Research On The Subspace-Based System Identification And Predictive Control Scheme During Anesthesia Process

With the rapid development of modern medical technology, there spring up various elaborate therapies, so the requirements of surgical accuracy and patient safety assurance become more and more rigorous. Under this circumstance, a favorable depth of anesthesia during operation seems especially important. On clinic, anesthesia depths are usually ensured by the anesthetists who should also keep an eye on the physical signs of the patients in real time, which burdens the anesthetists heavily. To reduce their pressure, the researchers came up with closed-loop anesthesia control conception and realized many effective closed-loop anesthesia algorithms, among which the model-based control methods exhibit the best perspective. However, due to the nonlinear characteristic, lots of experimental estimations of patient model parameters should be employed to eliminate the effect of nonlinearity in the model-based closed-loop control algorithms. But this treatment will cause weak specificity of patient model and consequently weaken the identification and control performance.To solve this problem, the thesis raises a kind of subspace-based Wiener model identification method. First of all, linearize the Propofol three-compartmental Wiener system, and then conduct the subspace orthogonal projection identification method on the linearized model, finally restore the original compartmental model according to the identified state-space system matrices. Next, simply process the identified linear model, construct an instrumental controlled variable, and conduct the model-based extended prediction self-adaptive control algorithm on the patient to achieve a desired hypnosis depih indirectly.The simulation results show that the proposed identification method designed to deal with nonlinear problems can preciously restore the system characteristics and also have perfect expandability. This algorithm requires little prior knowledge and gives out high identification accuracy. Meanwhile, the proposed instrumental variable can also indirectly describe the changing tendency of the Wiener system medial variable to some extent. The raised algorithm has firstly been applied to deal with the anesthesia hypnosis control problem, therefore it could definitely have major innovative and instructive meaning.