Nowadays, aiming at the defects commonly used Backward Propagation (BP) neural network Proportional-Integral-Differential (PID) algorithm, namely back propagation algorithm complexity, slow convergence in solving complex nonlinear problems, the network to converge to local minima, leading to failure of network training, while the neural network controller can meet the requirements in the software simulation run results, but in the field of engineering the actual application is difficult to adapt to the rapid and timely control of complex industrial field required. To solve the above problems, a new multi neuron PID neural network algorithm is proposed. Its principle by simplifying the PID neural network controller, its initialization was improved to achieve new neuron PID control process is fast and the timeliness of response, which would be applied to the hydraulic straightening machine four cylinder servo controller. With the new multi neuron PID controller of traditional neural network controller and the displacement response curve of this research compared, the new neuron PID servo controller with BP algorithm is simplified, and the advantages of fast convergence speed, good flexibility, and real time network weights, provide theoretical support and technical reference for the application of multi neuron PID controller engineering.