Research on Faults Symptoms Extraction and Neural Network Diagnosis for Direct Air-cooled Condenser

Based on the structural characteristics of direct air-cooled condenser, the typical faults that may occur in the operation of condenser are summarized, including insufficient tightness of vacuum system, dust accumulation of condenser, freezing of condenser and so on. Some process parameters that can accurately reflect the signs of the faults are determined to set up the fault knowledge library, and are divided into five stages according to the magnitude of parameter changes, i.e. low 2 value, low 1 value, normal, high 1 value and high 2 value. By using the BP neural network method, a typical faults diagnosis model of direct air-cooled condenser has been developed. The result of an simulation test shows that the model can accurately diagnose the direct air-cooled condenser's faults.