In the steel making process, element concentration is controlled to determine the intended properties of steel products. Especially, phosphorus leads to brittleness so the process to reduce its concentration to the desired level is conducted[1]. Phosphorus is removed by transferring it from molten steel to molten slag. Slag is a complexed oxides composed of CaO, SiO2, FeO and other oxides. The ratio of the amounts of CaO and SiO2 is called basicity, which is one of the factors to control the efficiency of removing phosphorus. Therefore, real-time analysis of the basicity of molten slag and its control are important.

Analyzers, such as handheld X-ray fluorescence (on-site analysis) or spark discharge atomic emission spectrometer, can quickly analyse the solid sample. However, due to the high temperature of molten slag (over 1,000°C), these analyzers cannot be placed near the molten slag. Furthermore, there is not enough time to take sample, so above analyzers cannot be employed. For these reasons, Laser Induced Breakdown Spectroscopy (LIBS), that can analyse samples located at a certain distance, was considered in this research.

 As a preliminary experiment, the accuracy of predicted basicity was investigated using molten slag samples with various basicity levels in melting pots (referred to as lab-slag in this research). Partial Least Square regression (PLS) was conducted to calculate basicity based on the obtained spectra. The results demonstrated that the intended accuracy was achieved. In the next step, the basicity of molten slag falling from a converter (referred to as factory-slag (Fig.1 (a))) was analyzed. PLS analysis was conducted using spectra obtained from lab-slag as training data. As shown in Fig.1 (b), the desired accuracy was not achieved due to the defocus of laser and dust or steam from the factory. Defocus occurred because the molten slag always fluctuated. Therefore, some spectra obtained from factory-slag that was used as training data to include the influence of above. As a result, the accuracy was dramatically improved, and the intended value was achieved (Fig. 1 (c)).

 This research shows that efficiency of controlling phosphorous can be improved by on-line analysis of the basicity of molten slag using LIBS.