This thesis deals with grain refinement under the influence of ultrasonic-driven cavitation in aluminium casting processes. Three major goals of this research were: (1) to identify the mechanism of the cavitation-aided grain refinement at different stages of solidification; (2) to reveal the conditions of the stable grain refinement effect in different alloying systems; and (3) to apply the knowledge gained as a result of an experimental work on a small scale to direct chill (DC) casting process. The research is experimental. Experiments were performed both in water and aluminium melts with specially designed ultrasonic installation, including an 5 kW ultrasonic generator, a 5 kW magnetostrictive transducer at a resonance frequency of 17.5 kHz, sonotrode and a selection of melting and casting facilities. DC casting was used to study the influence of cavitation treatment on a pilot scale. The general description of experimental procedures are highlighted in Chapter 2. To increase the efficiency of ultrasonic melt treatment (UST) both solidification conditions and parameters of sonication should be optimized, i.e. cooling rate, treatment temperature, amplitude and frequency of vibrations, treatment and holding time. The influence of parameters of UST on the effects of sonication in liquids (acoustic streaming and cavitation), were studied both in water and model Al-Cu alloys. The fluid flow patters produced by ultrasonic treatment and mixing were discussed based on the experiments in water, while the impact of cooling conditions and amplitude of vibrations is studied on aluminium alloys. The main conclusions that can be drawn from the results are: (1) the amplitude of vibrations should be high enough to promote cavitation, according to our study in aluminium melts it should be 20 ?m or higher; (2) the efficiency of UST increases with prolonged treatment time; (3) the smaller the treated volume, the finer is the grain size; (4) the effect of UST is quite stable, two minutes between UST of the same volume and casting result in only marginal grain coarsening. Based on the results of experimental work the criteria of efficient UST were suggested in Chapter 3. The influence of cavitation melt treatment performed from temperatures above liquidus until almost complete solidification (75 % solid fraction) on structure formation were studied on pure aluminium and model Al-Cu alloys. In all cases studied UST led to formation of equiaxed grain structure with four to eight times smaller grain size. During such a treatment two mechanisms are involved, namely cavitation-induced heterogeneous nucleation and dendrite fragmentation. The possible mechanisms are discussed in Chapter 4 based on the results of experiments performed isothermally in the semi-solid state and in the temperature range from the temperatures above liquidus until complete solidification. The up-scaling of the UST to direct chill casting can be done more efficiently if the processing occurs outside the primary solidification range of aluminum, when the alloy is still fluid. Thus, in order to be able to apply UST for commercial casting techniques (DC-, investment casting), grain refinement should be achieved after the treatment in the liquid state (Chapter 5). One of the mechanisms explaining enhanced nucleation under cavitation is related to activation of nonmetallic inclusions. To investigate the influence of non-wettable particles on the efficiency of cavitation-aided grain refinement in aluminium alloys, we performed a set of simple experiments. In our experiments aluminium films were manually introduced in the liquid metal by mixing the oxide film into the melt during two minutes, in order to increase the amount of oxides. As a result moderate grain refinement has been achieved, which is an indirect evidence of the cavitation-induced heterogeneous nucleation through activation of oxides. In Chapter 5 it has been shown that UST applied during 10 s above liquidus does not change the grain size and morphology of pure aluminium. When applied in Al-Cu alloys, it results in moderate grain refinement (~20 %). Additions of transition metals (Zr, Ti etc.) can significantly improve the efficiency of cavitation treatment. Many commercial wrought aluminium alloys contain these elements because they also prevent recrystallization. Chapter 5 describes the results on the influence of Zr and Ti on structure formation during ultrasonic melt treatment (UST) in the liquid state. Ultrasonic processing is performed in the temperature range of the primary solidification of an intermetallic phase, i.e. normal casting temperatures of aluminum alloys. Based on the investigations of the influence of UST on formation of primary intermetallics in aluminium alloys with high amount of Zr and Ti and an X-ray and EDX analysis of formed phases, a mechanism was proposed for grain refinement in Al–Zr–Ti alloys produced with cavitation treatment in the temperature range of solidification of primary intermetallics. Ultrasonic melt treatment promotes fragmentation of Al3Zr particles, decreasing the size of potential solidification sites and increasing their number, while Ti substitutes Zr in Al3(Zr1-xTix) particles, which leads to the delay of nucleation to larger undercooling, when the smaller and more numerous intermetallic particles become active. Based on the results of Chapters 3 to 5 the criteria of efficient cavitation-assisted grain refinement of aluminium alloys were formulated and used for: (1) production of commercial aluminium alloys on a small scale in permanent volume (90 cm3), (2) UST in the launder, and (3) DC casting of Al-Cu alloys with small additions of either transition metals (Zr. Ti) or Al-3Ti-B. Effects of ultrasonic melt treatment on structure formation and macrosegregation of aluminium alloys were studied. Five billets 195 mm in diameter were cast at a pilot vertical DC caster. In separate experiments UST was applied in the sump and in the launder. Three different launder designs were used. It was shown that UST during DC casting can refine the structure, if applied either in the hot top or in the launder. Further work is required to optimize the process in terms of treatment time, volume and casting speed. A possible solution might be either optimization of current launder design or introduction of an intermediate chamber, where cavitation treatment can be realized in a larger volume with prolonged treatment time.