تصميم و تحليل مفصل الكاحل الأصطناعي الذكي

Design and Analysis of a Powered Ankle Joint Prosthesis

Abstract

Artificial limbs are an essential device in the life of amputees. The demand for artificial limbs was increased drastically due to the spread of wars, catastrophes, and road accidents. The conventionally available prostheses are passive. The passive prosthesis does not produce a positive network, unlike the natural limbs, which have a positive net work by the muscles-tendon system. Powered prosthesis tries to mimic the function of the natural limb through the use of the powered system. This work aims to investigate, design, and build a powered ankle-foot prosthesis. The optimal design of the powered ankle-foot prosthesis is explored in this work with its valid mathematical model and the design parameters variation. The maximum force that the actuator must exert to maintain a normal gait cycle is calculated in this work This work covers the design of a powered ankle-foot prosthesis model using Solidworks software based on the chosen mathematical model.
The Gaussian-based mixture model is used to optimize the muscle
contraction datasets, corrupted with noise to be applied to the regression process.
The highest regression performance will produce a more accurate angle prediction response of the powered ankle-foot prosthesis. Several Regression techniques were benchmarked to get the optimal regression approach to fit the ankle joint angle dataset. The electromyographical signals were filtered using root means square rectifier, low pass filter, and Gaussian smoothing filter. The proposed method that works based on the
Gaussian mixture model improved the performance of the regression process from 55% to 82% with the use of both the Lowpass filter and the root mean square rectification in the polynomial regression algorithm. The highest regression performance was achieved using Gaussian smoothing filter and the root means square rectifier by the polynomial regression to 84% and by K-nearest neighbor regression to 97%. This means that the created statistical model can predict the ankle joint angle based on the electromyographical signal with 97% accuracy.
Ankle joint angle regression feature selection to get the most effective muscle on the plantarflexion and dorsiflexion movements were studied. The linear correlation coefficient is used to get the most influential muscle on the ankle joint angle. Both Lateral and Medial Gastrocnemius muscles produce the highest correlation coefficient with the ankle joint angle.