Parametric modeling of optimal aircraft seat surface and seat pressure distribution

HAL Id: hal-01770183

https://hal.archives-ouvertes.fr/hal-01770183

Submitted on 18 Apr 2018

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Parametric modeling of optimal aircraft seat surface and

seat pressure distribution

Xuguang Wang, Michelle Cardoso, Georges Beurier

To cite this version:

Xuguang Wang, Michelle Cardoso, Georges Beurier. Parametric modeling of optimal aircraft seat

surface and seat pressure distribution. 1st International Comfort Congress 2017, Jun 2017, SALERNE,

Italy. 8 p. �hal-01770183�

Salerno, June 7th and 8th, 2017 1st _{International Comfort Congress}

Parametric modelling of optimal aircraft seat surface

and seat pressure distribution

Xuguang Wang

_{Michelle Cardoso}

_{Georges Beurier}

1 _{1Univ Lyon, Université Lyon 1, IFSTTAR, LBMC UMR_T 9406, F-69622, LYON, France}

* Corresponding author. Tel.: +33-4-72142451 ; fax: +33-4-72376837. E-mail address: [email protected]

Abstract: Pressure distribution is well accepted as the objective measure that is the most related to subjective

rating of seat discomfort. However, very few studies have examined ideal seat pressure distribution for guid-ing seat design. In this study, to improve airplane seat comfort, the compressed seat surface which optimizes pressure distribution was studied experimentally using a newly built multi-adjustable experimental seat. One of the innovative aspects of this new experimental seat was that seat pan surface was simulated by a matrix of 52 cylinders, each with a freely rotatable circular flat head of 60 mm in diameter. Each cylinder was equipped with a tri-axial force sensor, enabling the measurement of both normal and tangential forces. The height of each cylinder was adjustable with a maximum stroke length of 40 mm, making it possible to control pressure distribution underneath the soft tissue of the buttocks and thighs. A uniform law relating normal force and displacement for each cylinder was used, permitting a pressure distribution as uniform as possible for a given maximum displacement of the cylinders. The backrest consisted of adjustable lumbar, thoracic and neck pan-els. Seat pan length, height and inclination were also adjustable. Thirty-six participants (18 males, 18 females) of varying weight (healthy BMI, obese BMI) and stature (small, medium and tall) were recruited. Participants were instructed to sit in 40 seat positions (each lasting approximately 1 minute) that simulated an economic airplane seat. Varying seat positions included: backrest angle (100°, 110°), seat pan angle (0°, 5°, freely ad-justable), stroke height of seat pan cylinders (20mm, 40mm) and task (watching forward without neck sup-port, relaxed sitting with neck support). A principal component (PC) analysis was first proceeded to reduce the dimensionality of the data. A linear regression was then performed in the reduced PC space so that the seat surface and corresponding contact force distribution were related to main anthropometric dimensions (stature, BMI, sitting height to stature ratio) and seat parameters (seat pan and back angles). It was observed that the profile of the “optimal” compressed seat pan surface in the symmetric plane showed little variability between the different body sizes, while the profile in the frontal plane showed higher variability.. Contact force distri-bution was highly dependent on the sitter’s BMI. The parametric models obtained will provide quantitative guidance to the design of airplane seats ( i.e. determining the “optimal” seat pan compressed surface and pres-sure distribution) that would best accommodate the general population.

Keywords: Seat, Aircraft, Pressure distribution, Discomfort, Parametric modelling.

1 Introduction

Among the 22 features in the aircraft cabin which have an influence on passenger comfort, the seat was the most frequently mentioned one according to an online survey on a sample of 158 people who just had a long-haul trip (>4 hours). Aircraft seat suppliers are faced with two strong requirements from airline companies:

2

seat weight must be reduced while simultaneously improve its comfort. According to an extensive review of scientific publications, De Looze et al (2003) found that pressure distribution is the objective measure which has the clearest association with the subjective ratings of seat comfort. From a more recent review on the stud-ies of office chairs; Zemp et al (2015) suggested that peak pressure and pressure distribution on the back rest could be considered as reliable measures for quantifying comfort or discomfort. However, Zemp et al pointed out that very few studies have examined the optimum pressure distribution while sitting in an office chair. This is probably due to high number of factors affecting pressure distribution, such as sitter’s body dimension, posture, seat geometry, material proprieties etc., and to their interaction. Mergl et al. (2005) studied the rela-tionship between discomfort and contact force distribution experimentally and suggested an ‘ideal’ seat pan distribution for automotive seats. In addition to the limitations by experimental conditions, recommended pressure distribution is not necessarily applicable to aircraft passenger seats due to large differences in space requirements and sitting activities.

The present study aimed to develop design guidelines for airplane seats, based from the analysis of both optimal compressed seat surface and pressure distribution. The optimal compressed seat surface which dis-tributes the seat contact force as uniformly as possible was experimentally investigated using a multi-adjustable experimental seat. 3D compressed surface and contact force distribution data were collected for 36 men and women of varying body dimensions. Participants sat in 40 different positions. Collected data were used to build a parametric model for predicting optimal compressed seat surface and seat pressure distribu-tion.

2 Materials and methods

2.1 Data collection

The data was collected through use of a multi-adjustable experimental seat. The experimental seat (Figure 1) was composed of four main structural components: the supporting frame (A), seat back frame (B), seat pan frame (C) and foot support (D). The supporting frame (A) was mounted on four wheels and its orientation ranged from -5° to 5° (relative to the ground) with help of an electric actuator. The backrest frame (B) articu-lated with the (A) frame around a lateral axis (y-axis) passing through the reference point of the experimental seat, named PRC (‘Point de Référence du Conformateur’). The experimental seat had thirteen adjustable pa-rameters (Figure 1b) directly controlled by a computer. Adjustable features included: fore-aft (x) and vertical position (z) of the foot support, seat pan and three back supports; rotational angle of the seat pan, backrest and global inclination of the whole experimental seat. Two armrests (E) were also available and adjusted manual-ly. Force sensors were mounted to measure contact forces in xz plane on the foot support, seat pan, three back supports and two armrests. The seat pan surface was composed of a matrix of 52 cylinders, each was freely rotatable with a circular flat head of 60 mm in diameter. Each cylinder was equipped with a tri-axial force sensor, enabling the measurement of both normal and tangential forces. The height of each cylinder was ad-justable with a maximum stroke length of 40 mm and pressure distribution could be controlled by changing seat surface. Figure 2 shows a subject sitting in the experiment seat and the matrix of 52 cylinders simulating seat surface. A more detailed description can be found in Beurier et al (2017).

Pressure distribution on the seat pan surface was controlled using a uniform coupling law relating normal force and position for each cylinder. The coupling law enabled us to distribute normal contact force as uni-formly as possible among the 52 cylinders (given the maximum displacement of the cylinders). For a given normal contact force on the seat surface (𝐹𝑛𝑆𝑃), a target mean force (𝐹̅𝑛𝑆𝑃) was estimated as 𝐹𝑛𝑆𝑃/(0.75*52)

con-sidering that approximately ¼ of the seat surface was not in contact with the buttock or thighs. Each cylinder lowers its height once its contact force (𝐹𝑛𝑖) reaches to the target force 𝐹̅𝑛𝑆𝑃, while it maintains its position

when 𝐹𝑛𝑖≤ 𝐹̅𝑛𝑆𝑃. The movement of the cylinders had a limitation of 40 mm in stroke length, therefore the

3

(a) (b)

Fig. 1. Main structure of the experimental seat (a) and definition of thirteen adjustable parameters and different coordinate

sys-tems (b). The global coordinate system (GCS) is defined with X being directed rearward parallel to the foot support surface, Z di-rected upward perpendicular to the foot support surfaceas follows. Its origin is located at the mid of the lateral axis Y

(experi-mental reference point named PRC) which articulates the backrest frame (B) with the main supporting frame (A). Two local systems of coordinate (LCS) are defined from the global coordinate system (GCS), one attached to the seat back frame (B) after a

rotation of seat back angle (SBA), another attached to the seat pan frame (C) after a rotation of seat pan angle (SPA).

(a) (b)

Fig. 2. A participant sitting on the experimental seat (a) and the matrix of 52 cylinders (b)

Thirty-six participants (18 males, 18 females) were recruited based on their BMI (healthy 18.5-25, obese >30) and stature (small, medium and tall). Three stature groups were 154-157 cm, 162-166 cm and 170-175 cm for females; 168-171cm, 176-180 cm and 185-190 cm for males. A total of 12 groups were formed after considering gender, stature and height (3 individuals per group). Prior to the experiment, participants were screened using a health questionnaire. Participants who experienced any back injury or pain in the previous 3-months were excluded. The experimental protocol was approved by IFSTTAR (French Institute of Science and Technology for Transport, Development and Networks) ethics committee and informed consent was giv-en prior to experimgiv-ent.

Participants were instructed to test a total of 40 seat positions that simulated an economic airplane seat. The H-point location of an existing airplane seat was used to define the x position of the middle support with MS_X being fixed at 135mm, and the z position of the seat pan support with SP_Z=98 mm (Fig.1b). Two backrest angles (BA) from the horizontal (BA=100°, 110°) and three seat pan angles (SPA=0°, 5°, preferred) were used to define 6 different SPA/BA combinations. For each combination, 5 conditions were tested suc-cessively in the following order:

1. Reference position with the initial cylinder height of 20 mm (CH=20mm). This position was used to determine seat pan length, foot support height and armrests position for each participant. The three backrest panels were positioned at specific anatomical points (occipital bone, T9 and L3). Their

posi-X Z PRC A B C D E PRC SPA FS_Z FS_X Z X Zsb Xsb Zsp Xsp SBA

4

tion in x (protrusion) was fixed at 135 mm in the seat back LCS. The seat pan length (SP_X, Figure 1b) was fixed until there was approximately 70mm (hand width) between the popliteal (behind the knee) and the front of the seat pan. Participants were asked to keep their back in contact with the lower and middle supports. The foot support was adjusted (FS_Z, Figure 1b) until the knees were set at ap-proximately 90 degrees. Participants were also asked to place a rectangular foam of 100 mm (in thick-ness) between the knees to reduce postural variation. The armrests were self-positioned by subjects. Once participants were fitted to the seat, they were instructed to step off the experimental seat to zero all the force sensors. They were then asked to reposition themselves back on the experimental seat and look forward without use of the upper support. Measurements were recorded at a rate of 20 Hz for 1.25 seconds.

2. Watching a movie with CH=20mm. Participants were asked to re-place the 100mm foam between their knees and maintain a comfortable sitting position while looking forward. The uniform law was applied to distribute the seat contact force to the 52 cylinders. Once the pressure was distributed, the middle (T9) and lower (L3) supports needed to be repositioned. The level of lumbar protrusion (MS_X) was self-selected. Participants were required to remain in contact with the middle (thoracic) panel. Once the middle and lower back supports were appropriately positioned, participants were then instructed to step off the experimental seat to perform a complete zero of all the force sensors. Participants were then asked to reposition themselves back on the experimental seat and look forward without use of the upper support. Seat parameter and contact forces were then recorded.

3. Relaxing with CH=20 mm. Step 2 was repeated with the addition of the upper support. The upper sup-port was positioned at the occipital bone and its protrusion was self-selected.

4. Relaxing with CH=20mm. Step 3 was repeated with now the cylinders set at a stroke height at 40mm. 5. Watching a movie with CH=40mm. Step 2 was repeated with now the cylinders set at a stroke height

at 40mm.

In summary, in addition to seat pan (SPA) and back (BA) angles, cylinder stroke length (CH) and siting ac-tivity (task) were also considered as independent variables for the present experiment. The preferred seat pan angle was self-selected by participants for the reference position and kept unchanged for the 4 other test con-ditions with a fixed backrest angle. The five test concon-ditions of the combination with SPA=5° and BA=110° were repeated three times to measure intra individual variability. In total, 40 sitting positions were measured for each participant. The test order of six SPA/BA combinations was randomized.

2.2 Data processing

The medians of the measurements of each trial were calculated at first. After eliminating all inconsistent trials, 1383 over 1440 (36 participants x 40 conditions) trials were retained. As the sum of the external forces applied to the body had to equal to zero, the total sum of forces in the horizontal (X) and vertical (Z) had to be respectively smaller than 11 and 25N.

A principal component analysis (PCA) was used to reduce the dimensionality in data. The measurements of the dependent variables, such as foot support height, seat pan length, location of the three back supports, position of the 52 cylinders after contact force distribution, all contact forces, were gathered in a matrix Ψnxp

with n observations and p dependent variables. A smaller set of ordered variables, called principal component (PC) score, were obtained with PCA, so that the first PCs retained most of the variation in the original dataset. Assume that m main PCs µj (j =1, m) are retained. Then the i

th

observation (Ψi) containing p dependent varia-bles can be approximated:

𝛹𝑖(1: 𝑝) ≈ 𝛹̅(1: 𝑝) + ∑𝑚𝑗=1𝑐𝑗𝜇𝑗(1: 𝑝) (1)

where Ψ̅ is the average from the sample data sets and 𝑐𝑗 is the jth PC score. A linear regression can be per-formed between the m PC scores [𝐶]𝑛×𝑚 and k predictors (see Allen et al, 2003)

5

where [P]nxk is the matrix containing k predictors for n observations,

[𝑃]𝑛×(𝑘+1)= [

1 𝑃1,1 , ⋯ 𝑃𝑘,1

⋮ ⋱ ⋮

1 𝑃1,𝑛 , ⋯ 𝑃𝑘,𝑛

]

Knowing the predictors from a new observation P=[1, P1, P2, …,Pk] , m PC scores cj (j=1,m) can be obtained by

𝑐𝑗 = 𝑎0𝑗+ ∑ 𝑎𝑘𝑖 𝑖𝑗𝑃𝑖 (3)

From (1), p dependent variables of the new observation can then be predicted.

3 Results

Depending on selected dependent variables and predictors, different parametric models can be obtained. This paper will focus on seat surface and contact force distribution. Gender, stature and BMI were used as an-thropometric predictors. The seat configuration with SPA=5°, BA=110°, CH=40 mm and the task “relaxing position” was used to illustrate the parametric model obtained from the collected data. 105 observations corre-sponding to this test condition were extracted. Seat pan length; position and contact force (three components) of the 52 cylinders were selected as dependent variables. More than 90% of variance was explained by less than 20 PCs. For building parametric models, number of PCs was selected for accounting 99% cumulative variance.

Six target sitters were defined in Table 1 to show the effects of gender, stature and BMI. Small, average and tall sitters were defined as 5th_{, 50}th _{and 95}th _{percentiles in stature from National Health and Nutrition} Ex-amination Survey (NHANES III), conducted from 1988 to 1994. For each target seat occupant, the com-pressed seat surface and corresponding contact force at each cylinder can be predicted. An example is provid-ed in Figure 3 for the prprovid-ediction of a 3D seat surface and distribution of contact force for an average healthy male. Note that the maximum normal contact force was located on the cylinders under the pelvis and was ap-proximately three times of that of the first two rows of cylinders. Figure 4 compares the seat profile in XZ of the six target sitters and contact forces applied on each row of cylinders. Seat pan length was highly depend-ent on stature, a difference of 72 mm was obtained between the small healthy female and tall healthy male. Seat surface XZ profiles in the GCS were quite similar for the six target sitters. Noticeable differences could only be observed at the rear part when x>0. Force distribution was found highly dependent on BMI. Though almost the same seat profiles were obtained for the three tall target sitters, the force under the pelvis (around x=0) was much smaller for the sitter with high BMI (‘TallMaleObese’) in comparison to healthy BMI (‘TallMaleHealthy’), and the contact force was much more evenly distributed. A comparison of the YZ sur-face profiles in the seat pan LCS (at x=50, 0, -50 and -100 mm) between the six target sitters can be found in Figure 5. Compared to XZ profiles, YZ profiles in the frontal planes had greater variability between the six sitters.

Table 1. Six target seat sitters Predictor Small female

healthy Tall female obese Average male healthy Tall male underweight Tall male healthy Tall male obese Stature (mm) 1517 1737 1761 1879 1879 1879 BMI (kg/m²) 24 35 24 18 24 35

6

(a) (b)

Fig. 3. 3D representation of deformed seat surface in the seat pan LCS after interpolation (a) and corresponding normalized

nor-mal contact force Fn (1000/body weight) applied on each cylinder for the average nor-male with healthy BMI (Table 1)

Fig. 4. Seat surface profiles and contact forces in the GCS XZ plane for the six target occupants. They are represented by the

lowest position of the cylinders of each row and the sum of forces applied on the cylinders of each row. Contact forces were normalized body weight and multiplied by 1000. Grey lines are raw experimental datasets.

7

(c) (d)

Fig. 4. Seat surface YZ profiles in the seat pan LCS of the 6 target sitters at X=50, 0, -50, -100mm

4 Discussions and conclusions

3D compressed seat pan surface was experimentally investigated using a multi-adjustable experimental seat (equipped with a matrix of 52 movable cylinders on the seat pan surface). ‘Optimal’ seat pan surface was defined for this study as a compressed surface that distributes normal force as uniformly as possible with min-imal displacement of the initially flat surface. Parametric models were obtained from the data of 36 partici-pants of varying anthropometries. Main effects of gender, stature and BMI on compressed seat pan surface and pressure distribution were illustrated for a specific seat configuration (seat pan angle of 5° and back angle of 110°), results are summarized as follows:

 Similar surface XZ profiles in the seat symmetric plan (body sagittal plane) were obtained for differ-ently sized sitters, although seat pan length was highly dependent on stature.

 Surface YZ profile in the frontal plane had more variability especially for seat pan surface contact ar-ea.

 Pressure distribution was highly dependent on BMI. Pressure was less uniformly distributed for partic-ipants with a smaller BMI than those with higher BMI.

An optimal rigid seat surface is only optimal for one sitter and posture. The central question for seat sup-pliers is how to propose a unique seat that can accommodate multiple sitters of large varying body dimensions while performing different tasks. Predicting the effects of body size and posture is a first step which aims to provide quantitative guidance regarding optimal compressed seat pan surface and pressure distribution. Work is currently in progress to propose different target seat design solutions to improve the seat geometry and stiffness.

Acknowledgments: The work is partly supported by Direction Générale de l'Aviation Civile (project n°2014 930818).

References

1. Ahmadpour, A., Lindgaard, G., Robert J.M. and Pownall, B., 2014. The thematic structure of passenger comfort experience and its relationship to the context features in the aircraft cabin, Ergonomics, 57:6, 801-815, DOI: 10.1080/00140139.2014.899632

2. De Looze MP, Kuijt-Evers LF, van Dieen J, 2003. Sitting comfort and discomfort and the relationships with ob-jective measures. Ergonomics;46:985-97.

3. Zemp R, Taylor WR, Lorenzetti S, 2015. Are pressure measurements effective in the assessment of office chair comfort/discomfort? A review. Applied Ergonomics 2015;48:273-82.

8

4. Mergl C., Klendauer M., Mangen C., Bubb, H., 2004. Predicting Long Term Riding Comfort in Cars by Contact Forces between Human and Seat. SAE Technical Paper 2005-01-2690. Warrendale, PA: SAE International. 5. Beurier, G., Cardoso, M., and Wang, X., 2017. A New Multi-Adjustable Experimental Seat for Investigating

Biomechanical Factors of Sitting Discomfort. SAE Technical Paper 2017-01-1393, doi:10.4271/2017-01-1393. 6. Allen, B., Curless, B., and Popovic, Z. 2003. The space of all body shapes : reconstruction and parameterization