Single Neuron PID+Feedforward Controller of a Direct Drive Pump System

The linear motor driven piston pump (direct drive pump) system has strong nonlinearity. According to different flow rate demands and load conditions, the motion state of linear motor in the system needs to be changed during the running process. Manual adjustment of the control parameters is always needed in the traditional PID control strategy, and the control parameters do not change with the change of working conditions. Owing to the shortcomings of traditional PID control algorithm, a single neuron PID+ feedforward control algorithm based on programmable multi-axes controller (PMAC) was proposed. The output of the single neuron module which was used as the compensation dosage of the control variable, was added to the control algorithm. This algorithm has certain learning ability, adaptability of neural network algorithm and wide adaptability of PID control. The manual adjusted configuration was used to improve the neuron proportional coefficient K₁. As a result, the K₁ automatically changed with the absolute error value, and it was limited to a suitable range. Based on the modeling and simulation through MATLAB/Simulink, the "open servo" function of Turbo PMAC motion controller was used to experimentally validate the control strategy in the direct drive pump control system platform. Results show that compared with the control strategy built by the controller, the single neuron PID+ feedforward control strategy has a faster system response, lower overshoot and smaller following error. Besides, the dynamic performance and robustness of the system are also further improved.