Research On Evaluation Method Of Craniocerebral Injury Based On Neural Network And Finite Element Theory

In recent years,whether in traffic accidents or other aspects involving human injuries,head injuries have always accounted for a large proportion.And compared with other parts of the injury,the fatality rate and disability risk of head injury is greater.There are many evaluation methods for studying head injury,among which the finite element method is the most respected.However,the finite element method also has disadvantages such as high threshold,long time-consuming and high hardware requirements.In order to solve these shortcomings,this paper uses a combination of neural network and finite element technology to study and evaluate head injury.In order to obtain the maximum strike speed range of stick-like blunt objects as the boundary conditions of finite element simulation,and to predict the maximum strike speed of blunt objects of different groups of people.This paper used a high-speed camera to capture videos of the testers using different types of blunt tools to hit sandbags of different heights,and used the collected video to calculate the maximum speed of the blunt tools during the attack.Based on the collected data,the relevant factors affecting the maximum strike speed were statistically analyzed,Such as subject’s height,weight,age and gender.In the follow-up process,the factors related to the maximum blow speed and the maximum blow speed were respectively used as input and output,and the BP neural network was trained by the ten-fold cross-validation method.Finally,the trained and verified BP neural network can meet the needs of predicting the maximum speed of blunt force for people of different ages,genders,heights,and weights.In order to explore the fast evaluation method of stick-like blunt instrument-induced craniocerebral injury,this paper used the finite element method to simulate five different parts of the 50^thadult head finite element model hit by different stick types with different speeds.and based on the simulation experiment results,the von Mises stress of the outer plate of the skull,the plate camphor,and the inner plate,as well as the maximum principal strain（MPS）of the brain,cerebellum,and corpus callosum were extracted.Subsequently,the blunt-strike speed parameter and the brain injury parameters were used as the input and output of the convolutional neural network,and the ten-fold cross-validation method was used to train the convolutional neural network.Different types of convolutional neural network structure were selected,and the best model was selected based on the performance on the test set.Inputting the blunt-strike speed parameter into the final convolutional neural network model can quickly obtain the biomechanical response parameters of the brain,and use this to judge the result of the brain injury.In order to study and quickly evaluate the craniocerebral injury caused by the collision between the rear occupants and the front seat in a traffic accident,this paper used a 50^thpercentile finite element head model to carry out simulation collision experiments with different postures and different speeds.According to the experimental simulation results,the maximum principal strain,von Mises stress and shear stress values of brain tissue were extracted.And using the BP neural network model that has undergone ten-fold cross-validation to predict the von Mises stress,shear stress,and maximum principal strain of the brain tissue of the rear occupants when they collide with the front seat under different postures and different speeds.This can be used to judge the degree of brain damage.