Research On Human Knee Joint Modeling And Data Classification By Finite Element Method

Knee joint is one of the most important joints in lower limb activities.At present,the pathological diagnosis of knee joint is only judged by the clinical experience of medical staff.There is no cheap medical testing equipment for knee joint,and there is no relatively scientific and effective method to assist diagnosis.Above these problems,this paper takes human knee joint as the research object,and carries out knee joint system simulation research based on dynamic modeling,Kalman filter,hardware platform construction and machine learning SVM data classification.The main contents are as follows:By analyzing the structure of the tibiofemoral and patellofemoral joints in the knee joint,the human lower limb system model is simplified to a rigid two-degree-of-freedom model,the kinematics equation of the knee joint was established by a simplified human lower limb model,at the same time,the dynamic model of the human knee joint is established by using the Lagrangian dynamic equation.To obtained the mathematical models of the forces of patellofemoral joint tibiofemoral joint and quadriceps femoris in the knee joint.According to the established knee joint dynamics model parameters to be measured,the hardware detection platform of knee joint was designed.Due to the noise affects the data collected by the sensor,the error between the measured value and the real value is large.The kalman filtering algorithm is proposed.Two sets of attitude angles of the controlled object are obtained by means of gyroscope,accelerometer and magnetometer,and data fusion is carried out,contrast before and after denoising treatment.The Kalman filtered knee data acquisition platform is more accurate.Using the designed hardware platform for the experimental analysis of knee joint,and combined with the dynamic model of knee joint,to obtain tibiofemoral joint,patellofemoral joint and quadriceps femoris forces when human body squat and walking at different speeds.Three-dimensional human body model is used to scan the three-dimensional reconstruction of human knee joint,and the actual biological force obtained by combining the hardware detection platform with the dynamic model is applied to the reconstructed three-dimensional knee joint model through finite element analysis to simulate the internal stress of human walking and squatting in real situation.In order to form an effective method to assist the diagnosis model,the original data set was established by combining the dynamic model,the knee joint data obtained by the motion acquisition platform of parameter identification and the internal stress calculated by finite element analysis.The original data sets are preprocessed by means of the separation of the continuous original data,the processing of invalid values and invalid feature items;the preprocessed data sets of knee joint are classified by SVM method,and at the same time,the preprocessed data sets of knee joint pathology are classified by experiments.Two other machine learning methods were used for comparative reference experiments.Experiments show that the classification accuracy of the knee joint data sets SVM are better.The accuracy of the data set which combines the internal force of the joint as the feature item added by the dynamic model is relatively good,which verifies the scientificity of the research method and can form a relatively scientific and effective auxiliary diagnosis model.