You are on page 1of 8

Int J Interact Des Manuf DOI 10.

1007/s12008-007-0034-0

ORIGINAL PAPER

Robust design of car packaging in virtual environment


Antonio Lanzotti

Received: 14 September 2006 / Revised: 10 October 2007 / Accepted: 15 October 2007 Springer-Verlag France 2007

Abstract This paper presents a statistical methodology to improve the car packaging setup in the rst phase of a new mini-car design. An original procedure for comfort assessment using virtual manikins is formulated. The Robust Design approach enables to identify the optimal level for the main design factors of the new car packaging. The optimal solution is the most insensitive to anthropometric variability. The case study of a new mini-car packaging setup is exploited. The experimental results in virtual environment are obtained using the virtual manikin Jack by UGS. On the base of adequate comfort indexes, the proposed methodology allows dening a car packaging which is, on average, more comfortable than that obtainable as initial setting by applying the Enhanced SAE Packaging Guidelines and the Posture Prediction algorithms proposed by the UMTRI (UMI-USA). Keywords Virtual ergonomics Robust design Human variability modelling

1 Introduction The evaluation of the car comfort is a complex and structured task, directly inuencing safety [1]. Modern cars offer to occupants ergonomic seats and handy controls. In order to improve good performances in the long run, ergonomic requirements and constraints have to be taken into account since the briefing of a new car [2]. The rational ergonomic design requires the management of a large number of design parameters and variables [3]. In the last 10 years, the use
A. Lanzotti (B) Dipartimento di Ingegneria Aerospaziale, University of Naples Federico II, P.le Tecchio 80, 80125 Naples, Italy e-mail: antonio.lanzotti@unina.it

of the Digital Mock Up (DMU) of cars and virtual manikins, that allow to simulate drivers and passengers tasks is extended at the beginning of the life cycle [413]. In this way, it is possible to reduce the activities based on physical prototypes, being virtual design validation more and more advanced. Furthermore, the SAE Packaging Guidelines [1418] make useful reference to the choice of packaging parameters for the main market segments. The validation of the packaging design requires methods based on anthropometric database and 3D virtual manikins. The virtual manikin is a quite realistic human model, indeed it consists of a fully jointed skeleton. The dimensions of body segments can be simulated in order to reproduce the anthropometric variability for selected populations [1,1921]. This paper presents a statistical methodology to improve the car packaging setup in the rst phase of a new mini-car design. The case study of a new mini-car packaging setup is exploited. The experimental results in virtual environment are obtained using the virtual manikin Jack. On the base of some adequate comfort indexes, the ergonomic robust design approach enables, in the rst phase, to identify the optimal level for the main design factors of the new car packaging and, in the second phase, to dene a design setting of car packaging which is, in average, more comfortable than that obtainable as initial setting by applying the Enhanced SAE Packaging Guidelines and the Posture Prediction algorithms proposed by the UMTRI (UMI-USA). The optimal solution is the most insensitive to anthropometric variability and improves robustness minimizing adjustment requirements.

2 Ergonomic robust design Figure 1 shows the p-diagram of the robust design approach to comfort improvement. The main aim is to dene a car

123

A. Lanzotti

Fig. 1 Ergonomic robust design approach of car packaging

packaging that improves the postural comfort of the target population, being the most insensitive to anthropometric variability and minimizing adjustment requirements. In this approach (see Fig. 1), the anthropometric variability can be considered a noise factor. Design factors of car packaging are suggested by the SAE guidelines and the level of each design factor can be dened together with designers. Following a cross array design, virtual experiments can be realized using virtual manikins. The response is a set of quantitative measures of static postural comfort. The quality evaluation of each design setting is obtained on the base of both a loss function and some comfort indexes. 2.1 Mini-car packaging definition The definition of mini-car packaging starts from the SAE packaging guidelines, that are implemented in the Enhanced SAE Packaging. The design parameters, reported in Table 1, are the most important ones and, for this reason, are taken into account in the experimental pre-design phase (see Fig. 2). The mini-car packaging starting levels, reported in Table 2 are chosen on the base of designer remark, the evaluation of the best in class and initial trials in virtual environment realized in order to avoid unrealistic posture, [4,5]. In particular, the high H-point follows from the requirements of improving accessibility and comfort, since in Europe, the 53% of user are elderly persons. In order to
Fig. 2 Design factors used in the pre-design experimental phase Table 2 Initial settings of design factors Design factor A. Lumbar Prominence (LB) B. Vertical SgRP to Hell Point (H30) C. Seat Cushion Angle (L27) D. Steering Wheel to Bof (SWB) E. Seat Track Angle (STA) Starting level 1.3 40.0 13.9 35.7 0 . 9 Dimensions cm cm degrees cm degrees

Table 1 Definition of car packaging design factor used in the pre-design experimental phase Lumbar Prominence Vertical SgRP to Hell Point Seat Cushion Angle Steering Wheel to Bof Seat Track Angle (LB) (H30) (L27) (SWB) (STA)

improve cabin accessibility and seat comfort, the trend is to make the car body higher. In this way, the designer can make the H-point higher without loosing space between the driver/passenger heads and the car ceiling. In this way, an elderly person can minimise the efforts to get into and out of a car. Briey, the H-point growth realizes the following advantages: good steering wheel grasp; safer position of legs; driver space reduction (about 3040 mm); attening car platform in the feet area. The car accessibility, with the previous mentioned design choices, is improved; a trunk rotation of 45 degrees, lifting and rotation of legs are required to reach and

123

Robust design of car packaging in virtual environment

leave the seat. Furthermore, the higher H-point can improve visibility and, as a result, safety. 2.2 Noise factor: anthropometric variability

Table 3 Experimental results about comfortable joint angles values [24] J 1 Joint angle Head exion Arm exion Elbow Trunk-thigh Knee included Head exion Observed minimum 10 19 86 90 99 80 Male mode 7 50 128 101 121 93 Female mode 4 40 113 99 117 92 Observed maximum 26 75 164 115 138 113

The driver seat comfort must be the most robust to anthropometric variability of the target population. This variability is a noise factor and has to be taken into account in the experimental design. The results could be parametrically expressed as a function of the population mix. The population mix can be dened as the percentage of females ( pF ) and males ( pM ), that constitutes the potential user population, being: pF + pM = 100% In this paper, the noise random variable is the height of target population, Hpop , that is dened as a mixture of the Normal random variable (r.v.) of female and male heights ( HF , HM ): f (h pop ) = pF . f (h f ) + pM .(h M ) (1)

2 3 4 5 6

2.3 Postural comfort evaluation The evaluation of seat driver comfort is a complex task; in fact it depends on both subjective and objective point of views. A good interpretation of driver comfort, based on objective elements, is given by the measures of the joint angles reported in Table 3 and proposed in [24]. Starting from [25], the asymmetrical quadratic comfort loss is dened: L Q ( Y j ) = k 1 j ( Y j y T j )2 (4)

where f (h f ) and f (h M ) are the pdf of Normal r.v. with parameters: F = 162.7, F = 6.69; M = 175.5, M = 6.78; Starting from the outer array definition and level choice, proposed in [22] for a Normal distribution, six percentiles, (i.e. 5-th, 50-th and 95-th for females and 5-th, 50-th and 95-th for males) can approximatively be chosen. In this way, it is possible to parametrically dene, a generic response as weighted average depending on the mix: Z w ( pF ) = pF
1 4 1 6 Z 5,F + 6 Z 50,F + 6 Z 95,F 4 1 + PM 1 6 Z 5,F + 6 Z 50,F + 6 Z 95,F

(2)

where k1 j and k2 j are obtained, setting equal to the conventional value one, the loss at ymin j and ymax j (reported in Table 3), as k1 j = (Ymin j yT j )2 and k2 j = (Ymax j yT j )2 . Assume that from the population of potential users we select one manikin which represents a specied body height percentile for a given sex, e.g. the 5-th percentile female. Assume that we measure the joint angles of the manikin, while it is accommodated in the driver seat, and denote with y j (5,F) the observed value of the joint angle Y j for the 5-th percentile female. Then the total joint angle comfort loss is:
m

In order to reduce the experiments, the following generalized weighted average is proposed:
( p ) = (w ( p ) Z Zw F 1 F 5,F + w2 ( pF ) Z 50,F +w3 ( pF ) Z 50,M + w4 ( pF ) Z 95,M )

CL5,F =
j =1

L Q (Y j (5,F) )

(5)

(3)

where the new weights w j ( pF ), j = 1, 2, 3, 4 can be calculated by satisfactorily approximating the results obtained using Eq. 2. The weights can be xed, in an original way, by generalizing the results in [22] for the noise factor design. In this way, the experimental runs are reduced by one third without loosing the quality of results. The weight evaluation, w j ( pF ), j = 1, 2, 3, 4, depending on the population mix is fully developed in [23]. For example, for a 50% mix, the weights are: w1 = 0.103; w2 = 0.389; w3 = 0.425; w4 = 0.083

where L Q (Y j (5, F ) ) is evaluated using the (4) for the 5-th percentile female. In the same way, the total joint angle loss for the 50-th percentile female, 50-th percentile male and 95-th percentile male can be evaluated. So, starting from (3), the weighted comfort loss (WCL) can be dened: WCL( pF ) = w1 ( pF )(CL5 + w2 ( pF )(CL50,F ) +w3 ( pF )(CL50,M ) + w4 ( pF )(CL95,M ) 2.3.1 Joint angle measure using Jack For the virtual experimentation, the virtual manikin Jack by UGS, developed in the late 1980s at the Center for Human (6)

123

A. Lanzotti

Modeling of the University of Pennsylvania, is used. A tool of the software Jack, the Enhanced SAE Packaging Guidelines, gives interactively the possibility to choice the car packaging for the market segment of potential users. The packaging setting enables to generate reproducible posture of the manikin through the Posture Prediction tool, based on research of the University of Michigan Transportation Research Institute (UMTRI) [11,2730,26]. After sex, percentile, and posture of the virtual manikin are xed, the joint angles are measured through the Comfort Assessment tool, based on the Porter and Gyi results too [24]. It is not possible to realize these steps automatically. So, each experiment in virtual environment (VE) requires time and needs interaction with a computer operator. The experiment in VE is, now, reproducible because the posture prediction tool ensures that only one posture satises the imposed constraints. 3 First experimental phase On the base of ve design factors, dened in Table 1, at two levels the inner array is a full factorial and, on the base of (3), the outer array denes only four runs. The cross array requires 128 experimental runs. Table 4 shows the level chosen for each of the ve design factors; the levels have been chosen on the base of the following designer remarks: 1. The Lumbar Prominence (LB) levels are xed following the guidelines and taking into account the actual range of variation of the car seat. Thus the value 1.3 (i.e. the minimum value suggested by the guidelines) is assigned to level 0 whereas the value 10 (i.e. the maximum value calculated by taking into account the actual range of variation of the car seat), is assigned to level 1, 2. The Vertical SgRP to Heel Point (H30) levels are xed following some preliminary trial results. Thus the value 40 (i.e. the minimum value obtained from the trials) is assigned to level 0, whereas the value 50 (i.e. the upper bound to realistic posture) is assigned to level 1,

3. The Seat Cushion Angle (L27) levels are xed so as to reduce the over-pression of thigh against seat. Thus the value zero is assigned to level 0, whereas the value13.9 is assigned to level 1, 4. The Steering Wheel to Bof (SWB) levels are xed following results of preliminary trials on visibility. Thus the value 35 (i.e. the minimum value obtained from the trials) is assigned to level 0, whereas the value 60 (i.e. the maximum value obtained from the trials) is assigned to level 1, 5. The Seat Track Angle levels are xed so as to improve comfort with reference to leg position. Thus the value 0.9 (i.e. the datum-point) is assigned to level 0, whereas the value 6 (i.e. the maximum in order to improve comfort) is assigned to level 1.

3.1 Experimental results Figure 3 shows the Pareto diagram as a synthesis of experimental results. This analysis highlights that the main factors

Fig. 3 Main factors and interactions affecting the global weighted comfort loss

Table 4 Design factors and levels dened for the rst experimental phase Design factor Levels 0 1. 2. 3. 4. 5. Lumbar Prominence (LB) Vertical SgRP to Heel Point (H30) Seat Cushion Angle (L27) Steering Wheel to Bof (SWB) Seat Track Angle (STA) 1.3 40 0 35 0 . 9 1 10 50 13.9 60 6 cm cm deg cm deg Fig. 4 Plot and mean effects of interaction BD Dimensions

123

Robust design of car packaging in virtual environment Table 5 Results of the rst experimental phase (limit cases) Run CL( p, S ) 5th, F Initial setting Worst case (run16) Best case (run19) 10.40 44.07 7.25 50th, F 11.25 29.68 6.25 50th, M 11.25 19.29 7.32 95th, M 11.27 19.35 7.21 11.16 24.89 6.89 CL(50)

rst phase, the outer array denes only four runs. The levels are chosen following the classical DOE approach. So, in this case, the cross array requires 108 virtual experiments. Table 6 shows the levels chosen for each of the ve design factor, following on the base of the experimental results and designer remarks: The factor B (H30), reduces the comfort loss at the maximum level due to the interaction with D-; instead, the absolute contribution is negative, since the more is the comfort the less is the level. In this case, it is not possible to extend the range so the new level is dened in the mean point in order to discover if the optimal value can be inside the interval, The factor D improves the comfort at the minimum level; in this case, the new levels are chosen in order to discover if this trend is hold outside the previous experimental interval,

are the interaction BD and the simple factors D, C and B. The significance of mean effects is evaluated applying the methods of Lenth [31]. In Fig. 4 the interaction plot and the mean effects of the interaction BD are shown (Table 5). 4 Second experimental phase On the base of the results of the rst experimental phase, the inner array is a three level full factorial (33 ) and, like in the

Table 6 Design factors and their levels for packaging settings at the second experimental phase Design factor Levels 1 B C D Vertical SgRP to Heel Point (H30) Seat Cushion Angle (L27) Steering Wheel to Bof (SWB) Fixed design factor A E Lumbar prominence (LB) Seat Track Angle (STA) 1.3 0 . 9 40 2 30 0 45 2 40 Dimensions cm degrees 1 50 6 50 40 13.9 35.7 cm degrees cm Initial setting Dimensions

Fig. 5 Plots of mean effects of design factors

123

A. Lanzotti

The factor C (L27) improves the comfort at the minimum level and the new levels are xed taking into account design constraints, The factors A (LP) and E (STA) are both xed at the optimum level from the rst experimental results.

4.1 Experimental results Figure 5 shows the mean effects of the three factors on the comfort loss, highlighting that the main factors are D (SWB) and B (H30) in decreasing order of importance. Factor C (L27) is not inuent on the response and can be xed at the more economical level. Figure 6 shows that the interaction BD increases robustness when the factor B is at the higher level and D is at the middle level. This combination is better than B at the middle level and D at the higher level for the main effect of B and D. Figures 7 and 8 show the sensibility of Comfort Loss to level choice of B and D. Table 7 shows the experimental results in terms of CL and WCL(50). The best run improves the good results based on previous experimental phase (about 10%). Further, the worst run points out the goodness of the initial setting.

Fig. 8 Discomfort variation for the four standing height for the design factor SWB Table 7 Results of the second experimental phase (limit cases) Run CL( p, S ) 5th, F Initial setting Best case (run20) 10.40 7.14 50th, F 50th, M 95th, M 11.25 12.87 5.71 11.25 13.97 6.43 11.27 14.11 6.46 11.16 13.37 6.22 WCL(50)

Worst case (run7) 12.17

5 Concluding remarks Figures 9 and 10 show the change in the driving posture from the starting settings to the optimal ones for the mini-car; it is possible to evaluate qualitatively differences and goodness of results.

Fig. 6 Mean effects of interaction BD Fig. 9 Initial setting

Fig. 7 Discomfort variation for the four standing height for the design factor H30

Fig. 10 Best run of second phase

123

Robust design of car packaging in virtual environment Table 8 Comfort improvement obtained through two experimental phases CL( p, S ) 5th, F Initial setting I exp. phase: Best Run II exp. phase: Best Run 10.40 7.25 7.14 50th, F 11.25 6.25 5.71 50th, M 11.25 7.32 6.43 95th, M 11.27 7.21 6.46 11.16 6.89 6.22

Run

WCL(50)

Table 8 and Fig. 11 show global improvement of postural comfort from the initial setting to the best setting suggested after the two experimental phases. The average improvement, measured by means of WCL(50), is greater than 40%. This result is good and can be generalised to all population

mix. Figure 12 shows how the proposed packaging is robust against the variation of the population mix ( pF ) of potential users. The proposed procedure can enable to nd more robust and more comfortable packaging settings for the mini-car

Fig. 11 Comfort improvement obtained through two experimental phases

Fig. 12 Comfort robustness in function of the population mix for the initial settings and the two experimental phases

123

A. Lanzotti

market segment. Furthermore, the result show how the starting settings based on SAE Packaging Guidelines and designer opinion can be improved by means of the proposed robust design experimental procedure. The main advantage of this procedure is the use of the information available in the Comfort Assessment tool of Jack and otherwise lost. Finally, being joint angles data normally available as output of virtual manikins used for ergonomic anlaysis, the proposed procedure is independent from the virtual manikin used for the experimental phases.
Acknowledgments The present work has been developed with the contribute of MIUR-Italy performing the activities of PRIN-2005: Statistical design of continuous product innovation.

References
1. Pheasant, S.: Bodyspace Anthropometry, Ergonomics and the Design of work. Taylor & Francis, London (1996) 2. Peacock, B., Karkowski, W.: Automotive Ergonomics. Taylor & Francis, London (1993) 3. Schneider, L., Robbins, D.H., Pug, Snyder, R.G.: Anthropometry of motor vehicle occupants. Ann Arbor, MI: University of Michigan Transportation Reasearch Institute, 1, December 1993 4. Barone, S., Carbone, F., Lanzotti, A.: Progettazione del posto guida di una mini-car basata su esperienza del designer e sperimentazione virtuale. In: XVII Ingegraf-XV ADM, Siviglia, giu (2005) 5. Barone, S., Fittipaldi, F., Lanzotti, A.: Improving comfort of a new city vehicle by means of parameter design in virtual environment. In: Proceedings of the First Annual Conference of the European Network for Business and Industrial Statistics, Oslo, 1718 September (2001) 6. Caputo, F., De Amicis, R., Di Gironimo, G., Stork, A.: Ergonomic driven design in augmented reality. In: Proceedings of XIII ADM - XV INGEGRAF International Conference, Naples, Italy (2003) 7. Caputo, F., Di Gironimo, G., Papa, S.: Realization of a virtual environment for ergonomics and usability validation of equipment controls (in italian). In: Proceedings of XV ADM - XVII INGEGRAF International Conference, Sevilla, Spain (2005) 8. Di Gironimo, G., Lanzotti, A., Vanacore, A.: Concept design for quality in virtual environment. Comput. Graph. 30 (2006). doi:10.1016/j.cag2006.08.002 9. Di Gironimo, G., Lanzotti, A., Vanacore, A.: Quality feature design of a minicar dashboard using doe in virtual reality. In: Proceedings of 5th International Conference on Advanced Engineering Design (AED 2006), Prague, Czech Republic (2006) 10. Di Gironimo, G., Martorelli, M., Monacelli, G., Vaudo, G.: Use of virtual mock-up for ergonomic design. In: 7th International Conference ATA, Florence (2001)

11. SAE J1100. Motor vehicle dimensions. May (1995) 12. Bubb, H., Estermann, S.: Inuence of forces on comfort feeling in vehicles. SAE Technical paper 2000-01-2171 (2000) 13. Grandjean, E.: Sitting posture of car drivers from the point of view of ergonomics. Human Factors in Transport Research (part 1), pp. 20213 (1980) 14. Flannagan, C.A.C., Manary, M.A., Reed, M.P., Schneider, L.W.: An Improved Seating Accommodation Model with Application to Different User Populations. SAE T.P. 980651 (1998) 15. Manary, M.A., Schneider, L.W., Flannagan, C.C., Eby, B.H.: Evaluation of the sae j826 3D manikin measures of driver positioning and posture. SAE T.P. 941048 (1994) 16. Manary, M.A., Flannagan, C.A.C., Reed, M.P., Schneider, L.W.: Atd positioning based on driver posture and position. In: Proceedings of 42nd Stapp Car Crash Conference (1998) 17. Manary, M.A., Flannagan, C.A.C., Reed, M.P., Schneider, L.W.: Development of an improved driver eye position model. SAE technical paper 980012 (1998) 18. Reed, M.P., Manary, M.A., Flannagan, A.C., Schneider, W.L.: New concept in vehicle interior design using aspect. SAE Technical Paper 1999-01-0967 (1999) 19. NASA.: Anthropometric source book. A Handbook of Anthropometric Data. Vol. II, NASA. Reference Publication 1024, Houston (1978) 20. Miller, H.: The anthropometrics of t. Zeeland Michigan O.MS2700-2 (2002) 21. Rebiffe, R.: The driving seat: its adaptation to functional and anthropometric requirements. In: Proceedings of a Symposium on Sitting Posture (1969) 22. Derrico, J., Zaino, R.A.: Statistical tolerancing using a modication of taguchis method. Technometrics 30(4), 397405 (1998) 23. Barone, S., Lanzotti, A.: On the treatment of anthropometrical noise factors in Robust Ergonomic Design, DallIdea al Prodotto: la rappresentazione come base per lo sviluppo e linnovazione, Edizioni ETS, Pisa, pp. 359367 (2007). ISBN 978-884671841-9 24. Porter, M., Gyi, D.E.: Exploring the optimum posture for driver comfort. Int. J. Vehicle Des. 19(3), 255266 (1998) 25. Barone, S., Lanzotti, A.: Quality engineering approach to improve comfort of a new vehicle in virtual environment. In: Proceedings of ASA, SCS. AI, USA (2002) 26. Reed, M.P., Manary, M.A., Flannagan A, C., Schneider, W.L.: Comparison of Methods for Predicting Automotive Driver Posture. SAE Technical Paper 2000-01-2180 (2000) 27. SAE J1517: Driver selected seat position. March (1990) 28. SAE J1052: Motor vehicle driver and passenger head position. April (1997) 29. SAE J287: Drive hand control reach. June (1988) 30. SAE J182: Motor vehicle ducial marks and three-dimensional reference system. August (1997) 31. Lenth. Quick and easy analisys of unreplicated factorials. Technometrics 31(4), (1989)

123

You might also like