Abstract A non-intrusive method of two-phase flow identification is investigated in this paper. It is based on image processing of data obtained partly from dynamic neutron radiography recordings of real two-phase flow in a heated metal channel, and partly by visible light from a two-component mixture of water and air. Classification of the flow regime types is performed by an artificial neural network (ANN) algorithm. The input data to the ANN are some statistical moments of the wavelet pre-processed pixel intensity data of the images. The pre-processing used in this paper consists of a one-step multiresolution analysis of the 2-D image data. The investigations of the neutron radiography images, where all four flow regimes are represented, show that bubbly and annular flows can be identified with a high confidence, but slug and churn-turbulent flows are more often mixed up in between themselves. The reason for the faulty identifications, at least partially, lies in the insufficient quality of these images. In the measurements with air-water two-component mixture, only bubbly and slug flow regimes were available, and these were identified with nearly 100% success ratio. The maximum success ratio attainable was approximately the same whether the raw data was used without wavelet preprocessing or with a wavelet preprocessing of the input data. However, the use of wavelet preprocessing decreased the training time (number of epochs) with about a factor 100.

This paper deals with the analysis of experimental positron lifetime spectra in polymer materials by using various algorithms of neural networks. A method based on the use of artificial neural networks for unfolding the mean lifetime and intensity of the spectral components of simulated positron lifetime spectra was previously suggested and tested on simulated data [Pžzsitetal, Applied Surface Science, 149 (1998), 97]. In this work, the applicability of the method to the analysis of experimental positron spectra has been verified in the case of spectra from polymer materials with three components. It has been demonstrated that the backpropagation neural network can determine the spectral parameters with a high accuracy and perform the decomposi-tion of lifetimes which differ by 10% or more. The backpropagation network has not been suitable for the identification of both the parameters and the number of spectral components. Therefore, a separate artificial neural network module has been designed to solve the classification problem. Module types based on self-organizing map and learning vector quantization algorithms have been tested. The learning vector quantization algorithm was found to have better performance and reliability. A complete artificial neural network analysis tool of positron lifetime spectra has been constructed to include a spectra classification module and parameter evaluation modules for spectra with a different number of components. In this way, both flexibility and high resolution can be achieved.