Supported by the NSF-Career
As the baby-boom generation advances into old age, more research is being conducted on the health problems that older adults face. One such problem is hip injury resulting from the loss of balance. Currently $7-10 billion dollars a year is being spent on care for fall-related injuries in the United States alone.
Simulations of single step recovery in a forward fall provide a means to study muscle properties important to balance recovery that remain unknown in experimentation. Interest was placed on how the model performed during the step and on impact when certain model parameters were varied. This related to age effects in that parameters such as muscle contraction velocity and muscle strength are reduced with age. Identifying which muscles experience greater loading on impact than on initiation was also of interest. With this information further investigation can be done on the specific muscles.
The single step simulations were conducted using a seven-segment human model created in A.D.A.M.S. (Automatic Dynamic Analysis of Mechanical Systems). The model includes 15 muscles in each leg that are divided into 9 muscle groups according to the flexion or extension motion they create when contracted. Model height and weight were set to represent a 50-percentile male (1.75m and 75kg). Three sets of excitation times were used to simulate the initial, swing, and impact phases of the step. Experimental EMG data was used to determine the initial settings for the switch times. Plots of joint torque and joint angle were used to match the model to experimental data.
Two spring-damper forces were added under the step foot to simulate ground reactions on impact. The horizontal and vertical forces had coefficients c= 200 N/(m/s), k= 450000000 N/m and c= 2500 N/(m/s), k= 450000000 N/m respectively. These forces were activated on impact. Impact was simulated for 0.2 to 0.3 seconds.
Experimental impact force data was taken when young males stepped onto an AMTI force plate upon release from a forward lean. Forces were normalized to the model's body weight. Initial comparisons of this data to the model show very similar vertical impact forces that peak at approximately 2000 N and have a duration of 0.09 seconds. Model data shows a smooth single peak, while experimental data shows two sharp peaks. We suspect these two peaks are a result of a more heel to toe landing seen in the experimental data. The theoretical horizontal impact force displays an initial peak of -1500 N, which is 3 times larger than that seen in the experimental data. We suspect this force results from the rapid knee extension that the model displays just prior to impact. Qualitative data suggests a deceleration in knee extension prior to impact resulting in lower horizontal impact forces that damp out quickly.
Model parameters related to muscle speed, strength, and activation time were varied to study age-effects. Performance was measured by the change in step length relative to the model's center of mass. This function was chosen because the further the subject's foot is in front of their center of mass the easier it is for them to regain balance in a forward fall. The model parameters identified as most affecting performance during a 10% increase were Af (the force-velocity curve coefficient), Vmax (the maximum contraction velocity of the muscles measured in fiber lengths per second), and STRESSmax (the maximum strength of the muscles per unit area). These were identified because they produced between a 2 to 10 cm increase in step length relative to center of mass in the model. Experiment shows a 30% decrease in muscle strength in the elderly as well as a decrease in contraction velocity. Increasing muscle stiffness also showed a decrease in performance on the order of 2 cm.
Several problems were encountered throughout this study due to simplifications in the model's construction. The head arms and trunk segment of the model does not include any back muscles and is lumped into one body segment. The model also does not account for the evident lateral motion observed so often in experiment. To simplify the step sequence muscles were only 100% active or 1% active. In reality muscles can vary in percent activation. We also encountered a lot of evidence that suggested the model wasn't producing sufficient torque about the knee and hip joints. The muscles in the model were created from point to point. In reality the muscles in the knee and hip wrap around the joints. For example, when the knee is in flexion the VAS muscle will wrap around the knee joint. In the model because the muscle is a straight line it will actually become closer to the joint producing a smaller moment arm for the torque about the joint. This smaller moment arm cannot produce sufficient torque about the joints to create the desired motions in the model.
Specific muscles that are important to balance recovery were not yet studied in full. We were not confident enough with our final simulation to move ahead to lookat the impact forces up through the leg. The next step is to correct the weakjoint torque. This can be done in a relatively simple way by adding a limit to controlthe moment arm radius. However, this will mean a complete reconfiguration of everyswitch time from initiation to impact. Further research must also be done on impactforces and joint torques to obtain better confidence in predicting muscle forces on impact