Applied Ergonomics 1974, 5.3, 153-160
Dummies for crash testing motor cars Christine M. Haslegrave Motor Industry ResearchAssociation The paper describes the design and development of the Ogle-MI RA anthropomorphic dummies over the past five years. In the motor industry these dummies are used in safety testing of vehicles and the design of restraint systems. The work started at MIRA in 1969 to produce a dummy which was sufficiently realistic to test steering columns and safety belts for vehicle occupants. The first dummies were manufactured by David Ogle Ltd in 1972 and an advanced dummy (the OPAT Dummy) has now been designed in collaboration with the Transport and Road Research Laboratory.
In recent years, vehicle accidents have been recognised as a major problem throughout the world. While attention must be paid to eliminating their causes it is also necessary to investigate ways of protecting vehicle occupants and other persons involved when an accident occurs. Legislation in many countries now requires vehicles and vehicle components to undergo various impact tests and in a complete vehicle test this usually involves the simulation of the occupant in order to test the restraint system. These simulations may represent a complete pedestrian or vehicle occupant, and are then known as anthropomorphic dummies, or may just represent a part of a body, which is known as a body form (eg, a head form to test facia panels or windscreens). Various types of impact tests are shown in Fig 1. In these tests measurements are made of the loads and accelerations experienced by the device under controlled impact conditions, which may then be used to assess the potential of safety measures.
Historical background One of the earliest simulations was designed by the Royal Aircraft Establishment at Farnborough in the early 1950's to test parachute harnesses. This was a relatively simple dummy which was intended to reproduce gross body movement and to give a correct distribution of weight. An RAE dummy with instruments fitted is shown in Fig 2. This dummy was in widespread use in the UK for vehicle crash testing until about 1968. Meanwhile, in the USA, two commercial dummies were produced by Aiderson Research Laboratories and the Sierra Engineering Co and later versions of these dummies are still in use. At the Motor Industry Research Association (MIRA) we began to use these later designs in 1968 both for routine vehicle crash testing and for research work on safety systems.
Both types of dummy were used extensively in the testing of cars to American safety regulations, and in particular they were used in testing to Federal Motor Vehicle Safety Standard 208 which specified certain performance requirements measured in terms of accelerations of a dummy during a 30 mile/h (48 kin/h) impact into a concrete barrier. However, in 1972 a court action was brought jointly by the Ford and Chrysler companies who objected to this Standard. Evidence was presented to show that the dimension and weight specification for the design of these dummies was too broad in its scope, and that different designs of dummy, each complying with this specification, could produce very different results. In order to comply with the ruling of the Court, in 1973 the National Highway Transport Safety Authority (NHTSA) replaced the dummy specification with their own design specification which combined in one dummy components from both Alderson and Sierra designs. The resulting dummy, known as the G.M. Hybrid II, is to be used in Federal tests for several years, after which the US Government hopes to replace it with one based on biomechanical studies.
Development of the Ogle-MI RA dummies In the late 1960's, MIRA was using both the RAE and Sierra dummies in crash tests. The Sierra 95th percentile dummy shown in Fig 3 was obtained initially for work on seat belt systems, but the performance of the dummy was found not to be sufficiently realistic to obtain meaningful results. The proportions of this dummy appear strange due to the problems of derermining 95th percentile dimensions, but this has been discussed at length elsewhere (Searle and Haslegrave, 1969 and 1970). Therefore a programme of work was initiated which produced on a one-off basis improved components (thorax, shoulders, pelvis and femurs), and the modified dummy was then used extensively
Applied Ergonomics September 1974
Fig 1. Dummies
in various impact
Fig 3. Sierra 292 - 895 dummy
for research programmes seat belt systems.
Fig 2. RAE Mark VB Dummy Caliber
on collapsible steering columns and
In 197 1, Ogle Design Ltd, who are well-known for their design work in the automobile industry and in other areas, became interested in the field of anthropomorphic dummies. They joined MIRA in a project for the complete design of a new dummy, to be known as the Ogle-MIRA M50/71 (indicating a male 50th percentile dummy). By the following year this dummy (shown in Figs 4 and 5) was in production, and a subsidiary company David Ogle Ltd had been formed for its manufacture. The Ogle MlRA M50/71 dummy was first marketed in the summer of 1972 and sold to many of the countries which are involved in the vehicle safety effort.
Fig 4. Ogle-MIRA M50/71 dummy (Photo by courtesy of ICI PlasticsDivision)
Fig 5. Ogle-MIRA M50/71 dummy with skeleton
With the general interest in biomechanical simulation of vehicle occupants, new information from volunteer and cadaver testing is constantly being produced. During 1973, the Transport and Road Research Laboratory (TRRL), as part of its programme to encourage the production of safer road vehicles, participated with Ogle and MIRA in a joint programme to produce an advanced crash test dummy incorporating the latest advances in biomechanics. The result is the Occupant Protection Assessment Test (or OPAT) dummy. A seated form is used for most vehicle crash testing, but a standing version has also been produced for research work into pedestrian impacts. This dummy is able to stand unsupported, but upon impact behaves as a freely jointed body. David Ogle Ltd have now extended their range of dummies to include child forms and specially adapted dummies for aerospace testing.
other bones to measure the loads imposed on these areas. Potentiometers are used to measure chest deflection under load. Gyroscopes have been used in research work to estimate the dummy trajectories, but these are more frequently obtained from photographic records.
Design of an anthropomorphic dummy When subjected to an impact, various types of measurements can be made on the dummies. Accelerometers are mounted in cavities in the head, chest and possibly pelvis to take measurements in three different axes in order to obtain the resultant accelerations. The load transmitted from the knee impact through the femur is obtained from a load cell mounted along the axis of the bone. These are the measurements required by legislative standards, but it would also be possible to put strain gauges on the ribs or
It can be seen that, in order to obtain realistic results from so many complex measurements, the dummy must accurately represent all the following: dimensions, weight distribution, centre of gravity positions, ranges and modes of joint movement, and stiffness of joints as well as shape and impact characteristics in possible areas of contact with the vehicle or restraint system. The areas which were least developed and also most critical in the commercial dummies were the shape and dynamic force/deflection characteristic of the chest, the structure of the shoulder area and the shaping of the pelvis; it was on these that the initial work at MIRA was concentrated.
The Ogle-MIRA M50/71 Dummy The first dummy to be produced as a result of the research work at MIRA was the Ogle-MIRA M50/71, which is shown in Fig 4 and compared with a human skeleton in Fig 5. The dummy chest was designed to simulate the forces applied when it hit a steering wheel. The results of a cadaver test under similar conditions were kindly made available by
Applied Ergonomics September 1974
Fig 6.(a-d). Prototype dummy thorax with skeletal thorax
Prof Patrick in America and the performance of the thorax was matched to these results. The structure of the rib cage was duplicated as closely as possible with the positions of all non-floating ribs reproduced, except for the top rib which is effectively covered by the clavicle and which was omitted. The curvatures of the human rib cage were also repeated in order to give a correct distribution of load. The ribs were constructed with an upsweep of 10 ° from front to back, as this had a significant effect on the way in which the chest deflected: the whole rib cage moved downward and backward upon impact. This feature was entirely new to dummies, which normally had rib cages of an upright cylindrical form. The final version of the prototype chest is shown in Fig 6. When the form of the chest was decided, the major problem remained of simulating the forces generated upon impact. A theoretical model, (described in detail in Searle and Haslegrave, 1970) indicated that with a chest of humanlike static stiffness, the load/time history of an impact could be represented by summing the contributions from several elements (as in Fig 7): the spring of the rib cage, the inertial loading of the chest contents, the inertial loading of the shoulder and arms as these move forward and upward, and the inertial loading of the head. These elements were therefore all reproduced in the dummy. The internal organs of the chest were simulated by three lead sheets interleaved with polyurethane foam which permitted them to continue to move forward when the chest wall was 156
Applied Ergonomics September 1974
Contributions of components Predicted from sum of components
I I m
-5,, r I0
Tirn¢, ms Fig 7. Force time characteristics of theoretical model of the thorax
stopped. The front sheet was intended to simulate the mass of musculature and tissue associated with the chest wall, the other sheets simulating the mass of the internal organs. The total weight (3.7 kg) and forward movement (50 ram) were representative of the organs of a 95th percentile human thorax. The best estimate at that time for the static stiffness of the rib cage was obtained from cadaver data measured by Prof Patrick in America (Patrick et al, 1965): the value of 44 x 103 N/m chosen for the dummy fell within his range
of data. (Subsequent test results have indicated that this value should be even lower.) The dynamic performance of the dummy chest was measured by supporting the torso, head and arms on a pendulum which reached a speed of approximately 6 m/s and released the torso which then struck an unyielding load cell. The test is shown in the high speed film record of Fig 8 and the force/time history measured in Fig 9 with Patrick's cadaver results for comparison. It is important that the contributions to this load history occur at the correct time intervals and this was one
reason for the redesign of the shoulder joint mechanism. The more important reason was that the dummy was to be used to test seat belt systems, and the lie and loading of the belt is considerably influenced by the shape and movement of the shoulder and underlying bone structure. In the redesigned dummy, the top of the sternum was joined to each shoulder by a steel clavicle (shaped and sloped as is the human), the sterno-clavicular joints being represented by ball and socket joints. The shoulder construction was designed to give the very wide range of movement permitted by the human shoulder, which is an under-centre ball joint
Fig 8. Torso impact at 5.9 m/s
12000 Range of cadaver data (Patrick, 1965) M I R A chest
4000 200C O 0
30 40 Time, ms
I i i [ I i lil I ,. L~-
Fig 9. Force time characteristics of chest striking unyielding load-cell
constrained by ligaments and musculature. The range of movement could not therefore be represented by a more stable over-centre ball joint and the arrangement shown in Fig 10 allowed three separated rotational modes. This also provides for correct setting of the angle of the humerus so that the flail of the arms at impact is along a diverging path as in man. The outer end of the clavicle is pin-jointed to the acromion process of the spine of the scapula. In the human, the movement of the scapula around the rib cage is controlled by musculature, the clavicle acting as a radius arm. In the dummy, the scapula was constrained to move over the chest wall by means of a guide rod with a stop to prevent excessive forward movement. The advantage of this design over other dummies is that the shoulder is able to shrug forwards around the seat belt, which is less likely to slip off, ie, it is retained in the same manner as by a human wearer.
The last area to be redesigned was the pelvis. No dummy at that time had a pelvis of human form and the design for a dummy of 95th percentile size presented some interesting problems'. By definition, people of this size are not common in the population and among skeletons available for study they are even less common: most skeletons in medical departments are of Asian rather than Caucasian origin and body proportions will differ between these populations. As a solution, it was decided to use a pelvis from a skeleton of 50th percentile size and to scale the dimensions. The key dimensions for this scaling were however not obtained from the skeleton but from a small sample of 8 men of approximately 95th percentile size. In addition, correlations between these dimensions and height, weight or derivatives of these were obtained from a larger sample of 33 covering the whole range of male population size. From this information, the shape of the scaled-up pelvis was obtained. The key dimensions were chosen between points that could be easily located and provided references in two planes. They were the bi-spinous width and the height of the anterior-superior iliac spines in the seated posture (see Fig 11); both are related to the way the correctly fastened seat belt lies across the body since the lap belt should be retained by the anterior-superior iliac spines. The shaping of the ischial tuberosities is also important, as these partly determine the movement of the body on the seat.
The design of the hip joint is complicated since this too is an under-centre ball joint with a slightly more restricted
Fig 10 (a, b). Shoulder design
(a) Fig 12. OPAT dummy present a smooth surface all facial sculpture has been omitted. Similarly, the knee form is of hemispherical shape, and is mounted on the lower leg, which is considered to approximate most closely to the action of the human patella.
(b) Fig 11 (a, b). Pelvis design range of movement than the shoulder. The smaller range however permitted representation by a specially designed intemaUy retained ball joint giving a continuous range of movement. The angle at which the joint was inclined was determined from a study of the skeletal pelvis, and the femur too was set at an appropriate angle to the axis of the leg. As a result, this is the only dummy able to cross its legs in a humanlike manner.
The OPAT dummy Improvements have been made to most areas of the Ogle-MIRA M50/71 dummy in developing the OPAT dummy shown on a test sled in Fig 12. The two major aims were to make the dummy humanlike in response and to ensure that the results are repeatable over many tests (which is even more essential for compliance testing to legislation than for research work). Areas of the body which are likely to contact the vehicle (in impacts with windscreen, facia panel, door, etc) have been designed to present a smooth geometrically shaped surface and to have a uniform depth of covering. This ensures that minor variations in the setting up of the dummy and in its trajectory wilt not cause large differences in the results measured. The body areas affected are the head, femur, knee and lower leg. The arm also contacts various parts of the car when it flails, but no measurements are at present made to assess arm injuries. The head has been made of an aluminium skull of a simplified geometric shape (Fig 13) and covered with a vinyl skin of uniform thickness. In order to
The thorax has been matched to the latest biomechanical data, although the design concept has not been altered. The pelvis has been retained in its original design, but is now joined to the thorax by a lumbar spine simulation which is constructed of a solid rubber cylinder attached to the base of a rigid thoracic spine and to the top of the pelvis. The spine is very similar in construction to the neck, but is of larger diameter and stiffer rubber. The abdominal sac is simulated in flexible solid plastic material, which is of the correct weight, and the internal organs of the chest are now represented in a similar way. The hip joint has been altered to a more advanced design. This consists of an accurately ground ball attached to the femur and set inside a split bearing which is clamped in the acetabulum of the pelvis. Friction is controlled by a pinch bolt through the split bearing. The shoulder design has also been up-dated to a simpler form of the Ogle-MIRA M50/71. Shoulder movement is obtained by means of a variable length link connecting the outer end of the clavicle to the thoracic spine. This link also supports the simulation of the scapula, which is necessary because the onset of the inertial loading of the shoulders and arms plays a critical part in the force history of a chest impact. Joints at the shoulder and elbow are split threaded plastic bushes retaining turned steel pivots. The frictions are controlled by pinch bolts. The skin over the limbs is simulated by vinyl covered foam, but the torso is covered by a one-piece suit with zipped access to accelerometers and other instrumentation. The suit is designed so that areas contacted by a seat belt are unbroken. The dummy and suit are coloured either light blue or orange, as these colours provide the best contrast for the high speed film used to record vehicle impacts.
Discussion of dummy design Dummies have several conflicting requirements. The most important is a realistic simulation of the human under
Applied Ergonomics September 1974
Fig 13. OPAT dummy headform impact, although the extent of the simulation is determined by the type of test for which it is used. For many tests a highly complex device is needed and this makes the other requirements less easy to achieve. These are an accurately repeatable response (.particularly for legislative compliance testing), strength to withstand severe test conditions, and easy maintenance and adjustment. Various approaches to these problems have been tried. Simple body blocks with minimum articulation have been used for routine testing in a simplified impact situation, as for example in the BSI test for seat belts. An extension of this is the use of body component simulations such as the aluminium hemispherical headform for facia panel impact tests. However, in tests of more complicated environments such as new types of restraint systems the kinematics of the vehicle occupant may also need to be simulated. One dummy was designed with a further refinement: the Sierra frangible dummy. This had an internal skeletal structure intended to break at the typical fracture loads of human bones. However, the human body is relatively fragile and this type of dummy not only makes testing extremely expensive in replacements, and repeatability impossible to assess, but also wastes information. There is no indication at the end of the test of how much improvement is needed in the system to prevent any fractures. The same information which is used to design the frangible bones can be used to set injury criterion levels, and measurements against these can then provide additional information on the degree by which such levels are exceeded.
AppliedErgonomics September 1974
Therefore the most usual form of dummy for vehicle tests will for the present be one simulating the human shape, weight distributions and kinematics, but designed in addition to be robust and repeatable. Strength and repeatability are two problems receiving much attention in the design o f advanced dummies. The first requirement however is the simulation of the human, but only a limited amount of test data is available. Impact tests on human volunteers can only be carried out at non-injurious levels and further data must be obtained from tests on cadavers. There is not therefore a complete description of the characteristics of any area o f the body and the maximum use must be made of the available information. As an example, the response of the chest is available in the form of force histories and force/deflection curves from cadaver tests for frontal impact, but not for any other direction of impact. In designing the OPAT dummy, it was felt that where this information was limited, the dummy structure should be kept as close to the human as possible, as this would give the greatest chance of simulating the total human response. This is especially important where the dummy is to be used for a wide range of types of tests or for testing in situations where the critical parameters are unknown, such as tests of new types of restraint systems. A further problem remains in the correlation of the measurements made on dummies with the results of real life accidents. The types of measurements and injury criteria have been standardised to a large extent by legislation. Unfortunately follow-up studies to assess the relevance of these tests, and their resultant designs of safety systems, to the accident situation on the roads, have been few. The TRRL are intending that the next stage in the development of the OPAT dummy should be a series of experiments to correlate the measurements made on the dummy in reconstructed accidents with data obtained in the study of the accident vehicles and victims. The final results should be a dummy which will enable us to make realistic assessments of the performance of new vehicles and restraint systems.
References Patrick, LM., Kroell, C.K. and Mertz, H.J. 1965 Forces on the human body in simulated crashes. Proceedings of the 9th Stapp Car Crash Conference, October 20-21, Figures 8 & 11.
Searle, J.A. and Haslegrave, C.M. 1969 Anthropometric dummies for crash research. MIRA Bulletin No 5, 5 - 30.
Scare, J.A. and Haslegrave, C.M. 1970 Improvements in the design of anthropometric/ anthropomorphic dummies. MIRA Bulletin No 5, 10-23.