Because,of incomplete historica} records for the transient operation of steam turbine in power plant, data such as temperature of the steam, inner metat of HP, power output and time interva[ of a transient process, were not enough to accurately calculate the stress and Life-loss of a turbine unit, The general way to evaluate the life-loss of an old turbine unit was te count transient process times under cool and warm as well as hot state to get accumulative total Life-loss. Such general

more »

... on way was rough because of the ne.ylects of the transient details in starting up or shutting down process. ] t was possib]e to use onIine [it'e management system for more accurate evaluation of life loss ofa turbine. The paper proposed a new method cembining statistics and artificial neural network ba$ed on online lifetlme rnanagement system, to evaluate past life-loss of the thermal turbine parts. First, the statistics of alt previous typical transient operation and the concerned parameters for a year were made to be an operating sample, Second, an artificial neural network was applied to deal with these data. which need not set up accurate math mode] and carry out mapping between input and output. The ]it'e-]oss could be output after historica] data inputting and artificial neural network training. The new method overcatne the constant life-loss caused by transient operation. This method was apptied in a l25MW unit and basicall>, satisfied results were achieved,