A model-based approach for the analysis of aircraft seating comfort

Abstract

BACKGROUND:

Aircraft seating comfort has a significant impact on passenger on-board experience. Its assessment requires the adoption of well-designed strategies for data collection as well as appropriate data analysis methods in order to obtain accurate and reliable results.

OBJECTIVES:

This paper focuses on the assessment of aircraft seating comfort based on subjective comfort responses collected during laboratory experiments and taking into account seat features and passenger characteristics.

METHODS:

The subjective comfort evaluations have been analyzed using a model-based approach to investigate the relationship between overall seating comfort and specific seat/user characteristics.

RESULTS:

The results show that the overall seating comfort perception is significantly influenced by the thickness of the seat pan, the backrest position (upright or reclined), the age of the passenger and the passenger perception of being comfortably supported at the lumbar region.

CONCLUSIONS:

The adopted model-based approach allows the analysis of subjective seating comfort data taking into account their ordinal nature as well as the dependency between evaluations provided by the same subject.

Keywords

Seat comfort laboratory experiments ordinal regression repeated evaluations

1 Introduction

In the last decades, commercial air traffic and number of passengers have been constantly increasing and the competition among airlines has intensified, resulting in airlines seeking ways to differentiate their products and services and employing market penetration strategies based on price, point-to-point connections, timeliness, flight frequency and service quality [1 –3].

Being strictly related to passenger satisfaction and willingness to pay, the improvement of on-board comfort has become a strategic goal for the airline management [4].

Literature studies highlight that passenger comfort experience mainly depends on sitting comfort [e.g. 5–9] and that, to attract passengers, seats should take into account the diversity of passengers (e.g. an-thropometry, state of mind and expectations) and the activities they want to perform during travel [10, 11].

In recent years the number of papers touching com-fort knowledge continues to expand, but the theo-retical foundations for comfort research remain underdeveloped [12]. Several theoretical models have been proposed to explain and describe (dis-)comfort [e.g. 11, 13–15], but none of these models is able to predict either comfort or discomfort.

A recent literature review [16] evidences the necessity of more research to enable a better prediction, especially in the field of passenger seat (dis-)comfort (as opposed to driver’s (dis-)comfort). The main factors related to sitting (dis-)comfort (i.e. human, context and seat characteristics) have only been considered in separate studies and the relationships between them and passenger (dis-)comfort remain unclear due to a lack of statistical evidence and large differences in research set-ups.

In this study, the focus is on comfort data collected during a laboratory experiment planned to explore the effects of seat design parameters and passenger characteristics on overall perceived comfort.

The data have been modeled using a cumulative link mixed model (CLMM) that is an extension of linear mixed models for ordinal data. The higher computational complexity of the adopted CLMM is counterbalanced by its higher flexibility that accounts for both the ordinal nature of the comfort evaluations and the dependency between evaluations provided by the same subject [17, 18].

2 Overview of the experiment

The experiment involved 17 frequent flyers (Table 1) with no history of back problems. Each participant was involved in 5 seating test sessions. Ethics approval was arranged and participants signed an informed consent.

Table 1
Main anthropometric characteristics of participants

Num Age [year] Weight [kg] Height [m] BMI [kg/m²]

[min-max] [min-max] [min-max] [min-max]

Males 9 [27–41] [73–101.8] [1.60–1.90] [22.8–34.7]

Mean (SD) 35 (4.4) 88(8.53) 1.77 (0.08) 28.03 (3.46)

Females 8 [26–44] [55–75] [1.55–1.73] [21.15–27.55]

Mean (SD) 34 (5.9) 66 (5.4) 1.66 (0.05) 24.1 (2.08)

	Num	Age [year]	Weight [kg]	Height [m]	BMI [kg/m²]
Males	9	[27–41]	[73–101.8]	[1.60–1.90]	[22.8–34.7]
Mean (SD)		35 (4.4)	88(8.53)	1.77 (0.08)	28.03 (3.46)
Females	8	[26–44]	[55–75]	[1.55–1.73]	[21.15–27.55]
Mean (SD)		34 (5.9)	66 (5.4)	1.66 (0.05)	24.1 (2.08)

Participants assessed the comfort of 2 typical double-seats identified as “baseline configurations” (denoted seat A and seat B) and 1 lightweight double-seat (denoted C; confidential). The three double-seats under study differ in weight, reclining, headrest and dimensions of seat pan and backrest (Table 2).

Table 2

Seat dimensions

	Seat A	Seat B	Seat C
Height of seat [mm]	1087	1152	1136.6
Height of seat pan [mm]	435	440	453.2
Width of seat pan [mm]	424.9	424.9	457.2
Backrest configuration	Hybrid	Hybrid	Full composite
Height of backrest [mm]	707	772	780
Width of backrest [mm]	427	450	444.5
Reclining	No	Yes	Yes

Seat A, being not reclining, was tested only in upright position while seat B and seat C were tested both in upright and reclined position.

The test sessions were planned using a cross-over design [19] and were carried out following a detailed experimental protocol. At the end of each test session, lasting about 40 minutes, with fixed posture and task (reading/playing a game with the smartphone), the participant evaluated the comfort of specific seat features using a scale with three ordered categories (i.e. 1: low comfort, 2: medium comfort, 3: high comfort) and scored the overall seating experience using an ordinal scale ranging from 0 (i.e. no comfort) to 8 (i.e. extreme comfort).

3 Methods

Comfort ratings have been analyzed in a regression setting using a set of covariates representing: 1) objective seat features (viz. height of seat, height of seat pan, width of seat, backrest configuration, height of backrest, thick of backrest, backrest reclining); 2) participant anthropometrical characteristics (viz. gender, age and BMI); 3) participant feelings of comfort with specific seat features (viz. seat pan, backrest, seat pan padding, backrest padding, lumbar support, lumbo-sacral support).

The CLMM relies on the idea that a subjective evaluation expressed on an ordinal scale (e.g. comfort rating) is actually a categorized version of an unobservable (latent) continuous variable.

Let Y_i the outcome category selected by subject i for the response variable. Given a set of p covariates, x₁,..,x_k, . . . x_p, the CLMM can be formulated as follows [17]:

$\begin{matrix} logit [P (Y_{it} \leq j)] = u_{i} + α_{j} + β_{1} x_{1 it} \\ + β_{2} x_{2 it} + \dots + β_{p} x_{pit} \end{matrix}$ (1)

The parameter β _k measures the impact of x_k on Y; the parameters α _j are the category cut-points on a standardized version of the latent variable; u_i is the random effect due to subject i for response categories j = 1, 2, . . . , J-1; it is assumed normally distributed and centered at zero (u_i ∼ N (0, σ _u ² )).

The intra-class correlation (ICC) is a way to look at the correlation of observations within a group, it is calculated as follows: $ICC = \frac{σ_{u}^{2}}{{\hat{σ}}_{u}^{2} + σ_{u}^{2}}$ (2)

where ${\hat{σ}}_{u}^{2}$ and σ² represent the estimated variance of the random effect and the residual variance, respectively. Values of ICC near one indicate that observations within a group are very correlated; whereas values of ICC close to zero indicate that the random effect can be neglected since observations within a group are nearly independent [17].

4 Results

The optimal fitted CLMM for the analyzed comfort data has 4 significant covariates: age (X _a ), comfort feeling with lumbar support (X _l ), height of seat pan (X _h ) and backrest position (X _r ). The estimated coefficients of the significant covariates ${\hat{β}}_{k}$ are reported in Table 3 together with their asymptotic standard error.

Table 3
Coefficients of the significant covariates for the optimal CLMM (asymptotic standard error, in parentheses)

${\hat{β}}_{X_{a}}$ ${\hat{β}}_{X_{l}}$ ${\hat{β}}_{X_{h}}$ ${\hat{β}}_{X_{r}}$

–0.82 –1.48 0.83 2.01

(0.41) (0.35) (0.29) (0.47)

${\hat{β}}_{X_{a}}$	${\hat{β}}_{X_{l}}$	${\hat{β}}_{X_{h}}$	${\hat{β}}_{X_{r}}$
–0.82	–1.48	0.83	2.01
(0.41)	(0.35)	(0.29)	(0.47)

The coefficients ${\hat{β}}_{X_{a}}$ and ${\hat{β}}_{X_{l}}$ are negative indicating that the probability of low overall comfort ratings increases with the age of the participants and the passenger feeling of a comfortable support in the lumbar area. On the contrary, the coefficients ${\hat{β}}_{X_{h}}$ and ${\hat{β}}_{X_{r}}$ are positive indicating that the probability of higher overall comfort ratings increases for seats with higher seat pan and for seats in reclined position.

The ${\hat{σ}}_{u}^{2}$ = 0.003 for the random effects model implies a low participant effect; ICC equals to 0.0009 confirms the substantial independency of comfort evaluations provided by the same participants for different seat conditions.

5 Discussion

Despite the small size of the sample, the study provides significant results. The optimal model has been identified with 4 significant covariates: two objective seat features, height of seat pan and backrest position, and two subjective covariates, comfort feeling with lumbar support and age. In the following, these findings will be compared to the field literature.

The significance of the height of seat pan, measured as the distance between the top of the seat pan and the floor, confirms the criticality of this parameter for the design of the seats. In a recent experimental study investigating the effects of seat parameters and sitter anthropometric dimensions on seat profile and optimal compressed seat pan surface, Peng at al. [20] found that height of seat pan was dependent on seat pan angle, stature, sitting height to stature ratio and BMI.

Some attempts have been already made to derive the optimal value of the height of the seat pan from the dimensions of the popliteal height of the end user [21]. However, it is clear that, depending on the amount of adjustability, it will be difficult to define dimensions that include the entire population. Referring to a population of passengers aged between 20 and 60 with a distribution made up of 50% of male and 50% of female passengers [22], Hiemstra-van Mastrigt concludes that a 10 mm increase in the height of the seat pan from 420 to 430 mm will include an additional 11% of passengers, but an increase of the same width, from 470 to 480 mm, will include only an additional 0.4% of passengers [21]. Therefore, a careful selection of this dimension is necessary in order to achieve an optimum trade-off between including people, increasing the comfort experience and an efficient use of space.

The significance of the backrest position confirms that a tilted sitting position improves passenger comfort. Several literature studies have investigated the benefits of tilted position with respect to the reduction of sitter discomfort. The tilt angle determines body posture, which is related to the interface pressure [11, 16]. Vos et al. [23] found that an increased torso–thigh angle reduced the pressure coefficient value (peak pressure, average pressure), which in turn reduces subjective discomfort. Lueder’s study [24] demonstrates that a more dynamic seat, with the possibility of varying the posture adopted (for example by the reclining of the seat), reduces the perceived discomfort.

At first sight, the most surprising result of our study concerns the negative relationship between the overall seating comfort perception and the feeling of being comfortably supported at the lumbar area. The impact of the lumbar support on subjective feeling of comfort is largely unknown. The lumbar support design is usually motivated by the idea that a seat to be comfortable should preserve the curve in the low back (i.e. low back lordosis). It is widely understood that lumbar lordosis decreases as the angle between the trunk and hip approaches 90 percent, as in an erect sitting posture [25]. However, a recent study on office chairs has discussed a paradoxical behavior: the seat designed to ensure correct lumbar curvature recorded the highest level of discomfort and pain for the lower back by the evaluators who participated in the tests [26]. Our findings seem to confirm this paradoxical behavior and thus suggest another interesting point that deserves more research investigation.

Finally, with respect to the covariate age, our study evidences that the probability of low comfort rating becomes higher as the age of participant decreases. Despite the narrow range considered (26 to 44 years), this result seems to confirm the findings obtained in other comfort studies investigating the effect of age on (dis-)comfort perceptions.

Kyung et al. [27], in a study in the automotive field, compare the perceptions of (dis-)comfort and the pressure distribution of two groups of drivers: young (between 20 and 35 years) and elderly (with an age greater than 60 years). No significant differences between the two groups emerge in terms of overall comfort, whereas, a significant difference emerges from the point of view of local comfort assessments (i.e. relating to specific areas of the body analyzed during the study): younger drivers reported lower levels of local comfort than the elderly. Kyung et al. [27] also highlight that the younger drivers appeared to be more sensitive to discomfort than older drivers. A similar result emerges from a study conducted by Lijmbach et al. [28] about the aircraft seat in-and egress differences between elderly (with a mean age of 75 years) and young adults (with a mean age of 22 years). Most elderly who participated in the research were extremely positive about the comfort of the seat. This hindered the research and made it very difficult to compare the comfort evaluations provided by the two groups of participants.

Participant effect resulted negligible in our study, this finding could be related to the involvement of expert and trained assessors (i.e. frequent flyers) who may show less individual psychological biases in the evaluation task. However, since psychological and physiological biases generally affect the subjective assessment in a sample set, assessor’s effect cannot be disregarded.

Besides the above interesting results, it is worthwhile to note that the small number of participants does not allow a generalization of the obtained results.

6 Conclusions

This paper focuses on the assessment of aircraft seating comfort based on subjective responses collected during laboratory experiments. The adopted data analysis strategy, based on CLMM, allows to investigate the strength and direction of association in subjective comfort data taking into account their ordinal nature as well as the potential grouping structure of replicated observations, overcoming the hypothesis of independency that is often unrealistic in experimental settings.

The proposed strategy is appropriate for the analysis of discomfort ratings as well. In future research, it would be interesting to investigate which factors significantly influence overall discomfort perception and also to extend the investigation by including for example the impact of aircraft vibrations on the perception of passenger (dis)comfort.

Conflict of interest

None to report.

References

Chen

, Peng

, Hackley

. Evaluating service marketing in airline industry and its influence on student passengers’ purchasing behavior using Taipei–London route as an example. Journal of Travel & Tourism Marketing. 2008;25(2):149–60.

Atalık

, Bakır

, Akan

. The Role of In-Flight Service Quality on Value for Money in Business Class: A Logit Model on the Airline Industry. Administrative Sciences. 2019;9(1):26.

Delbari

, Ng

, Aziz

, Ho

. An investigation of key competitiveness indicators and drivers of full-service airlines using Delphi and AHP techniques. Journal of Air Transport Management. 2016;52:23–34.

Patel

, D’Cruz

. Passenger-centric factors influencing the experience of aircraft comfort. Transport Reviews. 2018;38(2):252–69.

Vink

, Van Mastrigt

. The aircraft interior comfort experience of 10,032 passengers. In Proceedings of the human factors and ergonomics society annual meeting 2011 Sep (Vol. 55, No. 1, pp. 579-583). Sage CA: Los Angeles, CA: SAGE Publications.

Ahmadpour

, Robert

, Lindgaard

. Aircraft passenger comfort experience: Underlying factors and differentiation from discomfort. Applied Ergonomics. 2016;52:301–8.

Bouwens

, Hiemstra-van Mastrigt

, Vink

. Ranking of human senses in relation to different in-flight activities contributing to the comfort experience of airplane passengers. International Journal of Aviation, Aeronautics, and Aerospace. 2018;5(2):9.

Zhao

, Yu

, Miller

, Ghulam

, Li

, Wang

. Predicting aircraft seat comfort using an artificial neural network. Human Factors and Ergonomics in Manufacturing & Service Industries. 2019;29(2):154–62.

Brindisi

, Concilio

. Passengers’ Comfort Modeling Inside Aircraft. Journal of Aircraft. 2008;45(6):2001–8.

10.

Vink

, Bazley

, Kamp

, Blok

. Possibilities to improve the aircraft interior comfort experience. Applied Ergonomics. 2012;43(2):354–9.

11.

Groenesteijn

, Hiemstra-van Mastrigt

, Gallais

, Blok

, Kuijt-Evers

, Vink

. Activities, postures and comfort perception of train passengers as input for train seat design. Ergonomics. 2014;57(8):1154–65.

12.

Mansfield

, Naddeo

, Frohriep

, Vink

. Integrating and applying models of comfort. Applied Ergonomics. 2020;82:102917.

13.

De Looze

, Kuijt-Evers

, Van Dieen

. Sitting comfort and discomfort and the relationships with objective measures. Ergonomics. 2003;46(10):985–97.

14.

Kuijt-Evers

, Groenesteijn

, de Looze

, Vink

. Identifying factors of comfort in using hand tools. Applied Ergonomics. 2004;35(5):453–8.

15.

Vink

, Hallbeck

. Comfort and discomfort studies demonstrate the need for a new model.

16.

Hiemstra-van Mastrigt

, Groenesteijn

, Vink

, Kuijt-Evers

. Predicting passenger seat comfort and discomfort on the basis of human, context and seat characteristics: a literature review. Ergonomics. 2017;60(7):889–911.

17.

Agresti

. Analysis of ordinal categorical data. John Wiley & Sons; 2010 Apr 19.

18.

Christensen

, Brockhoff

. Analysis of sensory ratings data with cumulative link models. Journal de la Societe Francaise de Statistique & Revue de Statistique Appliquee. 2013;154(3):58–79.

19.

Jones

, Kenward

. Design and analysis of cross-over trials. CRC press; 2014.

20.

Peng

, Wang

, Denninger

. Effects of anthropometric variables and seat height on automobile drivers’ preferred posture with the presence of the clutch. Human Factors. 2018;60(2):172–90.

21.

Hiemstra-van Mastrigt

. Comfortable passenger seats: Recommendations for design and research. TU Delft, Delft University of Technology.

22.

DINED2004. Website: www.dined.io.tudelft.nl.

23.

Vos

, Congleton

, Moore

, Amendola

, Ringer

. Postural versus chair design impacts upon interface pressure. Applied Ergonomics. 2006;37(5):619–28.

24.

Lueder

, Rice

, editors. Ergonomics for Children: Designing products and places for toddler to teens. CRC Press; 2007 Jul 25.

25.

Vink

. Aircraft interior comfort and design. CRC press; 2016 Apr 19.

26.

Santo

, Araújo

. Assessment of changes in spine curvatures and the sensations caused in three different types of working seats. Motriz: Revista de Educação Física. 2017;23(3).

27.

Kyung

, Nussbaum

. Age-related difference in perceptual responses and interface pressure requirements for driver seat design. Ergonomics. 2013;56(12):1795–805.

28.

Lijmbach

, Miehlke

, Vink

. Aircraft seat in-and egress differences between elderly and young adults. In Proceedings of the Human Factors and Ergonomics Society Annual Meeting 2014 Sep (Vol. 58, No. 1, pp. 520-524). Sage CA: Los Angeles, CA: Sage Publications.

A model-based approach for the analysis of aircraft seating comfort

Abstract

BACKGROUND:

OBJECTIVES:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1 Introduction

2 Overview of the experiment

Table 3 Coefficients of the significant covariates for the optimal CLMM (asymptotic standard error, in parentheses) β ˆ X a β ˆ X l β ˆ X h β ˆ X r –0.82 –1.48 0.83 2.01 (0.41) (0.35) (0.29) (0.47)

6 Conclusions

Conflict of interest

References

Table 3
Coefficients of the significant covariates for the optimal CLMM (asymptotic standard error, in parentheses)

${\hat{β}}_{X_{a}}$ ${\hat{β}}_{X_{l}}$ ${\hat{β}}_{X_{h}}$ ${\hat{β}}_{X_{r}}$

–0.82 –1.48 0.83 2.01

(0.41) (0.35) (0.29) (0.47)