Abstract: Aiming at the difficulties in parameter setting and poor pressure control accuracy of lowpressure casting machine liquid level pressurization system, the composition, process and mechanism of liquid level pressurization system were studied. The PID parameters adjusted online by fuzzy neural network (FNN) was proposed and the fuzzy neural network with 2 input and 3 output was designed. The shortcomings of BP learning algorithm were analyzed and the fuzzy neural network training method was improved. The fruit fly algorithm（FOA）was used as the outer loop and the BP algorithm was used as the inner loop to train fuzzy neural network. The fuzzy neural network was trained by selecting the appropriate objective function. The control effects of traditional PID, fuzzy PID and FNNPID were simulated and analyzed in MATLAB. The results indicate that compared with the traditional PID control, the maximum error of the liquid level pressure is reduced by 35.6% and the average error is reduced by 21.6% when using FNNPID controller, which effectively improves the control accuracy of the liquid level pressure.
Key words: low pressure casting; liquid pressurization; fuzzy neural network（FNN）; PID; fruit fly algorithm（FOA）