A singular pencil model and neural network cascade fault diagnosis strategy for the lumber-drying process

Abstract

This paper presents some results that were obtained in fault diagnosis for a wood-drying process. It is difficult to detect and diagnose faults or failure in the drying process due to complex dynamic nonlinearity, coupling effects among key variables, and process disturbances caused by the variation of lumber sizes, species and environmental factors. In this paper, a real-time fault diagnosis algorithm is developed based on a singular pencil model and neural network classifier. Inputs of the network are the process I/O data, such as moisture and temperature, and estimated parameters and states, while outputs of the network are process fault situations. A wood-drying kiln is studied as a test case, which is with two actuators and 23 sensors, six estimated parameters and states, and 11 fault situations. The simulation results show that the strategy appears to be better suited to diagnose faults of such industrial processes.