Research On Methods Of Fault Diagnosis And Simulation Experiment Of Shaft Furnace Roasting Process Based On Reconstruction And Contribution Plots

The immense amount of hematite in china is generally in a low concentration grade. And it is rather difficult to perform a magnetic separation to get iron ore concentrate with a satisfying concentration grade using conventional methods. A shaft furnace is often adopted in the process of high temperature reduction roasting in advance to transform the weakly magnetic ore into strongly magnetic one. And this process is quite essential in the derivation of crude material of the process of magnetic separation.The shaft furnace roasting processes share the synthetic complexities with other process industries, such as multi-variables, strong coupling, strong nonlinearity, difficulties with continuously measuring key technological parameters and frequent changes of working condition. Under the condition when an operator cannot react properly to corresponding working conditions, faults like fire-emitting, flame-out, ore-melting, under-reduction and over-reduction are highly likely to occur. And if the fault is not detected, identified and dealt with in time, technological parameters of the process would be degraded and a hazard for the whole system of breaking down would emerge. Fault diagnosis of shaft furnace roasting process is key to ensure the safety and stability of this process. Therefore, the research on the fault diagnosis of shaft furnace roasting process has intrinsically important value for practice.Supported by the National 973 project "The research of several technologies and the partially material object simulation platform in integrated control system with the property of security, synergism and convenience", with a basis of existing methods of fault diagnosis, aiming at shaft furnace roasting process, the thesis designs and develops a fault diagnosis method based on reconstruction and contribution plots. The major contributions of this thesis are as follows.(1) With the basis of existing dynamic mechanism model of shaft furnace roasting process on the partially material object simulation platform, we have designed and developed a simulation experimental system of shaft furnace roasting process and accomplished the simulation of faulty operation conditions. First, the physical and chemical processes which include combustion process, heat transfer process and reduction process are analyzed, along with the mechanism of the faulty operation conditions. Then according to the special distribution inside the furnace, dynamic mechanism models are modified in combustion chamber, preheating zone, heating zone, reducing zone and moving out process based on the characteristics of four typical fault working conditions, and a tag to set simulation fault off has been introduced and simulation of normal and fault working conditions has been completed.(2) The thesis has designed a diagnosis method for shaft furnace roasting process to monitor and diagnose the working condition based on reconstruction and contribution plots, which is composed of three parts, fault detection based on principal component analysis, fault reason diagnosis based on reconstruction and fault variables diagnosis based on contribution plots. They are achieved according to the following steps:1) Historical data in the normal process of roasting of shaft furnace is used to develop a model for relationship between variables via principal component analysis. Fault detection of shaft furnace roasting process is accomplished.2) Historical data in the fault processes of roasting of shaft furnace is collected to develop a library of fault directions. Reconstruction along each pre-defined direction is applied to diagnose the fault reason.3) Contribution plot is designed to diagnose the faulty variables dealing with fault working conditions that never happened before.(3) A partially material object simulation platform has been built. The platform is composed of five parts, i.e. shaft furnace virtual object, virtual detecting devices and actuators, PLC controller, monitoring system of shaft furnace roasting process and fault diagnosis system. A virtual object of shaft furnace has been built on the integrated simulation platform, a monitoring system of shaft furnace roasting process has been developed in RSView32 and a fault diagnosis system for shaft furnace roasting process has been developed on the platform of operation control. This simulation platform for fault diagnosis has an algorithm-friendly structure which helps to testify and maintain the algorithms.(4)Experiments have been carried out on the simulation system to testify the feasibility and validity of the methods of fault diagnosis that have been proposed in this thesis. The data in processes of normal working condition, fire-emitting, flame-out, under-reduction and over-reduction is used in the experiments in order to test the diagnosis results in each operation conditions. And the results show that the proposed method can diagnose the fault reason and fault variables when fault working conditions occur. False alarm rate in a normal condition is 7.5%o and the accurate rate of diagnosis in fault working conditions is 100%.