During roasting in a shaft furnace (used for the deoxidizing roasting of ore), work-situation faults (WSFs) arise as a result of variations in process conditions and off-spec operation. These work-situation faults can be potentially disastrous and can lead to a total collapse of the control system if they are not detected and diagnosed in time. Furthermore, by their very nature they have to be distinguished from the results addressed by existing methods of diagnosis and tolerance control. This paper presents an innovative work-situation fault diagnosis (WSFD) and fault-tolerance control (FTC) strategy for a control system where a combination of neural networks, expert system, and case-based reasoning is used. As such, a system is established that consists of a magnetic tube recovery rate (MTRR) prediction model, a work-situation fault diagnosis unit, and a fault-tolerance controller. The proposed system diagnoses imminent work-situation faults, and then the fault-tolerance controller adjusts the set-points of the control loops. The outputs of the lower-level control system track the modified set-points, which makes the process deviate gradually from work-situation faults with an acceptable product quality. The proposed system has been applied to the shaft-furnace roasting process in the largest minerals processing factory in China and has reduced the frequency of all work-situation faults by more than 50 per cent, with the ratio of furnace operation increased by 2.98 per cent. It has been proven to provide many benefits to the factory.
A Novel Approach for Fault Tolerance Control System and Embedded System Security
