Analysis and Evaluation of Ride Comfort on Standing Posture
T. KOIZUMI, N. TSUJIUCHI, M.OKAMURA, and Y.HASHIMOTO
Department of Mechanical Engineering
Doshisha University
1-3 Tatara-Miyakodani, Kyotanabe, Kyoto 610-0321, JAPAN
Nomenclature
A0
Ai
N
Xi
X 11
X 12
X 21
X 22
X 23
Y
: Constant term
: Partial regression coefficient
: Number of independent variable
: Independent variable
: Acceleration of the knee in the vertical direction at 12.5 Hz [m/s2]
: Acceleration of the shoulder in the lateral direction at 20 Hz [m/s2]
: Acceleration of the waist in the vertical direction at 2.0 Hz [m/s2]
: Acceleration of the knee in the fore-and-aft direction at 1.0 Hz [m/s2]
: Height of subjects [m]
: Dependent variable
Abstract
This paper presents an investigation into the dynamic characteristics of the human body in relation to ride comfort
of trains in a standing posture.
Many researches have been conducted on ride comfort of vehicles; however, most of them focus on a sitting
posture, e.g., in a car or on a high-speed train [1]-[6]. In addition, ride comfort has been evaluated only from the
acceleration of floors or sheets [5]-[7]. Also, however, ride comfort should be evaluated by taking into account the
swaying of body parts induced by floor vibration. Therefore, for better accuracy, ride comfort should be evaluated
considering human dynamics.
This study aims to clarify the relation between ride comfort and dynamic characteristics of the human body in a
standing posture. For this purpose, frequency responses of the human body were measured by exciting sinusoidal
wave and reproducing train floor vibration. The subjects evaluated ride comfort of the reproduced vibrations.
As a result, some features of the dynamic characteristics in a standing posture and the relation of ride comfort with
the dynamic characteristics were clarified. The evaluation equations were proposed using a multiple regression
analysis.
1. Introduction
Although people are often exposed to vibration in a standing posture, e.g., in a commuter train or bus, few studies
have been conducted on ride comfort. Here, ride comfort in a standing posture on vehicles is evaluated by the
same method as in a sitting posture, e.g., ISO 2631 (1995) and the JNR Riding Comfort Level methods (see 3.5).
The dynamic characteristics of the human body in a standing posture differ greatly from a sitting one; therefore,
another method for evaluating ride comfort specialized for a standing posture seems to be necessary.
Another problem is that ride comfort of vehicles has been evaluated only from the acceleration of floors or sheets.
Ride comfort is also supposed to be affected by the swaying of body parts induced by floor vibration because some
parts of the body, such as the head, are particularly sensitive to swaying. Therefore, to attain better accuracy, ride
comfort should be evaluated considering human dynamics.
The purpose of this study is to clarify the relation between ride comfort and the dynamic characteristics of human
body in a standing posture from questionnaires and frequency responses of some of the body parts. For this
purpose, excitation experiments are conducted employing sinusoidal waves and reproduced train floor vibrations.
2. Excitation Experiment under a Single- Dimensional Swept Sinusoidal Wave
2.1 Experimental Apparatus
Fig. 1 shows the experimental apparatus for excitation. Frequency and acceleration of input vibration are controlled
by a PC connected to this two-dimensional shaker, and a subject stands on the shaker’s moving platform. The
frequency response of the head, shoulder, waist, and knee to the floor are measured by charge accelerometers,
which are attached with a metal bite-bar for the head measurement and with rubber bands for the other points.
Fig.1: Experimental apparatus
2.2 Experimental Conditions
A male, 22 years old, 58 kg in weight, and 1.72 m tall subject was exposed to a single-dimensional swept
sinusoidal wave in three directions: vertical, lateral, and in the fore-and-aft direction. To minimize variability, the
subject was requested to look straight ahead, maintain an erect posture, and try not to change it during the
excitation. Separation between the feet was set to 30 cm. He was exposed to the wave seven times in each
direction.
Acceleration of the swept sinusoidal wave was set to 1.0 m/s2 and the frequency range was set to 1-30 Hz in the
vertical, 1-20 Hz in the horizontal with consideration given to the dominant frequency range of train floor vibration,
specifications of the shaker, and a report that ride comfort deteriorates mainly due to vibration in the frequency
range of 0.5-20 Hz in the vertical, 0.5-10 Hz in the horizontal [8]. The wave was swept in steps of 1 Hz for 10 s/Hz.
2.3 Result
Figs. 2 shows the median of the frequency response, with each sub-figure showing the transmissibility of each
body part. The first resonance peak for the vertical direction is observed at around 4-5 Hz in all body parts, while
second resonance peak is observed at around 10 Hz at the waist and 18 Hz at the head and knee. As for the
lateral direction, a resonance peak is observed at 9 Hz at the waist and the knee, indicating that the resonance
frequency of the torso is 9 Hz. For the fore-and-aft direction, the first resonance peak is observed at around 4 Hz in
all parts except for the head, and the second resonance peak is observed at around 8-9 Hz at the knee.
A significant fall in the head and shoulder transmissibility in the lateral direction and the head transmissibility in the
fore-and-aft was observed compared to the result of other body parts and of the vertical direction. The cause of this
fall is the wide mobility range of the torso and the legs in the lateral and fore-and-aft direction. With this mobility
range, the torso and the legs absorbed most of the vibration.
－index－
head
shoulder
waist
knee
Transmissibility
2.5
2.0
1.5
1.0
0.5
0.0
0
5
10
15
20
25
30
Frequency [Hz]
1.2
1.2
1.0
1.0
Transmissibility
Transmissibility
(a) Vertical
0.8
0.6
0.4
0.2
0.8
0.6
0.4
0.2
0.0
0.0
0
5
10
Frequency [Hz]
15
20
0
5
10
Frequency [Hz]
(b) Lateral
(c) Fore-and-aft
Fig.2: Frequency response under sinusoidal wave
(a) Vertical
(b) Lateral
Fig.3: Average frequency response of sitting posture [1]
15
20
Figs. 3 shows an average frequency response of 26 subjects in the sitting posture, the data taken from Ueda [1].
By comparing Figs. 2 and 3, we see a significant difference in the dynamic characteristics in the lateral direction.
Here, a resonance peak is observed at around 1-2 Hz at each body part, with the head transmissibility exceeding
2.0 in the sitting position. In comparison, a clear resonance peak is observed only at the waist and knee at 9 Hz,
and the head transmissibility is less than 0.1 in the standing posture. The cause appears to be the difference of
vibration mode due to the posture change. In the sitting posture, while damping is observed in the shoulder and
waist that were near the contact plane with the chair, resonance phenomena is observed in the head and knee that
were comparatively distant from the chair. Furthermore, in the sitting posture, the transmissibility decreases from
the knees toward head, meaning that most of the vibration is absorbed by the legs and waist, as mentioned before.
The cause of the damping may be a characteristic of the skinfold under sheering force. Fig. 3(b) shows the
frequency dependence of the damping effect. However, because this experiment employed the swept sinusoidal
wave, the range of the movement narrows as the frequency rises. That is to say, the effect of damping depends on
the range of movement. This statement agrees with the fact that the mobility range of the skinfold has a limit and it
doesn’t operate as a buffer between the skin and bones when the mobility range is exceeded.
3. Excitation and Evaluation Experiment under a Reproduced Train Floor Vibration
Experiments to evaluate ride comfort were conducted using two-dimensional reproduced train floor vibrations. Due
to the restriction of the shaker, two experiments were conducted: vertical and lateral input (Exp.Ⅰ), and vertical
and fore-and-aft input (Exp.Ⅱ).
3.1 Experimental Method
The same apparatus was employed for these experiments (see 2.1); however, two accelerometers were attached
to each point for taking measurements in the vertical and horizontal direction.
All experiments employed six types of two-dimensional reproduced train floor vibrations. For the evaluation not to
be difficult, these vibrations were amplified by 1-3 times, except for the Reference by 3-5 times. Table 1 shows the
r.m.s. values and standard deviations of the accelerations of these vibrations. The procedure involved exposing
subjects to the 30-sec Reference vibration and one of the 250-sec Sample vibrations randomly chosen from the
waveform A-E with a short interval between them. Subjects then evaluated ride comfort of the Sample vibration in
the range from 0 to -3 in increments of 0.1 points, where 0 is equivalent to No sensation and –3 is to ride comfort
under the Reference vibration. This procedure was repeated five times to per subject, with different Sample
vibrations each time. The physical characteristics of the group of subjects (male university students) are
summarized in Table 2, and instructions provided to the subjects are shown in the Appendix.
Table 1: Vibration magnitude
r.m.s.±standard deviation [m/s2]
Table 2: Physical characteristics of subjects
Exp.Ⅰ
17 subjects
Exp.Ⅱ
18 subjects
Waveform
Vertical
Horizontal
A
0.168±0.002
0.222±0.004
B
0.167±0.001
0.440±0.007
Minimum
49
1.64
49
1.64
C
0.276±0.002
0.328±0.004
Maximum
80
1.83
80
1.83
D
0.344±0.002
0.139±0.004
Mean
60.8
172.9
61.4
172.8
E
0.342±0.002
0.250±0.003
S.D.
8.9
5.2
9.6
5.2
Reference
0.704±0.008
1.319±0.018
Weight [kg] Stature [m] Weight [kg] Stature [m]
3.2 Multiple Regression Analysis
This study employed a multiple regression analysis to relate human dynamics to the subjects’ ride comfort. We
assumed that ride comfort could be explained by the following regression equation:
N
Y ＝ A0 ＋ ∑ Ai X i
(1)
i =1
Here, height [m], weight [kg], and frequency analysis data of a 1/3-octave band applied to accelerations [m/s2] of
the head, shoulder, waist, and knee with the frequency range of 0.63-40 Hz are used as independent valuables X i ,
while the evaluation point of waveforms is used as a dependent valuable Y . The relation was assumed to be linear.
Prior to analysis, some trials whose evaluation points were unique were excluded, e.g., the case of feeling better
even when the vibration increased. Six trials were excluded out of 85 in Exp.Ⅰ and seven were excluded out of 90
in Exp.Ⅱ. To select the independent valuables, a test of uncorrelation (p < 0.05) was performed first, and a
stepwise method was used in the multiple regression analysis.
3.3 Result of Frequency Response
The median of the frequency response for the same subject as in Fig. 2 is shown in Figs. 4 and 5. For easy
comparison, the results of Fig. 2 are also plotted.
－index－
1.2
-0.8
head
0
10
shoulder
waist
knee
Dot: sinusoidal wave, Line: reproduced vibration
20
1.2
2.5
1
Transmissibility
Transmissibility
2.0
1.5
1.0
0.5
0.8
0.6
0.4
0.2
0
0.0
0
5
10
15
20
25
0
30
5
10
15
20
Frequency [Hz]
Frequency [Hz]
(a) Vertical
(b) Lateral
Fig.4: Frequency response under sinusoidal wave and reproduced vibration of Exp.Ⅰ
－index－
1.2
-0.8
head
0
10
shoulder
waist
knee
Dot: sinusoidal wave, Line: reproduced vibration
20
1.2
2.5
1
Transmissibility
Transmissibility
2.0
1.5
1.0
0.5
0.8
0.6
0.4
0.2
0
0.0
0
5
10
15
20
Frequency [Hz]
(a) Vertical
25
30
0
5
10
15
Frequency [Hz]
(b) Fore-and-aft
Fig.5: Frequency response under sinusoidal wave and reproduced vibration of Exp.Ⅱ
20
－wave type index (see Table1)－
A
B
C
－index－
D
E
Reference
0.5 m/ss
1.0 m/ss
2.0 m/ss
Dot: measured value, Line: average
2.0
Transmissibility
Transmissibility
3.0
1.5
1.0
0.5
2.5
2.0
1.5
1.0
0.5
0.0
0.0
0
5
10
15
20
25
30
0
Frequency [Hz]
(a) Vertical head motion under reproduced vibration
5
10
15
Frequency [Hz]
(b) Vertical shoulder motion under sinusoidal wave
Fig.6: Verification of amplitude dependency
When comparing with Fig. 2, it is clear that increases in the transmissibility and resonance frequency occur,
especially in the vertical direction. These are the effects of the human body’s amplitude dependence. Indeed, the
r.m.s. value of the sinusoidal wave was 0.707 m/s2, whereas r.m.s. values of the Sample vibrations ranged from
0.167-0.344 m/s2 in the vertical direction. Fig. 6(a) shows the frequency response of the head in the vertical
direction for every waveform: as the magnitude rises from waveform A to Reference, the transmissibility decreases.
Fig. 6(b) shows the result of an additional experiment using sinusoidal vertical waves with amplitudes of 0.5, 1.0,
and 2.0 m/s2. The resonance frequency falls from 5 Hz in 0.5 m/s2 to 4 Hz in 2.0 m/s2. Through these figs., the
amplitude dependence of both resonance frequency and the transmissibility is verified.
3.4 Result of Analysis of Ride Comfort
The following equation was obtained for Exp.Ⅰas a result of the analysis:
Y = － 5.947 X11 － 9.585 X12＋ 0.056
(2)
The coefficient of determination adjusted for degrees of freedom of this equation is 0.465 and this equation is
significant (p < 0.001). Regression coefficients are shown in Table 3. Fig. 7 shows the observed experimental
values and the suggested values by the regression equation in each trial. Despite using only two independent
variables, this figure indicates good predictability of this model.
As for Exp.Ⅱ, the following Eq. (3) was obtained.
Y = －19.03 X21 － 32.89 X22 － 3.174 X23 ＋ 6.142
(3)
The adjusted COD is 0.442 and this equation is significant (p < 0.001). Regression coefficients are shown in Table
4. The analytical result indicates the greater the height, the worse the ride comfort becomes, a result that agrees
with the characteristic of the human body to easily lose balance, especially in the fore-and-aft direction and the fact
that the higher the center of gravity, the easier it is to lose balance. Fig. 8 shows the observed values and the
suggested values in each trial, and despite using only three independent variables, it clearly indicates good
predictability.
Table 4: Result of analysis of Exp.Ⅱ
Table 3: Result of analysis of Exp.Ⅰ
Partial
regression
coefficient
Standard partial
regression
coefficient
vertical knee 12.5Hz
-5.947
-0.5730
lateral shoulder 20Hz
-9.585
-0.3167
0.05581
---
Constant term
－index－
Partial
regression
coefficient
Standard partial
regression
coefficient
vertical waist 2.0Hz
-19.03
-0.5372
fore-and-aft knee 1.0Hz
-32.89
-0.3699
height
-3.174
-0.3146
Constant term
6.142
---
－index－
observed value
suggested value
observed value
Trial
0
0
10
20
30
40
50
60
70
0
-0.5
10
20
30
40
50
60
70
80
-0.5
Evaluation
Evaluation
Trial
0
80
suggested value
-1
-1.5
-2
-1
-1.5
-2
-2.5
-2.5
-3
-3
Fig.7: Predictability of Exp.Ⅰ
Fig.8: Predictability of Exp.Ⅱ
3.5 Comparison with the Conventional Method
In this section, the effectiveness of considering human dynamics was tested using the same regression model as
Eq. (1). As in the comparison contrasts, the “JNR” and “Floor Only” models were employed. In the JNR model, the
Riding Comfort Levels* in two directions [m/s2] were used as independent variables, i.e., the level of the vertical
and lateral direction for Exp.Ⅰ, that of the vertical and fore-and-aft for Exp.Ⅱ. The Riding Comfort Level was
actually proposed by the now-defunct Japan National Railways in 1980 and it is now commonly used to evaluate
ride comfort of trains in Japan. It is calculated for each direction by filtering the floor vibration with the amplitude
shown in Fig. 9. The Floor Only model is our original model. Frequency analysis data of 1/3-octave band applied to
floor accelerations [m/s2] with the frequency range of 0.63-40 Hz were used as independent variables.
Table 5 presents the result of the analysis. For Exp.Ⅱ, the two cases that consider height and the one that does
not are shown to be comparable under the same conditions with a number of variables. The proposed evaluation
methods give the best adjusted COD and clarify the effectiveness of considering human dynamics.
Amplitude
Table 5: Adjusted COD
vertical vibration
horizontal vibration
Frequency [Hz]
Fig. 9: Ride Comfort Filter by JNR [9]
Evaluation
method
Exp.Ⅰ
2 variables
Proposed
JNR
Floor only
0.4650
0.3066
0.3247
Exp.Ⅱ
with height
3 variables
0.4419
0.2775
0.3627
Exp.Ⅱ
w/o height
2 variables
0.3475
0.1708
0.2456
* The original JNR’s Riding Comfort Level is represented
2
in dB units. However, the m/s unit was used in the
analysis because the predictability and the correlation
with observed values were higher than the dB unit.
4. Conclusion
Investigations on the dynamic characteristics of the human body and ride comfort of trains in a standing posture
were conducted. The following conclusions are obtained.
1. The transmissibility of the head, shoulder, waist, and knee in a standing posture was measured, and it was
revealed that under lateral and fore-and-aft vibration, the waist and legs absorbed most of the vibration.
2. The difference in the dynamic characteristics of the human body between standing and sitting postures was
clarified.
3. The amplitude dependence of resonance frequency and transmissibility was observed.
4. Experimental evaluations of ride comfort were conducted, and with a multiple regression analysis, evaluation
equations were proposed and the body parts and frequency that strongly affect ride comfort were identified.
5. Results indicated that the proposed evaluation methods showed good predictability.
6. The proposed evaluation methods clarify the effectiveness of considering human dynamics.
Acknowledgements
This work is partially supported by Grant-in-Aid for Scientific Research (c) (15560210).
References
[1] K. Ueda, A study on the evaluation of ride comfort due to human dynamic characteristics and modeling of
human body, Doshisha Univ. master’s thesis, pp2-6, (1999)
[2] T. Koizumi et al., Analysis of the behavior of human body under biaxial vibration, Proceedings of IMAC XXI,
#128, (2003)
[3] Yoshihiko Ozawa et al., Analysis of the ride comfort, Proceedings of JSAE 842, pp359-364, (1984)
[4] K. C. Parsons and M. J. Griffin，The Effect of the Position of the Axis of Rotation on the Discomfort Caused by
Whole-Body Roll and Pitch Vibrations of Seated Persons，Journal of Sound and Vibration, 58(1), pp127-141,
(1978)
[5] Shunichi Doi et al., Evaluation of Ride Comfort on Vehicle Harshness, Proceedings of the 1st JSME transport
and logistics division conference, No.920-98, pp267-272, (1992)
[6] Hiroaki Suzuki et al., Ride Comfort of Train, RRR 1995-2, pp6-13, (1995)
[7] Hideyuki Takai, Changes in Evaluation Methods for Riding Comfort, RTRI Report, vol.9, No.8, pp61-66, (1995)
[8] Hiroaki Suzuki, WG9(Train Vibration), Noise Control, Vol.25, No.6, pp363-366, (2001)
[9] JSME, Dynamics of Railway Car, Electric Car Workshop, pp72, (1994)
Appendix: Subject Instruction
The following are the instructions that were given to the subjects taking part in the evaluation experiments.
•
•
•
•
•
•
This experiment consists of three parts: 30-sec Reference vibrations, 250-sec Sample vibrations, and
Evaluation of the sample vibrations. Each trial is to be repeated five times under different Sample vibration.
During the excitation, concentrate on evaluating ride comfort with the vibrations and think about nothing else.
Look straight ahead, maintain an erect posture, and try not to change it.
Evaluate only the vibration (do not take discomfort due to noise and the equipment into consideration).
Do not be prepossessed with any local strength of vibration, just evaluate the total vibration.
Mark your evaluation in the range of 0 to -3 (Unit: 0.1 point), where 0 is equivalent to no sensation and -3 is
equivalent to the ride comfort under the Reference Vibration.