Investigation of noise and vibration impact on aircraft crew, studied in an aircraft simulator

Transcription

1 The 33 rd International Congress and Exposition on Noise Control Engineering Investigation of noise and vibration impact on aircraft crew, studied in an aircraft simulator Volker Mellert, Ingo Baumann, Nils Freese, Reinhard Weber Oldenburg University, Institute for Physics, Oldenburg, Germany Abstract [664] Noise, vibration, air quality and other environmental parameters in an aircraft cabin affects comfort, well-being, performance etc of the flight and cabin crew, as well as passengers. The various physical, psychophysical and intrinsic factors (like health status) are linked to overall output parameters representing e.g. crews health and performance. The mutual interaction of all factors must be investigated in order to establish a human response model which allows to quantify the impact of measures to improve the cabin environment. Experiments to reveal input-output relations are difficult to perform in real flights and are therefore conducted in simulated flight conditions using virtual reality. A selection of noise and vibration, of temperature and of humidity levels are investigated with respect to numerous output parameters determined by psychophysical methodology, by questionnaires and by physiological measurements. The experiments are performed in two simulator facilities. 1 INTRODUCTION Health, performance, safe operation are key issues for flight and cabin crew as well as for passengers, who additionally demand well-being and comfort in a supportive travel environment. European research aims therefore to enhance the friendliness of aircraft transport in the framework programs of the past and future. One objective is to improve the cabin and cockpit environment. Impact from the cabin environment on cabin crew is obviously caused by numerous factors like sound and vibration, motion, air quality, work space etc., but also by psychological and health status, work load, passenger demand and last but not least by the complex field of manmachine-interaction which affects not only the performance of the crew and but influences the overall stability of the complex safety-critical system aircraft. To improve cabin and cockpit environment it is necessary to study the relation between physical factors and intrinsic moderators on one side with the set of relevant output parameters on the other side, which reflect e.g. wellbeing, or comfort, or health, or performance. A human response model provides a quantitative summary of these relationships and could offer tools to optimize the cabin environment on the design or re-engineering level. Part of the reported investigation is based on results from EU projects, [IDEA PACI (identification of a passenger comfort index, Co-ordinator: CIRA, closed; HEACE (health effects in aircraft cabin environment, Co-ordinator: Oldenburg University, running; and FACE (friendly aircraft cabin environment, Co-ordinator Alenia, running] [1] and from a national research program. Experiments in simulator facilities were conducted together with 1/8

2 the British Research Establishment BRE in the ACE simulator and in air-quality measurements [2], the Medical University Vienna in the AUA emergency trainer for data acquisition of medical, physiological, performance parameters [3], EADS-CRC, CIRA, University of Patras, Paragon Ltd., and itap GmbH in data acquisition and evaluation. 2 CONCEPT AND RESEARCH DESIGN For obvious reasons it is only possible to change environmental conditions in a real flight within rather restricted limits. Therefore, tests are designed in a simulator facility, which is tuned to reproduce real flight data in virtual reality. Two simulators with certain vicinity to a real aircraft are used (based on parts of an A 320 and an A 300): The AUA trainer in Vienna [3] and the Aircraft Cabin Environment ACE at BRE in Watford [2]. Virtual reality is created in the simulators based on real flight data of sound and vibration, humidity and temperature. The AUA trainer allows for adjustment of acoustics and motion, but only limited change in climate is possible. This can be adjusted down to very low humidity in the ACE, including sound and some vibration, but no motion is provided. Up to date, no simulator facility allows to change air pressure. Both simulators have only very limited possibility to support virtual reality with visual simulations. Figure 1: Emergency trainer of Austrian Airlines (left) and test rig at BRE (right). It is not sufficient to establish only realistic physical parameters in the simulator. The experimental setting has to resemble a real flight situation as much as possible including briefing of the crew, check-in of passengers, realistic task load for the crew with in-flight service etc, and a realistic duration of the simulated flight (i.e. for investigations of long-haul impact it is necessary to run the simulation for several hours). Provided that the environment is configured as optimal as possible for a virtual flight the response of the crew (and passengers) to changes of the environmental parameters is monitored. This human response is recorded with the help of questionnaires, psychological tests and with methods to trace physiological quantities. By revealing the relationship between independent input parameters and dependent output items a so-called human response model is derived. Input and output entities are either identified by the structure of the model or by definition, and not all are obvious. The relations between the different parameters are investigated with the help of (linear and non-linear) correlation and other statistical methods. 2/8

3 3 EXPERIMENTS One of several experiments is reported now, based on a full-factorial design: 3 levels of temperature, 3 levels of humidity, 3 levels of sound (including vibration). Each combination is investigated, i.e. 27 different flight situations of 1 hours duration. The experiment is designed as a 3 hour flight (per day) with boarding and de-boarding time, i.e. crew and passengers perceive one flight of about 4 hours (incl. start and landing) without any break. Climate is kept constant during each flight, sound level is very smoothly changed after one hour, respectively. This transition is not perceived subjectively, neither by crew nor by the test supervisors in the simulator. About 100 crew members participated in this test. Table 1 gives the test condition with the absolute range of the parameters as determined within the cabin in the workplace area of the crew (cockpit not included). Temperature t [ C] rel. humidity rh [%] sound level s/ v [db(a)] 70 ± 1 73 ± 1 76 ± 1 Table 1: Target values for parameter selection with actual range in the cabin. + t1 t1 rh1 s/v1 s/v2 s/v3 rh1 s/v3 s/v2 s/v1 rh2 s/v1 s/v2 s/v3 rh2 s/v3 s/v2 s/v1 rh3 s/v1 s/v2 s/v3 rh3 s/v3 s/v2 s/v1 t2 t2 rh1 s/v1 s/v2 s/v3 rh1 s/v3 s/v2 s/v1 rh2 s/v1 s/v2 s/v3 rh2 s/v3 s/v2 s/v1 rh3 s/v1 s/v2 s/v3 rh3 s/v3 s/v2 s/v1 t3 t3 rh1 s/v1 s/v2 s/v3 rh1 s/v3 s/v2 s/v1 rh2 s/v1 s/v2 s/v3 rh2 s/v3 s/v2 s/v1 rh3 s/v1 s/v2 s/v3 rh3 s/v3 s/v2 s/v1 Table 2: Parameter combinations: Sound level is either increasing (+) or decreasing (-) during one simulated flight. Each combination lasts for one hour, each row represents one flight. Table 2 gives the complete series of combination of the three parameter levels including one repetition, summing up to 18 flights. Due to the experimental conditions to become and stay stable after take-off it is not possible to randomize the climate conditions. This could be done in principle for the sound level, but is not realized in the present experiment. The order of presentation is therefore an additional environmental parameter. The objective of the design is to study each mutual interaction of the selected parameters in the human response. The variations in Table 1 reflect the fact that the target values are only provided for certain fixed locations in the simulator. Real values deviate at different places in the cabin in the indicated way. The target values are adapted to real flight data and do by purpose not include extreme values. The crew members are wired to pick up physiological data. During service breaks they have to fill in questionnaires which contain items on health and well-being (30 items) environmental conditions (45 items) control over environment (8 items) 3/8

4 relative comfort contribution (18 items) effect of the environment (18 items) ability to work (8 items) alertness and mood (9 items) Figure 2: Cabin crew during service in simulated flight (left). Detail of wired pilot (right). 4 DATA ANALYSIS AND EVALUATION An analysis of the variance of answers from the questionnaire gives insight into those parameters which might contribute to a model output, i.e. in simplified words: Those parameters, which do not seem to vary in response to environmental change or other variability of the input (including intrinsic, individual items) cannot contribute to a model by definition. Figure 3 gives a typical example for pre-processing questionnaire data. SYM_01 (left panel) gives the frequency of answers to experiencing dry eyes, SYM_09 (right panel) to experiencing dry skin or lips. More than about 20 answers (i.e. 10% of all answers) are distributed in only 2 bins for SYM_01 but in 5 bins for SYM_09. It is concluded that experiencing dry eyes represents the same perception for all crew embers and is independent from the change of the environmental factors in this experiment, and therefore will not contribute to any variation of output parameters. SYM_01 is omitted from further analysis. Figure 3: Frequency of about 200 answers to questions related to dry eyes and skin on a 6-level scale (see text). 4/8

5 The analysis of variance reduces some 150 items from the questionnaire to about 50. This number might be even further reduced to a half or third by applying more strict (statistical) criterions. Additionally, a principal component analysis reveals the similarity of different items if they cluster well. The PCA expands a two-dimensional space of perception which explains about 80% of the variance in the answers of the cabin crew. 5 RESULTS Figure 4 gives the two-dimensional space of perception representing the result of the PCA. The inner ellipse indicates a correlation of about 0.4, i.e. really important items are located outside this boundary. Questions related to air are marked in dark blue, to noise and vibration in margenta, to different kinds of feeling in red. A clear clustering is observed. Since the polarity of questions change, some clusters are separated by point symmetry. One dimension is correlated with items related to sound and vibration, the other to the air quality and factors related to feeling. This result is of course to be expected by the experimental setting, if any environmental impact is assumed at all. The over-all comfort is fed by both factors, while the annoyance is more correlated with the noise. Assessment of work space is well correlated with the over-all comfort in the answers of the crew. Figure 5 contains much more information, which is not discussed in this paper. 1,0 0,8 Air quality in the cabin: freshness Air quality in the galley: freshness 0,6 0,4 overall comfort effect on comfort from noise in the galley -3=negative effect, +3=positive effect 0,2 light level of distraction / annoyance 0,0-1,0-0,8-0,6-0,4-0,2 0,0 0,2 0,4 0,6 0,8 1,0 working space effect on comfort from temperature in the galley "feel" -0,2-0,4-0,6-0,8 How do you feel at the moment (temp)? How would you change the temperature in the cabin? How would you change the temperature in the galley? c-air c-noise c-light cc-space c-comf g-air g-noise g-light gc-spac g-comf feel -1,0 Figure 5: Two-dimensional space of perception of cabin crew, representing 80% of variance. Since annoyance is obviously related to factors of noise, an analysis is carried out to test for order effects (cf. Table 2). The perceived annoyance is plotted against the target sound levels separately for the cases of increasing level and decreasing level during one simulated flight. The result is shown in Figure 6. For increasing sound level a significant increase in annoyance to noise is observed (green). But if the order of representation of the same sound levels is reversed, the annoyance is not decreased. This might be due to an increase of annoyance with flight duration, which compensates for the decreasing sound levels. 5/8

6 7,0 95% categorial units high 6,0 5,0 4,0 3,0 level change: decrease 2,0 level change: increase 1, sound level [db (A)] Figure 6: Order effect in annoyance perception: Increasing level during experiment is perceived increasingly annoying while decreasing level does not affect perception of annoyance. Numerous correlations between different sets of input variables with subjective answers and with physiology parameters have been carried out up to now. These findings are summarized in a flow diagram of 5 input vectors of a preliminary human response model, representing task load noise and vibration air quality psychological status physiological and health status The basic concept of the model is sketched as follows: Each input vector consists of several components (e.g. noise and vibration as a multispectrum representation). The inputs vectors are evaluated and transferred into an internal representation (sensation) by functional models. The underlying functions are partly accessible by the questionnaires and by physiological measurements. These internal layers are mutually linked (like in an artificial neural network) and may additionally include control loops. The output of the different assessments, which have to reflect different, physiologically motivated control loops of the human body, are weighted and combined to three output vectors in the present stage of investigation, representing performance comfort/ well-being subjective and objective state of health Each output vector has a low-dimensional base (of about 2 to 5). 6 LIMITS OF PRESENT INVESTIGATION In principle the number of adjustable variables is small due to the limits of test time. The range of variables is limited by the physical test facility. A response model is therefore only valid within the given variance of the test set-up. Related to sound and vibration impact, only the level of noise was 6/8

7 adjusted in the present experiment, the level of vibration was depending on the noise. Known tradeoff between noise and vibration was not taken into account [5]. Figure 7 shows the measurement of a spectrum during cruise in a with most annoying tonal components, which do not contribute to the overall level (which is in the order of 70 db(a)) Cruise (3) [R900] 90 left right Level [db] tonal components Frequency [Hz] Figure 7: Real-flight spectrum inside cabin of a with strong tonal components (due to some defect in a pressurized appliance). Spectral changes are not investigated in the present experiment. The spectral shape affects wellknown psychoacoustic parameters like sharpness, which have impact on the perceived annoyance [4]. Figure 7 clearly shows the necessity to include spectral changes in simulator experiments in order to enhance the human response model. Spectral components at very low frequency (infrasound) are not reproducible in the present test facilities, though fundamental modes are often excited in the real cabin as shown in the measurement Figure 8. The impact of infrasound on the output vectors of the crew-model is not investigated, yet (nor for passengers). Another important environmental factor, not include in the present investigation, is the pressure in the cabin. The mutual interaction with the other parameters is not known. The qualitative comparison of tests in the AUA-simulator with those in the ACE test rig show another deficit of the present investigation: Motion is a very important component of virtual reality. Additionally, there is a continuous transition from the perception of vibration to perception of motion. The appropriate frequency weighting to model the perception of low-frequent vibration is still under investigation. The impact of stronger motion on the output vectors of the human response model is known for limits of high amplitudes but moderate motion is not included in the present model. It is therefore necessary to include realistic time histories of motion into the development of the human response model, and to parameterize the vibration and motion input [6]. 7/8

8 Cruise (max) [TCD-7] 90 left right Level [db] Frequency [Hz] Figure 8: Real flight spectrum inside cabin of a with excitation of fundamental acoustic modes. 7 SUMMARY Since real-flight investigations are only feasible under strictly limited control of the variables which might affect the human response, it is necessary to set up ground-based test beds which provide a simulation of virtual reality as close to real flight as possible. The adjustment and variation of a large set of control parameters is necessary since the subjective perception and assessment is driven by a multidimensional input. Mutual interaction is observed between different inputs, which must be taken into account in a quantitative human response model. The quantitative human response model is derived from the statistical relation between different grades of processed input parameters, which is only possible, if the internal representations of the input are measured by psycho-physical and physiological means. Since the output vectors of the model depend on the application (crew performance, passenger comfort, etc.), different models are expected to serve the respective request. Not all input parameters have been included in the investigation, in particular the control of the cabin pressure, and spectral shape of noise, vibration and motion was not involved. [1] and [2] [3] and [4] E. Zwicker, H. Fastl: Psychoacoustics. Facts and Models. Springer, 1999 [5] J. Quehl, Comfort studies on aircraft interior sound and vibration, Diss., Univ. Oldenburg, 2001 [6] M. A. Bellmann: Perception of Whole-Body Vibrations: From basic experiments to effects of seat and steeringwheel vibrations on the passenger`s comfort inside vehicles. Diss., Univ. Oldenburg /8