Development And Optimization Of Blood Purification System Based On Soft Sensing

The blood purification equipment developed in this work integrated various functions including hemodialysis, hemofiltration, hemodiafiltration, hemoperfusion and so on. Blood purification equipment is a very complicated system, which need to measure and control hemodynamic and machine varriables. Thare are also intricate reationships between those varriables. As a result, it is very hard to develop a blood purification machine with high security and reliability. In this paper, signal processing and peristaltic pump control are mainly investigated and dissucssed here. Generally, Butterworth filter is applied in the denoising of the signal from the loadcell, and the output signal from the filter inevitably has some delay. Calibration of peristaltic pumps is usually carried out by low-oder polynomial curve fitting, which may limit its application. Follow-up control of peristaltic pumps hasn’t been studied yet. For multiple peristaltic pumps, methods implemented in the blood purification equipment can only adjust the single pump, which means the pumps don’t cooperate with each other to reach an expected ultrafiltration rate.Error analysis is first conducted in this research. It is found that the major types of the error are random error and systematic error generated by the loadcell. Further analysis finds that the random error obeys Gaussian distribution, and it can be eliminated by finding the mean value or by filters; the systematic error is nearly constant that its sign and absolute value is stable during measurement, which can be reduced by adding corrections.In order to overcome the limitation of the commonly used peristaltic pump calibration methods, a soft sensing method based on the artificial neural network is proposed in this work. The relationship between the input voltage and the output flow rate of the pump is close to linear relationship when the flow rate is raletively small and becomes nonlinear as the flow rate rises. The algorithm applied in this study can describe the relationship well both in low flow rate and high flow rate.Some mode of blood purification therapy expects the peristaltic pump is able to fllow-up the changing input so as to reduce the occurrence of complication during blood purification treatment. To achieve this goal, the random error from the loadcell that is very important for flow rate measurement is eliminated by real-time wavelet filter instead of the Butterworth filter. The delay of the signal caused by the Butterworth filter is minished and the precision of the measured flow rate is enhanced. Then the follow-up controller is designed and tested. The result shows the peristaltic pump can follow-up the irregularly changed input with a satisfying accuracy. By replacing the irregularly changed input with regularly changed input, the peristaltic pump can export flow rate that is lower than its minimum flow rate when it is running continuously.Ultrafiltration rate is also a parameter that should be carefully controlled. Its requirement for precision varies according to different mode of treatments. Therefore fuzzy PID cooperative controller and a double closed loop controller are designed separately for low and high precision demands of ultrafiltration rate. Those two controllers are tested and compared with conventional methods. It is found that the fuzzy PID controller can make the three pumps cooperating with each other according to the current ultrafiltration rate. The error of both the pumps and the ultrafiltration rate can be kept at a low level when the fuzzy PID is applied. The double closed loop add a patient’s weight closed loop to the flow rate control loop. Compared to the fuzzy PID controller, the ultrafiltration rate can be maintained at a lower level. This is significant in the blood purification treatment.