Abstract: Based on a brief overview of the current development status of computer-integrated production systems in coal preparation plants, this paper proposes an Intelligent Control and Maintenance Management Integrated System (IICMMS) for coal preparation plants, based on multi-agent architecture and integrating intelligent control of the production process, equipment maintenance, and production management. To meet the needs of establishing an IICMMS for coal preparation plants, a distributed network control system based on field intelligent nodes is studied, along with its implementation strategy to overcome the weaknesses of the widely used PLC control systems in coal preparation plants. The software framework of the multi-agent-based intelligent control and maintenance management integrated system is investigated, and its implementation methods are discussed.
Keywords: Intelligent control, multi-agent system integration, production system, coal preparation plant
1 Introduction
Coal washing and processing can significantly reduce ash and sulfur content, and reduce emissions of pollutants such as dust and SO2. Currently, the raw coal washing rate in developed countries is 50% to 90%, while the washing rate in my country is only 20% to 30%. Moreover, factors such as small average plant size, poor equipment reliability, and low degree of automation lead to high coal washing costs, which are the main reasons restricting the development of coal washing in my country [1].
Since the 1990s, my country has systematically researched and applied advanced technologies in centralized computer control of coal preparation plants, intelligent control of washing and beneficiation processes such as jigging and flotation, integrated production of coal preparation plants, and CIMS of coal preparation plants. A number of demonstration and application projects have been established, which has raised the production process control and production management of coal preparation plants in my country to a new level.
However, in the current system, the three subsystems of intelligent control of coal preparation plant process, equipment maintenance and production management are designed and implemented independently, and are separated, which makes hardware resources and operation control lack optimization. The computer control system of coal preparation plant is still a highly centralized structure. Usually, centralized control adopts PLC network system, while flotation, jigging, heavy media and other process control adopt their own independent microcomputer system. The independence of software and hardware creates data islands. Although database interconnection technology (ODBC, JDBC) can be used to realize data access, such isolated data still cannot realize process optimization. The integration of these three subsystems is the key to improving the comprehensive production efficiency of coal preparation plant, and is also the basis for further development of coal preparation plant CIMS[2].
Multi-Agent Systems (MAS) offer unparalleled advantages for complex systems. Coal preparation plant production is a complex industrial process with multiple input and output variables. Most parameters exhibit nonlinearity, uncertainty, and spatial distribution. Establishing an integrated intelligent control, maintenance, and management system for such a production process has always been a technical challenge. However, the autonomy, distribution, and adaptability of agents offer new approaches to solving this problem. This paper utilizes multi-agent technology to implement a distributed intelligent control and maintenance management system (IICMMS-CP) for coal preparation plants based on Functional Management Systems (FMS). Specifically, it studies the distributed measurement and control network, the overall structure, and implementation methods of the distributed intelligent control, maintenance, and management system for coal preparation plants, aiming to decentralize process control functions and achieve intelligent, networked, and integrated process control, equipment maintenance management, and production management.
2. IICMMS-CP Architecture
The IICMMS-CP system is functionally divided into three parts: Process Control (IPC), Equipment Maintenance (BAM), and Production Management (PM). Process Control refers to the use of intelligent control theory to control coal preparation processes such as heavy media, jigging, and flotation, taking into account the multivariable, multi-objective, and nonlinear characteristics of the coal preparation process. The Production Management Agent handles coal preparation task planning, target optimization, process scheduling, and statistical analysis. Equipment Maintenance includes equipment status monitoring, process parameter monitoring, intelligent fault diagnosis, and equipment maintenance management.
In a multi-agent system, each agent performs different functions and has its own scope of action. The coordination, cooperation, and other interaction methods between agents are complex problems. Based on the characteristics of the coal preparation plant production process, in the intelligent control and maintenance management system of the coal preparation plant based on multi-agent, it is not only necessary to handle the behavior of each agent related to control, maintenance, and management, but more importantly, to coordinate their interactions in a distributed environment. In 1997, Bonasso R. P et al. proposed a three-layer multi-agent architecture in the development of autonomous robots. In 1998, Schreckenghost D et al. applied it to the intelligent control of the life support system of the space station, proving that it is suitable for distributed process control systems and decision-making systems. [3-5]
The multi-agent-based IICMMS-CP system adopts a three-tiered hierarchical organizational structure as shown in Figure 1. The bottom layer is the closed-loop control and data interface layer agent (DACM Agent). For process control (IPC), equipment maintenance (BAM), and production management (PM), the most basic tasks are closed-loop control and data acquisition. Therefore, the bottom-layer agent implements basic control of production process parameters and data acquisition of equipment status and process parameters. The second layer is the task layer agent (TA Agent). This layer schedules the bottom-layer agents to achieve predetermined goals based on the tasks set by the upper layers and the specified models. The top layer is the planning and coordination layer agent (MD Agent). For the overall goal, the top layer decomposes the task into sub-tasks and, during the scheduling of lower-level tasks, promptly modifies the target values and system model based on the global system status, execution results, and manually set values. For example, for the density and level control of the suspension in the heavy media production process of a coal preparation plant, based on the production management plan and equipment status, the MD layer agent sets the control target, specifies the suspension circulation system control model, and determines and schedules the TA layer agent to execute the density and level control tasks. Based on the established model, the TA layer Agent decomposes this control process into tasks such as density and level fuzzy control, level and density data acquisition, and diversion valve opening control, and schedules the underlying Agent to execute them.
The IICMMS-CP User Interface (UI Agent) provides an information interaction interface for human intervention. Through the user interface, operators can set control parameters, modify control models, fault diagnosis models, production planning models, and equipment management plans, and directly control equipment.
[align=center] Figure 1. Organizational Structure of the IICMMS-CP Agent System[/align]
In the aforementioned hierarchical organizational structure, each layer of intelligent agents is only responsible to the intelligent agents above it, forming a tree-like branching logical structure. However, in a distributed system, each intelligent agent utilizes a unified communication protocol and message response mechanism to achieve information interaction, coordination, and negotiation within a fieldbus control network environment.
3. IICMMS-CP Distributed Control Network Environment
Distributed intelligent control and maintenance management systems require a real-time, distributed computing environment. In multi-agent-based IICMMS-CP, each agent possesses a high degree of autonomy and self-governance, while the distributed environment provides an interactive mechanism for coordination and cooperation among these agents. Fieldbus-based control systems (FCS) provide such a real-time, fully distributed control computing environment. Fieldbus communication protocols and fieldbus technology standards provide a cooperative and interoperable interaction mechanism between agents. Agents are implemented by network devices, and the interaction and collaboration of these network devices achieve the overall system objectives.
To construct a distributed control network environment, in the project "Computer Integrated Production Platform and Comprehensive Automation System for Coal Mines and Coal Preparation Plants," a distributed computer control network platform for coal preparation plants was researched and developed based on the production environment requirements and current equipment technology status of coal preparation plants, utilizing standard fieldbus and a distributed architecture. This platform consists of fieldbus network interfaces, intelligent I/O substations, gateways, intelligent substations (IS), control substations (CS), and an IICMMS-CP central station (CPS), as shown in Figure 2.
Figure 2. Fieldbus control network of the coal preparation plant: IS - Intelligent Substation, CS - Control Station, NG - Gateway, CPS - IICMMS Main Station.
The platform divides the network into two segments—CAN and Industrial Ethernet—based on real-time requirements. Industrial Ethernet uses gateways to forcibly isolate nodes with frequent and high-volume data transmission, reducing the uncertainty of Ethernet data transmission time. Within the system, agents are mapped to various intelligent substations, control substations, and the IICMMS-CP central station. Message passing is accomplished via CAN-bus fieldbus and Industrial Ethernet. Typically, logic control agents are handled by the PLC, data acquisition agents by intelligent I/O, other task agents by the IS (Intelligent Information System), and coordination and scheduling agents by the control substations. The IICMMS-CP central station executes the IICMMS-CP coordination agent. A single station can run multiple agents of various types.
4. Implementation of Agent-Based IICMMS-CP System
A large coal preparation plant processes 5 million tons of raw coal annually, employing screening, heavy media, and jigging washing processes, followed by coal slurry water concentration and pressure filtration. The IICMMS-CP system monitors and maintains the status of key equipment. It enables fault detection and alarms for major equipment, equipment main/standby rotation, accumulated working time, maintenance prompts, and management. Production management includes optimizing product structure and production targets, developing a reasonable product structure, and monitoring production process parameters. Daily production indicators are analyzed, providing early warnings of quality exceeding standards. Management and operational indicators are analyzed against national and industry standards, providing analysis results and recommendations. Intelligent process control enables intelligent control of the heavy media and jigging processes.
4.1 Three-layer Agent Structure Implementation Mode
[align=center]Figure 3 Three-layer control structure[/align]
For problems related to IICMMS-CP intelligent control, equipment maintenance and fault diagnosis, and production management decision-making, based on the aforementioned three-layer structure, its implementation is divided into the following three parts [6,7], as shown in Figure 3. Planning and Decision-Making (PD): When completing control tasks, a hierarchical task planning network is used to decompose an overall goal into several sub-goals, and each sub-goal into several tasks. Resources needed to complete these tasks are allocated, and the execution process of these tasks is guided and monitored. In fault mode discrimination, based on fuzzy knowledge and inference rules, the fault type is determined, and maintenance or online adjustment tasks are proposed.
Task Execution Sequence (TS): A sequence of tasks that are broken down from planning or operation into subtasks and then used to implement those tasks.
Execution Management (SM): Dynamically connects the closed-loop control module and the data acquisition module into an execution network to complete the tasks specified by the previous layer.
The above implementation method is highly robust, capable of promptly detecting and correcting errors at each step. The SM (Search Engine Controller) determines whether control commands have been successfully executed based on instrument data and reacts accordingly. The TS (Search Engine Controller) can select different strategies based on the real-time system status. For example, to reduce the ash content of jigging process products, adjustments can be made based on the current coal particle size and washability, such as adjusting bed thickness, coal feed rate, and changing the ventilation and water system, and the control strategy can be adjusted based on the control effect. When a target fails, the PD (Product Development Controller) can re-divide sub-targets and plan new tasks.
In IICMMS-CP, three TS Agents—heavy media, jigging, and concentration—are used to achieve fuzzy control of liquid level and density in the heavy media process, control of coal feeder speed, fuzzy control of bed thickness in the first and second stages of jigging, and control of air valves and water valves, respectively.
4.2 Communication between Agents
For multi-agent systems, communication between agents is fundamental to collaboration. We utilize a predefined messaging system to implement communication between various agents, and transmit data via CAN and TCP/IP protocols.
We have defined the following messages: SM Update Message: When an SM-class Agent detects a change in the state of the monitored system, it broadcasts this information to the system. Fault and Production Process Discrimination Agents use this information to infer and determine the state of system equipment and processes.
Fault Message: Fault and production process identification agent sends fault information to the system, including information about the faulty equipment or process, fault type, and handling suggestions.
System task setting messages: These notify the system of the goals to be achieved or the tasks to be completed. Currently, these messages are sent manually. They enable parameter setting, running specific programs, and fault recovery. Fault recovery messages: These allow for online correction of specific faults or further verification of a particular fault. The fault identification agent sends this message, and the agent and the management system (SM) execute specific procedures.
The above are four types of broadcast information. In addition, during operation, there will be query and response messages between Agents at different layers. We have defined six types of query and response messages: HA-PM, HA-TS, HA-SM, PM-TS, PM-SM, and TS-SM. Because each layer handles problems and presents data differently, we have defined the structure and format of the information. For example, the execution Agent directly corresponds to the instruments and actuators, processing digital or switch information. However, the PA layer uses fuzzy numbers and fuzzy inference to determine faults. Therefore, when the execution Agent answers a PA query, it needs to convert the digital data into fuzzy values. For example, a specific liquid level value might be converted into "extremely low," "low," "medium," "high," or "extremely high." To achieve this information conversion, an information interface Agent was designed.
5. Conclusion
The multi-agent intelligent control and maintenance management integrated system for coal preparation plants is proposed based on research and applications in distributed control networks, intelligent process control, and production information management. Its distributed hierarchical structure, collaborative and parallel computing features meet the needs of integrated control in coal preparation plants with large spatial distributions, close process connections, and complex systems. Its application further improves the reliability and accuracy of process control, enhances the accuracy of fault diagnosis, and improves the intelligence of production scheduling, laying a technical foundation for system integration and flexible production in coal preparation plants. The system's main functions have passed on-site commissioning, and application demonstrates that the system has good effects on stabilizing product quality, improving production efficiency, and reducing consumption.
The innovation of this paper lies in proposing an integrated model for intelligent control, equipment maintenance, and production scheduling management of coal preparation plants based on agent technology, taking into account the production characteristics of coal preparation plants. The paper studies the system structure and implementation technology, completes the process control and optimization functions of some coal preparation processes based on agent technology, and proves the feasibility and advancement of the method.
References
[1] Wang Puzhou. Development Trends and Suggestions for Automation in Coal Preparation Plants [J]. Coal Preparation Technology, 2000(6): 45-47
[2] Shi Youqun, Tan Chao, Gong Dunwei. System integration of CIMS in coal preparation plants [J]. Journal of China University of Mining and Technology, 1999, 28(5): 479-481
[3] Wang Jinyun. Multi-Agent Distributed Collaborative Management Mechanism for Handling Information Security Emergencies [J]. Microcomputer Information, 2005, 26(36): 32-36 [4] Ouelhadj D. A Multi-Agent System For the Integrated dynamic scheduling of steel production [EB/OL]. http://www.asap.cs.nott.ac.uk/publications/pdf/Ouelhadjthesis.pdf, 2003 [5] Han W, Jafari M A. Component and Agent Based FMS Modeling and controller synthesis [J]. IEEE Trans. syst., Man, and cybernetics-part C: applications and reviews, 2003, 33(2): 193-206
[6]Ahmad MR, Shamsudin HM, Amin R M. proc. of Seventh international Conference on Control, Automation, Robotics And Vision (ICARCV'OZ) [C], IEEE Computer Society Press, 2002: 220-227
[7]Brennan RW, Fletcher M, Norrie DH. An Agent-Based Approach to Reconfiguration of real-time Distributed control system[J]. IEEE Trans. On Robotics and Automation, 2002, 18(4):444-451
