| Left ventricular assist devices (LVADs) were first used as a bridge to heart transplant. In recent years the application of LVADs has been extended to many other cardiovascular disease therapies, such as coronary revascularization, aortic valve replacement and myocardial repair. The pulsatile LVAD system consists of an actuator, the transmission mechanism and the cardiovascular system. The optimization of pump design aims to acquire the best output performance with limited power input. Hence an integrated model being able to reflect the coupling between the three parts is required. The objective of physiological control is to adaptively regulate LVAD's output, as the the natural heart, according to different physiological states. The varying of cardiovascular parameters reflects the blood demand of body, so the design of physiological controller should be based on the cardiovascular model. In order to track the time-varying parameters, the controller must contain a parameter estimator which requires a predetermined model structure. Therefore, model structure identification and parameter estimation are key technologies to implement the control system. Before clinical applications an LVAD and its control system must be elaborately tested. Considering cost and ethics issues, animal experiments are strictly limited. To solve this problem, a mock cardiovascular system should be developed as an experimental platform.Main research contents in this paper are as follows:1. Models of the left ventricle and systemic arterial system were numerically simulated. A mock circulatory loop (MCL) was developed. Important physiological variables, such as heart rate, left ventricular pressure, aortic pressure and aortic flow are tunable. Important hydraulic parameters, such as total peripheral resistance (TPR), arterial compliance, aortic resistance and aortic inertance, can be quantitatively calculated.2. A dynamic model for the cam was deduced. The friction coefficient of cam groove was calibrated by a trial method using a prototype pump. Combining the model with the torque-speed curve of an ultrasonic motor and the cardiovascular model yields an integrated model of the pulsatile LVAD system. A novel method to enhance the pump's output (pressure and flow) and efficiency was proposed by means of improving the cam profile curve. The integrated model was numerically implemented and the optimal curve was selected from a set of simulated candidates.3. Under the nonlinear Bayesian filtering framework, parameter estimation per-formance of the extended Kalman filter (EKF) and the unscented Kalman filter (UKF) for the mock systemic arterial system was compared. Parameter identifiability was evaluated by differential algebraic methods. An estimator capable of tracking the time-varying systemic arterial parameters was designed by augmenting the process noise covariance of the extended variables. The physiological controller was developed by combining the estimator with the PID algorithm. The mean arterial pressure was maintained by regulating pulsatile rate, simulating the heart rate regulation during exercise. The controller was implemented in an embedded real-time system and tested in the MCL.4. Dynamic identification methods for the outlet model of a pulsatile pump were studied. The subspace model identification (SMI) algorithm was proposed to identify the model structure of the outlet tube by means of some model structure invariable terms. The method was also used to identify the mock systemic arterial system. Parameters of the outlet model were estimated by UKF. The optimal time-varying resistance model was selected by comparing modeling accuracy. Replacing the tradionally used constant aortic resistance with the time-varying one yields a new four-element windkessel model. Effects of the time-varying resistance on other parameter estimates were studied. |
PDF Full Text Request