Critical Heat Flux – CHF – occurence is the main thermo-hydraulical factor that restrains the energy produced in Pressurized Water Reactor – PWR – nuclear plants. The usual method of determining CFCH is based upon numerical simulation performed by computer programs such as COBRA, which were developed considering the reactor core sub-channel analysis. These programs implement a correlation, or empirical function, wich interpolates the results obtained through experimental simulation, acocomplished on test sections – TSs – for the sake of obtaining CHF in a wide core operational range. This work investigate and analyze an alternate method of detrmining CHF using, as a correlation, artificial neural networks – ANNs. In this method, the ANNs are obtained through trainning, making use of backpropagation paradigm, against the same experimental data set that came from the TSs.