Research On Fault Working Situations Diagnosis And Fault-tolerance Control For Shaft Furnace Roasting Operation

Ore dressing industries in China are often required to process the material of hematite, from which the useful parts are difficult to separate by conventional methods because of low concentration grade. Then, shaft furnaces are often adopted to roast the hematite ore before it is separated since high temperature reduction roasting process can enhance the magnetism of hematite. Therefore, the working situation of operation process of shaft furnace roasting is directly related to the metal recovery processing and concentrate grade. Shaft furnace roasting is a complex semibatch production process, where chemical reaction occurs to the hematite ore under high temperature. In this process, ore’s size, category and ingredient change frequently. Especially, the operational index and the control object of the roasting process, namely, magnetic tube recovery ratio (MTRR), cannot measure online in real time, thus the operational control system often generates improper setpoints for combustion chamber temperature (CCT), reducing gas flow (RGF) and ore discharging time (ODT). Consequently, some fault working situations (FWS) appear, they are named as follows:FE:Fires emit out of the combustion chamberFO:Flames reach out of the top of the furnaceSE:Iron ores stick inside the furnace so that further entry of ores is difficultUD:Under deoxidizationOD:Over deoxidization Therefore, it is significant for the normal operation of shaft furnace to diagnose the FWS correctly in real time and then change the setpoints of CCT, RGF, and ODT, which are tracked by control system so that the shaft furnace can keep away from FWS.As a part of the project "Research and development of ore dressing industry process comprehensive automation system" supported by national863plans, this thesis investigated the fault working situation(FWS) diagnosis and fault-tolerance control(FTC) for shaft furnace roasting process in view of the key problems of MTRR soft sensing, FWS and FTC for shaft furnace, as well as how the FWS and FTC can be realized, with the following achievements:1. Aiming at the problems about MTRR, including that it cannot measure online in real time, it performs fuzzy dynamic relationship with CCT, RGF, and ODT though without mechanism model, it is obtained after great time-lag that cannot meet the requirement on real-time by FWS diagnosis, this thesis presented a hybrid MTRR soft sensing model, which was combined by a fuzzy master model and a neural network compensation model. The fuzzy rules in master model was gained with the method of online clustering based on equal temporal interval, while the neural network compensation model was provided with high training speed by adopting search-then-converge approach. Simulation showed that this method can reach better soft sensing precision than other models;2. In terms of the fact that the dynamic characteristic between FWS and MTRR soft sensing value, output of control loop, control quantity, roasting process variables and boundary conditions cannot be formulized as a mechanism model, along with the problem when the operators judge FWS with their experience during observation, this thesis proposed a FWS diagnosis method using rule-based reasoning and a heuristic search strategy based on Analytic Hierarchy Process(AHP). This method extracted the expert knowledge from the operators to form a rule base that can fulfill FWS diagnosis according to the MTRR soft sensing value, the output of control loop, the control quantity, the process variables, as well as the bounary conditions;3. Considering that the dynamic process of FWS elimination by modifying CCT, RGF, and ODT is closely related to but cannot be formulated by category of FWSs, output of control loop, control quantity, process variables and bounary conditions, plus that the FWS is currently disposed by the operators who change the setpoints of CCT, RGF, and ODT experientially, a case-based reasoning method was proposed for FTC of shaft furnace in this thesis. The FTC takes the category of FWSs, the output of control loop, the control quantity, the process variables and the bounary conditions as its input parameters, and then generates the correction values for setpoints by the technology of case-based reasoning. The revised setpoints can help the furnace step out from FWSs and keep the MTRR within a small range around its target value. As the weights for case descriptions are unknow in the reasoning process, this thesis suggested an method for optimization of these weighs by combination of ant colony algorithm with correlation analysis, which was proved to be effective to the FTC for shaft furnace;4. By integrating the above mentioned MTRR soft sensing model, FWS diagnosis and FTC method, a FWS diagnosis and FTC strategy for shaft furnace was suggested while a software that can realize this strategy was developed in this thesis. This software contains the modules of data collection, data display, data saving, test timing, MTRR soft sensing, FWS diagnosis and FTC, AHP search manner, ant colony algorithm and correlation analysis; and5. All the methods proposed in this thesis was testified through the data from the real-world shaft furnaces. The test result indicated that the system can implenent FWS diagnosis for shaft furnace roasting process when the setpoints of CCT, RGF, and ODT turn to improper because of the variation of boundary conditions and FWSs happen. Then, the setpoints under FWS can be modified to a proper ones by FTC. During the process of the whole control system tracking the modified setpoints, the shaft furnace operates further and further from fault working situations, with the MTRR being controlled well around its target.