[ Abstract ] Combining the time-varying, lag-dependent, and nonlinear characteristics of ash content control, this paper proposes a controller based on a combination of artificial neural networks and fuzzy control. Utilizing the self-learning, adaptive, and parallel processing capabilities of artificial neural networks, fuzzy control rules are transformed into learning samples for the neural network, which are then memorized using the backpropagation (BP) learning algorithm of the ANN. Experiments show that the controller exhibits fast response speed, high accuracy, and robustness. [ Keywords ] Temperature, Fuzzy Control, Artificial Neural Network, Frequency Converter. The coal blending system in a coal mine preparation plant is a nonlinear, time-varying process with large lag. Ash content is a crucial control parameter in this process. Currently, PID control devices are commonly used, but this type of PID control has poor adaptability and is highly sensitive to high-frequency interference. Moreover, since ash content is a nonlinear time-varying parameter, coupled with the randomness of the working environment, it is difficult to establish an accurate mathematical model. Fuzzy and neural network control, on the other hand, is a nonlinear control method that can achieve satisfactory control results for systems where a mathematical model is unavailable or is quite coarse. 1. Analysis of Controlled Objects and Overall System Framework Based on the process flow of the coal preparation plant and the analysis of the coal blending system, the following control characteristics of the coal blending system can be summarized: (1) There is a serious nonlinearity between the coal feeder's feed rate and the frequency of the inverter; (2) There is a large time lag. Due to the certain reaction period of the online ash meter and the change in the coal feeder speed, the coal flow rate changes only after the conveyor belt is used, resulting in a time lag. (3) There is grayness. Information such as the particle size, shape, density, moisture content of the coal in the coal bunker and the friction between the coal and the outer sleeve cannot be obtained, and these factors affect the flow rate of the coal outlet. (4) When the coal storage bunker is blocked, penetrated, or empty, the feeder baffle needs to be manually adjusted, and the characteristics of the object change. In summary, due to the constant changes in ash content and quantity of the coal being blended, the lag in coal flow transportation, and the detection lag in ash content measurement, we cannot simply follow the flow control based on the expert system-based coal blending method described in the previous chapter in actual coal blending operations. The main reason for this is that expert reasoning in coal blending control strategies is based on superficial knowledge and cannot fully utilize the implicit knowledge such as unmeasurable factors and their changes contained in the real-time sampled data of the system. Therefore, the adjustment of coal blending flow in the expert system can only be intermittent, i.e., diagnostic adjustment is only made when the coal flow changes significantly (such as when the silo is empty or blocked). We consider using the ash content or calorific value of the online ash analyzer as a feedback parameter, and the coal feed rate of the finished product coal (which can be finely adjusted) as the adjustment method. Since the quality assessment of the loaded coal is based on the test value of the average sample mixed from each wagon, the cumulative ash content (calorific value) is the main criterion for measuring the loading quality. With ash-based feedback control, it is not necessary to know the exact ash content, coal flow rate, and other complex factors of each coal bunker to achieve the above requirements. The system control structure diagram based on ash feedback is shown in Figure 1. [align=center] Figure 1 System control structure based on ash feedback[/align] In Figure 1, is the ash setpoint, is the actual output value, and are the proportional coefficients, x1 and x2 are fuzzy input values, FNNC is the fuzzy neural network controller, and are the fuzzy value increment, actual value increment, previous actual value, and current actual value, respectively. 2. Coal preparation plant coal blending process Jining No. 2 Coal Mine Coal Preparation Plant is a large-scale mine coal preparation plant with an annual processing capacity of 4 million tons of raw coal. The coal washing process involves mixed jigging of the raw coal and treatment of coal slime by sedimentation centrifuge and filter press. The main products include: screened raw coal, lump clean coal, fine clean coal, moving screen clean coal, and washed mixed coal. The heavy media separation system for lump coal was added later. Materials of 300-50mm size are separated using an inclined wheel separator. The resulting lump refined coal can be sold separately as lump coal or as general thermal coal. The coal preparation plant's blending and loading process is shown in Figure 2 (the No. 2 coal preparation plant currently has two loading lines, 539 and 540, with the same loading process). [align=center] Figure 2 Coal Preparation Plant Blending and Loading Process[/align] The Jining No. 2 Coal Mine's coal preparation plant currently has four types of coal, stored in eight coal bunkers: three raw coal bunkers, two washed fine coal bunkers, two lump refined coal bunkers, and one high-ash refined coal bunker. The high-ash refined coal bunker is planned for future construction and is included in this plan. There are two tracks for train loading, corresponding to the 539 and 540 loading conveyor belts respectively. Each raw coal bunker has 4 feeders, 2 of which supply 539 belts and the other 2 supply 540 belts; each of the two fine coal bunkers has 1 feeder, which supplies two belts respectively; each of the two lump fine coal bunkers has 6 feeders, which supply two belts respectively; the high-ash fine coal bunker has 2 feeders, which supply two belts respectively through belts and chutes (control gates). Thus, each belt has 14 feeders feeding coal. 3. Technological problems affecting coal blending effect Under the current technological conditions, there are several technological problems affecting the coal blending effect in coal blending operation: (1) The ash content of the blended coal is usually constantly changing, and the ash content of some coal types (such as raw coal) fluctuates within a large range. (2) The coal feed rate of each blending bunker is also constantly changing. Whether it is gate adjustment or frequency conversion adjustment, for the same opening degree or the same frequency, the coal quantity may be different at different times. (3) The instantaneous value of the ash analyzer has a delay of about 1 minute, and there is a certain transmission time for the coal flow on the conveyor belt, resulting in a significant lag in the coal blending system. In view of the above process problems affecting the coal blending effect, we have implemented the following improvement measures: adopting ash feedback control; online monitoring of coal quantity; and using frequency converter to regulate the coal quantity. The modified coal blending and loading process is shown in Figure 3: [align=center] Figure 3 Modified Coal Preparation Plant Coal Blending and Loading Process Flow[/align] 4. Working Principle of Automatic Coal Blending System First, the target ash content and coal type are manually selected. The system determines the number of coal feeders. After calculation, the ash content of the product to be loaded is qualitatively determined. Based on the target ash content of the blended coal and the ash content of each coal bunker, the feed flow rate of each coal type is calculated. The number of coal feeders is used for coarse adjustment of the feed rate. By real-time monitoring of the ash content after blending on the loading belt and the actual ash content loaded, it is compared with the target ash content. Then, the system calculates and adjusts the instantaneous flow rate of the belt scale and guides the operation of the frequency converter. The frequency converter is used for fine adjustment of the feed rate, so that the blended coal is as close as possible to the target ash content. The working principle diagram of the automatic coal blending system is shown in Figure 4. The frequency converter uses Rockwell Automation AB frequency converters, with a voltage of 380V and a power of 45KW, and a total of 28 units for two belts. 5. Conclusion This paper provides a detailed introduction to the coal blending system process of a coal preparation plant, analyzes the process factors affecting the blended coal products, and modifies the coal blending process based on this analysis. A fuzzy neural network-based ash feedback control strategy is proposed, providing a new method for implementing intelligent coal blending control. [align=center] Figure 4 Working principle diagram of the automatic coal blending system[/align] References 1 Li Shiyong. Fuzzy Control. Neural Control and Intelligent Control Theory. Harbin: Harbin Institute of Technology Press, 1998 2 Zhang Weiguo, Yang Xiangzhong. Fuzzy Control Theory and Application. Xi'an: Northwestern Polytechnical University Press, 1999 3 Zhang Liming. Models and Applications of Artificial Neural Networks. Fudan University Press, 1992 Author Biography: Wang Zhijie, male, born in 1964, from Weifang City, Shandong Province, graduated from the Department of Automation, Shandong University of Science and Technology in 1985, professor, PhD, mainly engaged in research on industrial microcomputer control, fuzzy control and networks. He has published 30 papers and achieved numerous scientific research results.