International journal of computer integrated manufacturing , tập 23, số 3, 2010

International Journal of Computer Integrated Manufacturing Vol 23, No 3, March 2010, 195–215 Integrated approach to modelling human systems as reuseable components of manufacturing workplaces Joseph Ajaefobi, Richard Weston*, Bilal Wahid and Aysin Rahimifard Manufacturing Systems Integration (MSI) Research Institute, Loughborough University, LE11 3TU, UK (Received 31 May 2007; final version received September 2009) A new approach to modelling human systems as reusable components of manufacturing workplaces is described Graphical and computer executable models of people competences and behaviours are created which are qualitatively and quantitatively matched to equivalent models of process networks, decomposed into roles and dependencies between roles To enable model creation and reuse, coherent sets of role, competence and dynamic producer unit (DPU) modelling concepts have been defined and instrumented using enterprise modelling (EM), simulation modelling (SM) and causal loop modelling (CLM) techniques This paper reports on an application of the modelling approach to create related models of ‘process oriented roles’ and ‘candidate human systems’ so as to systemise matching of role requirements to resource systems attributes and to inform aspects of strategic and tactical decision making in an SME making composite bearings Keywords: dynamic producer unit; enterprise modelling; human systems modelling; process modelling; simulation modelling Introduction Manufacturing enterprises (MEs) are designed and engineered by people to achieve a wide range of goals Normally, MEs comprise structured and technology enabled systems of people who process physical and informational workflows to add value to specific products in timely and cost effective ways Globalisation of product markets, product customisation and shortening product lifetimes all impact in terms of increased and more frequent customer requirement changes (Krappe et al 2006) To cope with customer and other environmentally induced product dynamics, MEs operating in most industry sectors require enhanced competences (provided by human systems) and improved capabilities (from technical systems) to realise their process-oriented roles effectively and in a timely manner Generally, in MEs, people have collective and ongoing responsibility for: (1) deciding what an enterprise should do, (2) deciding how the enterprise should be structured and use available supporting technology to achieve those desired goals and constrain unwanted behaviours and (3) most of those product realising activities in a structured and technically-supported way as defined by (1) and (2) (Weston et al 2004.) Essentially, people-centred organisations such as MEs are complex in terms of their composition, structures and operations Consequently, effective and timely realisation of value adding activities *Corresponding author Email: R.H.Weston@lboro.ac.uk ISSN 0951-192X print/ISSN 1362-3052 online Ó 2010 Taylor & Francis DOI: 10.1080/09511920903527846 http://www.informaworld.com requires selection and matching of appropriate resource system competences and capabilities to ME requirements, on an ongoing basis, to change resource system solutions as ME requirement changes (Skyttner 2005, Swarz and DeRosa 2006) To achieve desired responses to ME requirement changes, enterprise systems and their human and technical components are often recomposed, reconfigured and reprogrammed As necessary, enterprise system change can give rise to emergent behaviours in MEs with resultant changes in operational scope and role requirements for people and their supporting technical systems To remain competitive however, MEs need constantly to develop their (human and technical) systems so that their competences, capabilities and capacities remain aligned to emergent business and environmental requirements With increased business fluidity comes a growing need for change capable manufacturing organisations, namely organisation that possess an ability to ‘recompose’, ‘reconfigure’ and ‘reprogram’ their system components rapidly and effectively In turn this requires improved understandings about how business and environmental change can be realised via suitable change to process structures and how this is related to required resource systems structures, attributes and behaviours (Zhen and Weston 2006, Weston et al 2007) Such understandings can be gained by studying ME ‘requirements’ and related ‘resource components 196 J Ajaefobi et al and their configurations’ in model views Models of MEs and their component parts can be captured in different views, at different levels of granularity, via alternative methods and by deploying appropriate modelling tools and modelling languages; thereby making it easier to represent, visualise, analyse, understand and possibly predict behaviours of viable configurations of enterprise components and to inform management decisions Many ‘method-based’ approaches to engineering change in MEs have been conceived and are becoming widely adopted by industry including: just in time and lean manufacture, agile manufacturing and postponement and mass customisation (Womack et al 1990, Womack and Jones 2003) However, it is observed that typically in industry the application of these change methods: behaviours and performances that continue to match explicitly defined but changing ME requirements (i) is ad hoc, constrained and piecemeal; (ii) supports qualitative, rather than quantitative analysis; (iii) does not facilitate an ongoing externalisation and reuse of organisational knowledge and data (iv) is techno-centric, with limited characterisation of impacts of people system roles, competences, behaviours and cultures However every day we all generate and use simple models of ourselves, our fellows, colleagues, companions, etc and of related situational impacts Research reported in this paper is concerned with understanding and characterising problems and constraints associated with modelling people in ME workplaces Definitive foci of reporting is on creating and using models of ‘human systems’ in relationship to common roles performed in MEs Here the term ‘human systems’ is used to infer either: competent individuals working systematically; loosely affiliated ‘workgroups’; or closely coupled teams of people deployed to interact in a structured work environment Generally though, resourcing value adding roles in MEs involves the use of (a) systems of competent people, (b) suitable technical systems and (c) combination of (a) and (b) The choice and deployment of the above resource system types depends on the nature of the requirements, i.e the nature of the work to be done, which often dictates the extent of human involvement and possible extent of automation, the costs of deploying particular resource system types and the nature of expected outcomes To enable human systems to function effectively in MEs, various organising structures that impact on their actions and behaviours are commonly deployed including: human organising structures (such as hierarchy, roles, responsibilities and authority) (Ashkenas 1995, Hendrick 1997), work organising structures (such as processing routes, batching and prioritising rules and ‘job’ and ‘task’ assignments) (Bennis 1996, Vernadat, 1996, Medsker and Campion 1997) and enterprise cultures (including corporate beliefs and values) Previous research findings by the present authors had observed a key differentiation between human and their technical system counterparts which centre on a common human ability to reflect on job performance outcomes and thereby as necessary to (1) develop new Hence the present authors propose the use of a model-based approach to underpinning manufacturing organisation design and change In conjunction with enterprise modelling (EM), causal loop modelling (CLM) and simulation modelling (SM) can usefully be deployed to achieve ME requirements specifications and capture, resource systems (solution) design, and the ongoing matching of emerging requirements to changes in solution design Since early 2000, the authors modelling research have specified developed and case tested systematic uses of state of the art EM, CLM and SM technologies to develop virtual models of large and small scale manufacturing systems Essentially, their approach unifies the use of: (a) decomposition principles defined by public domain EM methodologies especially CIMOSA (AMICE ESPRIT Consortium 1993), (b) causal and temporal relationship modelling notations, provided by CLM technologies, (c) discrete event and continuous SM tools to computer exercise behaviours of selected configurations of work loaded process segments and their underpinning resource systems and (d) mixed reality modelling based on the use of workflow modelling techniques that enable interaction and information interchange between simulation models and real resource systems In that context, this paper reports on progress made with respect to developing models of people in their manufacturing work places for the purpose of realising enhanced enterprise ‘Modelling people at work’ in MEs Evidently it is difficult for humans to model themselves for a number of reasons which include the following: (1) People are complex entities that generate various (individual and collective) behaviours that are often context dependent (Ajaefobi et al 2006); (2) People acting as modellers often have constrained understanding, knowledge and data about themselves, about the modelling context and about related causal impacts International Journal of Computer Integrated Manufacturing competences and/or (2) develop new structures; thereby modifying their behaviours and the behaviour of the entire system leading to improved performance (Weston et al 2003) People are therefore more flexible than most technical systems because they have the ability to reflect on (and develop) what they do, their work patterns and behavioural relationships To realise a prime objective of this research (i.e to systemise and support with models aspects of matching people to roles in the context of specific and changing ME work places), it is assumed that people and technical systems and the process-oriented roles they realise in manufacturing workplaces need to be modelled in a coherent manner Also it is assumed that in conformance with established general systems engineering practice, flexible ‘interconnection’ is required between developed models of process-oriented roles (that explicitly define work requirements) and developed models of solution configurations of human and technical resource components To identify common ME requirements (things MEs to create values for their customers), the authors have adopted a process view of MEs thereby modelling specific ME requirements as a specific network of dependent processes and their derivative roles Previous authors have classified and characterised processes commonly found in most MEs (Pandya 1997, Chatha et al 2007) A consensus view is that MEs typically deploy people and technical systems to realise the following process types: (1) processes that realise products and services for customers; (2) processes that ensure that those product and service realisations are well managed, such that they remains aligned to established business and manufacturing policies and strategic goals of the ME; and (3) processes that structure and enable ongoing change as the ME systematically renews and reconfigures itself, developing and implementing new strategies, policies and processes in response to external change While conceiving, specifying, developing, realizing and changing enterprise processes; people exercise different role types, namely interpersonal, informational, decisional and operational roles (Mintzberg 1989, Steers and Black 1994) The term role can further be described as: functions (tasks or activities) that need to be performed by role incumbents; identity created (or positions occupied) by incumbents in a social structure while performing the role and behaviours that people or stereotypical people can/will bring to roles and the management of role 197 dependencies (Wagner and Hollenbeck 1992, Steers and Black 1994, Ashfort 2000) The focus of this paper is on functional roles; which represent sets of functional activities and operations that are resourced by people and their supporting technology during product (service) realisation (Zaidat et al 2004) To effectively satisfy role requirements, i.e attainment of specific results required by the role through specific actions while maintaining or being consistent with the policies, procedures and conditions of the organisational environment (Boyatzis 1982), role incumbents need to bring to bear upon assigned roles work-related attributes especially competences The term competence is presume to mean those work-related attributes, including: natural traits (underlying attributes), acquired traits (knowledge, skills, education, training, experience, etc) and consistent performance outcomes for which a given resource system is known It follows that in resourcing ME roles, two types of competences are evident: (1) competences required by roles and (2) available competences possessed by potential role incumbents (Harzallah and Vernadat 2002) Selection and matching of people to roles involves ‘mapping’ between available competences and required competences Any such mapping is naturally constrained by factors such as: (a) Does the selected candidate system (person or people) have all the competences required by the role(s)?; (b) If the answer to (a) is affirmative, what is the capacity i.e (how much in case of quantifiable outcomes) will the system deliver in a given time frame?; (c) Can the selected candidate system cope with changing workload requirements, including changes related to production volumes and product variances?; (d) What aspects of the required competences are lacking in the solution provision?; (e) Can such deficiencies (as identified in (d)) be remedied by training or upgrading the supporting technology so as to enhance achievable performance of the deployed candidate system? Addressing the above questions requires modelling concepts with analytical and dynamic features to support data capture on requirements, representation and analysis of the available and required competences In subsequent sections, data capture and modelling of requirements from such captured data, modelling of candidate solution and how simulation modelling can be used to match specified requirements to alternative solutions in both structural and dynamic behavioural terms are discussed 198 J Ajaefobi et al Need for new modelling concepts To develop an approach to modelling human systems as reusable components of MEs, it was observed to be necessary to realise the specification and selection of a modelling method with capabilities to: (a) represent and abstract generic and specific ME requirements, in terms of the required network of processes used by any specific ME to realise products and common workflows through different segments of that process network Here it was envisaged that such a modelling method would facilitate and systemise the decomposition of process-oriented requirements (explicitly modelled as process segments and their needed workflows) into well-defined roles that are themselves can be decomposed to enable their explicit modelling at different levels of granularity; so that later the roles defined can be flexibly matched to ‘work centres’ that can be physically realised by suitable (human and technical) resource systems; (b) represent, decompose, abstract and structure models of human systems and their component elements in terms of work-related attributes, and especially functional competences; (c) facilitate qualitative and flexible matching between models of people (competences) and models of process oriented roles; (d) enable the selection and testing of alternative role-people ‘couplings’ in a simulation environment, so that quantitative comparisons can be made between the behaviours of alternative candidate role-people couples when they are subjected to historical and/or possible future changing ME requirements To satisfy the modelling requirements listed and to achieve the envisaged benefits, a suitable modelling approach needed to be specified and selected In principle, any of the state of the art EM methods that have been successfully tested and usefully applied in industry (such as CIMOSA, IDEF, PERA and ARIS) could have been chosen as the foundation modelling method However, the authors chose to deploy the open system architecture for computer integrated manufacturing (CIMOSA) (Kosanke 1995, Zelm et al 1995, Vernadat 1996) In spite of its known modelling strengths, CIMOSA has some notably weaknesses including (a) models developed using CIMOSA are essentially static and hence cannot be used to reason about changing requirements and the impact of such changes on selected and alternative resource systems and (b) CIMOSA does not have specific modelling constructs to represent human systems competences To address the limitations observed, a modelling framework incorporating role modelling concepts, competence modelling concepts and the use of SM modelling tools to instrument rolepeople couplings was proposed The unified modelling framework proposed uses CIMOSA (AMICE ESPRIT Consortium 1993) as the main modelling foundation but it has extended the modelling capabilities of CIMOSA (by exploiting its eclectic nature) to incorporate competence and SM modelling concepts Furthermore, to reflect the fact that human systems execute assigned roles while being supported by technical systems, the dynamic producer unit (DPU) concept previously proposed by the authors and their research colleges was employed to further systemise human systems modelling DPU modelling construct was proposed for the purpose of abstract description of enterprise resource units comprising people, machines, computers and or a structured combination of those; that is a reconfigurable, reusable and interoperable component of complex organisations such as a manufacturing enterprise (Weston et al 2009) Modelling methodology conceived The modelling framework shown in Table was proposed and developed to enable the capture of coherent models of ‘process-oriented roles’ and ‘human systems commonly found in specific manufacturing work contexts’ Section of this paper describes a case study application of this integrated approach to modelling human systems, as reusable components of manufacturing workplaces The case study company and its modelling requirements 5.1 Company background A composite bearing manufacturing company which makes to order a wide range of composite bearing products was chosen as the subject of a case study For reasons of confidentiality the authors will refer to the company as ComBear Ltd ComBear Ltd is a rapidly growing UK based SME with a customer base which extends beyond Europe ComBear manufactures different composite bearing products suitable for agricultural, marine, mechanical, pharmaceutical and food processing applications Essentially, all ComBear products are manufactured from reinforced plastic laminates composed of synthetic fabrics impregnated with resins and lubricant fillers Final products are delivered to customers in tube and sheet forms as well as fully finished components 199 International Journal of Computer Integrated Manufacturing Table The modelling stages of the ‘integrated approach to modelling human systems’ Purpose of each modelling stage and the approach to modelling adopted Stage Stage 1: ‘Context Modelling’ Stage 2: ‘Role Specification’ Stage 3: ‘Shortlisting of Candidate Systems’ Stage 4: ‘Modelling Dynamic Behaviours’ Stage 5: ‘Overall ME Function and Behaviour Modelling’ Example entries modelled Enterprise Modelling is used at this stage to decompose and graphically represent relatively enduring aspects of the specific network of Business Processes (BPs) used by the ME under study Stage modelling is focused on characterising properties of the process logic currently used by the subject ME Various groupings of enterprise activities (that constitute specific process segments and realise dependencies between process segments) are analysed with respect to their ‘functional’ and ‘behavioural’ requirements Various grouping rules (based on research findings from work design, human science and the process modularisation literature) are used to: identify and specify viable roles and role relationships for human systems A shortlist of candidate (human and technical or DPUs) resource system designs is established in terms of their potential to match: (1) competences and characters possessed by candidate human systems to (2) required functions and behaviours of viable roles and role relationships defined during stage Dynamic behaviour of process segments resourced by viable candidate human systems are modelled using CLM and SM technologies in a unified way The computer executable SMs so produced reuse specific structures and data about the business content and ME process network defined previously by the EM during stage modelling Thereby the SMs encode: (1) specific process logic (and embedded role requirements); (2) alternative attributions of short listed resource systems (to embedded roles and roles relationships) and (3) ME specific workflows through viable [process logic – resource systems] couples The purpose of so doing is to optimise the choice of resource system and methods of achieving workflow control based mainly on cost and lead-time criteria The predicted functional and behavioural properties of specific process segment-resource system couples are reviewed with reference to (needed and achievable) overall (functional and behavioural) properties of the ME Various measures from the literature on process performance and motivation can be utilised such as structural bearings, washers, wear rings, sphericals, wear pads, wear strips, rollers, and bushes Figure shows some of ComBear’s current product range 5.2 Reasons for modelling The objectives of the research funded by the UK’s EPSRC are described in detail in the EPSRC case for Network of BPs – used to realise products and services Segments of the process network – that must be resourced by suitable human (and technical) systems Process segments are modelled in terms of activity, information, material, control and exception flows Required ‘functional operations’ and ‘functional entities’ Viable role behaviours and role relationships Functional and behavioural specifications for viable roles Relative performance levels and costs of resources are tabulated A short list of candidate resource systems – with potential to realise specified roles Process routes, embedded roles, op times, etc Alternative assignments and organisational groupings of human resources to roles Work entry points, inter-arrival times, workflow controls Relative process segment behaviours Comparative quality measures Motivational factors and measures Overall throughput, value stream and cost measures support (Weston 2005) In the ComBear case more specific modelling goals were agreed with the company management and are listed as follows: (1) to document ComBear’s current network of processes formally, identifying who does what and with what and at what time; (2) for selected segments of ComBear’s current network of processes, to create computer 200 J Ajaefobi et al Figure (3) (4) (5) (6) ComBear product samples executable models that predict dynamic impacts on (current and possible future) process performances (e.g lead-times, throughput, bottlenecks, inventory, value generation and processing costs) of alternative work patterns and workloads; use documented and computer executable models to suggest potential beneficial changes to organisational structures, management philosophies and culture; use integrated models to identify where processes can become more lean or agile so that the firm can gain business benefits; use integrated models to suggest ways of improving the deployment and performance of (human and technical) resources; use integrated models to improve the planning, scheduling and control of workloads placed on primary ‘operational’ process segments To realise the stated modelling goals, the present authors adopted use of the integrated approach to modelling human systems, as reusable components of manufacturing workplaces; the modelling stages of which are described by Table 5.3 Formal documentation of ComBear’s process network Stage of the integrated modelling approach involves the capture of a specific CIMOSA conformant ComBear’s enterprise model The main purposes here were to: (1) provide the university team with means of externalising and reusing knowledge (formerly only distributed amongst the minds of various personnel) about the firm’s ‘operational’ (day to day), ‘tactical’ (sometimes daily or weekly and sometimes episodic) and ‘strategic’ (longer term) activity flows; (2) understand how ComBear’s operational activity flows enable the company to generate short term values and profit, whilst remaining competitive in the medium and longer term; (3) assist ComBear management and workforce to develop a big picture of the firm’s activity flows, so that individuals can identify impacts of their roles on the business performance of the company; and (4) enable the University team (as directed by ComBear managers) to use the knowledge externalised as base level of company specific data which enables the development of reusable computer executable models of selected activity flows (so as to realise goals (2) to (6)) Before embarking on CIMOSA modelling the present authors spent approximately man days (i) discussing activity flows with its managers and (ii) observing technical and manufacturing personnel perform their various roles Figure provides an overview of the full range of ComBear processes identified, namely: International Journal of Computer Integrated Manufacturing Figure 201 Key processes identified in ComBear Strategic processes that operate as required to envision, conceive and realise improved competitiveness through day to day management, leadership, financial and fiscal policy management and control, and adapting business rules and manufacturing policies in response to changing customer requirements, environmental and government regulations; Tactical processes resourced by ‘technical and mid-management teams’ that : (1) obtain and process customer orders, (2) develop process plans and job cards for product manufacture, (3) design new products or improve the design of existing products, (4) control and manage production materials, and (5) plan, schedule and control production operations; Operate processes that produce and deliver composite bearings and other products manufactured in three shops located within the production facility namely: (a) raw material processing shop, (b) sand and saw shop and (c) machine shop Although causal, temporal and structural links were observed between most of the key processes, the chosen focus of case study modelling was on the ‘operate processes’; instances of which need to regularly be resourced by human and technical resources, so that products are realised for customers and ComBear profit is generated Four types of modelling template were used to describe ComBear’s EM in a graphical form, namely: context diagrams, interaction diagrams, structure diagrams and activity diagrams Context Diagram: describe in overview how various ‘domain actors’ (i.e departmental sections and their supply chain partners) work together within the business context under study In this case the context modelled was the day to day production of composite bearings, of types and in quantities needed to satisfy orders from a variety of customers Figure shows an example context diagram captured in respect of ComBear Interaction Diagrams: describe various (relatively enduring) entity flows between processes (which in CIMOSA terms are modelled as Domain Processes (DPs) and their elemental Business Processes (BPs), Enterprise Activities (EAs) and Functional Operations (FOs)) Structure Diagrams: depict relatively enduring structural relationships between DPs, BPs and EAs This class of diagram is designed to code specific process decompositions consisting of ordered set of activities linked by precedence relationships, execution of which is triggered by events such as arrival of customer’s orders 202 J Ajaefobi et al Activity Diagrams: are used to depict specific process segments of concern, in terms of standard activity flows needed to create products Activity flows Figure ComBear top-level context diagram Figure ComBear interaction diagram were represented for each product family manufactured by ComBear Figures and are examples of structure and activity diagrams International Journal of Computer Integrated Manufacturing Figure 203 Activity diagram for making materials for round products Other kinds of modelled entity (such as events, information flows and precedence relationships) can be attributed to activity flows In Figure 5, the activity flows illustrated explicitly depict processing activities carried out in ComBear’s raw materials processing shop to create so called ‘round products’ 5.4 Roles identification and specification Having represented ComBear production processes using standard CIMOSA formalisms: domain processes (DPs), business processes (BPs), enterprise activities (EAs), and functional operations (FOs) (see Figures 3–5 (section 5)), step of the integrated modelling method was implemented by specifying and defining viable roles executed by ComBear resource system elements at various work centres The term role was used to refer to functions performed by role incumbents Three classes of role were identified in ComBear namely: (a) management roles representing those management and coordination functions, (b) technical/support roles that realise functions such as specifying process plans and procedures for products manufacture, planning and controlling production, designing new products; etc, and (c) operational roles, which represent direct products realising functions Matching human systems to operational roles is the focus of this paper and those roles in ComBear comprised: (1) raw material processing-related roles; that produced materials for making different ComBear products, (2) machining roles; that shape processed materials into components and where applicable assemble them into products, (3) sanding roles; that put finishing touches to outputs from machining roles and (4) packaging and delivery roles Figure shows some roles identified in raw materials processing and sanding operations In satisfying the raw materials processing functions for instance, five different roles (R1.11–R1.15), which are groupings of operations executed at different work centres, were identified When grouping operations into roles, considerations were made about: (1) precedence relationships between activities and their functional operations (bearing in mind closely coupled and non-separable activities); (2) dependencies between possible candidate ‘as-is’ and possible ‘to-be’ role incumbents (e.g availability of people and supporting machine resources) In general, roles specified were explicitly defined in terms of functional competences that potential candidate solutions need to bring to realise those roles It was observed that ComBear’s operators perform different roles while realising specified activity 282 M Sadeghi et al Habtemariam, E., Marwala, T., and Lagazio, M., 2005 Artificial intelligence for conflict management In: Proceedings of IEEE International Joint Conference on Neural Networks, Montreal, Canada, Vol 4, pp 2583– 2588 Hicks, D.L., 2004 Towards multi-granular versioning support for collaborative design applications In: Proceedings of the IEEE International Conference on Computer Supported Cooperative Work in Design, Xiamen, China, Vol 2, pp 108–112 IPPOP, 2008 Available from: http://projects.opencascade org/IPPOP, visited on July 2009 Liang, C and Guodong, J., 2006 Product modelling for multidisciplinary collaborative design International Journal of Advanced Manufacturing Technology, 30 (7–8), 589–600 Liu, D.T and Xu, X.W., 2001 A review of web-based product data management systems Computers in Industry, 44 (3), 251–262 Lottaz, C., Sam-Haroud, D., Faltings, B., and Smith, I., 1998 Constraint techniques for collaborative design In: Proceedings of the IEEE International Conference on Tools with Artificial Intelligence, Taipei, Taiwan, pp 34–41 Noel, F., Brissaud, D., and Tichkiewitch, S., 2003 Integrative design environment to improve collaboration between various experts CIRP Annals Manufacturing Technology, 52 (1), 109–112 Noel, F and Roucoules, L., 2008 The PPO design model with respect to digital enterprise technologies among product life cycle International Journal of Computer Integrated Manufacturing, 21 (2), 139–145 Okada, Y and Tanaka, Y., 1998 Collaborative environments of IntelligentBox for distributed 3D graphics applications The Visual Computer, 14 (4), 140–152 Petit, T., Regin, J.C., and Bessiere, C., 2000 Metaconstraints on violations for over constrained problems In: Proceedings of the IEEE International Conference on Tools with Artificial Intelligence, 358–365, Vancouver, BC, Canada Poon, W., 1994 Modelling a distributed database system In: Proceedings of the conference of the Centre for Advanced Studies on Collaborative research, Toronto, Ontario, Canada, p 54 PPR, 2008 Dassault Syste`mes’ unique Product, Process, Resource model for PLM solution Available from: http://www.3ds.com Ram, D and Vivekananda, N., 1997 Constraint metaobject: a new object model for distributed collaborative design IEEE Transactions on System, Man, and Cybernetics- part A: Systems and Humans, 27 (2), 208–221 Riboulet, V., Marin, P., Gowers, S., and Wurtz, F., 2004 A framework supporting collaborative optimisation for multi professional design teams In: Proceedings of the International Conference on Integrated Design and Manufacturing in Mechanical Engineering, Bath, UK Roller, D., Eck, O., and Dalakakis, S., 2002 Advanced database approach for cooperative product design Journal of Engineering Design, 13 (1), 49–61 Sadeghi, M., Noel, F., and Hadj-Hamou, K., 2009 Development of control mechanisms to support coherency of product model during cooperative design process Journal of Intelligent Manufacturing doi: 10.1007/ s10845-009-0237-2 Available online: 20 January 2009 STEP, 2008 Available from: http://www.steptools.com, visited on July 2009 Strom, R and Banavar, G., 1998 Concurrency control and view notification algorithms for collaborative replicated objects IEEE Transactions on Computers, 47 (4), 458– 471 Tichkiewitch, S., 1996 Specification on integrated design methodology using a multi-view product model In: Proceedings of the Joint Conference on ASME Engineering Systems Design and Analysis, Montpellier, France, pp 101–108 Wang, Y and Nnaji, B.O., 2004 UL-PML: constraintenabled distributed product data model International Journal of Production Research, 42 (17), 3743–3763 International Journal of Computer Integrated Manufacturing Vol 23, No 3, March 2010, 283–296 Metrology-assisted robotic processing of aerospace applications Nirosh Jayaweeraa* and Phil Webbb a Department of Mechanical, Materials and Manufacturing Engineering, University of Nottingham, UK; bDepartment of Systems Engineering and Human Factors, University of Cranfield, UK (Received 15 May 2008; final version received 16 November 2009) Modern aerospace structures tend to consist of large numbers of geometrically complex structural components which by their very nature tend to suffer from significant levels of physical distortion and are thus difficult to assemble One solution to this problem is to use large complex jigs which physically control the shape of the parts These jigs are usually loaded using either direct manual labour or manual labour assisted by cranes or lifting devices The use of manual operations represents a significant health and safety risk and increased likelihood of damaging components during assembly The application of automation in the processing of such structures has so far been confined to small product specific cells owing to difficulties in pre-defining and fixing the exact geometry and positioning of parts within the work volume The use of specially designed jigs, fixtures and aids such as drilling templates can be adapted to support automation but are expensive, have long manufacturing lead times and cannot be economically modified for use on other aircraft types The paper proposes a solution to the above problems using standard industrial robots and an advanced control and non-contact metrology systems The developed methodology is generic and has been evaluated and demonstrated in a number of different applications The application described in this paper is the assembly of regional jet fuselage panels Results are presented along with an analysis of the accuracy and repeatability of the system and its elements Keywords: airframe manufacture; automated assembly; non-contact metrology; robotics Introduction Aircraft fuselages are constructed from a series of subassemblies referred to as panels (Figure 1) For a typical regional jet the assembly process for these involves precisely positioning structural components (frames and stringers) onto the outer skin and then drilling, countersinking, sealing and riveting them together Conventionally, dedicated tooling is used for the assembly of the components to ensure that the final assembly meets design requirements These tools are expensive, require long lead times to manufacture and cannot usually be economically modified for use on other aircraft types The majority of the existing automated solutions are based around large dedicated fixed systems, often called ‘monuments’ owing to their large size Typical examples are, the Automated Spar Assembly Tool (ASAT) (Hartmann 1998), Automated Spar Drilling and Fettling System (ASDAFS) (Siegel et al 2003), Gear Rib Automatic Wing Drilling Equipment (GRAWDE) (Hogan et al 2003), Horizontal Automated Wing Drilling Equipment (HAWDE) (Calawa et al 2004), Gemcor’s All Electric Rotary Head fastening system (Mangus and Grenier 2003), and the BRO¨TJE iGantry Riveter (Schwarze and Mayla¨nder 2003) These systems are inflexible and have a high capital cost as well as long lead-time and can lead to capital bottlenecks One way of introducing flexibility is by incorporating robots Background literature suggested that there has been limited work on the use of flexible-robot-based systems in the aerospace assembly applications However, a number of robotic systems have been developed with varying degrees of success These include the ICAM system for drilling and routing system (Groover et al 1986), Grumman Aerospace Corporation Integrated Robotic Cell (IRC) (McCluskey 1984) and dual robot assembly cell (Barone 1989), Saab-Scania Aircraft robotised production cell (Madesater 1993), Airbus floor-grid structure assembly cell (Roche 1995), Dassault Aviation assembly cells (Dacosta 1992, 1996) and ONCE (One-sided Cell End-effector) robotic drilling system (DeVlieg et al 2002) All of these systems required accurate positioning and fixturing of the components within the cell However, a number of articles describe robotic applications in the aerospace industry and demonstrate that there is a significant amount of interest in the adoption of external metrology systems to improve the *Corresponding author Email: nirosh.jayaweera@nottingham.ac.uk ISSN 0951-192X print/ISSN 1362-3052 online Ó 2010 Taylor & Francis DOI: 10.1080/09511920903529255 http://www.informaworld.com 284 Figure N Jayaweera and P Webb Regional jet fuselage panel capability of robotic systems Most of the existing research such as the Automated Wing Box Assembly (AWBA) project (Rooks 2001), (Anderson 2002), the TI2 system for the manufacturing of airframe subassemblies (Eastwood et al 2001, 2003), (Webb and Eastwood 2004) and metrology assisted robot calibration (Kihlman and Loser 2003, Kihlman et al 2004) utilise metrology systems specifically laser tracker, photogrammetry and vision systems to provide vision-guided assembly and manufacturing operations These installations have all supported greater levels of automation but they all required sophisticated and expensive software and hardware for their implementation including training and familiarity with the technique Each of the metrology systems has its own particular merits and limitations The majority of the work published on robotic installations in the aerospace industry is based on drilling and riveting applications for aero-structure sub-assemblies and very little work has been reported on automated assembly The research described in this paper aims to provide a low-cost system for the automated assembly of fuselage panels by integrating an off-the-shelf laser metrology system, standard lowcost industrial robot, re-configurable end-effector and a mathematical ‘best-fit’ algorithm The metrology system used in this research is a low-cost laser seam finder with a standard interface to a robot controller The work described in this paper is a part of an ongoing research project to develop an Adaptive Robotic Assembly Manufacturing Cell (ADRAM) (Webb et al 2005, 2006) at the University of Nottingham Methodology A feature-based assembly technique was adopted for the assembly of the components described in this paper since a feature contains much more information than just geometry During the assembly process accurately positioned features such as part to part holes were used to provide alignment for individual components within the structure A metrology system was used to locate the component and measure the position of the assembly features relative to the robot to identify the actual and required position of the component before assembly Similarly, the metrology system was used to locate features on the counter-part The acquired data were processed using a mathematical algorithm to calculate the relative component positions required for optimal assembly The data can also be used to check for gross distortion of the components and to reject those outside the specification limits The resulting metrology data could also be stored and analysed to produce quality assurance and process verification information The system developed calculates the actual relationship between the robot TCP (Tool Centre Point) and component to be measured relative to the robot base coordinate system and uses this relationship to generate the required robot programs by manipulating a pre-programmed path This optimises the robot used since it and the part and counter part are locally calibrated to each other The robot also operates in a relatively small working area relative to the workpiece thus ensuring that maximum repeatability and accuracy are achieved If the part cannot be placed in the exact position required, due to misalignment of the measured features, the software performs a ‘best-fit’ operation relative to features on its counterpart, along with an overall tolerance check, to ensure that the part can be assembled within the required specifications To increase the flexibility of the methodology it has been realised using open cell architecture The cell can consist of any number of robots or resources such as metrology systems and end effectors linked together using Ethernet The core of the cell is a software PLC and a mathematical processing unit built around Matlab This allows the system to be re-configured for different applications The next section describes the design and implementation of robotic cell for the assembly of regional jet fuselage panels System design and development To realise this aim a number of core technologies have been developed and combined to form an integrated cell These include: (1) Non-contact metrology: Advanced noncontact metrology systems to allow the realtime measurement of component position and International Journal of Computer Integrated Manufacturing geometry Metrology data can also be stored and analysed to produce quality assurance and process verification information (2) Mathematical toolbox: A mathematical toolbox has been developed to allow the ‘best-fit’ assembly processes to be performed (3) End-effector design: Robot end-effectors have been developed to handle a number of different types of component The core of the system is the mathematical toolbox, this processes the data from the metrology system to find the actual position and orientation of the part and counterpart and then computes the optimal placement position for the assembly operation The positional data are obtained by scanning part features using a metrology system mounted on an industrial robot In addition, these algorithms are also capable of compensating for some of the error due to sensor uncertainty and inherent robot errors by incorporating additional measurement points and constraints in the programs Moreover, these algorithm are used to analyse mating hole misalignment after the assembly The mathematical tool box also provides a user interface (Figure 2) and a database for the storage of measurement and process data for process control and monitoring purpose as well as co-ordinating the transfer of data to and from the robot controller Figure System user interface 3.1 285 Cell construction To demonstrate the developed technologies practically a demonstrator cell was constructed at the University of Nottingham as shown in Figure Care was taken when developing the cell to ensure that only generic ‘of the shelf’ systems were used to so that the resulting cell would be low cost and robust A Comau S2 industrial robot was chosen as the manipulator and equipped with a purpose designed reconfigurable end-effector The Comau S2 robot has an articulated, six axis anthropomorphic structure AC brushless motors control the movement of all axes with internal resolvers and brakes Motion is transmitted through gearboxes except for axis 6, which uses a ‘harmonic drive’ Its maximum wrist load capacity is 78.45 N (8 kg) and electrical and pneumatic services are available at the forearm The repeatability of this robot is + 0.1 mm The robot is controlled by an open version of the C3G controller unit with a communication link to a PC The use of a reconfigurable end-effector (see Figure 4) allows the system to handle a number of different types of component In this application it was designed to handle two different sizes of stringer as well as curved frames The developed end-effector incorporates both a gripping and sensing system and was designed to handle all of the stringers and frames of the Bombardier CRJ 700 commercial aircraft fuselage 286 Figure Figure N Jayaweera and P Webb System architecture Reconfigurable end-effector panel However, it can also be used for other fuselage panel assemblies with slight modifications to the endeffector The end-effector control system consisted of a digital input-output (I/O) board, Relays, DC power supply unit, solenoid valves, and an industrial personal computer Pneumatic cylinders were used in the endeffector to reconfigure it according to the component type The measurement system chosen was a MTF (Meta Technology Finder) Seam Finder Meta MTF single stripe sensor is relatively low cost, lightweight and can be directly mounted on an industrial robot The laser sensor contains a charge coupled device (CCD) camera and a single laser diode The laser acts as a structured light source and produces a stripe on the surface under the sensor From the shape of the stripe seen on the workpiece, The MTF control unit processes the picture from the camera and the software and uses the settings from the seam type to divide the stripe into lines that form that seam From the position of the lines it can detect the location of the seam Measurements from the picture are then converted into measurements in millimetres to give the seam’s position under the sensor This conversion is performed using calibration data stored in the sensor head The settings for different seams are stored under different seam numbers and the seam number is shown on the control unit display If the sensor detects an error it will display a code number for the particular error The tool box was realised using Matlab software Communications between the laser seam finder and the robot controller were established using RS232 serial links Owing to the nature of the system a large amount of data needs to be handled and stored for further analysis To support this requirement a database was developed using Microsoft Access software It consisted of nine tables of related data each one containing its own set of data The database was used to store robot coordinates, laser offset data, and Matlab generated data These stored data can then be used for quality assurance and process verification information International Journal of Computer Integrated Manufacturing 3.2 Cell operation After initial validation and testing using simple flat sheet components (Jayaweera and Webb 2007a) the system was ready for testing on real production components A 1.2 m stringer was chosen for the main validation test as this was considered to be of a reasonable size and mass and had distinct datum features In the application described here three ‘part to part’ holes were used as the reference features on both part and counterpart These are accurately drilled tooling holes that are normally used to support jigless manual assembly In conventional assembly an operator would line them up and insert temporary fasteners into them to hold the parts in place whilst they are drilled and riveted The temporary fasteners are then removed and permanent ones installed Although they are used as the data in this application, any convenient feature can be used The operating procedure is illustrated in the flow chart shown in Figure In the case of the stringer shown in Figure 6, which has three ‘part-to-part’ holes, Figure Method for measuring a hole feature Figure Test stringer 287 the part was placed in an approximate pick up location on a flat table and the holes (C1, C2, C3) and edges (P1, P2, P3, P4) were scanned to give their exact position relative to the robot’s TCP (Tool Centre Point) The toolbox then uses the measurements to calculate the optimum pick up point before driving the robot to that position and collecting the part, Step in Figure The measurements taken are shown in Figure and consist of two lines cutting across the tooling hole at 908 to each other Equations (1) and (2) can then be used to calculate the diameter of the hole The centre may also be calculated from the intersection of the centre points of the measured gaps between each hole’s edges The diameter (D) of the measured hole can be calculated from Equations (1) or (2): pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D ¼ GH2 þ GE2 ð1Þ D¼2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi CH2 þ AC2 ð2Þ The process is repeated for each tooling hole on the part and thus a series of reference points can be calculated for the component Next the robot moves to the panel (counter part) and the corresponding part-to-part hole positions are measured these measurements are then used to calculate the optimum drop off position before the part is placed on the fuselage panel, Step in Figure For a simple rectangular part such as a stringer a pair of best-fit lines is calculated using a least squares method The resulting lines are then matched and the placement co-ordinates converted to the robot’s coordinate system The actual best-fit algorithm used can be tailored to the geometry of the components being assembled The final step is a quality check in which the system re-measures the position of the stringer tooling holes for identification of the true assembly positions to verify that the placement has been successful 288 Figure N Jayaweera and P Webb Operating flow chart Error analysis This section briefly covers the source of errors, which may be present in the robot and laser measuring system There are four main types of errors (Greenway 2000) that can occur during robotic applications These are geometric, dynamic, thermal, and system, each category has unique causes Geometrical errors in robot geometry are inevitable as inherent inaccuracies occur during the manufacturing and assembly of individual structural components of the robot Dynamic loads associated with robot movement create dynamic errors, which result from both the internal loading and structural resonance excited by the motion The expansive properties of the materials used to make the robot causes thermal errors The effect of temperature changes on a serial robotic structure will amplify the effect of any deviation throughout the structure Systematic errors are caused by deficiencies in the physical and computational model used to process measurements Errors can be eliminated from the data during post-processing by applying correction values to the measurements These errors are caused by improper calibration, sensor inaccuracies, drive train backlash, and poorly tuned 289 International Journal of Computer Integrated Manufacturing servos These errors can be managed once they are identified 4.1 Measurement of performance In addition to the practical evaluation carried out in the previous sections the two main elements of the system (Robot and Laser Measurement systems) were evaluated and their actual performance compared with that stated by the manufacturers Experimental trials included laser and robot combined accuracy and repeatability, analysis of Z height variation when rotating 90 degrees, robot accuracy and repeatability, and effect of different surface conditions on laser measurements 4.2 line for the average data were plotted using least square curve fitting method It can be seen from this that magnitude of the robot error has a positive correlation (r ¼ 0.99) with the distance measured Laser and robot combined repeatability An experiment was devised to analyse both laser and robot repeatability when measuring a hole’s centre coordinates The experiment was repeated ten times and the results obtained analysed The results showed that the robot and laser combined linear repeatability are equal to 0.27 mm Following this, an experiment was carried out to analyse the robot and laser combined repeatability for measuring between a pair of part-to-part holes The robot was used to measure the first hole centre and then the second hole and thus calculate the distance between the two holes Results revealed that for two holes robot and laser combined repeatability was equal to 0.46 mm This was approximately double the repeatability compared with the earlier case 4.3 Investigation of robot accuracy This experiment was used to analyse the robot’s accuracy There were two sets of experiments carried out to compare the robot measured distance moved against the actual distance moved The first set of experiment measured the accuracy for a single movement whilst the second one was used to analyse the accuracy under auto-run for repeated movements The experimental set-up (as shown in Figure 8) consisted of a dial-gauge mounted on the robot and a length gauge This experiment was conducted for four sets of length gauges The results were further analysed (average values) and are presented in Table to give a summary of the robot-measured distance against the actual distance and the corresponding robotic error for the manual method To find the correlation between robot error (robot accuracy) and the measured actual distance a graph was plotted, see Figure Variation bars and a trend Figure Experimental set-up for analysing robot accuracy Table Analysis of robot error compared with distance measured Actual distance (mm) 300 500 700 1000 Figure Avg robot measurement (mm) Robot error (mm) 299.485 499.21 698.76 998.32 0.515 0.79 1.24 1.68 Variation of robot error with actual distance 290 N Jayaweera and P Webb According to the results, the robot indicates a value of 70.17 mm for every 100 mm of actual movement 4.4 Investigation of robot repeatability This experiment was set-up to analyse robot repeatability The experimental set-up was similar to that used for accuracy In this experiment the robot was repeatedly moved a known distance and a dial gauge used to measure the deviation This experiment was carried out for four different distances and data were recorded According to above repeatability experiments the overall repeatability of the S2 robot was 0.02 mm 4.5 Robot and laser combined accuracy measurements This experiment was set up to compare the distanced measured using the Laser mounted on the robot with a known distance The set-up consisted of the laser attached to the S2 robot and a length gauge mounted on a bench To perform the experiment the robot was programmed to move to one edge of length gauge and measure the position of the edge relative to the robot TCP using the laser The robot was again moved to next edge of the length gauge its position measured The difference between these two readings gives a measurement of the length gauge using a combination of the robot and laser Four sets of experimental trials were conducted for different size of length gauges The effect of the laser value when measuring a distance is illustrated in Table Normally, from previous experiments, the robot measured value would be less than actual value When using a laser sensor to measure the distance, it increases approximately 0.80 mm than the robot measured value 4.7 Reliability of the laser measurement system The accuracy of the sensing system is dependant on target reflectance, temperature, ambient light, and correct standoff height chosen Repeatability is a measure of sensor stability over time and tends to worsen through long-term drift in the components The main aim of this experiment is to analyse the laser measurements on different surface conditions The experimental set up consisted of a Laser seam finder, S2 robot, and five test-pieces The test-pieces from right to left are matt surface, shiny surface, white surface, green surface, and matt black surface as shown in Figure 10 Laser measurements were taken for each hole at a fixed stand off value and side value The experiment was repeated three times and the images captured from a video monitor the resulting images are shown in Figure 11 The laser scanner relies on the laser being reflected back from the object surface and the strength of the returning signal is influenced by the reflective properties of the surface It has been observed that different levels of reflectivity can result in systematic errors in the laser readings In this case the white and green painted surfaces gave the strongest reflections whereas reflection was weak from the black and the shiny surfaces The matt surface gives a better reading but it was less accurate than green and white In summary the overall repeatability that can be expected from the system is 0.27 mm when measuring a hole centre from a known point 4.8 System accuracy and repeatability The accuracy of the robot can be defined as how closely the robot TCP can be programmed to hit a 4.6 Analysis of Z-height variation when rotating 90 degrees This experiment was set-up to analyse any Z variation when rotating 908 degrees about that axis There was a 0.42 mm variation when robot rotating 90 degrees This was purely attributable to robot error Table Effect of laser value when measuring a distance Actual Length (mm) 300 500 700 1000 Average Robot þ Laser (mm) Robot (mm) Laser (mm) 300.27 500.06 699.64 999.02 299.485 499.21 698.76 998.32 0.785 0.85 0.88 0.70 0.80 Figure 10 Laser measurements on different surfaces International Journal of Computer Integrated Manufacturing qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi PASystem ¼ Æ PA2Robot þ PA2Sensor 291 ð3Þ Therefore, the system measured positional accuracies are: In the horizontal plane ¼ +0.180 mm In the vertical plane ¼ +0.315 mm This compares with the system measured positional accuracy for a centre of a hole of 0.27 mm Experimental results demonstrated that the system positional accuracy is within the manufacturer’s quoted accuracy According to the experimental results, the repeatability of S2 robot is 0.02mm and the absolute accuracy of S2 robot is 70.17 mm for every 100 mm actual measurement (Jayaweera and Webb 2007b) Figure 11 Laser measurements on different surfaces desired point Accuracy can be affected by both the speed of movement and the weight of the payload The repeatability is concerned with the ability to position the TCP at a point in space that has previously been taught to the robot Good repeatability is more desirable than accuracy as inaccuracies are easier to correct These especially true if the inaccuracies are consistent for all moves If inaccuracy is consistent for all movements of robot, then the programmer can compensate for this error Adjustment of poor repeatability is more difficult than consistent one Repeatability can change with use, especially when robot performs the same task day after day This is because mechanical components are subject to wear, thus increasing mechanical inaccuracies, this reduces the repeatability The individual sources of error can be combined to give an estimate of the overall system error The magnitudes of the empirically obtained accuracies for the system components are show below: Robot positional accuracy PARobot ¼ +0.1 mm Horizontal measurement accuracy of the sensor head (PASensor) ¼ +0.15 mm Vertical measurement accuracy of the sensor head (PASensor) ¼ +0.30 mm The total system positional accuracy (PASystem) can then be calculated from: Cell Testing and Evaluation As already mentioned in section 3.2, after the placement has been completed the placement position can be re-measured and the acquired data processed to show the relative alignment between the tooling holes on the part and counter part To verify the accuracy and repeatability of the system a stringer was placed ten times and the robot positions and laser offset values were recorded The data obtained from the assembly were then processed and the results are shown in Table The notations used in the table are illustrated in Figure 12 This shows a typical post assembly arrangement, in this diagram the holes are deliberately shown out of alignment for clarity The distance d1, d2, d3 of part and panel were calculated using Equations (4) to (9) to give the actual distance between the hole centres on both the part and counterpart The tooling hole misalignments cd1, cd2 and cd3 were calculated using Equations (10) to (12) to show the distance between the measured tooling hole centres on the part and counterpart Angle y was calculated from Equation (13) to give the angle between the respective best-fit lines through the centres of the tooling holes on the part and counterpart Centre offset values were obtained from Equations (14) and (15), these are the actual relationship between respective tooling hole centres in Y and Z co-ordinates The measurements shown in the diagram are as follows: A, B, C: Centre of part tooling holes (yA, zA), (yB, zB) and (yC, zC) D, E, F: Centre of panel tooling holes (yD, zD), (yE, zE) and (yF, zF) G, H: Mid points of part and panel tooling holes (yG, zG) and (yH, zH) cd1: First set of mating hole axes difference cd2: Second set of mating hole axes difference 292 N Jayaweera and P Webb Table Experimental measurements and calculated positions for ten stringer placements d1 (mm) d2 (mm) d3 (mm) cd (mm) Panel Part Panel Part Panel Part Y Z Angle y (deg) cd1 (mm) cd2 (mm) cd3 (mm) 206.9 206.9 206.9 207.2 207.2 207.2 207.2 207.2 207.2 207.2 206.9 207.1 207.0 206.9 206.9 207.1 207.1 207.0 207.0 207.0 230.0 230.1 229.9 229.6 229.8 229.8 229.8 229.8 229.8 229.8 230.1 229.9 229.9 229.7 230.0 229.9 229.9 229.9 229.9 229.5 436.9 437.0 436.9 436.8 437.0 437.0 437.0 437.0 437.0 437.0 437.0 437.0 437.0 436.6 436.9 437.0 436.9 436.9 436.9 436.9 70.4 70.3 0.3 70.4 70.1 70.3 70.3 0.1 0.1 0.1 70.2 70.3 0.3 70.3 70.4 70.3 70.3 70.1 70.4 70.4 0.03 0.03 0.04 0.03 0.02 0.04 0.03 0.06 0.02 70.01 0.41 0.36 0.21 0.60 0.35 0.42 0.49 0.12 0.31 0.36 0.42 0.55 0.51 0.26 0.43 0.30 0.30 0.28 0.50 0.46 0.55 0.53 0.56 0.55 0.46 0.48 0.59 0.42 0.43 0.46 Figure 12 Generic view of final hole misalignment cd3: Third set of mating hole axes difference d1: Tooling hole distance between 1st and 2nd holes d2: Tooling hole distance between 2nd and 3rd holes d3: Tooling hole distance between 1st and 3rd holes cd: Centre point offset Y and Z values y: Angular deviation between part and panel hole axes Formulae for calculating distances d1, d2, d3 of panel: d1 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðyD À yE Þ2 þðzD À zE Þ2 ð7Þ d2 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðyE À yF Þ2 þðzE À zF Þ2 ð8Þ d3 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðyD À yF Þ2 þðzD À zF Þ2 ð9Þ Formulae for calculating distances d1, d2, d3 of part: Formulae for calculating distances cd1, cd2, and cd3 are: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d1 ¼ ðyA À yB Þ2 þðzA À zB Þ2 ð4Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cd1 ¼ ðyA À yD Þ2 þðzA À zD Þ2 ð10Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d2 ¼ ðyB À yC Þ2 þðzB À zC Þ2 ð5Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cd2 ¼ ðyB À yE Þ2 þðzB À zE Þ2 ð11Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi d3 ¼ ðyA À yC Þ2 þðzA À zC Þ2 ð6Þ cd3 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð yC À yF Þ þ ð z C À z F Þ ð12Þ International Journal of Computer Integrated Manufacturing The angle y between two ‘best-fit’ lines of gradient m1 and m2 can be calculated from Equation (12) À1 y ¼ tan  m1 À m2 þ m1 Ã m2  ð13Þ Centre offset values Y and Z can be calculated using Equations (13) and (14) Y ¼ yG À yH ð14Þ Z ¼ zG À zH ð15Þ Figure 13 shows a graphical representation of part-topart hole misalignment after the robotic assembly process After analysis the experimental results revealed that the aero-structure components could be assembled within 0.6mm of the required position and this is within a typical aerospace production tolerance of 1.588 mm (Jayaweera and Webb 2007a) which is based on whether the ‘part to part’ holes can be drilled out 293 to full size without leaving any trace of the original holes A final visual check was also made which proved that in all cases ‘daylight’ could be seen through the holes 5.1 Assembly of components with misaligned tooling holes To test the ‘best-fit’ capability of the control system stringers were produced with deliberately misaligned part-to-part holes and the resulting alignment measured In this experiment part-to-part hole centres were misaligned in such a way that the mid hole purposely deviates 0.5 mm from the centre line as illustrated in Figure 14 Five experimental trials were performed to test the ‘best-fit’ assembly technique each following the operational procedure described in Figure The placement accuracy was assessed by measuring the offset between the centre of the part-to-part holes on the part and counter part The results obtained from the experimental trials are tabulated as shown in Table Figure 13 Hole misalignment after assembly Figure 14 CAD drawing of a flexible part with misaligned tooling holes 294 N Jayaweera and P Webb of components The metrology system used in this research was originally designed to measure the position of seam in welding applications In this research the laser seam finder was used to measure holes and edges in structural components In this case the laser stripe is placed across the hole and a laseroffset reading taken The sensor mounted end effector is then rotated 90 degrees to take a second measurement The experiment results revealed that when the sensor head rotates 90 degrees, the sensor stand off height was varying by þ0.42 mm This gives a less accurate reading for a hole centre measurement To solve this problem a non-contact laser cross sensor or any other system that does not require rotation can be used A laser cross sensor consists of a pair of laser stripes at right angles to each other and does not require rotation of the sensor to find a centre hole Alternatively two single stripe sensors can be used to find a hole without 90 degrees rotation by installing these sensors perpendicular to each other Moreover, handling of parts using vacuum gripping may reduce the assembly accuracy as vacuum Figure 15 represents the mating hole misalignment after the assembly operation According to results, part-to-part assembly with three tooling holes has been successfully demonstrated for assembling of flexible parts within typical aerospace production tolerance limit This indicates that the process is suitable for real stringer placement operation since part-to-part alignment is sufficient to leave a single hole when the parts are drilled through to insert a rivet Discussions A number of factors could be the result of inaccuracies in the assembly These include the accuracy and stability of the end-effector, accuracy of the laser seam finder (refer Section 4.1), robot errors such as geometric, dynamic, thermal, and system errors, setting of sensor and end-effector TCPs, errors in component pick-up and drop-off orientations, vibration, surface condition of features, manufacturing tolerance on parts and counter parts and distortion Table Results of best-fit assembly with tooling holes are not aligned d1 (mm) d2 (mm) d3 (mm) cd (mm) Panel Part Panel Part Panel Part Y Z Angle y (deg) cd1 (mm) cd2 (mm) cd3 (mm) 174.7 175.0 174.9 175.0 174.9 174.0 174.1 174.0 174.0 173.9 275.3 275.3 275.2 275.2 275.2 275.5 274.4 275.4 274.3 274.4 450.0 450.3 450.1 450.2 450.1 449.5 449.5 449.4 449.3 449.3 0.0 70.3 70.1 70.3 70.3 0.7 70.1 70.1 0.2 0.2 70.08 70.05 70.02 70.11 70.12 0.5 0.7 0.5 0.8 0.7 0.6 0.6 0.6 0.2 0.5 1.0 0.3 0.3 0.7 0.8 Figure 15 Mating hole misalignments after assembly International Journal of Computer Integrated Manufacturing cups have rotational and translational flexibility during operation It was also necessary to set up the laser TCPs and end-effector TCP before starting the experimental work Setting up of the TCP is not 100% accurate and it may contribute to inaccuracy in the individual measurements This can be overcome by using more precise measuring systems such as a Laser tracker to calibrate the robot TCP However, this system is only used to measure the final assembly position, along with an overall tolerance check, to ensure that the part can be assembled within the required specifications It is important to emphasis that the results obtained from the experiment were not absolute placement accuracy but are a combination of this and the results of the best-fit process Conclusions An automated assembly method has been successfully demonstrated for the assembly of typical aircraft parts using a standard industrial robot integrated with a laser stripe sensor Experiments have been performed to prove that it is possible to automatically assemble parts, in this case stringers, using the system within aerospace production tolerances The post assembly measurement results were all within 0.6 mm of each other and suggesting that the best-fit process had worked well as there was no particular difference in the average degree of misalignment between respective pairs of holes The developed system was highly flexible and re-configurable due to the reduced reliance on fixtures and the use of laser guided robots Acknowledgements This work was supported by the United Kingdom Engineering and Physical Sciences Research Council and Bombardier Aerospace Ltd References Anderson, J., 2002 Advanced robotic applications in aircraft component assembly In: Proceedings of Mechatronics, University of Twente, Netherlands Barone, P.A., 1989 Techniques for developing a multi-robot cell for production processing of complex airframe assemblies In: Society of Manufacturing Engineers Technical paper, Dearborn, Michigan Calawa, R., Smith, S., Moore, I., and Jackson, T., 2004 HAWDE five axis wing surface drilling machine In: Proceedings of Aerospace Manufacturing and Automated Fastening Conference & Exhibition 2004, St Louis, Missouri, USA, SAE Ref: 2004-01-2806 DaCosta, S., 1992 A new type of robotic cell for assembly in aeronautics In: Proceedings of Aerospace Automated Fastening Conference and Exposition 1992, Bellevue, Washington, USA, Society of Manufacturing Engineers DaCosta, S., 1996 Dassault adaptive cells Industrial Robot, 23 (1), 34–40 295 DeVlieg, R., Sitton, K., Feikert, E., and Inman, J., 2002 ONCE (one-sided cell end effector) robotic drilling system In: Proceedings of Automated Fastening Conference and Exhibition 2002, Chester, SAE Ref: 2002-012626 Eastwood, S.J., Webb, P., and McKeown, C., 2001 Practical evaluation of the TI2 manufacturing system for the aerospace industry In: Proceedings of the 11th International Conference on Flexible Automation and Intelligent Manufacturing FAIM, Dublin, Dublin City University, 881–890 Eastwood, S.J., Webb, P., and McKeown, C., 2003 The use of the TI2 manufacturing system on a double-curvature aerospace panel Journal of Engineering Manufacture, Part B, 217, 849–855 Greenway, B., 2000 Robot accuracy Industrial robot: An International Journal, 27, 257–265 Groover, M.P., Weiss, M., Nagel, R.N., and Odrev, N.G., 1986 Industrial Robotics – Technology, Programming, and Applications New York: McGraw-Hill, ISBN: 0-07024989-X Hartmann, J., 1998 ASAT4- Enhanced flexibility for the C-17 In: Proceedings of Aerofast Conference and Exposition, Long Beach, CA, USA, SAE Ref: 982126 Hogan, S., Hartmann, J., Thayer, B., Brown, J., Moore, I., and Rowe, J., 2003 Burrows automated wing drilling system for the A380–GRAWDE In: Proceedings of Automated Fastening Conference and Exposition 2003, Montreal, Canada, SAE Ref: 2003-01-2940 Jayaweera, N and Webb, P., 2007a Adaptive robotic assembly of compliant aero-structure components Robotics and Computer-Integrated Manufacturing, 23 (2), 180–194 Jayaweera, N and Webb, P., 2007b Automated assembly of fuselage skin panels Assembly Automation, 27 (4), 343–355 Kihlman, H and Loser, R., 2003 6DOF metrologyintegrated robot control In: Proceedings Automated Fastening Conference and Exposition, Montreal, Canada, SAE Ref: 2003-01-2961 Kihlman, H., Loser, R., and Cooke, A., 2004 Metrologyintegrated industrial robots – calibration, implementation and testing In: Proceeding of the 35th International Symposium on Robotics, Paris, France Madesater, A., 1993 Robot-based machine tool for the aerospace industry Industrial Robot, 20 (3), 11–13 Mangus, W and Grenier, D., 2003 All electric rotary head fasting system at Bombardier/Canadair In: Proceedings of Automated Fastening Conference and Exposition, Montreal, Canada, SAE Ref: 2003-012945 McCluskey, R.J., 1984 Robotic system cuts airplane parts American Machinist Cleveland, Ohio: Penton Media Inc., 71–73 Roche, N.R., 1995 Automatic riveting cell for commercial aircraft floor grid assembly Aerospace Engineering, 15 (1), 7–10 Rooks, B., 2001 Automatic wing box assembly developments Industrial Robot, 28 (4), 297–302 Schwarze, K and Mayla¨nder, S., 2003 BRO¨TJE-iGantry automated fastening concept allowing a progressive production flow In: Proceedings of Automated Fastening Conference and Exposition, Montreal, Canada, SAE Ref: 2003-01-2943 296 N Jayaweera and P Webb Siegel, S., Cunov, J., and Doyle, P., 2003 A380 Automated spar drilling and fettling system (ASDAFS); A culmination of enabling technologies In: Proceedings of Automated Fastening Conference and Exposition, Montreal, Canada, SAE Ref: 2003-01-2951 Webb, P and Eastwood, S.J., 2004 An evaluation of the TI2 manufacturing system for the machining of airframe subassemblies Journal of Engineering Manufacture Part B, 218, 819–826 Webb, P., Eastwood, S., Jayaweera, N., and Ye, C., 2005 Automated aerostructure assembly Industrial Robot, 32 (5), 383–387 Webb, P., Eastwood, S., Jayaweera, N., Ye, C., and McKeown, C., 2006 An automated fuselage panel assembly and riveting cell: validation and testing In: Proceedings of Aerospace Manufacturing and Automated Fastening (AMAF) Conference and Exhibition, Toulouse, France, SAE Ref: 2006-01-3142 [...]... Institute, Loughborough University Weston, R.H ., Rahimifard, A ., Ajaefobi, J.O ., and Cui, Z ., 2009 On modelling reuseable components of changecapable manufacturing systems Journal of Engineering Manufacture, Proceedings of IMechE, Part B, 22 3, 313– 336 Weston, R.H ., Byer, N ., and Ajaefobi, J.O ., 2003 EM in support of team system engineering In: R Jardim-Goncalves, et al ., eds Proceedings of 10th ISPE International. .. Industry, 4 1, 51–63 International Journal of Computer Integrated Manufacturing Vol 2 3, No 3, March 201 0, 229–236 Automated discrete-pin adjustment for reconfigurable moulding machine Y Wanga *, Z Wanga, N Gindya, R Tangb and X.-J Gub a Faculty of Engineering, Ningbo Campus, University of Nottingham, UK; bSchool of Mechanical, Materials and Manufacturing Engineering, University of Nottingham, UK; cCollege of. .. process Computers in Industry, 2 7, 123–142 Zhen, M and Weston, R.H ., 2006 Simulating dynamic behaviours in complex organisations: case study application of a well structured modelling approach In: T Nketsa, et al ., eds Proceedings of European Simulation and Modelling Conference, Toulouse, France, EUROSISETI, pp 390–395 International Journal of Computer Integrated Manufacturing Vol 2 3, No 3, March 201 0,. .. Zeitilinger International Journal of Computer Integrated Manufacturing Weston, R.H ., Chatha, K.A ., and Ajaefobi, J.O ., 2004 Process thinking in support of system specification and selection Advanced Engineering Informatics, 1 8, 217–229 Weston, R.H ., Guererro, A ., and Chatha, K.A ., 2007 Process classes deployed in manufacturing enterprises IJCIM, 20 (6 ), 505–523 Womack, J.P and Jones, D.T ., 2003 Lean... Choia,b, Tae Hyuck Yoonc and Sang Do Nohb* a Korea Institute of Industrial Technology, 35- 3, Hongcheon-ri, Ipchang-myun, Seobuk-gu, Cheonan-si, Chungnam, 331-82 5, Korea; Department of Systems Management Engineering, Sungkyunkwan University 300 Cheoncheon-dong, Jangan-gu, Suwon, Gyeonggido, 440-74 6, Korea; cProduction System R&D Team, Daewoo Shipbuilding & Marine Engineering Co ., LTD, 1, Aju-dong, Geoje-si,... Proceedings of the IEEE International Conference on Automation Science, Shanghai, China, pp 551–557 Medsker, G.J and Campion, M.A ., 1997 Job and team design In: G Salvendy, ed Handbook of human factors and ergonomics New York: Wiley, 450–489 Mintzberg, H ., 1989 Mintzberg on management: inside our strange world of organizations London: Colliar Macmillan Pandya, K.V ., Karlsson, A ., Sega, S ., and Carrie, A ., 1997... operations Computers in Industry, 4 8, 157–179 Hendrick, H.W ., 1997 Organisational design and macroergonomics In: G Salvendy, ed Handbook of human factors and ergonomics New York: Wiley, 595– 637 Kosanke, K ., 1995 CIMOSA overview and status Computers in Industry, 2 7, 101–109 Krappe, H ., Rogalski, S ., and Sander, M ., 2006 Changes for handling flexibility in the change management process of manufacturing. .. 2007] Lou, Y ., Lin, L ., and Dong, J ., 2007 Research on ££ Proceedings of the 2007 11th International Conference on Computer Supported Cooperative Works in Design, 26–28 April 200 7, Melbourne, 698–702 Lukas, U and Nowacki, S ., 2005 High Level Integration based on the PLM Services Standard ProSTEP iViP Science Days 200 5, 28–29 September 200 5, Darmstadt: Cross-Domain Engineering, 50–61 Moeck, H ., 1990 Exchange... Free Press Womack, J.P ., Jones, D.T ., and Roos, D ., 1990 The machine that changed the world New York: Rawson Associates Zaidat, A ., Vincent, L ., and Boucher, X ., 2004 Role and processes based modelling approach for organisation network In: L.M Camarinha-Matos, ed Virtual enterprises and collaborative networks Dordrecht: Kluwer Adademic, 35–42 215 Zelm, M ., Vernadat, F.B ., and Kosanke, K ., 1995 The CIMOSA... integration Computer- Aided Design, 3 3, 489–500 Choi, S.S ., et al ., 2009 MEMPHIS: new framework for realistic virtual engineering Concurrent Engineering, 17 (1 ), 21–33 Feltes, M ., et al ., 2002 Product Lifecycle Management Service Revised Submission (OMG document number mantis/200404-01) [online] Available from: http://www.prostep.org/ file/13580.plm-serv10 [Accessed 5 May 2004] Gong, P ., Yang, C ., and Shi, D.,