Designing cable harness assemblies in virtual environments

  • Published on
    14-Jul-2016

  • View
    217

  • Download
    2

Transcript

Designing cable harness assemblies in virtual environmentsF.M. Ng, J.M. Ritchie*, J.E.L. Simmons, R.G. DewarDepartment of Mechanical and Chemical Engineering, Heriot-Watt University, Riccarton, Edinburgh EH14 4AS, UKAbstractCable harness assemblies are amongst the most costly items in any electro-mechanical product. The domain is not widely recognised asan area for academic research. Internationally, some efforts have been made to automate or semi-automate the choice of cable harness paththrough the use of artificial intelligence (AI) via CAD systems, but with little success. Common themes voiced are that the problem is tooopen-ended and it is very difficult to capture the design intent of the activity. Human input is still very much required to guide the computersystems to reach an optimum solution. Case study investigations were carried out at five advanced manufacturing organisations todetermine the current industrial practice. The investigations revealed that the cable harness design and planning (CHDP) process isessentially sequential in nature and consists of lengthy activities carried out late in the overall product development cycle. It was also foundthat there has been little attempt to integrate any of the core activities involved. This paper describes work undertaken at Heriot-WattUniversity to research the effectiveness of immersive virtual reality for designing and routing cable harnesses by enhancing the expertise ofthe cable harness designer rather than by replacing the individual via an automated system. The new virtual cable design system developedin the course of this work has now undergone some pilot trials to test its usability. The system will subsequently be used to carry out fullindustrial trials in conjunction with a number of high technology equipment manufacturers. These pilot trials, combined with the casestudies of current practice carried out at the companies, have highlighted a number of issues regarding cable design, particularly thatimmersive VR has a potentially unique role to play in the integration of cable harness electrical and mechanical design activities.# 2000 Elsevier Science B.V. All rights reserved.Keywords: Immersive virtual reality; Cable harness design1. IntroductionCable harnesses are a vital part of all electro-mechanicalsystems from aircraft and automobiles to personal compu-ters and domestic appliances. In many instances the cableharness is one of the most costly items in the overallengineered system. In spite of this the detail design andplanning of cable harnesses are often only addressed almostas afterthoughts at the end of the product design process.Cable harness design and planning (CHDP) in fact cover aset of manually intensive, time-consuming and costly activi-ties. There is the obvious problem of determining satisfac-tory routes for bundles of cables in crowded spaces. Thewires themselves will vary in size depending on their duties.The stiffness and mass distribution of the bundle is deter-mined by the size and type of cables involved. Acceptablebend radii must be defined as well as the position anddistribution of the fasteners used to constrain the harness.One important concern for harness designers is that ofvoltage drop. Voltage drop is directly proportional to cablelength and inversely proportional to cable cross-sectionalarea. Ideally, the designer must find a routing configurationthat maintains a suitable voltage drop for all cables in thebundled harness. Fig. 1 shows an example of a completedcable harness ready for assembly into a final product.Current industrial practice, confirmed in case study inves-tigations at five leading UK companies, often requires thebuilding of a physical prototype of a new design beforeengineers are able to manually determine the correct cablelengths and routes, as well as the numbers and positions offasteners. Once a set of suitable cable paths have beenchosen and the associated components selected, the resultsare entered into a database that allows the production of two-dimensional drawings and parts lists together with assemblyinstructions. It is vital that this information is accurate andwell-proven since the actual manufacture of the harnessassembly is often carried out by an external specialistsupplier.The routing problem is further complicated by the vulne-rability of the cable harness to decisions made upstream. Thecable harness may have to be reconfigured after only minorchanges that affect, say, the chassis and the individualmodules within a prototype product. The routing processJournal of Materials Processing Technology 107 (2000) 3743* Corresponding author.E-mail address: j.m.ritchie@hw.ac.uk (J.M. Ritchie).0924-0136/00/$ see front matter # 2000 Elsevier Science B.V. All rights reserved.PII: S 0 9 2 4 - 0 1 3 6 ( 0 0 ) 0 0 7 2 5 - 1can even result in the late and expensive re-design of themachine chassis to allow the cables to reach their terminalpoints.2. BackgroundIn spite of its industrial importance, cable harness designis not widely recognised as an area for academic research.Most investigators who have explored the subject haveattempted to semi-automate or automate the choice ofharness path through the use of artificial intelligence (AI)in conjunction with CAD systems. Such systems are used asa review tool for use after the equipment has been designed.Park et al. [1] recognised that cable harness designrequires in depth three-dimensional spatial reasoning. Theyproposed the use of agents to produce different cableconfigurations that satisfy the pin-to-pin connections of atypical harness circuit layout and automate routine opera-tions such as moving a section of bundles from one positionto another. Conru and Cutkosky [2,3] report that they haveincorporated into Park et al.s system a set of algorithms thatattempt to automate cable routing in a 3D environment. Twogenetic algorithms were developed to route cable harnessesin a 3D environment. Finally, Petrie et al. [4] report thedevelopment of a harness design system called Next-Linkthat allows different designers to create different harnesslayout concurrently. Next-Link is essentially a manage-ment tool that uses a software agent to co-ordinate, updateand keep track of the work of individual designers, evaluat-ing all the routings developed by each designer based onsatisfying global constraints.Much more recently, Cerezuela et al. [5] carried out a casestudy on cable harness design at a helicopter manufacturingcompany. From the case study they found that harnessdesign is an iterative process involving schematic, routingand component design. It is postulated that harness design isa dynamic process and it is not feasible to automate theentire activity by computers. Thus, Cerezuela et al. proposea conceptual knowledge based decision support system toassist in the design of cable harnesses.In summary, the review of published academic literaturein the design and planning of cable harnesses shows thatmuch of the limited amount of research in the area has beenconcerned with developing automated or semi-automatedsystems for determining cable routings. The algorithmsdeveloped tend to be demonstrated in simple geometriclayouts of components and little evidence is provided thatthe work has been applied in industry.3. Industrial case studiesAs part of the present research, case study investigationswere carried out carried out at five UK advanced electro-mechanical technology businesses. These were carriedthrough extensive visits, discussions and meetings withpractitioners and managers. The results were documentedand returned to the companies involved for their verification.Taken together, the five case studies show that the CHDPprocess is essentially sequential in nature and consists oflengthy activities carried out late in the overall productdevelopment cycle. The investigations revealed that therehas been little attempt to integrate any of the core activitiesinvolved. It was also found that companies are increasinglyusing CAD based systems to support the design of harnesses.There was also no evidence to suggest the use of automatedor semi-automated harness design tools in use by thecompanies, confirming prior impressions obtained fromthe literature survey.The case studies results were used to create a genericmodel shown by Fig. 2 for the CHDP process; this providesFig. 1. Complete cable harness prior to assembly.Fig. 2. General stages in the harness design and planning process.38 F.M. Ng et al. / Journal of Materials Processing Technology 107 (2000) 3743an outline picture of how manufacturing companies in theelectro-mechanical sector address the cable harness designproblem. The model is of course subject to detail change inparticular cases dependent on the types of product manu-factured and the required electrical specifications.The contention of the work described in this paper is thatcompanies prefer to have cable harness design as an inter-active technique under the control of the designer. Theremaining sections of the paper describe a prototype immer-sive virtual reality demonstrator system, developed to assistdesigners in producing feasible virtual prototype harnessassemblies, and the corresponding pilot trial results.4. Cable layout using immersive virtual realityThe virtual design and planning cable routing system atHeriot-Watt University is implemented on a Hewlett-Pack-ard workstation with additional VR hardware and softwarefrom Division Ltd. CAD models of a prototype assembly canbe imported directly into the system which negates the needfor any extra component modelling. As illustrated in Fig. 3,the user interacts with the system by means of a headmounted display (HMD). This provides a stereo image ofthe virtual environment. A three-dimensional mouse (3D) isused as an input device.The ability to touch and feel objects in the real world isone that is taken for granted. However, the development ofviable systems to provide this haptic feedback in virtualenvironments is still the subject of much research [68]. Forthis reason, the system described here makes use of alter-native visual and audio cues to highlight collisions. A fullpolygonal collision detection algorithm is available in thesoftware. Thus, when a collision occurs, the system utilisesmessages sent from the algorithm to make images of objectsin the virtual world turn to wire-frame representations. This,along with a simple audio cue, informs the user that some-thing is amiss (Fig. 4).The virtual cable router has five key design tools in itsoperation namely, point-to-point, continuous path,way-point routing, rubber banding and size manage-ment. Collision detection is inherent within the first threefeatures. Point-to-point and continuous path are creationfunctions, whereas way-point routing allows the creation ofcable bundle assemblies along existing routes. Rubberbanding is normally used during editing and size manage-ment enables the user to amend the size of the model relativeto the system user. All the features are activated through avirtual toolbox as shown in Fig. 5.4.1. Point-to-pointThe point-to-point technique of routing cables providesthe capability to generate outline cable routes rapidly bypicking positions or nodes in the virtual environment. Theuser simply probes ports located on cable connectors, or apoint in space, and a section of cable appears between thisand the last node created as shown in Fig. 6. Once an existingnode has been picked in an operation it can be moved aroundin three dimensions, stretching or contracting the associatedcables as required. This editing facility within point-to-pointis called rubber banding and is described later. The pickingFig. 3. A user interacting in the virtual environment.Fig. 4. Wire-frame collision warning of a clash with a cable.Fig. 5. A virtual toolbox.F.M. Ng et al. / Journal of Materials Processing Technology 107 (2000) 3743 39of another node makes that node active and any subsequentpoint chosen in space will create a section of cablebetween it and the active node. By choosing existing nodes,multiple spliced or breakaway cable branches can emanatefrom a single node. Some examples of these are shown inFig. 7.4.2. Continuous pathContinuous path generates a cable route by extruding anew section from a user-selected node. Thus, by picking anexisting node, or a port on a connector, a new node is createdand attached to the virtual hand until the node is released.This method has rubber banding implicit within it also; thenew section changes in length and position as the virtualhand moves. In a fashion similar to point-to-point, multiplespliced or breakaway cable branches can be produced.The user can observe collisions and immediately takeaction to move the section and so as to avoid a clash. Again,this method allows for creating nodes with multiplebranches.4.3. Way-point routingHaving laid one cable, it is possible quickly to lay bundlesof cables along the same route by using way-points. This isachieved by simply choosing the relevant beginning and endnodes along the common length of an existing cable betweenwhich a new cable is to run.4.4. Rubber bandingOnce the entire cable layout is produced some modifica-tions may be required. The rubber banding facility allowsthe user to re-position either entire sections of cable orbundles that are knitted together simply by holding on to andsubsequently moving a node. Although this editing facilitystands alone, it has already been mentioned that it is avail-able in the point-to-point and, to some extent, the continuouspath tools.4.5. Size managementThe final VR cable layout tool developed and defined aspart of this research is size management. This provides theuser with the ability to enlarge or shrink the virtual prototypeto enable human-scale ergonomic access to either finegeometry details or large-scale geometric features withinthe virtual environment as well as deal with any scale ofproduct.5. System architectureThe set of nodes and cable sections created by the user arestored in a multi-linked graph structure containing a linkedlist of nodes and a further linked list of joins for each node[9] (Fig. 8).At the end of the routing session, the system generates atext file by traversing the graph structure and extractinguseful information which details the bills-of-materials andprocess planning information associated with the physicalcable harness. These outputs include the types of endconnectors and cable configurations selected as well asthe positions and liaisons that exist between the virtualnodes as shown in Fig. 8. The connector type and liai-sons/cable configurations indicated in the text file are spe-cified by the user during the immersive routing session. Thenumbers highlighted on the connector type list describe thephysical configuration of the connector, i.e. actual size,number of crimps found on the connector. The liaisons/cable configurations, on the other hand, describe the types ofbundles of wires that are specified for use within certainsections of the cable harness layout. A post-processor hasbeen developed to convert the data within the text file into atwo-dimensional layout of the cable harness in AutoCADDXF format. This drawing can be used in the manufacture ofthe physical cable harness.Fig. 6. Cables leaving a connector via ports.Fig. 7. A cable harness laid out on an assembly in the virtual world.40 F.M. Ng et al. / Journal of Materials Processing Technology 107 (2000) 37436. Pilot studyA pilot study was carried out to evaluate:1. the usability and robustness of the VR routing toolsdeveloped;2. the effects of learning by comparing repetitions for eachmethods and the key differences between the two cablecreation methods.Six participants took part in the pilot study, aged between23 and 30, all were male post-graduate students from theDepartment of Mechanical and Chemical Engineering atHeriot-Watt University. None of them had used an immer-sive virtual reality system before. A description of theexperiment and the results now follows.6.1. Experimental procedureThe participants task was to produce a cable route fromone side of the wall of a virtual component to the other sidewhile immersed in the virtual environment as shown inFigs. 9 and 10. They were asked to develop a route byfollowing the contours of the component. The six partici-pants were divided into two groups of three denoted as PTPand CP group. The first group used point-to-point and theother group used continuous path to develop the cable paths.Each participant performed the task for 10 consecutive trialsand the time to complete the task was measured for each one.6.2. ResultsIn order to detect improvement in performance for alltrials, one-way analysis of variance (ANOVA) was carriedout for both PTP and CP participants. The one-way ANOVAtests were applied to the task completion time (TCT) scoresof both groups. The TCT defines the total time required forcompleting a trial in an experiment. The analysis revealedsignificant differences between the trials within the PTPgroup based on the scores, F9; 20 7:19; p 0:0001,suggesting that there is overall improvement in performancebetween the trials. Statistical differences were not detectedbetween the trials for the CP group, F9; 20 0:89; p 0:55, confirming that the CP method of routing cable path ismuch more difficult to learn.Subsequently, student t-tests were applied to identifyimprovement in performance from the later trials whencompared with trial 1 for scores from TCTs both PTPand CP groups. The results from the t-tests on TCT dataare tabulated in Tables 1 and 2. The t-tests revealed sig-nificant differences within the PTP group for trials 35, 710Fig. 8. An example of the output from the system.Fig. 9. Layout of the assembly. Fig. 10. An example of a laid cable in place for the experiment.F.M. Ng et al. / Journal of Materials Processing Technology 107 (2000) 3743 41when compared with trial 1, suggesting that participantswere learning quickly. On the other hand, no significantdifferences were detected between trial 2 and 6 whencompared with trial 1. The trial 2 finding indicatesthat participants are still learning to use the method.From the video recordings, it was found that participantstend to explore alternative routes in trial 6, thus statisticalsignificance was not detected as participants were spendingmore time planning the routes. Direct observations alsoindicated that the participants were looking for alternativedesigns.The t-tests revealed no significant differences betweentrial 1 when compared with successive trials for the CPgroup, suggesting that there were no obvious improvementsover all subsequent trials. This finding suggests that the CPmethod is much more difficult to master and takes longer tolearn.6.3. Participants feedbackParticipants were asked about the functionality of theCHIVE system after performing the trials. In general theparticipants from the CP group found that the CP method forcabling is tiring to operate as the constant holding of thecable node is necessary when laying the cable sections.Participants from the PTP group using the PTP techniquefelt that the method was easy to use. Participants fromboth group felt that the two-dimensional virtual toolboxwas blocking their routing operations thus hampering theirperformances.To summarise, the pilot study found that: It is easier to learn PTP than CP. As the number of trials performed increases the TCTdecreases. There is a significant difference between the 10 trials inPTP indicating that the participants have learned themethod. There is no significance difference detected for CP groupsuggesting that participants are still learning the opera-tions of the CP method and that more trials are requiredbefore operators become fully proficient. It has been observed that participants in both groups tendto require more guidance in the operation of the VRcabling tools and also navigating in the VE during theearlier trials. Direct observations and feedback from participants alsofound that the CP group was having considerably moredifficulties in routing than the PTP group. All subjects experienced fatigue whilst conducting thevirtual experiments. Fatigue was experienced morequickly in the CP group than the PTP group. The pilot tests fulfilled their purpose of testing systemsoperation and usability. It was possible to develop feasible cable routes using bothPTP and CP methods.The pilot tests and the industrial case studies together alsopointed out a number of issues related to both the industrialtrials and future system development, namely: Future industrial trials will have to be restricted to pre-defined routing paths in order to compare the efficiency ofdifferent designers and alternative cable harness designsolutions, e.g. CAD. A generic assembly will be used which incorporatesfeatures that can be designed on both the virtual cabledesigner and the companies CAD systems. Longer training times are required to allow users tobecome proficient with the virtual cable design tools. A new paradigm can now be researched whereby adesigner could design feasible cable routes connections,using PTP or CP, for electrical connectivity on the virtualtable top (in, say, 212D). In parallel with this, these con-nections would be automatically mapped using straightline connections onto the 3D model in the same virtualspace. Rubber banding would be used to tailor the routesaround obstacles. This means that, potentially, cableharness electrical and mechanical design can be carriedout concurrently instead of sequentially as is the case atpresent. Thus the potential for reduction in design lead-time is substantial.Table 1Comparing subsequent PTP task completion times (TCTs) with trial 1: t-tests t- and p-valuesTrial No. t p2 2.24 0.093 3.54* 0.024 3.8* 0.025 3.63* 0.026 2.46 0.077 3.33* 0.038 3.20* 0.039 3.85* 0.0210 4.07* 0.02* Significantly different if p < 0:05.Table 2Comparing subsequent CP task completion times (TCTs) with trial 1: t-tests t- and p-valuesTrial No. t p2 0.79 0.473 0.53 0.624 0.87 0.445 1.03 0.366 1.18 0.307 0.87 0.438 1.25 0.289 1.01 0.3710 1.39 0.2442 F.M. Ng et al. / Journal of Materials Processing Technology 107 (2000) 37437. ConclusionThis paper has described a novel software tool to assistusers to perform cable routing in a virtual environment. Thesystem here has been successfully tested in pilot trials. Therecommendations made by the participants during the pilotstudy were noted and changes had been incorporated intovirtual cable routing system. Firstly to enable easier userselection of the cabling tools, the dimension of the virtualtoolbox for choosing cable routing methods has beenenlarged to the size of a billboard within the virtualenvironment and the virtual buttons in the toolbox werealso spaced widely to facilitate easier user selection. Toovercome obstruction caused by the virtual toolbox that wasblocking the routing operations, the user is now transportedto another position within the virtual environment with onlythe toolbox in view so that they can select the requiredcabling tools for subsequent cable layout routing. Once therequired tool or tools are selected, the user can return to theiroriginal position by hitting the return option on the toolboxin the original location of the user inside the assembly beforethe toolbox was invoked and continue the cabling operationas per normal.The pilot study also indicated that the CP method is moredifficult to learn since statistical analysis were unable todetect obvious improvements in performance for all trials.More repetitions would have to be carried out in order tobring participants to a level where they are confident in usingthe tool for laying cables. Direct observation and feedbackfrom participants also indicated that it was tiring to use theCP method when performing the routing experiment. Unlikethe PTP method, the CP cable creation method does not havethe editing facility rubber-banding available to it. To alterthe cable layout, users were required to select the rubber-banding function by invoking the virtual toolbox. Once themodification was completed, users were required to invokethe toolbox again so as to select the CP method and carry onlaying the cable as per normal. Thus from a user-friendlyinterface point of view CP has two extra redundant stepswhen modifications are required to be performed to the cablepath. The feedback from participants and the results from thepilot trials have suggested that training on the general usageof the 3D mouse for navigating and interacting with the VEmight be useful prior to the actual industrial trials. Thisexercise may make it easier for the participants to concen-trate on learning the VR cabling tools since the basicmethods of navigation will have been learnt through thetraining exercise.In the future full scale industrial trials will be carried outto investigate the viability of this approach to complete thecable harness routing task as compared to current commer-cial CAD systems. However, this work did show conclu-sively that CHDP will be possible in an immersive VRenvironment.Combined with the industrial case study investigations, anew and novel concurrent electrical and mechanical designparadigm has been recognised. The technology and applica-tion of immersive VR in this environment provides a newsolution to a traditionally difficult, costly and tail-end part ofthe overall product design process.AcknowledgementsThe authors are grateful to the five companies thatcollaborated in this research, for their support of thiswork and for access to their expertise and knowledge.The support of the EPSRC, through access to the equipmentprovided under grant GR/K41823, is also very gratefullyacknowledged.References[1] H. Park, H. Lee, M.R. Cutkosky, Computational support forconcurrent engineering of cable harnesses, in: Computers inEngineering, Proceedings of the International Computers in En-gineering Conference and Exhibit, Vol. 1, No. 1, San Francisco,USA, 1992, pp. 261268.[2] A.B. Conru, M.R. Cutkosky, Computational support for interactivecable harness routing and design, in: Proceedings of the 19th AnnualASME Design Automation Conference, Vol. 65, Albuquerque, USA,1993, pp. 551558.[3] A.B. Conru, A genetic approach to the cable harness routingproblem, in: Proceedings of the IEEE Conference on EvolutionaryComputation, Vol. 1, Orlando, USA, 1994, pp. 200205.[4] C.J. Petrie, T.A. Webster, M.R. Cutkosky, Using Pareto optimality tocoordinate distributed agents, Artificial Intell. Eng. Des. Anal.Manuf. 9 (4) (1995) 269281.[5] C. Cerezuela, A. Cauvin, X. Boucher, J.P. Kieffer, A decision supportsystem for a concurrent design of cable harnesses: conceptualapproach and implementation, Concurrent Eng. Res. Appl. 6 (1)(1998) 4352.[6] D.G. Caldwell, S. Lawther, A. Wardle, Tactile perception and itsapplication to the design of multi-modal cutaneous feedback systems,in: Proceedings of the IEEE International Conference on Roboticsand Automation, Vol. 4, Minneapolis, USA, 1996, pp. 32153221.[7] T. Tateno, M. Igoshi, Hierarchical processing for presenting reactionforces in virtual assembly environments, J. Jpn. Soc. Prec. Eng. 62(8) (1996) 11821186.[8] P. Taylor, Tactile and kinaesthetic feedback in virtual environments,in: Transactions of the Institute of Measurement and Control, Vol. 17,No. 5, 1995, pp. 225233.[9] Y. Langsam, M.J. Augenstein, A.M. Tenenbaum, Data Structuresusing C and C, 2nd Edition, Prentice-Hall, NJ, 1996.F.M. Ng et al. / Journal of Materials Processing Technology 107 (2000) 3743 43

Recommended

View more >