The interest of gaining on-line information related to non-metallic inclusions during the steelmaking process has recently increased due to the development as well as the promising results of the Pulse Distribution Analysis with Optical Emission Spectroscopy method (PDA/OES). Even though, the time from sampling to presented results on inclusions is only about 5-10 minutes, the method has also shown limitations with respect to the determination of some inclusion characteristics.Therefore, a first step was to perform a study on other methods such as the cross-section method (CS) on a polished sample surface, the cross-section after etching method (CSE), the bromine-methanol extraction method (BME), and the electrolytic extraction method (EE). This study focused on the evaluation of these methods with respect to the time consumption for preparation and analysis of a sample, the analyzed volume and the determination of inclusion and cluster characteristics such as size, number, particle size distribution (PSD) and composition. The CS and CSE methods were found to be suitable in the determination of the largest cluster in a sample which can be recommended in order to select proper extraction parameters for further studies. The BME method was considered to be fast with the possibility of analyzing a large volume. However, the used solution is chemically stronger compared to electrolytic extraction solutions, which can affect the results. In most aspects, the EE method was found to be the most stable, reliable and accurate method with some limitations regarding the time aspect.Based on this conclusion, the EE method was selected for a comparative study with the PDA/OES method. Reliably detected size ranges by using the PDA/OES method were defined for two low-alloyed steel grades. These are 2.0-5.7 μm and 1.4-5.7 μm for steel samples taken before and after a Ca-addition during the secondary steelmaking, respectively. Moreover, agreements between the EE and PDA/OES methods were observed in the average size and number of detected inclusions when only inclusions with the size > 2 μm were considered. Also, a theoretical minimum size and a maximum number ofinclusions present in the steel sample, which can be detected by using the PDA/OES method, were estimated.The work continued by successfully applying the EE method to study correlations between inclusions observed in the liquid steel samples and in a clogged nozzle (clogging material). It was found that the average sizes of spherical and non-spherical inclusions observed in the steel corresponded well with those observed in the clogging material. However, there were some differences in the frequencies of these inclusions. This was explained by a possible transformation of the present inclusions due to a reoxidation and a reaction with the nozzle refractory of the steel melt. The results of this study may contribute in the selection of proper process parameters or inclusion characteristics for future studies on the improvement and application of on-line methods.Finally, suggestions on how to present and interpret data obtained by the PDA/OES method during production of stainless steels were given in the present thesis. More specifically, the possibilities of defining operating windows with respect to inclusion composition and the use of a B-factor for Al (the total content of Al in inclusions detected by using the PDA/OES method) during the secondary steelmaking were discussed. In addition, a correlation study between B-factors for Al and numbers of inclusions (dV > 4 μm) obtained by using the PDA/OES method on process samples, and corresponding slivers indices from plate products was performed. The results showed a moderate correlation between these parameters as well as an increase of the slivers index with increased values of the chosen PDA/OES data. This indicates that it could be possible to predict when there is an increased risk of having slivers on the final rolled product at an early stage of the steelmaking process.