The basic goal of grain drying is to achieve the ideal drying quality of grains with low drying costs and energy consumption while maintaining a stable drying process. The process of grain drying is a typical nonlinear, multivariate, large lag, parameter related coupled non steady state heat and mass transfer process. Grain itself is a complex biochemical substance. In order to achieve the above goals, it is necessary to continuously adjust the drying parameters and control the working process of the dryer during the drying process. Automatic control of the drying process is an effective means to achieve high-quality, efficient, low consumption, and safe operation of the dryer. The automatic control of the drying process and the automatic control of the grain dryer are of great significance for ensuring uniform and consistent moisture content of the discharged grain, improving the quality of the dried grain, reducing the labor intensity of operators, and fully utilizing the production capacity of the dryer. According to the development goals set by the National Grain Administration in the "15th Five Year Plan" for the scientific and technological development of the grain industry, online monitoring and automatic control of the grain drying process have become key issues in improving the efficiency of China's grain drying processing technology and an important path to achieve the "15th Five Year Plan". With the increasing investment in grain depot construction in China, the grain processing industry is increasingly integrated with the international market. The automation of grain drying will lay the foundation for China's grain to enter the international circulation market.
1. Characteristics of Advanced Control
The research on automatic control of grain drying process began in the 1960s. At that time, traditional control methods such as feedforward control, feedback control, feedback feedforward control, and adaptive control were used. Traditional control theory uses differential equations or transfer functions to express the knowledge and existing information of the drying process system into analytical expressions. However, there are many difficulties encountered when using and designing a control system for a grain dryer that adopts the above control methods. The reasons are: (1) the grain drying process is complex, time-varying, and nonlinear; (2) Some drying process variables (such as grain quality and color) cannot be directly measured, and the measurement of some variables (such as grain moisture content) may be discontinuous, incomplete, or unreliable; (3) The process model of the dryer is an approximation of the actual process and requires a significant amount of computation time; (4) It is almost impossible to use an appropriate model to represent a nonlinear, lagging, and time-varying complex system like the drying process; (5) There is an interactive effect between the controlled variables and the control variables of the grain dryer; (6) The operating conditions of the grain dryer are complex, with a wide range of disturbance variables that are difficult to control.
Obviously, to overcome the above difficulties, it is necessary to continuously improve the traditional control methods of grain dryers, while exploring new and more effective control methods. In the 1970s, the progress of the electronics industry, especially the development of computer technology, enabled the widespread dissemination of the idea of advanced control. The goal of advanced control is to solve the control problems of complex industrial processes that are difficult or even uncontrollable using conventional control methods. In recent years, modern control and artificial intelligence have made significant progress, laying a strong theoretical foundation for the implementation of advanced control systems; The popularization of distributed control systems (DCS) and the rapid advancement of computer network technology have provided powerful hardware and software platforms for advanced control applications. In short, the needs of industrial development, the development of control theory, and the advancement of computer and network technology have strongly promoted the development of advanced control.
With the rapid development of computer technology, artificial intelligence control theory has begun to be applied in the control of drying machines, significantly improving the performance of the drying machine control system. Traditional control methods are not suitable for controlling grain dryers due to their large lag and nonlinear relationship with the grain drying process. The advancement of artificial intelligence technology is widely applied in the engineering field, and advanced control theory and methods are applied to the automation control of grain drying processes. The control methods are continuously improved, and the control effect is enhanced. After the 1990s, process control began to develop towards intelligence, and intelligent control theory was increasingly integrated with drying technology, using artificial neural networks to model and control the drying process^ The system is applied in grain quality prediction, drying process control, and management consulting.
Advanced control systems closely related to control theory, instrumentation, computers, computer communication and networking technologies, have the following characteristics:
(1) The theoretical basis of advanced control systems is mainly based on model-based control strategies, such as model predictive control, which fully utilizes information related to industrial process inputs and outputs to establish system models without relying on in-depth research on reaction mechanisms. Recently, knowledge-based control, such as ^ control and fuzzy logic control, is becoming an important development direction for advanced control.
(2) Advanced control systems are commonly used to handle complex and variable process control problems, such as large time delays, multivariable coupling, and various constraints between controlled and controlled variables. The advanced control strategy adopted is dynamic coordinated constraint control based on conventional single loop control, which can adapt the control system to the dynamic characteristics and operational requirements of actual industrial production processes.
(3) The implementation of advanced control systems requires high-performance computers as support platforms Due to the complexity of advanced controller control algorithms and the influence of computer hardware, advanced control algorithms for complex systems are usually implemented on the upper computer. With the continuous enhancement of DCS functions and the development of advanced control technology, some advanced control strategies can be implemented on DCS together with basic control circuits. The latter approach can effectively enhance the reliability, operability, and maintainability of advanced control.
2. Development Status of Advanced Control in Drying Process
Advanced control strategy is the core content of advanced control systems. Currently, there are many types of advanced control strategies, and the main advanced control strategies in the drying process include predictive control, fuzzy logic control, neural control, adaptive control, and system.
2.1 Model based control
2.1.1 Adaptive Control
The basic principle of adaptive control is to adjust the control parameters at any time according to the changes in drying process parameters and external disturbances, so that the dryer is in an ideal working state. Adaptive control has the advantages of being applicable to various grain drying machines, without requiring any data about the characteristics of the drying machine itself, without special requirements for environmental and grain conditions, with a fast response speed to interference from the controller, and the parameters in the control model can be automatically adjusted according to changes in external conditions. Nybrant (1985) from Sweden applied self correction technology to the control of cross flow grain dryers. The exhaust temperature of the dryer is used as the output variable, the grain discharge rate is used as the controlled variable, and an automatic regression moving average (ARMA) model is selected to represent the dynamic characteristics of the cross flow dryer. Validation tests were conducted on a cross flow dryer in the laboratory, and the standard deviation of control error was 0.13 ℃ during the last 50 sample periods. The results indicate that the adaptive controller can accurately control the exhaust temperature. Liu Jianjun [5] (2003) conducted research on the HTJ-200 dryer, quantitatively analyzed the system through online sample collection and intelligent optimization algorithms, established a process intelligent model determined by real-time detection data, and then called the artificial intelligence model through intelligent optimization algorithms to obtain the control rules of the system. The control program provided the control quantity, which was then converted into D/A and output to the executing components. Li Xiaobin et al. [3] (1998) studied the advanced control system of vacuum freeze-drying equipment. Two adaptive and self-tuning control methods, DRA algorithm and critical ratio method, were adopted to meet the process requirements of different freeze-drying materials, solving the lag problem of the main control parameter of the controlled object - temperature.
2.1.2 Model Predictive Control
The research field of process control theory is model predictive control, which is an optimization control algorithm based on models, rolling implementation, and feedback correction. It is particularly effective for controlling nonlinear and large lag processes.
Forbes，Jacobson，Rhodes， Sullivan (1984) and Eltigani designed a model-based drying controller, whose control behavior is based on a process model and a so-called fake inlet grain moisture content. The drying rate parameter is intermittently updated based on the difference between the predicted value of the model and the measured moisture content at the sensor outlet. The difference between Forbes and Eltigani controllers lies in the type of process model used in the control algorithm. Liu Qiang (2001) from the University of Michigan proposed a model predictive controller for a cross flow dryer. The simulation test was conducted on a Zimmerman VT-1210 tower cross flow grain dryer, and the controller established using Labview was able to operate successfully, achieving control of the corn moisture content at the outlet within 0.7% of the set point. The controller can effectively compensate for a wide range of changes in the moisture content of the grains entering the dryer, as well as large step changes in the temperature of the hot air.
In the research of model predictive control, a lot of work is focused on the establishment and solution of process models, and the issue of drying quality is considered in the models. France's P Dufour et al. (2003) extended model predictive control to system models using partial differential equations (PDES), enabling the large-scale application of PDES equations. They proposed a global model aimed at reducing the online computation time caused by PDE models based on optimization task solutions. Develop a universal MPC framework that combines with the IMC structure widely used in practical applications. Two feedback loops are used in the IMC MPC structure to correct process performance and simulation errors caused by model-based online optimizers. Helge Didriksen from Denmark [29] (2002) developed a dynamic first-order rule model for describing the conversion of mass, energy, and momentum in a rotary dryer, and applied it to predictive control in sugar beet drying. The results indicate that the model has good predictive ability with changes in operating variables and disturbances. By simulating and comparing model predictive control with traditional feedback control, model predictive control showed better performance. In 1997, I.C. Trelea, G. Trystram, and F. Courtois from France designed a nonlinear predictive optimization control algorithm for batch drying processes and tested it on a pilot scale dryer. Experiments have shown that the algorithm can handle important disturbances and failures. This control algorithm can be easily applied to other batch processes, such as freezing, sterilization, or fermentation. Some scholars use neural networks for modeling predictive control processes. Jay [32] (1996) first applied neural network models for predictive control of drying processes. France's J A. Hernandez Perez et al. (2004) proposed a mass and heat transfer prediction model based on artificial neural networks, which takes product shrinkage as a function of moisture and applies two independent feedforward networks with a hidden layer containing three neural cells to predict mass and heat transfer. In data device verification, simulation and experimental kinematic testing are consistent. The developed model can be used for online state estimation and control of the drying process.
2.2 Intelligent Control
Intelligent control is an emerging theory and technology, which is an advanced stage in the development of traditional control. This is a control theory characterized by model free thinking that is closer to the human brain's way of thinking. It is mainly used to solve the control of complex systems that are difficult to solve using traditional methods. The design of its controller breaks free from the constraints of system models, and the algorithm is simple and robust. At present, intelligent control technologies such as ^ control, neural control, and fuzzy control are becoming an important development direction for advanced control.
2.2.1 ^ Control
^System technology can integrate mathematical algorithms with the operational experience of control engineers, maximizing the use of existing knowledge to achieve control effects that are difficult to achieve with traditional control methods^ The control system operates in a continuous real-time environment, utilizing real-time information processing to monitor the dynamic characteristics of the system and provide appropriate control actions. Combining system technology with grain drying process control for grain production, management, and monitoring can improve the efficiency and benefits of grain production. Liu Mingshan (2001) developed a fuzzy control system for grain drying, and compared the simulation results with the measured data, which were basically consistent. Liu Shurong [13] (2001) combined system technology with drying process control and designed a fuzzy system for controlling the drying process of high moisture grains. He Yuchun [14] (2001) optimized the drying parameters during the drying process through intelligent control, and found the common points of energy consumption, efficiency, and quality in the design of the drying equipment and the drying process, so that the dryer can dry grains along the common line and maintain ideal operation of the equipment during the drying process; At the same time, temperature measurement and control technology will be interconnected with network technology to establish a simple and effective temperature based network measurement and control system.
2.2.2 Neural Network Control
Neural networks can provide effective methods for modeling complex nonlinear processes, which can be used in the design of process soft sensing and control systems. There are two main applications of neural networks in the drying process: modeling and control of the drying process.
France's J- L. Dirion (1996) et al. developed a neural controller for adjusting the temperature of a semi batch experimental reactor. The basic experiment formed a learning database for the neural network, and the neural controller can provide excellent set point tracking and interference elimination. Liu Yaqiu (2000) developed an adaptive PID controller based on a single neuron and designed a neural network model for a wood drying kiln. The BP algorithm was used to describe the input-output characteristics of the drying kiln and to learn and train the model. Experimental and simulation results showed that the conclusions obtained met the requirements of the error index. Zhang Jili [10] (2003) combined fuzzy control technology with neural network technology to design an online detection and intelligent predictive control system for grain drying process parameters. The range of variation in grain moisture content at the outlet of the dryer under intelligent control is smaller than that under manual control, with the former ranging from 13.6% to 14.4% and the latter from 12.4% to 14.2%; The fluctuation frequency of the moisture content of exported grains under intelligent control is smaller than that under manual control, with a fluctuation period of about 20 hours for the former and about 8 hours for the latter. Wang Pin [11] (2003) used an improved BP network algorithm to establish a neural network model for the drying tower, and established a neural network controller through the neural network model to achieve intelligent control of grain moisture drying in the arch drying tower system, improving the quality and efficiency of grain drying.
Liu Yongzhong [8] (1999) applied artificial neural network system theory to predict the characteristics of freeze-drying process. The drying process characteristic parameters such as drying time, the proportion of sublimation drying time, the productivity of dried products, and the sublimation interface temperature were used as the output parameters of the network model. The predicted results of the network were compared with the calculations of the mathematical model, and the predicted results were in good agreement with the calculated results. Zheng Wenli [7] (2000) used artificial neural networks to intelligently simulate the weight changes of freeze-dried materials during the freeze-drying process: learned the orthogonal experimental results of freeze-drying process conditions, and used the learned network to predict and optimize the process conditions.
2.2.3 Fuzzy Control
Fuzzy control is a rule-based control that directly adopts linguistic control rules. It is based on the control experience or relevant knowledge of on-site operators, and does not require the establishment of a mathematical model of the controlled object in the design. Therefore, the control mechanism and strategy are easy to accept and understand.
At present, the fuzzy control method is mainly used for drying process control both domestically and internationally. Zhang Qin et al. (1994) conducted a study on fuzzy control of a continuous cross flow grain dryer. The operation of the dryer was controlled by adjusting the power of the heater and the speed of the unloading agitator, and the experimental control success rate was verified to be 86.4%. Li Junming et al. (1996) developed fuzzy control rules based on the temperature of the hot air in the drying tower, using the observation and experience of a skilled operator in corn drying production through a sensory system. They used fuzzy control to adjust the speed of the displacement motor and proposed that the self-organizing fuzzy controller of the cross flow corn dryer should adopt an open-loop fuzzy control system to solve the problem of large lag in the corn drying process. Li Yede and Li Yegang (2001) designed a fuzzy intelligent controller with 89c51 microcontroller as the core. Through online drying experiments on wheat in a co current dryer, it was proven that the system has short response time, small overshoot, and high control accuracy. However, fluctuations in the moisture content of the imported grains can affect the drying process.
Many graduate students in China are engaged in research on fuzzy control of grain dryers. Meng Xianpei from Northeastern University [18] (2003) used fuzzy set theory and optimization algorithms to establish an intelligent model of the grain drying system and fuzzy rules for the fuzzy control system in the intelligent modeling and control of the grain drying tower, and designed a fuzzy controller for the system. Tang Xiaojian from Harbin Institute of Technology [20] (2003) studied the multivariable fuzzy control method of mixed flow grain drying tower based on TS model, conducted control simulation on the system, and compared it with manual control method and traditional fuzzy control method. Cao Yanming from South China Agricultural University [21] (2000) developed an automatic control system for a rice circulating drying machine based on the characteristics of the high humidity rice circulating slow drying process, using a design method that simulates human thinking through fuzzy control. Su Yufeng (2002) from Northwest Institute of Light Industry adopted a fuzzy algorithm based on workers' actual operating experience and used a microcontroller to control the freeze-drying system, improving the automation level of the equipment.
