Abstract: For a three-stage continuous heating furnace using heavy oil as fuel, automatic temperature regulation of the furnace was achieved by implementing a computer DCS system and employing expert control. Actual operation results show that the control effect is very good. Keywords: Heating furnace temperature, DCS, Expert control 1 Introduction A steel rolling mill heating furnace is a three-stage continuous heating furnace using heavy oil as fuel. Each stage has a row of burners at the top and bottom, but the burners at the top of the second heating stage are not used. Many factors affect the furnace temperature, and there are many uncertain interferences, such as whether the billet to be heated is cold or hot, the quality of the fuel oil, the tapping speed, the furnace pressure, and the air-fuel ratio. To ensure heating quality, increase output, reduce energy consumption, improve control performance, and effectively suppress various disturbances, the electrical control part was first modified using a DCS system, replacing the original decentralized manual control, and automatic temperature regulation of the heating furnace was achieved using expert control. 2 System Hardware Configuration The computer distributed control system (i.e., the computer DCS system) consists of a host system and a slave system. The upper-level system uses an industrial control computer and Siemens WinCC configuration software to complete the real-time display, storage, alarm processing, printing, and control parameter setting of field data. The lower-level system consists of Siemens PLCs and is connected to the field devices. Communication between the upper-level and lower-level systems uses Profibus, with a maximum transmission rate of 1.5 Mbit/s, fully meeting the requirements for real-time data monitoring. The DCS system composition is shown in Figure 1. 3 Expert Controller and Software Configuration 3.1 Expert Controller This heating furnace was originally manually controlled. Due to the untimely adjustment of manual control and its reliance on the operator's experience, energy consumption could not be reduced. The conventional PID automatic control was not adaptable enough, especially when there were large differences in fuel viscosity and calorific value, and the field equipment was aging, resulting in poor control performance. Therefore, based on the actual field conditions, we adopted the concept of expert control and designed an expert controller for the heating furnace temperature. The expert controller consists of four parts: a knowledge base, a control rule set, an inference mechanism, and information acquisition and processing. Its structure diagram is shown in Figure 2. [align=center]Figure 2 Expert Controller Structure Diagram[/align] 3.1.1 The knowledge base consists of a fact set, an experience database, and empirical formulas. The fact set mainly includes knowledge about the controlled object (heating furnace), such as that the heating furnace is a three-section (two heating sections, one heating section, and a soaking section) continuous steel pushing heating furnace, with temperature detection points on the east and west sides of each section. The temperature of each section of the heating furnace is controlled by adjusting the opening of the electric actuator valve and regulating the flow of fuel oil and air. The data in the experience database includes the parameter variation range of the heating furnace, the adjustment range and limits of the control parameters, the static and dynamic characteristics, parameters and thresholds of the sensors, the performance indicators of the control system, and empirical formulas summarized by experts. 3.1.2 The control rule set, based on the experience of experts (or skilled operators) in the characteristics of the controlled object and their operation and control, can use various methods such as production rules, fuzzy relations, and analytical forms to describe the characteristics of the controlled object. This allows for the processing of various qualitative, fuzzy, quantitative, and precise information. Through on-site investigation and observation of the operational control experience of skilled operators across three shifts, combined with theoretical knowledge from textbooks, we constructed a control rule set using production rules. During debugging, we initially found slow response and long adjustment times, so we considered a callback variable in the rules. However, we later discovered that the maximum deviation value increased (i.e., overshoot increased), so we decided not to add a callback when the deviation was large, but to add a callback when the deviation was less than a certain value, so that the system could reach steady state as quickly as possible. Here, input quantity 1 (temperature deviation) is divided into nine levels according to its magnitude range; input quantity 2 (change in temperature deviation) is divided into seven levels according to its magnitude. The output (incremental) of the control quantity (opening degree of the electric actuator valve) is divided into 13 levels. A limiting condition is added. Finally, 55 control rules were summarized, forming the control rule set of this expert controller. 3.1.3 Inference Mechanism Due to the relatively small size of the knowledge base and control rule set of this expert controller, the search space of the inference mechanism is limited, and a forward inference mechanism is adopted. The control rules are matched one by one from front to back until a match is successful (of course, when writing control rules, it is necessary to consider the possibility of loss of control). 3.1.4 Information Acquisition and Processing Information acquisition is mainly obtained through feedback information from closed-loop control and input information from the system. Through information processing, useful information for control, such as the error of the control system and the change of the error, is obtained. This heating furnace uses thermocouples to detect the furnace temperature of each section, and then compares it with the set value of the furnace temperature of each section to obtain the deviation, and calculates the change of the deviation. Segmented control is implemented. In addition, information processing also includes necessary data filtering measures. This system uses arithmetic mean filtering. 3.2 Software Configuration 3.2.1 Monitoring Interface Configuration The collected data is displayed in real time on the host computer, so that the operator can understand the working status of the heating furnace in a timely manner. The main data can be stored for up to one year, and can be easily queried and printed. Control parameters and alarm parameters can be set. The main monitoring interface configured with WinCC is shown in Figure 3. [align=center]Figure 3 Main Monitoring Interface[/align] In addition, according to the needs of the site, interfaces including steam drum screen, control parameter setting interface, historical trend curve interface, pressure and water level interface, alarm interface, and report printing interface were also configured. They are all linked to the measured data and dynamically displayed on the interface. At the same time, manual and automatic seamless switching can be performed on this interface. 3.2.2 Implementation of Control Algorithm The control algorithm of expert control is implemented on the lower-level system (PLC) and programmed using Step 7. Considering that the production process of the heating furnace is a slow-changing process with time delay, the output of control signals cannot be too frequent, otherwise oscillation will occur. Therefore, a control signal is output once every 5 sampling cycles. Its software flowchart is shown in Figure 4. [align=center]Figure 4 Software Flowchart[/align] 3.3 Application Effect After adopting the DCS system and expert control, the operation is convenient for the operator. The entire operation status of the heating furnace can be seen at a glance on the computer screen. In terms of control performance, the adjustment is timely, the overshoot is small, and the fluctuation is within ±5℃ during stable operation. From the perspective of energy conservation and consumption reduction, approximately 3 tons of heavy oil are saved daily. At 1100 RMB per ton, this translates to monthly savings of approximately 100,000 RMB and annual profits of approximately 1.2 million RMB. This brings economic benefits to the enterprise. 4. Conclusion Since its commissioning, this system has been operating stably. Its expert control algorithm is simple and practical, with good control effects, and is worth promoting for small and medium-sized steel rolling heating furnaces. References: 1. Li Shiyong, Fuzzy Control, Neural Control and Intelligent Control Theory, Harbin: Harbin Institute of Technology Press, 1996: 244-246. 2. Sun Zengqi, Intelligent Control Theory and Technology, Beijing: Tsinghua University Press, 1997: 264-270.