Abstract : This paper discusses the overall structure of a fieldbus control system. Based on this, it elucidates the role and implementation of soft measurement technology in the fieldbus control system. Finally, it uses a model fieldbus control system established in the laboratory to illustrate how the fieldbus control system and its soft measurement technology are implemented. Keywords : Fieldbus, Control System, Soft Measurement, Soft Instrument, FCS Fieldbus is attracting increasing attention due to its superior characteristics. Fieldbus control systems use digital signal communication and place the basic control level directly in the industrial field, which can greatly improve the real-time performance of process control and reduce the possibility of interference signals, thereby significantly improving the performance of the control system. Therefore, fieldbus-based control systems will be an inevitable trend in the future development of control systems. The existence of a control system depends on a measurement system. Soft measurement technology is an extension of measurement tools and a powerful supplement to traditional measurement methods. Soft measurement relies on indirect knowledge to estimate variables that cannot be measured by hardware instruments. When hardware cannot complete the measurement task, soft measurement technology can play a substitute role to a certain extent. Especially for key production parameters such as product output and quality, which are usually impossible to measure with hardware instruments, the use of soft measurement can provide a powerful means to improve production efficiency and ensure product quality. Soft measurement technology has been widely used in process industries and has achieved good results [1]. Fieldbus control systems provide convenient conditions for the real-time transmission of field detection data, making the widespread use of soft measurement technology possible. The combination of soft measurement technology and fieldbus control systems can make full use of the advantages of fieldbus and soft measurement technology, ensure the effective and correct operation of the production process, and bring significant economic benefits to factories and enterprises. The combination of soft measurement technology and fieldbus technology is an inevitable trend in the future development of process control. This paper discusses the overall structure of the fieldbus control system. On this basis, it elaborates on the position and implementation of soft measurement technology in the fieldbus control system. Finally, it uses the model fieldbus control system we established in the laboratory to specifically illustrate how the fieldbus control system and the soft measurement technology therein are implemented. 1 Overall Structure of Fieldbus Control System The fieldbus control system is a computer integrated control system that uses fieldbus as the communication medium. According to the CIMS (CIPS) viewpoint, it can be divided into three levels: management decision level, high-level control level, and basic control level. ① The management decision-making level includes functions such as decision analysis, marketing, planning, offline optimization, scheduling, and production management. This mainly refers to the integrated control system providing information services and decision support for these tasks. This includes proposing development goals and marketing strategies through historical data analysis and mining, adjusting production plans based on corresponding marketing strategies, achieving integrated management of production and business information, formulating and implementing annual, quarterly, and monthly comprehensive plans, completing production plan decomposition, generating scheduling instructions based on actual production conditions, organizing balanced daily production and handling abnormal events, and providing real-time production command. ② The advanced control level is the guarantee for stable production and optimized operation of the system, and it is also the level at which humans interact with the production process. The process monitoring system receives field status information from fieldbus intelligent nodes and scheduling information from the decision-making management layer. It uses soft measurement and data correction technologies to perform completeness and consistency processing on this data, forming a real-time process database. The system then uses data from this real-time process database to achieve functions such as operation guidance, dynamic optimization, advanced control, fault diagnosis, and real-time alarms. ③ The basic control level realizes routine detection and basic control of the production process. In the fieldbus control system, it is implemented by fieldbus intelligent nodes, including the integration of various functions such as sensing and transmitting parts and PID controller parts. In the intelligent node, there is generally an independent microprocessor chip that can complete relatively complex calculation tasks. Figure 1 is a schematic diagram of the complete fieldbus control system we developed. 2 General Soft Measurement Technology In the actual production process, it is necessary to detect and control the product quality and output parameters in real time. However, in the actual process, there is a large class of variables that cannot be directly detected by traditional measuring instruments [2]. The main reasons for this problem are as follows: ① Due to the limitations of process conditions, it is impossible to detect or it is not allowed to install measuring instruments; ② The current detection methods are not perfect and are not sufficient to complete the required detection tasks; ③ Advanced detection tools are too expensive; ④ The measurement lag caused by the measuring instruments makes the measurement results unable to meet the real-time requirements; ⑤ Traditional sensors can only sense a certain state variable. In some cases, it is necessary to fuse the measurement results of multiple sensors to obtain a more ideal result. Soft measurement technology relies on indirect knowledge to estimate variables that hardware instruments cannot measure. It can serve as a substitute to some extent when hardware cannot perform the measurement task. Soft measurement technology mainly includes four aspects: selection of secondary measurement variables; data processing; establishment of the soft measurement model; and online calibration of the soft measurement model. Selection of secondary measurement variables refers to determining the corresponding auxiliary process variables for a specific production process and selected output variables to be estimated. Data processing refers to how to detect, eliminate, and correct errors in the raw measurement data of auxiliary variables. Establishment of the soft measurement model refers to how to determine the process model through various information and data. Online calibration of the soft measurement model refers to how to adjust the model itself when the process state changes. Through long-term theoretical and applied research, we believe that soft measurement methods themselves have similarities, and we have proposed the concept of general soft measurement technology, forming a general productized soft measurement software package. The generalization of soft measurement technology needs to focus on solving two key problems: first, establishing a general soft measurement model; and second, solving the problem of selecting secondary variables for specific production processes. Only when these two problems are solved can the generalization of soft measurement technology be truly realized. To address these two problems, we have established a data-oriented solution. In this approach, the selection of quadratic variables and the establishment of the soft measurement model do not rely on the process's mechanistic model. Instead, they directly utilize historical process records from production equipment and the scheduling room. Correlation and factor analyses are performed on this data to identify the main measurable variables influencing the estimated variables. The historical data is then appropriately clustered and organized to form the sample data for building the soft measurement model. The soft measurement model employs a fast-learning, high-accuracy RBF neural network and uses an orthogonal least squares algorithm to automatically adjust the neural network structure based on the specific process, thus achieving generalization. Based on this approach, we have developed a data-oriented general soft measurement software package. The specific workflow when using this package in actual production is shown in Figure 2. The first two steps are completed manually for the specific production process, while the last four steps are completed by the soft measurement software package. 3. Soft Measurement Technology in Fieldbus Control Systems Soft measurement technology can be implemented in fieldbus control systems in two ways: First, it integrates data fusion-based soft measurement technology into the intelligent field nodes of the fieldbus, enabling multi-sensor integrated measurement of single or multiple process variables, providing a "hardware" form of soft instrumentation. Second, it establishes dedicated soft measurement nodes within the fieldbus control system to complete the soft measurement tasks for the entire system. As seen from the overall structure of the fieldbus control system, soft measurement technology can function as a dedicated node at the high-level control level to complete the soft measurement tasks for the entire system. This is a relatively convenient and universal form of soft measurement technology. In this approach, the fieldbus control system has dedicated soft measurement nodes. The soft measurement system obtains real-time production process data from the process real-time database, performs real-time soft measurement calculations according to the engineer's pre-defined soft measurement system configuration scheme, and promptly sends the soft measurement results to the process real-time database for use by other tasks. Other process control and optimization tasks use the results generated by the soft measurement, just as they use data collected from the field via sensors. In this approach, the soft measurement system communicates with the database of the entire system, fully utilizing information across the entire plant and ensuring the validity of the soft measurement results. This combination is illustrated in Figure 1. Besides this application method, soft measurement can also function like "hard" instruments, directly performing soft measurements on specific variables. In this approach, the soft measurement system is directly integrated into the intelligent node of the fieldbus control system, appearing at the basic control level. As mentioned earlier, unlike traditional field instruments which can only perform one function, the intelligent node in the soft measurement system can perform multiple functions simultaneously. The soft measurement system is a powerful supplement to the detection function of the intelligent node. The structure of an intelligent node with soft measurement technology is shown in Figure 3. The dashed box in the figure represents the fieldbus intelligent node using soft measurement technology. This type of fieldbus node, like traditional field transmitters, has a certain degree of versatility for a certain type of production process, and only requires changes to certain parameter settings to adapt to different production conditions, unlike traditional transmitters which require replacement with different models of instruments. This is very important and effective in modern production, which is increasingly trending towards high flexibility and small-batch production, where the products, scale, and quality requirements of the production process frequently change. In fact, due to the limitations of intelligent nodes in fieldbus in terms of storage capacity and computing speed, even when using soft measurement technology in intelligent nodes, it is generally necessary to establish dedicated soft measurement nodes in the fieldbus control system to complete the soft measurement tasks of the entire system. 4 Specific Implementation During the research and development of the National Ninth Five-Year Plan key project "Research on Integration Technology of Fieldbus Network Control System (96-749-06)", we successfully implemented an integrated system for process control, optimization and management based on LonWorks fieldbus network. This system includes the basic control layer and the advanced control layer shown in Figure 1. The general soft measurement software package constitutes the soft measurement node in the advanced control layer, completing the soft measurement tasks of the entire system [3]. The entire system includes a simulation model system running on a PC, A/D and D/A converters, three LonWorks nodes namely the Echelon TP/FT-10 and TP/FT-10F modules and a control module developed by ourselves using the Neuron 3150 chip, and four advanced control layer nodes. The communication medium uses twisted-pair cables. Three fieldbus nodes are connected to the model system via A/D and D/A converters to simulate field conditions. Each of the three nodes runs a fuzzy control algorithm, a self-tuning PID control algorithm, and a simplified soft measurement system, respectively. Four advanced control nodes are designated as a real-time process monitoring node, a multivariable constraint control node, a real-time optimization node, and a soft measurement node. Intellution Dynamics 2.0 is used as the platform software for the real-time process database, and it includes an interface driver for the LonWorks adapter, allowing direct communication with the fieldbus. The general-purpose soft measurement software package is developed using Visual C++ 5.0 and can run on Windows 95 and Windows NT 4.0 and later operating systems. The software package consists of a matrix operation section, an RBF neural network section, a data processing section, an input/output interface section, configuration software, and a demonstration section. All parts are uniformly scheduled by the system scheduling module, forming an organic whole. In the fieldbus control system built in the laboratory, a simplified version of the large-scale catalytic cracking unit (FCCU) in a combined workshop of Shijiazhuang Refinery was used as a prototype to form a field production process simulation model system. The soft measurement system utilizes our developed general-purpose soft measurement software package and was applied to on-site data collection during commissioning at the Shijiazhuang Refinery in May 1998. The main products of the FCCU unit are stabilized gasoline, light diesel oil, and liquefied petroleum gas (LPG). Their yields are important production indicators, while the dry point of crude gasoline, the pour point of light diesel oil, and the saturated vapor pressure of stabilized gasoline are important quality indicators that need to be estimated. An analysis of the FCCU production process reveals that the main set of measurement variables related to the above-mentioned variables to be estimated consists of the following variables: feed flow rate, recycle oil flow rate, and catalytic reaction temperature in the reaction regeneration system; top temperature of the main fractionation tower, light diesel oil extraction temperature, top reflux flow rate, top reflux temperature, top circulation flow rate, top circulation extraction temperature, top circulation return temperature, 19th vapor phase, first intermediate circulation flow rate, first intermediate circulation extraction temperature, first intermediate circulation return temperature; bottom temperature of the stabilizer tower, top pressure, feed flow rate, feed temperature, reboiler return temperature, and top temperature. The above information was input into the soft measurement system through its human-machine interface configuration interface. The system, through correlation and factor analysis of historical data, determined the use of four soft measurement modules: yield module, dry point module, freezing point module, and vapor pressure module. Each module corresponds to certain auxiliary variables. An RBF neural network soft measurement model was established using the historical data of these variables. Based on the set sampling period, the system retrieved the measured values ​​of the auxiliary variables from the fieldbus database and estimated the variables to be estimated in real time. This system was successfully applied in conjunction with multivariate constraint control in the catalytic cracking unit of Shijiazhuang Refinery during the implementation of the National Ninth Five-Year Plan project "Multivariate Constraint Control and Optimization of Large-Scale Catalytic Cracking Units," achieving significant economic benefits. It passed the appraisal of the Ministry of Education in October 1998. 5. Conclusion The generalization of soft measurement technology and its application in fieldbus control systems give fieldbus control systems significantly advantages over traditional control systems. Further research aims to develop a plant-wide fieldbus control system, integrating decision-making and management tasks into the fieldbus control system, thereby making fuller use of convenient information communication technology and the complete field status data provided by soft measurement technology. References 1 Yu Jingjiang, Zhou Chunhui. Soft sensor technology in process control. Control Theory and Applications, 1996, 13(2): 137-144 2 Gerg Martin. Consider Soft Sensors. Chemical Engineering Progress, 1997: 66-70 3 Shao Huihe et al. Research on integration technology of fieldbus network control system. Technical report of the National Ninth Five-Year Plan Project, 1999