KEYWORDS: Low Altitude Seat Ejection; Multi Objective Optimization; Spine Injury; Spine Dynamic Models; Injury Parameter; Artificial Neural
Network
Aircraft seat ejection systems are essential lifesaving equipment for pilots. They have
been used in military aircrafts from the time of IInd world war. Since then significant

research and development has been carried out on the safety of the seat ejection sys-
tems. Although seat ejections have improved in terms of success rate over the past few

decades with a success rate of 89-91% in altitudes above 150 m, ejections at low al-
titudes (below 150 m) remain a serious concern with a success rate of 51.4%. These

are time critical events and demand optimal process parameters to improve the odds of
successful ejection due to the lack of lifesaving height and various other factors. Also,
ejecting from an aircraft is a violent sequence that places the human body under an
extreme amount of forces causing injuries to the pilot. The need for excessive thrust is
(i) To overcome the strong windblast on the pilot to avoid colliding with the tail of the
aircraft and (ii) To reach a safe height for parachute deployment in low altitude cases.
Ejection forces are primarily in the upward direction ie., compressive along the spinal
axis. Thus the spinal compression injuries are one of the most common injuries faced
by the pilot ejecting from aircrafts. The primary objectives of this thesis are two fold, one to setup a Multi-Objective
Optimization (MOO) formulation to maximize the height reached by the pilot and to
minimize the incidence of spinal injuries to the pilot. In accordance to this objective, a
Multi Objective Optimization scheme is set up based on the height reached by the pilot
(solving for trajectory) and a spinal injury parameter, Dynamic Response Index (DRI)

which is used to quantify the chances of spinal injury. The DRI is a single degree of free-
dom model to estimate the lumbar spine compression for a given force and correlates

the compression with injury probabilities. Clearly, the two objectives are conflicting in

nature as a higher force means higher lifesaving height, but also increases the proba-bility of spinal compression injuries. The Pareto-optimal solutions are obtained thereof

for different scenarios of aircraft flight using suitable multi-objective optimization al-
gorithms. A fuzzy logic system based on injury levels and minimum height required for

safe ejection is used to handle the uncertainties due to the objective functions under dif-
ferent ejection scenarios of the aircraft; it is also used as a decision-maker to choose the

initial parameters for goal programming based on the severity of the ejection scenarios.
The results were compared with ideal solutions obtained from Pareto fronts using the
same fuzzy logic system as decision-maker. It was seen that, Goal programming gave
similar results at faster time making it advantageous over the conventional “generate
first choose later” methods. It was also seen that, 6 out of the 9 adverse scenarios listed
in military specifications were safe according the injury parameter values.
Further, the next objective is of formulating a new spine model, and a derived injury
parameter to address the shortcomings of DRI model. DRI, which is commonly used as
the injury parameter for under body loading scenarios (as prescribed by NATO (2007);

USAF (2014)) suffers from inherent disadvantages and has been reported to under-
predict the chances of injury. The main reasons are, due to the SDOF modelling the DRI

model does not account for bending motion of spine and posture of the spine. Thus a
novel lumped full spine model capable of modelling the spine in different posture along
the sagittal plane is formulated. The unavailable data for the model were obtained using
inverse parameter identification approach by eigen frequency matching using Genetic
algorithm. Each vertebra has three degrees of freedom: axial, shear and rotary motion
to model the flexion of spine. A new injury parameter is proposed based on the sum of
compressions caused due to axial and rotary springs at each vertebral level, to account
for wedge compression and burst fractures. The results indicate that the model was
able to predict the injury parameter values under different postures of spine according
to trends in literature.
A more realistic non-linear seat restraint handling method is discussed to include
the effect of seat belt intactness on the injury parameter. With the developed new spine

model and the seat restraint handling method, biodynamic considerations such as, pos-
ture of the pilot, intactness of the seat belt can be, included into the Multi-Objective

Optimization framework. These cases cannot be assessed by the MOO framework with
DRI as the injury parameter. Based on the results of this robust MOO framework, the
decision making is done based on the minimum height required for successful parachute deployment for various different cases. The solution obtained for various simulated
scenarios is used to train a two stage Artificial Neural Network (ANN) architecture
to choose optimal parameters for any ejection scenario. A two stage architecture was
needed to maintain the Pareto optimality of the predicted output parameters which is
necessary due to the highly critical situation in consideration. The ANN proposed gave

good results which were along the generated Pareto fronts for different simulated air-
craft conditions instantaneously.