Comparative analysis of artificial neural network and multiple linear regression models in predicting pressure transmission of soft pneumatic actuators used for active compression

Abstract

Compression therapy is a crucial treatment method for managing Chronic Venous Disease (CVD), a prevalent condition that affects the veins in the lower extremities. Active compression using soft pneumatic actuators was found to be effective in maintaining consistent pressure across the circumference of the lower limb. However, the optimum design parameters of the soft pneumatic actuator have not been established. Thus, this study analyzed the performance of predicting the pressure transmission percentage of soft pneumatic actuators via an artificial neural network (ANN) and multiple linear regression models (MLR) in establishing optimum design parameters. It was observed that the lowest MSE on training data was recorded from MLR, however, better performances were recorded for the ANN model on testing data. Moreover, the highest R-squared values were obtained from the ANN model. Hence it was concluded that the ANN model was superior in terms of establishing optimum design parameters for the soft pneumatic actuators which are used in compression textiles.