages to which the XML-based UIDLs can be translated. x Supported platforms: Specifies the computing platform on which the language can be ren- dered by execution, interpretation or both.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 24 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Table 2: Properties Comparison of UIDLs [19].

UIDL Models Tools Supported lan- Supported plat- guages forms DISL Presentation, di- Rendering engine VoiceXML, Java Mobile and lim- alog and control MIDP, Java Swing, ited devices Visual C++, GIML Presentation, di- GITK (Generalized C++, Java, Perl Not specified alog and domain Interface Toolkit) ISML Presentation, Under construc- Java, Microsoft Desktop PC, 3D task, dialog and tion foundation class, screen domain Java swing classes RIML There is no infor- There is no infor- XHTML, XFORMS, Smart phone, mation mation XEvents, WML pda, Mobile, Desktop Pc SeescoaXML Task, presenta- CCOM (BetaVer- Java AWT, Swing, Mobile, desktop tion, dialog sion 1.0 2002) Pa- HTML, java.mi- PC, Palm III coSuite MSC Edi- croedition, app- tor let, VoxML, WML Juggler SunML Presentation, di- SunML Compiler Java Swing, Desktop Pc alog, domain voiceXML, HTML, UIML TeressaXML Presentation, CTTE Tool for task Markup: Digital DigitalTV, Mobile, task, dialog Models Teresa TV, VoiceXML, Desktop PC XHTML/SVG, X+V, Programming: C# UIML Presentation, di- UIML.net, HTML, Java, C++, desktop PC, alog, domain VoiceXML rende- VoiceXML, QT, handheld device, rer, WML rende- CORBA, and WML tv, mobile rer, VB2UMIL WSXL Presentation, di- Not specified HTML PC, Mobile phone alog, domain XICL Presentation, di- XICL STUDIO HTML, ECMAS- Desktop PC alog cript, CSS e DOM XIML Presentation, XIML Schema HTML, java swing, Mobile, desktop task, dialog, do- WLM PC, PDA main UsiXML SketchiXML, HTML, XHTML, Mobile, Pocket GraphiXML, VoiceXML, PC, interactive ki- FlowiXML, Java3D, VRML, osk, wall screen, FlasiXML, X3D, XAML, Java, pda, QtkiXML, Inter- Flash, QTk, WML, piXML XHTML, X+V, C++,

The research leading to these results has received funding from Horizon 2020, the European Un- Page 25 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

3.4 Physical Modelling A model is usually seen as a representation of selected aspects of the reality resp. a particular domain of interest to the modeller [113]. A model has the following characteristics [114]:

x Simplification, abstraction: i.e. the model does not cover all aspects of reality but only those of special interest. x Representation of one or more concepts in the real world; therefore, it can be seen as a kind of “image” of reality. x Purposefulness, pragmatism: models replace reality for certain subjects, for a certain time intervals or for certain operations / functions.

A high level classification of models is based on the United States Department of Defense Modeling and Simulation (M&S) Glossary [115] where models are defined as a physical, mathematical, or logical representation of a system. Furthermore, models can be divided into formal vs. informal, descriptive vs. analytical and/or static vs. dynamic (executable). [113]

Following this argumentation, a physical model is a concrete representation that is distinguished from mathematical and logical models; whereas, the latter are more abstract [113]. This is also the subject of this section: physical modelling considers the elements, structures, processes, functionality of a system (product, production system, etc.) in a way that relevant characteristics are kept in the model so that the accessibility and comprehensibility is quite high. This distinguishes physical modelling from more abstract forms, where systems are represented by a system of equations or by logical state- ments.

Modelling and simulation (Figure 15) play an important role when designing and optimizing manufacturing systems. Models represent all the relevant objects and their hierarchy and interrelation in a manufacturing system and support a common language between participants in the planning process, the analysis and evaluation as well as the development of concepts for system design. Dynamic models can be used for investigating the behaviour of a system over time using stochastic distributions. The basic relationship between modelling and simulation is shown in Figure 10.

Figure 10: Relationship between planning modelling and simulating in manufacturing systems [49].

The research leading to these results has received funding from Horizon 2020, the European Un- Page 26 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

In a manufacturing system’s model, there are different object classes. A typical distinction is between organization, physical objects and information / control (Figure 11).

Figure 11: Framework for object models in manufacturing systems [49].

Physical objects can be further divided into parts/ products, equipment and human resources. Equip- ment usually consists of machining, assembly, storage and transportation means. Information/ control objects cover processes, plans, orders, etc.

Information objects (mainly processes), resources (human and hardware) and parts/ products are in- terrelated: orders define product demand, processes define which resources are needed when to ma- nipulate parts resp. semi-finished products and plans define when and where the processes are exe- cuted. For products, resources and processes sometimes sub-models are built. This is shown in Figure 12.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 27 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 12: Data flow scheme in the virtual factory [47].

In product sub-models the hierarchical product structure is modelled. In process models there is a specific process for each item in the product structure (final product, subassembly, parts) which consists of operations, which in turn may also be hierarchically organized [51]. Such hierarchies and inter- relations are also used for ontologies [54].

Furthermore the described sub-models or classes shall be used for standardizing modelling approaches (reference model) in the context of cyber-physical systems engineering (Figure 13).

Figure 13: Schematic class overview of a possible reference model [52].

For the hierarchy especially of resources there are already standards available (Table 3):

The research leading to these results has received funding from Horizon 2020, the European Un- Page 28 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Table 3: Reference activity model [50]

Control level ISA 88 Control level FDI Interpretation A manufacturing enterprise may consist of one or more Enterprise Enterprise business units Business unit, de- Site A business unit may consist of one or more factories partment Area Factory, plant A factory must consist of one or more manufacturing lines A manufacturing line must consist of one or more manu- Process cell Manufacturing line facturing processes Manufacturing pro- A manufacturing process must consist of one or more Unit cess pieces of equipment Equipment module Equipment, labour An equipment must consist of one or more components Equipment compo- An equipment component is the lowest element in this Control module nent functional breakdown

Modelling is done with the help of software tools. A possible classification is given through Table 4. Tools are mainly focused on a certain task in the product or equipment lifecycle. All tools dealing mainly with the product itself are subsumed under the term product lifecycle management.

Table 4: Classification of modelling tools (similar to VDI 4499 [56])

Production planning and factory planning Product development Digital Factory Virtual Reality and digital mock up - - u m C i g P s m n N i i A

Groupware, project manage- s n w d C s o n n , l c n

ment, knowledge management a a f i Q l l o s i A p m a c t i i C t o n a r t l , u n o n e o i u o t o o t M D b i M y a g a t m o l A A r E a i a L M u E l R s C C F

In practice there are numerous applications for the aforementioned classes. A rough (and not complete) overview is given though Table 5.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 29 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Table 5: Overview of typical software applications for modelling in (industrial) engineering

Material flow simu- Ergonomics simu- Robotics and NC-Sim- PLM Layout planning lation lation ulation Plant Simulation Ema (3D) (dynam- AutoMod (dy- Arena PLM by Microstation (3D) (3D) (dynamic) by ich) by imk auto- namic)(2D&3D) by Arena Solutions (static) SimPlan Siemens PLM motive visTABLE®touch Simul8 (2D) (dy- DELMIA Ergonom- Demo3D (3D) (dy- Autodesk by (2D) (static) namic) by SimPlan ics (3D) (dynamic) namic) by SimPlan Vault CLASS Warehouse Anylogic (dynamic) RAMSIS (3D) (dy- Caatia & DELMIA Ro- Teamcenter by Layout and Simula- (2D&3D) namic) by Human botics by Dassault Sys- Siemens tèmes (3D) (static) tion (dynamic) Solutions (2D&3D) taraVRbuilder (3D) Jack and Process Robcad (3D) (dynamic) Agile by Oracle Layout3D (3D) (dy- (dynamic) by tara- Simulate Human by Siemens PLM namic) by Simplan kos (3D) (dynamic) by Siemens PLM Visual Components Sim3D (3D) (dy- Visual Components SAP PLM by SAP tools namic) by Simplan tools Enterprise Dynamics Windchill by (3D)(dynamic) by PTC Simulation Solution ENOVIA by Das- sault Systèmes

The models themselves might be classified by the following characteristics: x Visualization, among others dimensions (2D, 3D) and degree of immersion. x Resolution level. x Behaviour (static/ dynamic). x Purpose and target audience. A possible taxonomy of visualization is given through Figure 14.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 30 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Criteria Specification Structure Primary function: Intention Identification Localization Correlation Association Comparison and pat- Grouping Classification terns Qualitative Type of information to be Qualitative Quantitative and quanti- represented tative Information relationship Independent Relational Circular Hierarchical Network Level of measure- Nominal Ordinal dis- Ordinal Interval dis- Interval Ratio Ratio contin- n None o i ment discrete crete continuous crete continuous discrete uous t g a n i m d r o Dimension of de- o c

f None 1D 2D 3D nD n n

I e pendent variables Figure 14: Taxonomy of visualization techniques from the information-related point of view [57]

The resolution level describes how detailed the model represents the real system (Table 6):

Table 6: Resolution levels [48]

ISA-95/ ISA-88 Key en- Time-scales Simulation Level Modelled phenomena Objects/ term tity modelled paradigm Production of batches of parts Part Minutes to Cell Work centre DES through a multi-step process plan batch hours Production of parts at a single ma- Seconds to Machine Work unit chine including batch and part setup Part DES / ABS minutes and ejection Physics of the transformation pro- Continuous Actual produc- Part fea- Milliseconds Process cess such as turning and milling for based on tion process tures to seconds individual parts equations

The behaviour of models characterises if variables in the models (and the model itself) can be changed based on a timeframe.

Figure 15: Classification of simulation models [55]

The research leading to these results has received funding from Horizon 2020, the European Un- Page 31 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Models can be used for different purposes. An overview and also the relation to different target groups are provided in Table 7:

Table 7: Relations between application purpose and targeted group [55] t e - - n s t c t n e c e n g n i n e n o d t r l i e t n o a k o m i s i t a e i g r n i n t s e t m c u a n g e i c t y a e g m i a b t t r c l o n u i a h i i n l o m m l r l a n u t h c d t i n i n t r s b p r d d c c o a s a a m i e l a u r u u e n n i a S d P p M T p F m P a D a C P System analysis z z z Modelling z Validation z z Experimentation z z Representation and z z z z z z z z interpretation of results Training z z Operation parallel z z Sales promotion and z z z z public relation Interdisciplinary com- z z z z z z z z munication z full applicable partly applicable

When modelling human resources, modelling approaches can be divided into task based modelling and inverse dynamic modelling (Table 8).

Table 8: Characteristics of modelling methods for humans [53]

Task based modelling Inverse dynamic modelling

Inputs Anthropometry information Anthropometry information Target location/ orientation Maker’s trajectories Human location/ orientation Ground reaction forces Outputs Joint trajectories Joint trajectories Joint absolute angels Joint absolute angels Joint relative angels Joint relative angels Joint forces, moments and powers Strength Easy to generate human postures and High accuracy in simulation motions Applicability for complex motions Weakness Low accuracy in simulation Experimental setup in the real world Applicability for simple reaching motions

The research leading to these results has received funding from Horizon 2020, the European Un- Page 32 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

3.5 User Behaviour Simulation User behaviour describes a users’ interaction with a system. A user’s behaviour consists of a set of actions undertaken by a user interacting with the system in order to reach a goal or complete a task (IGI Global Disseminator of Knowledge).

The understanding of human behaviour plays an important role in simulations, games, and human- computer interaction systems [74]. The objective is to offer better interaction and assistance. The ultimate objective of user modelling is the benefit of users. [65]

A user’s behaviour is the key to the performance of a whole (socio-technical) system. Hereby, performance does not only mean (technical or monetary) measures of effectiveness or efficiency, but also user satisfaction and health. [75]

One of the probably earliest user models was introduced by Wilson in the 1980s resp. 1990s. This model aims at explaining the information seeking behaviour of users (Figure 16).

Figure 16: Information behaviour model [73]

Another famous model was developed to explain the acceptance of an IT system – the Technology Acceptance Model (TAM). This model will be further analysed in Factory2Fit deliverable D1.5 related with “Design and evaluation framework and measuring tools” [108]. The original model (Figure 17) was introduced by Davis [63] and has undergone numerous extensions so far.

Figure 17: Original TAM [63]

Most of the early (and also modern) models are based on the Theory of planned behaviour and/or the Theory of reasoned action [59], which postulate that an actual (and observable) behaviour is only

The research leading to these results has received funding from Horizon 2020, the European Un- Page 33 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model shown if an intention was formerly developed. This intention is influenced by subjective norms, attitudes and perceived control. The process of forming the intention and also the role of norms, attitudes and perceived control is moderated by external variables like demographics, the personal value system, personality traits, etc. [69]

According to Yordanova [74] three main purposes for human behaviour modelling can be distinguished:

x Simulation, which is the process of imitating a real world situation or behaviour. Simulation does not take into account how probable the execution sequence is, it just gives a sample of a future trajectory. x Prediction, which defines a probability distribution across an action space. There is not only a sample of a future trajectory, but also the probability of this action sequence happening. x Inference, which not only tries to predict future behaviour, but also to find the reasons behind this behaviour.

There are numerous human behaviour models. In an extensive analysis, the following structures resp. basic approaches could be identified [74]:

x Process-based models

Process-based models describe activities through a set of actions with transitions that lead from one action to another in order to achieve a goal. Process-based models represent human behaviour as a composite structure of activities with transitions defining the next activity. Thus, when describing how to reach a goal, the process-based model will go through several activities until reaching the desired state.

Process-based models can either be grammar-based models, where the human behaviour is described in the form of grammar and rules, or process calculi, which represent a diverse family of related approaches for modelling of concurrent systems.

x Causal (state-based) models

Causal models do not specify a set of actions with which a goal can be achieved, but rather define the precondition for reaching it, the effects after the goal has been reached and a set of states through which one should go in order to reach the goal state.

For capturing causality, different statistical models have been used. The main models are linear models, TFIDF-based models, Markov models, neural networks, classification and rule-induction methods, and Bayesian networks [70] [76]. Recent extensions have been for instance the integration of emotions [62] or the use of user behaviour ontologies [68].

The simulation itself, especially following the process-based approach, is done for instance based on activity based approach, coloured Petri nets [71] or using multi-agent systems [64].

The research leading to these results has received funding from Horizon 2020, the European Un- Page 34 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Because the characteristics of a user (user profile) are not necessarily constant but change over time or are based on the interaction with the system or with other internal or external variables, an evolving systems approach is proposed [67]. Hereby, a distinction should be made if the system is adaptive or adaptable (Table 9).

Table 9: Comparison between adaptive and adaptable systems [65]

Adaptive Adaptable

Definition Dynamic adaptation by the system itself User changes (with substantial system sup- to current task and current user port) the functionality of the system

Knowledge Contained in the system, projected in dif- Knowledge is extended ferent ways

Strengths Little (or no) effort by the user, no special User is in control, user knows her/his task knowledge of the user is required best, system knowledge will fit better, success model exists

Weaknesses User has difficulties developing a coher- Systems become incompatible, user must do ent model of the system, loss of control, substantial work, increased complexity, user few (if any) success models exist needs to learn the adaptation component

Required mecha- Models of users, task dialogues, Layered architecture, domain models, nisms knowledge base of goals and plans, “back-talk” from the system, design ra- matching capabilities, incremental up- tionale date of models

Yordana [74] identified five groups of requirements for a human behaviour modelling:

x Requirements for procedural modelling (composition, hierarchy, sequences, loops, interleav- ing activities, choice, constraints, enabling, disabling, priority, independence, suspend, re- sume). x Requirements for parallel execution modelling (parallelism, synchronisation). x Requirements for probabilistic modelling (observation models, probabilities for action sequences, probable durations of activities). x Requirements for causal modelling (preconditions, effects, relation to prior knowledge). x Requirements for modelling purpose (simulation, prediction, causal inference, state estimation, parameter estimation, detecting errors, unknown actions detection).

Those should be taken into account when modelling and simulating user behaviours.

According to Tick [72] the modelling of user’s behaviour (here in the context of HCI design) includes two fundamentally different steps:

The research leading to these results has received funding from Horizon 2020, the European Un- Page 35 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

1. Definition of the initial model: this model is set up in the course of the implementation phase, before the design of the user interface, and it provides a base for the user interface design. 2. Definition of the adaptivity of the user interface and its integration into the system; this system possessing this adaptive feature continuously collects information about the user’s behaviour during its usage, it evaluates these information and finally on the basis of the results, it adapts to the user’s changed behavioural habits by changing the user interface.

One interesting approach with respect to a modelling strategy and a comprehensive usage of user models was proposed by Hlavacs and Kotsis [66]. They present a user behaviour model framework that is constructed in a top down manner, consisting of various layers. Those layers offer services to the next higher layer and require services from the layers below.

As user behaviour modelling has been mainly a discipline of information technology, HCI design etc. it has gained increased attention on other areas of manufacturing systems design.

In Industrial Engineering, two types of human performance models can be identified [60]:

x Models of high-level factors dealing with complex interactions of psychological mechanisms. For example job satisfaction is known to impact work performance, but it is mediated by psychological and environmental factors. Such models are inherently complex, context specific and dependent on individual differences between people. x Models of low-level factors represent basal physiological mechanisms. For example models of dehydration provide estimates of performance changes, mediated by environmental conditions. Such models are relatively simple and can be applied to any individual.

Baines et al. [60] propose a framework for human performance modelling which is shown in Figure 18.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 36 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 18: Framework for human performance modelling [60]

Research in this field aims to improve the accuracy and reliability of simulation prediction by including models of human factors [60].

The research leading to these results has received funding from Horizon 2020, the European Un- Page 37 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

4 Theories of work satisfaction and performance

The former chapter analysed the standards, methodologies, as well as user behaviour simulation frameworks that will be used for the development of worker and dynamic worker models for the Fac- tory2Fit project. This chapter introduces theoretical background for the Worker Model parameters from the job satisfaction and work performance perspectives that will complete the needed elements for the development of more compound models. 4.1 Definition of job satisfaction The two most common definitions for job satisfaction describe it as: “a pleasurable or positive emotional state resulting from the appraisal of one's job or job experiences” [90] and “the extent to which people like (satisfaction) or dislike (dissatisfaction) their jobs” [95]. Job satisfaction can be considered as an overall attitude towards work in general, or in relation to different aspects of work, such as, colleagues, salary, the nature of work itself, supervision, or working conditions [95]. The extent to which work properties meet or exceed the personal expectations of employees determines the level of job satisfaction [90]. Satisfaction with the nature of work, including job challenge, autonomy, variety and scope, predicts best the overall job satisfaction [92]. Job satisfaction may predict absenteeism [97] and the turnover intentions of employees [79] [92]. In addition, employees with high job satisfaction seem to return faster back to work after physical (e.g. musculoskeletal and back) injuries [82] [86]. 4.2 Factors influencing job satisfaction According to the Job Characteristics Model (JCM) [84], satisfaction with the nature of work, together with other important job related outcomes, occur when the work is intrinsically motivating. The degree of motivation depends on whether 1) the work is experienced as personally meaningful, valuable and worthwhile, 2) employees perceive themselves as personally accountable and responsible for the work outcomes, and 3) employees are aware how well they perform in their work (Figure 19). Fur- thermore, the model postulates five key job characteristics, which influence these motivational factors: skill variety (the degree of utilizing different skills and talents of employees), task identity (the degree to which identifiable pieces of work can be completed from the beginning to the end), task significance (the job having a substantial impact on other people), autonomy (independence and dis- cretion to influence the schedule and ways of working) and feedback (obtaining direct and clear information about the effectiveness of one’s work performance). Increased job autonomy seems to be also directly associated with job satisfaction [78].

The research leading to these results has received funding from Horizon 2020, the European Un- Page 38 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 19: Job Characteristics Model [84]

Other key factors for job satisfaction are work stress, affective state (mood and emotions) during work, experiencing work achievements and the opportunity to participate in decision making. In addition, person-specific and social factors influence job satisfaction.

Work-related stress

Work-related stress refers to the prolonged situation where the work demands exceed employees’ perceived resources i.e. the skills and capabilities to cope with the demands [80]. The Job-Demands- Control-Support Model is one of the most common theories of stress, according to which high job strain (stress) results from high job demands coupled with low decision latitude or control in terms of one’s work [87]. Job demands refer to the workload related aspects, such as, time pressure, unclear tasks or work responsibilities, and cognitive load resulting from multi-tasking, interruptions, communication overload, other people talking at the background and sounds, movement or light flashes catching one’s attention [98]. Job control refers to the employees’ decision-making freedom regarding work properties (tasks, effort and schedule), similar to the autonomy factor in the Job Characteristics Model [84]. The degree of job control employees’ have over their work and the social support they receive in it improves the abilities to cope in demanding situations [88]. Furthermore, both “high strain job” and “passive job” (i.e. low job demands and low job decision latitude) seem to be associated with job dissatisfaction [87].

Levi & Levi [89] have proposed an exhaustive list of various work-related stressors, which includes inadequate work scheduling and break (rest) cycles during work shifts; cognitive overload; lack of feedback on one’s work performance; low job autonomy; ambiguous goals and roles; fragmented, repetitive, monotonous work with low task variety; discrimination; environmental factors such as ex- cessive noise, heat, and vibration; unsafe working environment, and physical workload such as lifting heavy loads, working in static or difficult positions, continuous sitting and applying strength through hands. [98]

The research leading to these results has received funding from Horizon 2020, the European Un- Page 39 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Affective state

Emotions are affective states that are evoked by events, thus having a clear cause. Moods, on the other hand, are states that are vaguer, lacking a direct contextual provocation, but which may follow from emotions [96]. Emotions, mood influence attitudes and behaviours, the association between work-related affective states and job satisfaction have been observed in some studies [81] [96]. The frequency of feeling positive emotions at work seems to be a better predictor for job satisfaction than the intensity of the emotion [81]. Fisher [81] suggests that employers, who wish to enhance the job attitudes of their employees, should aim for providing work environments that are free from minor irritations and hassles, evoking mild but frequent negative emotions in the employees, and implement small, reinforcing events that take place frequently. Negative emotions can also signal about the learning needs of employees [94]. Perceived success at work

Reaching one’s work-related personal goals, i.e. performance goals which are valued by the person, increases job satisfaction. Success at work leads to self and task satisfaction, pride in performance and the sense of achievement. Thus, it is important that the worker has clear goals and receives feedback regarding their performance in relation to these goals, especially highlighting personal achievements. [91][98]

Participation in work design

The opportunities for employees to be involved in work related decision making promotes job satisfaction and is closely related to job autonomy [93]. For instance, workers appreciate the ability to participate in the planning of their work and apply their ideas in work, as well as, working in an environment which encourages open discussions regarding work practises [78]. When employees are able to influence the decisions that have impact on them, they are more likely to value the outcomes of the decisions, which reinforce job satisfaction. High-level participation is associated with high level satisfaction, occurring when employees are involved in generating alternatives, planning processes and evaluating results. [93] Individual factors Job satisfaction is to some extent influenced by personality, since people with a positive outlook are prone to be satisfied with various dimensions of their life, including their job, compared to people with negative dispositions [92]. A U-shaped relationship between age and job satisfaction have been observed in some studies. Moreover, as psychological well-being (e.g. the work-family life balance) and physical health influence satisfaction in life in general, these factors influence also work related attitudes. Social factors Social support provided by supervisors and co-workers is an important predictor for job satisfaction. The behaviour of supervisors is crucial in terms of supporting employees to complete their work du-

The research leading to these results has received funding from Horizon 2020, the European Un- Page 40 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model ties, fair treatment of employees, the fluency of communication with employees, providing construc- tive feedback, allowing employees’ to feel valued, encouraging personal development, and providing recognition for achievements. The forms of social support amongst co-workers involve the feeling of social cohesion, friendship, emotional support and frequent contact with others. 4.3 Models of work performance Theoretical framework for modelling human performance

Baines et al. [77] introduced a theoretical framework, which summarizes the principal factors influencing human performance in the manufacturing production involving mostly manual tasks. In the framework, human performance is described via six metrics: the delay time occurring before an operator actually starts to work with a task, after the conditions required to begin the task are met (dependability); the time of completing the task (activity time); accuracy and reliability in conducting the task (error rate); as well as, absenteeism rate, accident rate and staff turnover rate. The last three metrics influence performance in the long-term (Figure 20).

The identified factors influencing performance metrics are based on an extensive literature review into several areas, such as, manufacturing management, behavioural sciences, health and safety, organizational studies and economics. The authors identified altogether 65 factors, which they evaluated based on four criteria: general relevance for human performance relevant to manual and repetitive activities, specific relevance to manual production work or individual human performance metrics, robustness of the factor regarding its impact and credibility, and measurability of the factor. The factors were ranked based their overall scoring on these four dimensions, the score ranging from 1 to 15. Factors scoring 9 points or higher, were incorporated to the framework. Altogether 30 factors were included, divided into the categories of individual worker, physical working environment and organizational working structure. The work ethics and skill level, range and experience of an individual worker are also part of these factors, though these are missing from the Figure 20.

In addition, sufficient sleep and recovery are vital individual worker related factors to consider, since tiredness influences negatively work performance, such as the cognitive ability to process information, memory capacity, attentiveness, learning ability, mood and creativity [98]. Though sleep was not incorporated into the framework of Baines et al. [77], sleep patterns together with lifestyle scored 8 points in their evaluations for the identified factors, which is just below their cut-off point (9 points). Finally, the mediation box between the factors and performance measures in the Figure 20 refers to a relationship model, which describes how person-specific and environmental factors together predict human performance.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 41 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 20: Theoretical framework for human performance [85]

High performance cycle

According to the high performance cycle model introduced by Locke and Latham [91], employees who face challenging, but well-defined work-related goals, accompanied by high self-efficacy (i.e. personal expectancy of success or the confidence in being able to perform well) produce high performance results, given that the employees are personally committed to the goals, they receive feedback regarding their progress towards the goals and have adequate ability to approach the goals, and the environmental constraints are low. Goals may be set for various aspects of performance, e.g. physical effort, reaction speed, quality, production efficiency, time spent on a task, profit, and cost.

High performance is achieved through four mechanisms: high goals and high self-efficacy lead individ- uals to persist longer at tasks than do easy goals or low self-efficacy; high goals and high self-efficacy lead people to exert more effort (work harder) on tasks with time limits; goals direct attention and action toward goal relevant activities at the expense of other activities; and goals stimulate people to develop task strategies to attain their goals. High performance, when rewarding, leads to job satisfaction, which in turn facilitates commitment to the organization and its goals (Figure 21).

The research leading to these results has received funding from Horizon 2020, the European Un- Page 42 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 21: High performance cycle [91]

The research leading to these results has received funding from Horizon 2020, the European Un- Page 43 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

5 Factory2Fit Worker Modelling Methodology

A Worker Model represents various worker characteristics, which can be used to define a worker and his/her abilities [18]. Worker Model includes worker’s preferences, knowledge and attributes gener- ally, or for a particular task. The development of the models that will be introduced in this chapter derives from the standards, methodologies and theories provided in chapter 3 and chapter 4. The Factory2Fit Worker Modelling methodology is being built based on these three main building blocks:

x Worker Model x Adaptive Worker Model x Worker Profile

Worker Model is a static model including general worker characteristics, though Adaptive Worker Model, which will be analyzed in Chapter 6, is a dynamic Worker Model describing the relations between worker characteristics with multiple other elements in the working area. The Factory2Fit worker modelling methodology is based on the ontology based Factory2Fit adaptation architecture. Based on these components a methodology has been synthesized during the first months of the project. This methodology is described in the following sections.

As far as worker characteristics concerns, based on the theories of work satisfaction and performance, various important factors can be identified which influence the well-being of workers and their work performance, thus many of these are relevant also for the Factory2Fit Worker model. However, only the factors anticipated as ethically feasible, acceptable by the workers, and practically possible to measure in the Factory2Fit industrial use cases are included in the Factory2Fit Worker model. For instance, performing IQ tests is not foreseen to be accepted by the workers, which is why this factor is omitted from the model. The characteristics selected for the Factory2Fit Worker model are described in detail in this chapter and in chapter 6.

5.1 Factory2Fit Worker modelling methodology outline

Factory2Fit Worker Modelling methodology (Figure 22) follows a top-down approach where initially at the top-level the Worker Model is an abstract related model referring to different general worker characteristics, described via ontologies (section 3.1) that can be potentially used-extended. In the mid-level of the hierarchy, the Adaptive Worker Model refers to the interaction between worker characteristics and any other element in the working area. The results of this level could also be potentially used by the broader community for applications that are however relevant to the Factory2Fit applications scenarios. On the final level of the hierarchy, Worker Profile refers to specific characteristics a single worker has. This profile will be used in further WPs to associate the worker with the proper tasks and machines existing in the working area.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 44 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Worker Model

Worker Model is a work-related abstract model that refers to a high level description of several worker characteristics. It is developed with respect to several characteristics, which are broken down according to the characteristics category. A Worker Model, that will be stored in ontologies, will include multiple characteristic’s related parameters like characteristic’s description, metrics, etc.

Adaptive Worker Model

Adaptive Worker Model includes worker characteristics referred in Worker Model and other working area related characteristics. These elements refer to machine, task and contextual characteristics with which the worker interacts daily. Adaptive Worker Model is a dynamic and highly adaptable model.

Worker Profile

Worker Profile refers to a single worker and the set of characteristics he/she has. Worker Profile can be comprised of one or more Adaptive Worker Models. The difference between each worker profile refers to the different and specific characteristics a single worker has.

Figure 22: Factory2Fit Worker Modelling methodology

Characteristics’ analysis has an important part in the Worker Model definition; worker characteristics can be described as low level as possible, in order to find the best matching between worker and machinery / tasks. Each characteristic consists of one or more variables describing its content. Fur- thermore, every variable has a concept name and one or more parameters. The concept name is the generic description of the variable’s content, and parameter is the quantitative description of the worker’s properties and constraints with respect to the variable’s concept name (Figure 23).

The research leading to these results has received funding from Horizon 2020, the European Un- Page 45 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 23: Characteristic analysis into its components

The advantage of using ontology for specification purposes gives a great potential to the models. On- tology describes the basic conceptual terms, semantics and relationships among them and also provides a common basis for communication and collaboration between heterogeneous artefacts. A fully semantic description of the Worker Model will allow the use of inference engines in the simulation environment.

A Worker Model and Worker Profile introduction follows in the next section. Adaptive Worker Model, machine, task and environmental characteristics and the relations between them and the worker will be analysed further in chapter 6.

5.2 Work-related Profile Worker Model is ontology based model including a wide variety of characteristics that describe the abilities of a worker in multiple dimensions (physical, sensorial, cognitive). Worker Profile derives from the Adaptive Worker Model and refers to a set of characteristics a worker has. Work-related Profile is being used to personalize each worker. In order to make a more compound and structured Worker Model, all the worker characteristics were gathered and classified in four major categories (Figure 24). Those categories are: x Wellbeing x Generic Properties x Skills x Preferences/ Aversions Each one of these categories includes multiple worker characteristics. Those characteristics are further defined by their variables, in order to better describe their content. Every characteristic has one or more variables that describe further its multiple aspects and every category. Furthermore, each variable has one or more parameters defining the variable’s value (Figure

The research leading to these results has received funding from Horizon 2020, the European Un- Page 46 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

25). Some characteristics are lacking variables’ or parameters’ definition, because Worker Model development is at a very early stage and not all variables or parameters have been defined. Those variables and parameters will be explicitly defined in the next version of the deliverable (M18). Table 10 presents the worker characteristics and their categories.

Figure 24: Worker Model characteristic's categories

Table 10: Categories of worker characteristics and characteristics’ variables

Category Characteristics Variables Physiological activation Exhaustion Emotional states Boredom Frustration Tension Step count Physical activity Sedentary periods Standing periods

Social interaction – number of in- Wellbeing stances

Length of continuous work Work rhythm Timing of work periods Number of breaks Quality of sleep Sleep Length of sleep General work satisfaction Work satisfaction Perceived workload Work day specific satisfaction

The research leading to these results has received funding from Horizon 2020, the European Un- Page 47 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Category Characteristics Variables Perceived competence – job demand balance Year of birth Demographic background Gender Personality traits (big five) Psychological Learning traits Generic properties Specific mental disabilities Height Physical Walking speed Specific physical disabilities Knowledge on relevant tasks Comprehension of relevant Work Experience processes Job tenure Experience with tools Time to finish task Errors (number, quality) Number of complete tasks Performance per time Skills Recovery time from mistakes Machine usage time Ability to cooperate Communication skills Social skills Motivation Ability to deal with conflict

Language skills

Intercultural competences to be explicitly defined in the Tasks next version of the delivera- Co-workers ble (M18) Preferences/Aversions Working hours/shift

Shift model

The research leading to these results has received funding from Horizon 2020, the European Un- Page 48 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Category Characteristics Variables Tools

Level of participation

Preferences/Aversions Level of personal feedback

Level of personal adaptation

Figure 25: Worker characteristic analyses.

All the characteristics and the variables that compound the Worker Model can be classified, based on their description, in further two worker models, the Strong characteristics Worker Model and the Soft characteristics Worker Model (Figure 26). Strong characteristics Worker Model describes the worker functionalities that provide some kind of process. On the other hand, Soft characteristics Worker Model includes the properties that do not naturally relate to any simple functional capability, but are important in relating the proper worker with the machine and task. Each one of these Worker Models includes the worker characteristics that correspond to each category, based on their content.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 49 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 26: Worker Model division in Strong characteristics Worker Model and Soft characteristics Worker Model

5.2.1 Strong characteristics of the Worker Model Strong characteristics Worker Model includes mostly physical abilities showing worker’s functionality that provides some kind of process. This category also includes characteristics showing worker’s general knowledge upon a certain operation or task. Every characteristic has one or more variables defined by one or more parameters. Table 11 lists the strong characteristics derived after total worker characteristics classification.

Table 11: Characteristics included in Strong characteristics Worker Model

Characteristics Variables Parameters Physical activity Step count Counting worker’s steps Work satisfaction Perceived competence – job demand balance

Personality traits (big five) Type of traits Psychological Learning traits Type of learning traits Specific mental disabilities Walking speed Physical Specific physical disabilities Knowledge on relevant tasks Knowledge levels / Operational knowledge Work experience Comprehension of relevant processes Job tenure Years in former position

The research leading to these results has received funding from Horizon 2020, the European Un- Page 50 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Characteristics Variables Parameters Experience with tools Years in former experience Language skills Knowledge levels Tasks to be explicitly defined in the next version of the deliverable (M18)

5.2.2 Soft characteristics of the Worker Model Soft characteristics Worker Model includes general worker characteristics, such as height, weight etc. Every characteristic has certain elements defining its parameter, such as heart rate, blood pressure, etc. The parameter’s final value is defined by the individual values of each element in parameters. Table 12 lists the soft characteristics with their parameters.

Table 12: Characteristics included in Soft characteristics Worker Model

Characteristics Variables Parameters Respiratory patterns Facial expression Physiological activation Galvanic skin conductance EEG-signals Energy Costs Blood lactate level Respiratory Rate Emotional states Exhaustion Heart rate Blood pressure subjective feeling Frustration Facial expression EEG-signals Boredom Facial expression EEG-signals Heart rate variability galvanic skin response Blood lactate level Emotional states Tension EEG-signals Respiratory pattern Body temperature Physical activity Length of sedentary periods Number

The research leading to these results has received funding from Horizon 2020, the European Un- Page 51 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Characteristics Variables Parameters Length of standing periods Number Social interaction – number of in- Number of instances stances Length of continuous work Work rhythm Timing of work periods Number of breaks Quality of sleep Sleep to be explicitly defined in the next Length of sleep version of the deliverable (M18) General work satisfaction Perceived workload Work satisfaction Work day specific satisfaction Year of birth Worker’s age Demographic background Gender Worker’s gender Physical Height Worker’s height

Time to finish task Number of complete tasks to be explicitly defined in the next per time version of the deliverable (M18) Recovery time from mistakes Performance Number Errors Quality Idle time Machine usage time Production time Ability to cooperate to be explicitly defined in the next Communication skills Social skills version of the deliverable (M18) Motivation Ability to deal with conflict Language skills Intercultural competencies Co-workers to be explicitly defined in the next version of the deliverable (M18) Working hours/shift Shift model

The research leading to these results has received funding from Horizon 2020, the European Un- Page 52 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Characteristics Variables Parameters Tools Level of participation Level of personal feedback Level of personal adaptation

The research leading to these results has received funding from Horizon 2020, the European Un- Page 53 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

6 Adaptive Worker Model

Based on Factory2Fit Modelling methodology presented in chapter 5, after the development of the static Worker Model, a dynamic model must be introduced. Adaptive Worker Model, presented in this chapter, is a dynamic and highly adaptable model which includes worker characteristics referred in Worker Model and other working area related characteristics. It is formed by examining the possible interactions between worker and other elements in the working area. A proper Adaptive Model should define these elements, examine their capabilities and parameters, in order to find the best fit among them. The needed elements which form the Adaptive Worker Model are:

x Task characteristics x Machine characteristics x Contextual characteristics

Each one of these categories contributes in a different way in the development of the Adaptive Worker Model. In this chapter there will be an introduction in these categories and their characteristics, and also a brief analysis about the relations between these characteristics with the worker, in order to define the Adaptive Worker Model. Those characteristics were delivered form the theoretical framework for modelling human performance mentioned in chapter 4. Although, some of the factors mentioned in this chapter (e.g. conscientiousness and extroversion) cannot be included in worker characteristics, because they do not meet the ethical guidelines of the project.

6.1 Definition of the Adaptive Worker Model Adaptive Worker Model is an ontology based model that dynamically adapts to the worker’s capabilities, preferences and tasks, by examining all the possible actions/interactions of a worker in a working area with different entities. The definition of the Adaptive Worker Model consists of multiple aspects (Figure 27), including the worker characteristics, the machine characteristics that are related with the worker, the analysis of worker’s tasks and the analysis of the environmental characteristics of the working area, along with matching individual profiles with the current work [110]. Besides the worker characteristics included in Worker Model (section 5.2), Adaptive Worker Model includes one more category, the current work content (Figure 28). Characteristics and variables in this category refer to task and environmental characteristics that are directly linked to the worker. All the characteristics included in Adaptive Worker Model are presented in Figure 29. Table 13 presents the additional category with those characteristics and variables.

Figure 27: Adaptive Worker Model

The research leading to these results has received funding from Horizon 2020, the European Un- Page 54 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 28: Adaptive Worker Model characteristic’s categories

Table 13: Characteristics in Current work context category

Category Characteristics Variables Current task Required body postures Task characteristics Distances/Location Mandatory qualification level Temperature (Room, Local) Air pollution Noise level Environmental characteristics Lighting/brightness Vibrations Current work context Air circulation Humidity Coordination requirements Communication require- Social context ments Dependability on others Shift schedule Organisation Current work hours Work break schedule

The research leading to these results has received funding from Horizon 2020, the European Un- Page 55 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 29: Mind map with all the worker characteristics for the Adaptive Worker Model

6.2 Task characteristics Task characteristics describe certain features of one small part of a work process and make it possible to measure and control a task. Furthermore, the requirements and results of a task can be deducted

The research leading to these results has received funding from Horizon 2020, the European Un- Page 56 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model with its characteristics and this allows an evaluation of each task as well as an estimation of the outcome. In our case, the features of a task will be described in order to define the best fit between task and worker.

These characteristics can be structured into decisions, general task features, results, dynamics, ergonomics, collaboration, materials / products and decisions (Figure 30).

Figure 30: Task characteristics

Table 14 shows a list of task characteristics and the defining variables.

Table 14: Task characteristics and the defining variables

Parameter Defining variables number of (sub)-tasks Number time share(s) of sub-tasks time per sub-task/ overall time duration of (sub-)tasks Time task variance number of tasks that require different competencies task complexity level of required education and experience type of task classification based on a catalogue number of materials handled / used Number likelihood of disturbances OEE (machine); OLE (worker)

The research leading to these results has received funding from Horizon 2020, the European Un- Page 57 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Parameter Defining variables number and extend of interventions that can be ap- controllability of disturbances plied to influence disturbances predictability of disturbances availability of data and models predictability of results availability of data and models number and extend of interventions that can be ap- controllability of results plied to the process and the environment max allowed processing time / standard processing processing time limitation time allowed response time to disturbances Time number of changes / number of tasks processed likelihood of changes (within a given period) predictability of changes availability of data and models number and extend of interventions that can be ap- controllability of changes plied to the situation/ to the work system requirements for communication classification (direction, reason) requirements for coordination Number number of relations (predecessors); number of al- dependability on others behaviour/ results ternatives; number of decoupling points; power to implement guidelines for others percentage of - monitoring tasks level of cognitive regulation - manual tasks - decisions (rule, knowledge based) criticality of own decisions quantified implication of decisions availability of necessary information for decisions personal rating tact rate / cycle time Time weights to be handled Weight performed work weight and distance pressures / forces to be applied Force number of recreation intervals Number duration of recreation intervals Time mandatory / necessary qualification level classification based on a catalogue accuracy level needed Tolerance

The research leading to these results has received funding from Horizon 2020, the European Un- Page 58 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Parameter Defining variables buffer times Time predefined reach distances Distance perceived autonomy for decisions personal rating material supply Classification b (ordered, disordered, ..) required body postures Classification based on a catalogue EAWS evaluation EAWS points, based on catalogue

6.3 Machine characteristics In order to define the characteristics of the machine and to ensure a better match to the characteristics of the worker, for the machine similar features have been selected from the worker characteristics. Table 15 shows a list of characteristics to describe the machines.

Table 15: Machine characteristics

Parameter Defining variables Length of continuous work periods Number of downtimes Robustness characteristics / Work rhythm Timing of shut downs Type of downtime Length of downtime Generic properties / other specific disabilities / damages Functionality / language Languages Errors Functionality / performance machine usage time recovery time from down time ability to deal with conflict Functionality / collaboration Communication abilities Current operating hours Number of materials handled / used Type of machines Predictability of machine / process behaviour Controllability of machine behaviour Tact rate / cycle time to be explicitly defined in the next version of the de- Performed work liverable (M18)

The research leading to these results has received funding from Horizon 2020, the European Un- Page 59 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Parameter Defining variables Number of maintenance intervals Duration of maintenance intervals Mandatory / necessary qualification level Object / process temperature Level of automation Visibility Feedback on machine's own performance Feedback on current state to be explicitly defined in the next version of the de- Location requirements liverable (M18) Noise level noise emission of machine Vibrations vibration emission of machine Type(s) of personal protective equipment (PPE) to to be explicitly defined in the next version of the de- be used liverable (M18)

6.4 Contextual characteristics The contextual characteristics include all surroundings of the worker that arise from natural forces within the factory, the machine or the organization of the work itself. The factors, that need to be considered, are properties of the environment, e.g. temperature, humidity or noise, as well as conditions determined by the machine, e.g. ergonomic conditions. Furthermore, organizational issues de- termine the context as well as the shift model. The context is an important part of the Adaptive Capa- bility Model. Table 16 shows a list of important characteristics and the measures.

Table 16: Important contextual characteristics and their measures

Characteristics Measures Room temperature Temperature Temperature at workspace Humidity Humidity level Air circulation Air circulation level Noise Noise level/ decibel Vibrations Vibration level Dust Dust level Lightning Lightning level Ergonomic conditions Operating height Protective equipment Type of personal protective equipment

The research leading to these results has received funding from Horizon 2020, the European Un- Page 60 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Characteristics Measures Shift work model Type Actual shift Type Salary/bonuses Salary/bonuses level People around Number of people around

The research leading to these results has received funding from Horizon 2020, the European Un- Page 61 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

7 Online Worker monitoring and off-line measurements definition

This Chapter consists of a brief analysis of the Smart Sensor Network hardware that will be used, alongside with the definition of the measurements that define the worker characteristics. Both chapter 5 and 6 introduce the characteristics included in Worker Model and Adaptive Worker Model, with the parameters of every variable needed to define each worker characteristic. These parameters should be measured with the proper devices in order to develop a complete Worker Model. In this chapter there will be an analysis of the devices proposed to be used to measure and gather the values of these parameters. These parameters are classified, based on the kind of measurement, in the following categories:

x Online Worker monitoring: parameters included in online worker monitoring category are based on the measurements acquired from certain devices used in Smart Sensor Network. This kind of monitoring concerns the measurements that derive directly from a certain device, e.g. blood pressure. 1. Off-line measurements: measurements included in this category derive from questionnaires and interviews, e.g. Specific physical disabilities and knowledge on relevant tasks. 7.1 Smart Sensor Network Smart Sensor Network consists of smart devices used for data acquisition and processing. These devices can be either smart bands, or smart watches. The differences between those two categories are relative. Smart watches are more complex devices that can easily substitute smartphones. In most cases, smartwatches can receive notifications, data and incoming calls. Their operating system supports many applications and also the processing of incoming and acquired data. On the other hand, the functionalities of smart bands are much more limited. Typically, these devices are designed to function either as sports managers and / or control of sleep. Their main role is to receive information from all the sensors they are equipped and redirect them to an application on a smart device for further processing. The suggested smart bands and smart watches for Factory2Fit Smart Sensor network are analyzed in this section.

1. Microsoft Band 2

Microsoft Band 2 (Figure 31) is a state-the-art smart band from Microsoft. Microsoft Band 2 is an advanced fitness tracker that connects the user with his/her smart phone, allows its remote control and in parallel informs the user for personal activity, i.e. the quality of sleep someone has, user’s heartrate etc. Smart band is cross-compatible with android, iOS and Windows smart devices are the apps and data visualisation of Microsoft’s Health app.

This smart band is equipped with ten different sensors and a microphone, for immediate data acquisition. Some of the sensors give direct read-outs, such as the constant optical heart-rate sensor, the UV sensor and GPS. Others are used in combination to detect steps and activity, track workouts and other things.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 62 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Specifications:

x Connectivity: Bluetooth 4.0 x Compatibility: Android 4.3+, iPhone 4S or later, Windows Phone 8.1 or later x Sensors: Optical heart rate, 3-axis accelerometer, gyrometer, GPS, ambient light, UV, skin temperature, capacitive sensor, galvanic skin response, Barometer. 2. Samsung Gear 3

Samsung Gear S3 (Figure 32) is a smart watch developed from Samsung. This smartwatch can substitute user’s smartphone or any other smart device by receiving calls, having social media applications and acquiring personal data and processing them in several integrated apps. The Samsung Gear’s S3 operating system is Tizen-based wearable platform 2.3.2 and it is equipped with four main sensors and a microphone.

Specifications:

x Connectivity: Bluetooth 4.2 x Sensors: Accelerometer, gyro, optical heart rate sensor, barometer, GPS, altimeter.

Figure 31: Microsoft Band 2 device Figure 32: Samsung Gear S3 device. Figure 33: FitBit devices (a) Fitbit Surge,(b) Fitbit Charge HR and (c) Fit- bit Charge2

1. Fit Bit

FitBit devices (Figure 33) are also a fitness tracker smart band designed by Consumer electronics. They perform the basic actions of a fitness smart device, though they have also additional features i.e. email and text capabilities. They are equipped with several devices that measure data such as the number of steps walked, heart rate, quality of sleep, steps climbed, and other personal metrics involved in fitness. Alongside their trackers, Fitbit devices offer a mobile app and website that can be used with or without the Fitbit Tracker, although owning one is recommended.

More information about the Smart Sensor Network hardware that will be used, with expert review results will be introduced in D1.1 “Enabling technologies” [107].

The research leading to these results has received funding from Horizon 2020, the European Un- Page 63 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

7.2 Online Worker monitoring

Online Worker monitoring includes measurements that derive from a certain device used in Smart Sensor Network, e.g. blood pressure. These measurements will be acquired in specific times, while the worker is working, in order to check the worker’s state and to define the Worker Profile.

In addition to a number of smart devices, video analytics can also be utilized for monitoring the state of the workers. The video analytics combined with machine learning provide an alternative mean to monitor the various aspect of the work environment such as facial expression and the posture of the worker. Some of the latest researches in machine learning aim at calculating the heart rate of the person using an image of his/her face. It provides the possibility of measuring various modalities of the worker using just a single device.

Measurements included in online monitoring are:

1. Step count 2. Respiratory patterns 3. Facial expression 4. Galvanic skin conductance 5. EEG-signals 6. Blood lactate level 7. Respiratory Rate 8. Heart rate 9. Blood pressure 10. Body temperature 11. Length of sedentary periods 12. Length of standing periods 13. Length of continuous work 14. Timing of work periods 15. Number of breaks 16. Quality of sleep 17. Length of sleep 18. Perceived workload 19. Time to finish task 20. Errors (number, quality) 21. Number of complete tasks per time 22. Recovery time from mistakes 23. Machine usage time (Idle time, production time) 24. Pose of the worker

7.3 Off-line measurements Off-line measurements derive from questionnaires and interviews of workers. These data will be updated by each worker himself/herself in a lower frequency than online worker monitoring parameters.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 64 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

2. Perceived competence – job demand balance 3. Personality traits (big five) 4. Learning traits 5. Specific mental disabilities 6. Walking speed 7. Specific physical disabilities 8. Knowledge on relevant tasks 9. Comprehension of relevant processes 10. Job tenure 11. Experience with tools 12. Type of language 13. Everyday Tasks 14. Number of social instances 15. Energy Costs 16. General work satisfaction 17. Perceived workload 18. Work day specific satisfaction 19. Worker’s age 20. Worker’s height 21. Worker’s weight 22. Ability to cooperate 23. Communication skills 24. Motivation 25. Ability to deal with conflict

The list of online worker monitoring and off-line worker measurements is not complete, because Fac- tory2Fit project is at a preliminary stage. Although this deliverable D2.1 is submitted on M6, these measurements will be explicitly defined and updated in the next version of the deliverable (M18). 7.4 Definition and Classification of Physical, Sensorial and Cognitive abilities Except from characteristics, each worker has also multiple abilities, such as physical, cognitive and sensorial (Figure 34).

The research leading to these results has received funding from Horizon 2020, the European Un- Page 65 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Figure 34: Multiple dimensions- abilities a worker has (physical, cognitive, sensorial)

x Physical abilities: refer to abilities that influence strength, endurance, flexibility, balance coordination and psychomotility. x Cognitive abilities: abilities that influence the acquisition and application of knowledge in problem solving. x Sensorial abilities: abilities that influence visual, auditory and speech perception. Table 17 provides a sample/possible list of worker abilities referring to these categories. These abilities have an impact on Worker Model and help in its further development.

Table 17: Examples of Physical, Cognitive and Sensorial abilities of a worker

Physical abilities Cognitive abilities Sensorial abilities

Walk Perception Vision Run Attention Hearing Kneel Memory Speech Bend Visual and Spatial Processing Smelling Grasp Language Pull IQ Push

7.5 Calculation of Overall Labour Efficiency/Effectiveness (OLE) and Overall Equip- ment Efficiency (OEE)

OEE and related standardisation

OEE is a high-level equipment performance metric and is commonly used for monitoring the running performance of process plant or as well as shop floor equipment. It was developed as equipment effectiveness metric in Japan to measure the effectiveness of a manufacturing technique called Total Productive Maintenance (TPM). It was originally called Overall Equipment Effectiveness. The Semicon- ductor Equipment and Materials International (SEMI) Metrics Committee changed it to Overall Equip- ment Efficiency, and the methodology is standardised since 2004 and latest release is from 2014 (SEMI E10, Specification for Definition and Measurement of Equipment Reliability, Availability, and Main- tainability (RAM) and Utilization and SEMI E79, Specification for Definition and Measurement of Equip- ment Productivity [79105]).

There is also ISO standard released 2014. ISO 22400 defines key performance indicators (KPIs) used in manufacturing operations management (ISO 22400-2:2014 (en) Automation systems and integration — Key performance indicators (KPIs) for manufacturing operations management — Part 2: Definitions and descriptions). ISO 22400-2:2014 specifies a selected number of KPIs in current practice. The KPIs are presented by means of their formula and corresponding elements, their time behaviour, their

The research leading to these results has received funding from Horizon 2020, the European Un- Page 66 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model unit/dimension and other characteristics. ISO 22400-2:2014 also indicates the user group where the KPIs are used, and the production methodology to which they correspond. With reference to equipment, the KPIs in ISO 22400-2:2014 relate to work units, as specified in IEC 62264 (ANSI/ISA-95.)

VDMA 66412 series on production indicators exist as well:

x VDMA 66412-1. Manufacturing Execution Systems (MES) Kennzahlen (2009-10) x VDMA 66412-1/A1 Manufacturing Execution Systems (MES) Kennzahlen; Änderung 1 zu VDMA 66412-1:2009-10 (2010-05) x VDMA 66412-2 Manufacturing Execution Systems (MES) Kennzahlen-Wirkmodell (2010-11) x VDMA 66412-3 Manufacturing Execution Systems (MES) - Operational Sequence description of data acquisition (2014-04) x VDMA 66412-10 Manufacturing Execution Systems - Data for Production Indicators (2015-04)

OEE calculation

OEE is based on time analysis and classification of reliability (MTBF), maintainability (MTTR), utilisation (availability), throughput, and yield (Figure 35). It identifies losses due to equipment failure, set-ups and adjustments, idling and minor stops, reduced speed, process defects and start up. All of the above factors are grouped under the following three sub-metrics of equipment efficiency:

1. Availability (Availability Efficiency)

2. Performance efficiency (Operational Efficiency, Rate Efficiency)

3. Rate of quality (Quality Efficiency)

The three sub-metrics and OEE are mathematically related as follows:

OEE, % = availability x performance efficiency x rate of quality x 100

Total Time

Availability Non-scheduled Operations Operational Efficiency Time Time Efficiency

Equipment Equipment Uptime Downtime

Engineering Manufacturing Time Time Unscheduled Scheduled Downtime Downtime Productive Standby Time Time

Rate Minor Speed Scrap Rework Quality Efficiency Stops Loss Efficiency

The research leading to these results has received funding from Horizon 2020, the European Un- Page 67 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Total Time (365 days x 24 hours)

Non Operating Time Scheduled Equipment Uptime Equipment Downtime Unscheduled Scheduled Downtime Downtime Engineering e k Manufacturing Time Failures h c Time t e e t n n Standy Time A le i t h Productive Time Idling/Starving t c OEE and OEE losses o a - Setups/Adjustments B M Actual Uutput Reduced Minor Speed Stoppage - Breakdowns - Idling/Minor Stoppages GOOD UNITS Rework Scrap - Reduced Speed - Defects/Rework - Yield

Figure 35: OEE calculation

Definition of sub-metrics:

1. Availability Efficiency: The fraction of total time that the equipment is in a condition to perform its intended function = (Equipment Uptime)/(Total Time). 2. Performance Efficiency: The fraction of equipment uptime during which the equipment is processing actual units at theoretically efficient rates = (Operational Efficiency) x (Rate Efficiency). - Operational Efficiency - The fraction of equipment uptime during which the equipment is processing actual units = (Production Time) / (Equipment Uptime). - Rate Efficiency - The fraction of production time during which the equipment is processing actual units at theoretically efficient rates = (Theoretical Production Time for Actual Units)/(Production Time) 3. Quality Efficiency: The theoretical production time for effective units divided by the theoretical production time for actual units. - Effective Units=(Actual Units)-(Scrap Units+ Rework Units) The OEE provides identification of the big losses (Table 18).

Table 18: OEE Six big losses

OEE efficiency Major OEE losses Losses sub-category SEMI E10 efficiency cal- calculations culations Availability Non-scheduled Non-scheduled time Set-ups/adjustments Preventative maintenance Scheduled downtime Qualifications Set-up Breakdowns Equipment failure Unscheduled downtime Process failure Facility failure

The research leading to these results has received funding from Horizon 2020, the European Un- Page 68 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Performance Idling/minor Engineering Engineering time efficiency stoppages No product Standby/idle time No operator Other time losses Reduced speed Limited capacity Tool speed losses Slow speed Average batch size Stoppage/assist Productive Rate of quality Defects/rework Rework Quality losses Yield Scrap

OEE provides a systematic way to classify and study equipment efficiency and time related losses. It helps to identify the actual time the system is producing good units and is best used to identify scope for process performance improvement, and how to get the improvement. OEE can be used for evaluating different production work time and work shift arrangements and at the same time identify and evaluate the OEE losses (Table 18).

Adapting OEE to human workers; Overall Labour Efficiency/Effectiveness (OLE)

Following text is partially adapted from Kronos White paper: Overall Labor Effectiveness (OLE): Achiev- ing a Highly Effective Workforce (2007) [106]. The white paper is no more available at Kronos web.

Optimizing workforce performance requires insight, methods of quantifying, diagnosing, and predicting the performance of human workforce, important and highly variable elements in manufacturing. That insight can be provided by Overall Labour Efficiency/Effectiveness (OLE). OLE is the analysis of the cumulative effect three factors, similar to OEE, that have on productive output:

• Utilization: the percentage of time the workforce spends making effective contributions

• Performance: the amount of product delivered

• Quality: the percentage of perfect or saleable product produced

The familiar OEE factors are the basic elements used in measuring human workforce effectiveness. But in measuring the contributions human operators, it is useful to look deeper and consider additional factors (Table 19).

Table 19: OEE factors used in measuring human workforce effectiveness

OEE Element OLE Element Key Human Workforce Drivers Availability Availability Utilization and absenteeism Scheduling of indirect activities Performance Performance Availability of processes, instructions, tools, and materials Training and skills Indirect support staff Quality Quality Employee knowledge Proper use of instructions and tools

The research leading to these results has received funding from Horizon 2020, the European Un- Page 69 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Availability is a basic criterion, and utilization is the most important component of availability. There are many things that influence human availability, and therefore the potential output of equipment and the factory. For example:

• Absenteeism and utilization: Standard worker utilization measures: employee illness, approved or unapproved leaves, and times when people are unavailable due to training, meetings, or other company-defined activities.

• Scheduling: Involves having the right skill at the right time. Beyond merely providing a worker, we must consider employee skills and certifications, as well as flexible work schedules.

• Indirect time: Includes material delays, idle time, shift changeover, and machine downtime.

Performance is the recording of output, which determines whether producing or delivering a product or service took as long as company standard work time indicated it would (whether tangible units are manufactured or specific services are delivered). Performance output includes:

• Availability of processes, instructions, tools, and materials: Shop floor issues, such as worn or misplaced tools, material shortages, or missing processes or instructions, will slow production and limit output — and likely have a negative impact on quality.

• Training and skills: Do employees know how to do the tasks they are assigned? Certainly these factors affect the ability to deliver the expected output throughout a complete shift or job run.

• Indirect support staff: A workforce that is insufficiently trained or skilled will require additional support staff, including supervisors, maintenance technicians, and quality assurance personnel.

With OLE analytics, relationships and dependencies between factors can be brought to the surface. For example: Something disregarded on the shop floor as a minor availability issue shows up as a troubling performance shortfall. OLE analytics track the problem back to a failure to meet planned standard job times, which was caused by jobs that weren’t started on time because the right employees weren’t at their stations and ready for the work at hand. When performance consistently falls below expectations, OLE quickly highlights the root causes, including inaccurately set labour standards.

Quality, we need to know if the output of production met specified quality levels. While quality is certainly a function of the materials used, it is impacted by important human factors:

• Employee knowledge: Do employees understand the quality drivers of their specific operations? Employee skills directly affect the quality of output. Knowledgeable operators know how to measure their work and understand how the processes operate, how variability affects quality, and what adjustments keep processes to spec as they run. They also know when to stop production for corrective actions, should quality fall below specified limits. Applying this type of knowledge reduces the amount of wasted work and cuts scrap and rework costs.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 70 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

• Proper use of instructions and tools: Did workers use the right tools and follow the right proce- dures?

Correlation between planned on-line worker measurements (section 7.2) and other defined worker, task or machine characteristics and calculation of OLE are to be explicitly defined in the next version of the deliverable (M18).

7.6 (Persistence) Data storage mechanism

The data storage mechanism which will be developed in the Factory2Fit project is very important because it forms the basis for further analysis. This mechanism will enable access to profile data, will store data acquired from both sensors and the worker, and will ensure the usage of historical data for analysis in order to produce the worker’s performance.

A database is a structured collection of datasets which can be accessed by a database management system (DBMS). The kind of database management system defines the model of the database. Being able to store and analyse the huge amount of data resulting from the large number of variables and measures, a storage mechanism has been set up (Figure 36). The relevant data, gathered from the worker, machine or task as well as the environment, will be organized within a relational database management system. This allows new raw data to be integrated, modified or deleted and in search of relevant information. Through the queries, different kinds of data can be combined, displayed, analysed or used for the algorithms.

Figure 36: Data storage framework

The core component is the relational open-source database management system PostgreSQL 9.1. All tools for data pre-processing as well as the analysis tools access the DBMS using the Structured Query

The research leading to these results has received funding from Horizon 2020, the European Un- Page 71 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

Language (SQL). SQL is standardized by the ISO 9075, which allows for a common access procedure using various software products. Compared to other analysis tools, the relational database provides some advantages like a minimum of redundancy, improved data consistency and accessibility as well as program-data independence. The amount of data is only limited by hardware resources. Because of the client-server-principle, only one server needs extended hardware resources.

These database features are especially useful when dealing with large, heterogeneous and related datasets like sensor data with different frequencies, questionnaires, videos, interview data and other kinds of information gathered in an environment of a smart factory. The main operations inside the database concern structuring and aggregating data. Media files such as video, pictures or audio are not stored entirely inside the database. These files are located on a separate fileserver and the database tables only contain a reference, e.g. the filename and frame number for each video. This structure allows for keeping the database size smaller and therefore faster, whereas the link to the media files is always present.

Due to the client-server principle, the DBMS runs on one server and includes rights management, backup strategy, failsafe hardware configuration... etc. In PostgreSQL, the rights management is built by roles. Every user or user group is defined as a role. Every role is described by different privileges and accessibility can be restricted individually. This complies with the German data security law that states 8 commandments:

1. Access control (spatial) 2. Entry control 3. Access control (no unauthorized executions within the system) 4. Transfer control 5. Input control 6. Job control 7. Availability control 8. Segregation command

At the Chemnitz University of technology, every commandment is complied with through different measures, for example, a firewall, selective approval of access to the database server, necessary au- thentication and, as mentioned, the rights management within the database management system.

This enables the architect of the dataflow to control the access even to sensitive data of the worker, for example, the workers’ health. Furthermore, the worker himself can be enabled to control the use of his own profile data. [112]

The research leading to these results has received funding from Horizon 2020, the European Un- Page 72 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

8 Conclusions

The main aim of this deliverable was to provide an analysis of the existing standards, techniques and methodologies, in order to drive the work towards the development of a work-related abstract Worker Model, and a dynamic worker model called Adaptive Worker Model. Both models include worker characteristics, though Adaptive Worker Model consists not only from worker characteristics, but also from task, machine, and contextual characteristics. Finally, a Worker Profile will be delivered, that will include each worker’s characteristics. This Profile will be used for the following WPs, to help matching workers and tasks.

In order to develop two structured Worker Models that take into account all the proper parameters and characteristics needed, several methodologies and standards need to be studied. In addition to Worker Models, this document also presented an overview and a comparative analysis of the most widely known User Interface Description languages. Furthermore, existing methods for physical user modelling and theories of work satisfaction and performance were investigated.

The document concluded with the online worker monitoring and off-line measurements analysis, alongside with the description of the data storage mechanism that will be developed. More specifically, each characteristic is further analysed in variables and parameters. Based on the context and the frequency measured, online and off-line measurements categories derive. The list of online and off-line worker measurements is not complete and will be further developed and updated throughout the project.

The models and measures analysed in this deliverable were selected based on the usefulness for each use case and the expected acceptability of users and meet the ethical guidelines of the Factory2Fit project. For the definition of these characteristics we took into account all the important aspects that define a worker and kept in mind that worker as a human being is the most important thing (human - centered approach). The different levels of Worker Models developed derive from the need to create a more personalized profile that fits to the workers needs and defines his/hers abilities properly. This profile will be used in further WPs in order to match the worker with the appropriate task. However, the development of these models is at an early stage, therefore not every variable and measure in worker characteristics is included. These models will be further evaluated and updated in a later phase of the project, where it will be decided which of these characteristics and measures better fit to each pilot.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 73 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

9 References

1 Schaefer, R., Steffen, B., Wolfgang, M., Task Models and Diagrams for User Inter-face Design, Proceedings of 5th International Workshop, TAMODIA'2006 (Hasselt, Belgium, October 2006), Lecture Notes in Com- puter Science, Vol. 4385, Springer Verlag Berlin, 2006, pp. 39-53. 2 Kost, S. (2004), Dynamically generated multi-modal application interfaces. Ph.D. Thesis, Technical Univer- sity of Dresden and Leipzig University of Applied Sciences, Germany. 3 Crowle, S., Hole, L., ISML: An Interface Specification Meta-Language. DSV-IS 2003, Funchal, Madeira Island, Portugal. June 11-13, 2003. Lecture Notes I Com-puter Science 2844 Springer 2003, ISBN 3-540-20159-9. 4 Demler, G., Wasmund, M., Grassel, G., Spriestersbach, A. & Ziegert, T. (2003), Flexible pagination and lay- outing for device independent authoring, WWW2003 Emerging Applications for Wireless and Mobile access Workshop. 5 Luyten, K., Abrams, M., Vanderdonckt, J. & Limbourg, Q. (2004) Developing User Interfaces with XML: Ad- vances on User Interface Description Languages, Sattelite workshop of Advanced Visual Interfaces 2004, Gallipoli, Italy. 6 Picard, E., Fierstone, J., Pinna-Dery, A-M. & M. Riveill (2003). Atelier de composi-tion d'ihm et évaluation du modèle de composants. Livrable l3, RNTL ASPECT, Laboratoire I3S, mai. 7 Paternò, F. & Santoro, C. (2003), A Unified Method for Designing Interactive Sys-tems Adaptable to Mobile and Stationary Platforms, Interacting with Computers, El-sevier, 15, pp. 349-366. 8 Abrams, M., Phanouriou, C., Batongbacal, A.L., Williams, S. & Shuster, J. (1999), UIML: An Appliance-Inde- pendent XML User Interface Language. In A. Mendelzon, editor, Proceedings of 8th International World- Wide Web Conference WWW‘8 (To-ronto, May 11-14, 1999), Amsterdam, 1999. Elsevier Science Publish- ers. 9 Vanderdonckt, J. A MDA-Compliant Environment for De-veloping User Interfaces of Information Systems. In Proc. of 17th Conf. on Advanced Information Systems En-gineering CAiSE'05 (Porto, 13-17 June 2005), LNCS, Vol. 3520, Springer-Verlag, Berlin, 2005, pp. 16-31. 10 Arsanjani, A., Chamberlain, D. and et al. (2002), (WSXL) web service experience language version, 2002. Retrieved from: http://www-106.ibm.com/developerworks/library/ws-wsxl2/. 11 IBM (2002), WSXL specification, April 2002, retrieved on Janury 2nd 2009. 12 Gomes de Sousa, L. & Leite, J. C. (2003), XICL: a language for the user's interfac-es development and its components. Proceedings of the Latin American confe-rence on Human-computer interaction (Rio de Janeiro, Brazil, August 17 - 20, 2003), ACM Press pages, New York, pp. 191-200. 13 Eisenstein J., Vanderdonckt J., Puerta A. (2000), Adapting to Mobile Contexts with User-Interface Modeling, Proceedings of 3rd IEEE Workshop on Mobile Computing Systems and Applications WMCSA‘2000 (Monte- rey, 7-8 December 2000), IEEE Press, Los Alamitos, 2000, pp. 83-92. 14 Eisenstein J., Vanderdonckt J., Puerta A. (2001), Model-Based User-Interface Development Techniques for Mobile Computing, Proceedings of 5th ACM Int. Conf. on Intelligent User Interfaces IUI‘2001 (Santa Fe, 14- 17 January 2001), Lester, J. (Ed.), ACM Press, New York, 2001, pp. 69-76. 15 Puerta, A., & Eisenstein, J. (2002), XIML: a common representation for interaction data, Proceedings of the 7th international conference on Intelligent User Interfaces, ACM, pp. 214-215. 16 Puerta A.R. (1996), The Mecano Project: Comprehensive and Integrated Support for Model-Based Interface Development, Proceedings of 2nd Int. Workshop on Computer-Aided Design of User Interfaces CADUI‘96 (Namur, 5-7 June 1996), Presses Universitaires de Namur, 1996, pp. 19-35.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 74 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

17 Fabio Paternò, Carmen Santoro, Lucio Davide Spano (2008), Designing Usable Applications based on Web Services, Proceedings of the First Workshop on the Interplay between Usability Evaluation and Software Development,Pisa, Italy. 18 Andrejko, Anton, Michal Barla, Mária Bieliková (2007), Ontology-based user modelling for web-based information systems, Advances in Information Systems Development, Springer US, pp. 457-468. 19 K. Agyapong-Kodua, Csaba Haraszkób and István Németh (2014): Resource selection ontologies in support of a recipe-based factory design methodology. International Journal of Production Research, Vol. 52 (21), 6235–6253. 20 F. Ameri, L. Patil (2012): Digital manufacturing market: a semantic web-based framework for agile supply chain deployment. Journal of Intelligent Manufacturing 23 (5), 1817-1832. 21 F. Ameri, C. McArthur (2014): Semantic rule modelling for intelligent supplier discovery. International Jour- nal of Computer Integrated Manufacturing 27 (6), 570-590. 22 Farhad Ameri, Colin Urbanovsky, and Christian McArthur (2012): A Systematic Approach to Developing On- tologies for Manufacturing Service Modeling. 7th International Conference on Formal Ontology in Infor- mation Systems (FOIS 2012), Graz, Austria. 23 Aamer Baqai, Ali Siadat, Jean-Yves Dantan, Patrick Martin (2007): Integration of a Manufacturing Ontology with the Design of Reconfigurable Machine Tool using Function-Behavior-Structure Approach. 2nd Inter- national Conference on Digital Information Management, Volume 1, 473 – 478. 24 Stefano Borgo, Paulo Leitao (2004): The Role of Foundational Ontologies in Manufacturing Domain Appli- cations. In: Meersman R., Tari Z. (eds) On the Move to Meaningful Internet Systems 2004: CoopIS, DOA, and ODBASE. OTM 2004. Lecture Notes in Computer Science, Vol 3290. Springer, Berlin, Heidelberg. 25 Christian Büschera, Hanno Voetb, Moritz Krunkeb, Peter Burggräfb, Tobias Meisena, Sabina Jeschke (2016): Semantic information modelling for factory planning projects. Procedia CIRP 41, 478 – 483. 26 Haibo Cheng, Peng Zeng, Lingling Xue, Zhao Shi, Peng Wang, Haibin Yu (2016): Manufacturing Ontology Development based on Industry 4.0 Demonstration Production Line. 2016 Third International Conference on Trustworthy Systems and Their Applications. 27 A. F. Cutting-Decelle, R. I. M. Young, J. J. Michel, R. Grangel, J. Le Cardinal and J. P. Bourey (2007): ISO 15531 MANDATE: A Product-process-resource based Approach for Managing Modularity in Production Manage- ment. Concurrent Engineering: Research and Applications, Vol. 15 (2). 28 Ivan M. Delamer, Jose L. Martinez Lastra (2006): Ontology Modeling of Assembly Processes and Systems using Semantic Web Services. 4th IEEE International Conference on Industrial Informatics, 611 – 617. 29 Marco Garetti, Luca Fumagalli, Elisa Negri (2015): Role of ontologies for CPS implementation in manufacturing. Management and Production Engineering Review, Vol. 6 (4), 26–32. 30 Antonio Giovannini, Alexis Aubry, Herve Panetto, Michele Dassisti, Hind El Haouzi (2012): Ontology-based system for supporting manufacturing sustainability. Annual Reviews in Control, 36 (2), 309-317. 31 OndƎej Harcuba, Pavel Vrba (2015): Ontologies for Flexible Production Systems. IEEE 20th Conference on Emerging Technologies & Factory Automation (ETFA), 1 – 8. 32 Dimitris Kiritsis (2013) Semantic technologies for engineering asset life cycle management, International Journal of Production Research, Vol. 51, 23-24. 33 Severin Lemaignan, Ali Siadat, Jean-Yves Dantan, Anatoli Semenenko (2006): MASON: A Proposal For An Ontology Of Manufacturing Domain. Proceedings of the IEEE Workshop on Distributed Intelligent Systems: Collective Intelligence and Its Applications (DIS’06). 34 Matthias Loskyll, Jochen Schlick (2013): Cyber Physical Production Systems. Dynamic Orchestration of Mechatronic Field Device Functions. at – Automatisierungstechnik, Vol. 61 (10).

The research leading to these results has received funding from Horizon 2020, the European Un- Page 75 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

35 H.K. Lin, J. A. Harding, M. Shahbaz (2004): Manufacturing system engineering ontology for semantic interoperability across extended project teams. International Journal of Production Research, 42 (24), pp. 5099 – 5118. 36 L.F. Lin, W.Y. Zhang, Y.C. Lou, C.Y. Chu, and M. Cai (2011): Developing manufacturing ontologies for knowledge reuse in distributed manufacturing environment. International Journal of Production Research, Vol. 49 (2). 37 Jose L. Martinez Lastra and Ivan M. Delamer (2008): Ontologies for Production Automation. In: T.S. Dillon et al. (Eds.): Advances in Web Semantics I, LNCS 4891, pp. 276–289. 38 Nelson Ricardo Martins Rodrigues (2012): Development of an Ontology for a Multi-Agent System Control- ling a Production Line. Instituto Politécnico de Bragança, Setembro de 2012. 39 Viviana Mascardi, Valentina Cordì, Paolo Rosso (2007): A Comparison of Upper Ontologies. Dagli Oggetti agli Agenti. 8th AI*IA/TABOO Joint Workshop "From Objects to Agents": Agents and Industry: Technological Applications of Software Agents, 24-25 September 2007, Genova, Italy. 40 Natalya F. Noy and Deborah L. McGuinness (2001): Ontology Development 101: A Guide to Creating Your First Ontology. Stanford Knowledge Systems Laboratory Technical Report KSL-01-05 and Stanford Medical Informatics Technical Report SMI-2001-0880, March 2001. 41 Craig Schlenoff, Tsai Hong, Roger Eastman, Connie Liu, Sebti Foufou (2013): A Literature Review of Sensor Ontologies for Manufacturing Applications. IEEE International Symposium on Robotic and Sensors Environ- ments (ROSE), 96 – 101. 42 Walter Terkaja, Ludovico Danzab, Anna Devitofrancescob, Stefano Gagliardoc, Matteo Ghellereb, Franca Gianninic, Marina Montic, Giulia Pedriellia, Marco Saccoa, Francesco Salamone (2014): A Semantic Frame- work for Sustainable Factories. Procedia CIRP 17, 547 – 552. 43 George J. Tsinarakis and Chrisa J. Tsinaraki (2013): OntoMoPS: A Modular Production System Description Ontology. 21st Mediterranean Conference on Control & Automation (MED) Platanias-Chania, Crete, Greece, June 25-28, 2013. 44 R.I.M. Young et al. (2007): Manufacturing knowledge sharing in PLM: a progression towards the use of heavy weight ontologies. International Journal of Production Research, 45 (7), pp. 1505 – 1519. 45 J. Zhang and K. Agyapong-Kodua (2015): Integrated Ontologies in Support of Factory Systems Design. IFAC- Papers online 48-3, 2095–2102. 46 Andriacchi, I. and Hurwitz D. Gait Biomechanics and the Evolution of Total Joint Replacement, Gait and Posture, 1997, 5, 256-264. 47 Debevec, M.; Simic, M; Herakovic, N. (2014): Virtual Factory as an advanced approach for production process optimization. Interentional Journal of Simulation modelling, Vol. 13 (1), 66-78. 48 Sanjay Jain, David Lechevalier (2016): Standards based generation of a virtual factory model. In: T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huschka, and S. E. Chick (eds.): Proceedings of the 2016 Winter Simulation Conference. 49 Jia Chenhui, Ma Wensuo, Xia Xintao (2010): Research on modelling and simulating optimization for Digital Factory. 2010 International Conference on Measuring Technology and Mechatronics Automation. 50 Kiwook Jung, SangSu Choi, Boonserm Kulvatunyou, Hyunbo Cho, K. C.: Morris (2017): A reference activity model for smart factory design and improvement. Production Planning & Control, 28 (2), 108-122. 51 Ju Yeon Lee, Gun Yeon Kim and Sang Do Noh (2011): Integration Framework and PPR+H Hub for DiFac. In: Luca Canetta,Claudia Redaelli, Myrna Flores (eds.): Digital Factory for Human-oriented Production Systems, Springer. 52 Rainer Müller, Leenhard Hörauf, Matthias Vette, Christoph Speicher (2016): Planning and developing cyber- physical assembly systems by connecting virtual and real worlds. Procedia CIRP 52, 35 – 40.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 76 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

53 Joung Hwan Mun and Yong Hoon Rim (2011): Human Body Modeling. In: Luca Canetta,Claudia Redaelli, Myrna Flores (eds.): Digital Factory for Human-oriented Production Systems, Springer. 54 Walter Terkaj; Marcello Urgo (2014): Ontology-based Modeling of Production Systems for Design and Per- formance Evaluation. 55 VDI 3633 (2014): Simulation of logistics, material flow and production systems. 56 VDI 4499 (2008): Digital Factory – Fundamentals. 57 Sigrid Wenzel, Jochen Bernhard, Ulrich Jessen (2003): A taxonomy of visualization techniques for simulation in production and logistics. In: S. Chick, P. J. Sánchez, D. Ferrin, and D. J. Morrice (eds): Proceedings of the 2003 Winter Simulation Conference. 58 Calvary, G., Coutaz, J., Thevenin, D., Limbourg, Q., Bouillon, L., Vanderdonckt, J.: A Unifying Reference Framework for Multi-Target User Interfaces. Interacting with Computers, Vol. 15, No. 3 (June 2003) 289– 308. 59 I. Ajzen, M. Fishbein (1980): Understanding Attitudes and predicting Social Behavior, Prentice-Hall, Eng- lewood Cliffs, NJ. 60 T.S. Baines, R. Asch, L. Hadfield, J.P. Mason, S. Fletcher, J.M. Kay (2005): Towards a theoretical framework for human performance modelling within manufacturing systems design. Simulation Modelling Practice and Theory, Vol. 13, 486–504. 61 Tim Baines, Stephen Mason, Peer-Olaf Siebers, John Ladbrook (2004): Humans: the missing link in manufacturing simulation? Simulation Modelling Practice and Theory, Vol. 12, 515–526. 62 Chris Creed (2005): Using Emotion Simulation to Influence User Attitudes and Behaviour. RSMG Progress Report 3. 63 F.D. Davis (1989): Perceived usefulness, perceived ease of use, and user acceptance of information technologies, MIS Quarterly, Vol. 13 (3), pp. 319 – 340. 64 Jan Dijkstra, Harry J.P. Timmermans, Bauke de Vries (2001): Virtual Reality-based simulation of user behaviour within the built environment to support the early stages of building design. Construction Informatics Digital Library. 65 Gerhard Fischer (2001): User Modeling in Human-Computer Interaction. User Modeling and User-Adapted Interaction 11, pp. 65-86. 66 Helmut Hlavacs and Gabriele Kotsis (1999): Modeling User Behavior: A Layered Approach. Proceedings. 7th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Sys- tems. 67 Jose A. Iglesias, Plamen Angelov, Agapito Ledezma and Araceli Sanchis (2009): Modelling Evolving User Behaviours. IEEE Workshop on Evolving and Self-Developing Intelligent Systems, ESDIS '09. 68 Sujith Kanagarajan, Sivakumar Ramakrishnan (2015): Development of Ontologies for Modelling User Be- haviour in Ambient Intelligence Environment. 2015 IEEE International Conference on Computational Intel- ligence and Computing Research (ICCIC). 69 Daniel E. Montano and Danuta Kasprzyk (2015): Theory of Reasoned Acton, Theory of Planned Behavior, and the Integrated Behavioral Model. In: Karen Glanz, Barbara K.Rimer, K. Viswanath (eds.): Health Behav- ior: Theory, Research, and Practice. 5th ed.,Wiley, pp. 95-124. 70 Pavel Osipov, Arkadijs Borisovs (2011): Simulation of typical behavior user using Markov models. Baltic Congress on Future Internet and Communications 2011. 71 Tabak, V., B. de Vries, and J. Dijkstra (2004): User Behaviour Modelling, In: Van Leeuwen, J.P. and H.J.P. Timmermans (eds.) Developments in Design & Decision Support Systems in Architecture and Urban Plan- ning, Eindhoven: Eindhoven University of Technology, ISBN 90-6814-155-4, p. 141-156.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 77 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

72 Józef Tick (2003): Aspects of User Behaviour Modelling. Proceedings of the International Conference in Memoriam John von Neumann, Budapest, Hungary, December 12, 2003, pp. 269-274. 73 T.D. Wilson (2000): Human Information Behavior. Informing Science, Vol. 3 (2). 74 Kristina Yordanova (2011): Toward a unified human behaviour modelling approach. Tech Report CS-02-11, 2011. ISSN 0944-5900. 75 Gerhard Zimmermann (2007): Modeling and Simulation of Individual User Behavior for Building Perfor- mance Predictions. Proceedings of the 2007 Summer Computer Simulation Conference, pp. 913-920. 76 Ingrid Zukerman and David W. Albrecht (2001): Predictive Statistical Models for User Modeling. User Mod- eling and User-Adapted Interaction 11, pp. 5-18. 77 Baines, T. S., Asch, R., Hadfield, L., Mason, J. P., Fletcher, S., Kay, J. M. (2005). Towards a theoretical framework for human performance modelling within manufacturing systems design. Simulation Modelling Prac- tice and Theory, 13(6), 486-504. 78 Cabrita, J., & Perista, H. (2007). Measuring job satisfaction in surveys - Comparative analytical report. 79 Coomber, B., & Louise Barriball, K. (2007). Impact of job satisfaction components on intent to leave and turnover for hospital-based nurses: A review of the research literature. International Journal of Nursing Studies, 44(2), 297–314. http://doi.org/10.1016/j.ijnurstu.2006.02.004 80 Cox, Tom, Griffiths, A., Rial-Gonzalez, E. (2000). Research on Work-related Stress. Office for Official Publi- cations of the European Communities. http://doi.org/http://europa.eu.int 81 Fisher, C. D. (2000). Mood and Emotions while Working: Missing Pieces of Job Satisfaction? Journal of Or- ganizational Behavior, 21(2), 185–202. 82 Fisher, T. F. (2003). Perception differences between groups of employees identifying the factors that influence a return to work after a work-related musculoskeletal injury. Work, 21(3), 211–220. 83 FIOH - Finnish Institute of Occupational Health: web pages in Finnish: Työkykyinen työtekijä, https://www.ttl.fi/tyontekija/; web pages in English: well-being at work https://www.ttl.fi/en/research- and-development-projects/ 84 Hackman, J. R., & Oldham, G. R. (1975). Development of the Job Diagnostic Survey. Journal of Applied Psy- chology, 60(2), 159–170. http://doi.org/10.1037/h0076546 85 Hadfield, L., Fletcher, S., Mason, J.P., Baines, T., & Ladbrook, J. (2002, April 3-5). A Theoretical Framework for Human Performance Modelling. Proceedings of the Second International Conference on Systems Think- ing in Management (ICSTM02), Salford, UK. 86 Hoogendoorn, W. E., Bongers, P. M., de Vet, H. C. W., Ariëns, G. A. M., van Mechelen, W., & Bouter, L. M. (2002). High physical work load and low job satisfaction increase the risk of sickness absence due to low back pain: results of a prospective cohort study. Occupational & Environmental Medicine, 59(5), 323–329. 87 Karasek, R. A. (1979). Job Demands, Job Decision Latitude, and Mental Strainථ: Implications for Job Redesign Author (s): Robert A. Karasek, Jr. Sourceථ: Administrative Science Quarterly, Vol. 24, No. 2 (Jun., 1979), pp. 285-308 Published byථ: Sage Publications,. Administrative Science Quarterly, 24(2), 285–308. 88 Karasek R, and Theorell T (1990). Healthy work - stress, productivity and the reconstruction of working life. Basic Books, New York. 89 Levi, L., & Levi, I. (2000). Guidance on work-related stress. Spice of life, or kiss of death? Luxembourg: Office for Official Publications of the European Communities. http://doi.org/ec.europa. eu/employment_social/publications/2002/ke4502361_en.pdf 90 Locke, E. A. “The Nature and Causes of Job Satisfaction” in Dunnette, M. D. Handbook of Industrial and Organizational Psychology (1st Ed.), Chicago, IL: Rand McNally. 1976. 1297-1349. 1976. 91 Locke, E., & Latham, G. P. (1990). Work Motivation and Satisfaction: Light at the end of the tunnel. Psycho- logical Science, 1(4), 240–246.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 78 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

92 Saari, L. M., & Judge, T. A. (2004). Employee attitudes and job satisfaction. Human Resource Management, 43(4), 395–407. http://doi.org/10.1002/hrm.20032 93 Scott-Ladd, B., Travaglione, A., & Marshall, V. (2006). Causal inferences between participation in decision making, task attributes, work effort, rewards, job satisfaction and commitment. Leadership & Organization Development Journal, 27(5), 399–414. http://doi.org/10.1108/01437730610677990 94 Simone, S. De. (2014). Conceptualizing Wellbeing in the Workplace, 5(12), 118–122. 95 Spector, P.E. (1997). Job satisfaction: Application, assessment, causes, and consequences, London: Sage. 96 Weiss, H. M., & Cropanzano, R. (1996). Affective Events Theory: A theoretical discussion of the structure, causes and consequences of affective experiences at work. Research in Organizational Behavior. http://doi.org/1-55938-938-9 97 Ybema, J. F., Smulders, P. G. W., & Bongers, P. M. (2010). Antecedents and consequences of employee absenteeism: A longitudinal perspective on the role of job satisfaction and burnout. European Journal of Work and Organizational Psychology, 19(1), 102–124. http://doi.org/10.1080/13594320902793691 98 Finnish Institute of Occupational Health (FIOH): web pages in Finnish: Työkykyinen työtekijä: https://www.ttl.fi/tyontekija/; web pages in English: well-being at work https://www.ttl.fi/en/research- and-development-projects/ 99 Iiro Harjunkoski, Reinhard Bauer (2014). Sharing data for production scheduling using the ISA-95 standard. Front. Energy Res., 21 October 2014 | https://doi.org/10.3389/fenrg.2014.00044. Article is available at http://journal.frontiersin.org/article/10.3389/fenrg.2014.00044/full 100 German Standardization Roadmap on Industry 4.0: available at http://www.din.de/en/innovation-and-research/industry-4-0/the-road-to-the-future--77392 101 ISO catalogue: available http://www.iso.org/iso/home/store/catalogue_tc.htm 102 ISO/TC 159/SC 3 13.180 Ergonomics http://www.iso.org/iso/iso_catalogue/catalogue_ics/cata- logue_ics_browse.htm?ICS1=13&ICS2=180 103 Ergonomics and human factors at work: available at http://www.hse.gov.uk/pubns/indg90.pdf 104 Ergonomic guidelines for Manual material handling. The National Institute for Occupational Safety and Health (NIOSH), USA. available at https://www.cdc.gov/niosh/docs/2007-131/pdfs/2007-131.pdf 105 SEMI10: available at http://www.semi.org/en/node/50191 106 Kronos White paper: Overall Labor Effectiveness (OLE): Achieving a Highly Effective Workforce (2007). The white paper is no more available at Kronos web (https://www.kronos.com/) 107 Factory2Fit Deliverable D1.1 “Enabling Technologies”. 108 Factory2Fit Deliverable D1.5 “Design and evaluation framework and measuring tools”. 109 Technical: ISA-88 and ISA-95: available at https://isa-95.com/technical-isa-88-and-isa-95/ 110 Colucci, S., Di Noia, T., Di Sciascio, E., M. Donini, F., Mongiello, M., & Piscitelli G. (2004): Semantic-based Approach to Task Assignment of Individual Profiles. Journal of Universal Computer Science 10(6): 723-730, http://dx.doi.org/10.3217/jucs-010-06-0723 111 Abowd, G. D., Dey, A. K., Brown, P. J., Davies, N., Smith, M., & Steggles, P. (1999). Towards a better understanding of context and context-awareness. In International Symposium on Handheld and Ubiquitous Com- puting (pp. 304-307). Springer Berlin Heidelberg. 112 Factory2Fit Deliverable D7.6 “Ethics report”. 113 BKCASE Editorial Board. 2016. The Guide to the Systems Engineering Body of Knowledge (SEBoK), v. 1.7. R.D. Adcock (EIC). Hoboken, NJ: The Trustees of the Stevens Institute of Technology. Accessed DATE. www.sebokwiki.org. BKCASE is managed and maintained by the Stevens Institute of Technology Systems Engineering Research Center, the International Council on Systems Engineering, and the Institute of Elec- trical and Electronics Engineers Computer Society.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 79 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277 D2.1 – Dynamic worker model

114 Stachowiak, H. (1973). Allgemeine Modelltheorie. Wien: Springer. 115 United States Department of Defense Modeling and Simulation (M&S) Glossary, 1998.

The research leading to these results has received funding from Horizon 2020, the European Un- Page 80 of 80 ion’s Framework Programme for Research and Innovation (H2020/2014-2020) under grant agree- ment no 723277