With the continuous development of information technology and artificial intelligence technology, the current production process in various fields has initially realized the automation process. The application of automation technology can also significantly improve production efficiency and product quality. At present, automatic control can solve the problem of raw material waste and loss in the mechanical production and processing process. The production efficiency has also been improved, so the economic benefits of the enterprise have been greatly improved.
Application in temperature control of injection molding machine barrel
The barrel temperature of the injection molding machine is an essential parameter of the injection molding process. Effective control of the barrel temperature is an integral part of ensuring the quality of plastic products. However, the barrel temperature system of the injection molding machine is a complex system with multivariable, discrete, intermittent work, large hysteresis, nonlinearity, strong coupling, and manual participation. Due to the complexity of the heating process, the control system is accurate. The mathematical model is difficult to establish, which makes the temperature control of the barrel become a problematic point in the design of the injection molding machine controller. In addition, in the barrel temperature controller of the injection molding machine, the controller’s performance, such as good robustness and low algorithm complexity, are parameters that cannot be ignored. At present, the PID control method is mainly used. The PID control has many applications and is widely used in nonlinear or time-varying control processes. It is easier to start the three parameters Kd, Kp, Ki in the PID controller. Tuning is the simplest and sometimes the best controller. This article is based on the development process of the barrel temperature PID controller, from linear PID controller to nonlinear PID controller, and then to PID controller based on BP neural network. The advantages and disadvantages of each are explained respectively.
1. Barrel temperature linear PID controller
The PID controller does not use the precise model of the controlled object but only uses the error between the control target and the actual behavior of the object to generate a control strategy that eliminates this error. Because PID control technology is a process control principle based on the error to reduce errors, it has been widely and effectively applied in control engineering practice. This article mainly studies the temperature control system of the barrel of the Pauletta injection molding machine using a linear PID controller. A lot of engineering practice shows that the linear PID controller is applied to the temperature control system of the injection molding machine barrel. Although it can achieve good results, the temperature control system is a sizeable pure lag system, and the parameters of the controlled object vary greatly. There are many influencing factors, the unmodeled dynamic factors are significant, and the nonlinearity is severe. When the PID control method is aimed at the control object with more internal and external uncertain factors, the PID control method appears powerless.
2. Barrel temperature nonlinear PID controller
In the temperature control system of the injection molding machine barrel, the linear PID control only considers the external information of the controlled object (temperature), which determines the finiteness of the controlled object. For objects with more internal and external uncertainties, linear PID is powerless. In this case, it is necessary to obtain and consider the system’s internal information and external uncertainties and let this information participate in the system’s control and adjustment to improve its anti-interference ability. Therefore, the nonlinear PID controller came into being. It is based on the linear PID controller with the following improvements: According to the ability of the system to withstand, the rationality of the controlled change, and the ability of the system to provide control, it is set by the system. It is crucial to arrange a suitable transition process first. The transition process is realized by TD, which not only gives the arranged transition process signal but also gives the differential signal of the transition process. It can extract the differential signal of the error with TD, state observer, or ESO with a shallow noise amplification effect.
Different from linear PID control, a suitable nonlinear function is used to combine errors to form a new nonlinear error feedback control law. The nonlinear PID controller is applied to the barrel temperature control system of the injection molding machine. Multiple single-loop nonlinear PID controllers adjust the power supply voltage of each electric heating loop to control the temperature of each section of the barrel. Since the control system does not need to establish an accurate mathematical model and can attribute all the uncertain factors acting on the controlled object to “unknown disturbance,” it can be estimated and compensated by the temperature data monitored in real-time. To achieve the purpose of automatic anti-interference, and then realize the automatic real-time control of temperature. However, in realizing the temperature control strategy, the temperature control system still has a significant problem. That is, it is not clear about the unmodeled dynamic factors and is not predictive.
3. PID controller for barrel temperature based on BP neural network
For the requirements of injection molding machine barrel temperature control and the deficiencies of linear PID controller and nonlinear PID controller, especially for the unpredictability of nonlinear PID controller. The controller combines neural network and PID control technology, which can infinitely approximate the nonlinear system and have fast convergence and predictability advantages. Moreover, the BP neural network-based barrel temperature PID controller can effectively shorten the transition process, have better stability and fast response, and meet the barrel’s temperature control requirements of the Paulownia injection molding machine. The controller consists of two parts: conventional PID controller and neural network controller. Considering the influence of neighboring heaters, the input of the neural network controller introduces the feedback signal of neighboring heaters and the input signal and feedback signal of this section heater. The output of the neural network controller is the three parameters Kd, Kp, and Ki of the PID controller. According to the input/output of the control system, the three PID controller parameters are adjusted in real-time by the neural network to realize the high-performance control of the barrel temperature.
Compared with linear PID control and nonlinear PID control, PID control based on BP neural network has better stability and fast dynamic response characteristics. The temperature adjustment process is short. Because the BP neural network has strong learning ability, it can continuously extract the basic information contained in the training samples for temperature prediction. In addition, the controller can maintain good working performance in the case of temperature parameter changes, inaccurate mathematical models, and changes in the control environment. The system is robust, making the control method have excellent application prospects.
In summary, the heating barrel temperature control is a vital part of the injection molding machine control system. The article summarizes three PID control methods: linear PID control and nonlinear PID control and PID control based on BP neural network to achieve high-performance temperature control functions. Among them, the best performance is the PID control method of the BP neural network, which combines the superior performance of the nonlinear PID control method. It can achieve better temperature control under the condition of more internal and external uncertain factors. It has good anti-interference performance; simultaneously, the method realizes temperature data training through the neural network’s learning ability, which can realize real-time prediction of future data, is predictable, has stronger robustness, and has higher practical value.
