During the last years, an increased interest in in situ-damage characterization during ultrasonic fatigue testing evolved. Different non-contact in situ-surveillance systems have been developed or applied to the present ultrasonic fatigue testing systems, e.g. analysis of the nonlinearity parameter and the resonant frequency [1, 2] as well as thermal measurements [3–5]. In addition, Heinz et al. [6] performed noncontact strain measurements during ultrasonic fatigue testing. The aim of the present investigations is to evaluate the fatigue crack initiation in the cast aluminium alloy AlSi7Mg (A356) in sand and die cast condition under high cycle and very high cycle fatigue loading conditions. Therefore, a combined analysis of the development of the specimen temperature, the resonant frequency fres and the nonlinearity parameter rel was performed. For such investigations it is essential to keep the ambient temperature strictly constant and the increase of temperature due to the resonant conditions as low as possible. Therefore, a spot cooler using a vortex tube was applied to the ultrasonic fatigue testing system. With this special cooling device it was possible to keep the ambient temperature constant. An infrared thermo-camera system was used for the in situ-examination of the specimen temperature and the surveillance of the ambient temperature during ultrasonic fatigue testing. The determination of the resonant frequency and the nonlinearity parameter was realized online during fatigue tests according to Mayer et al. [2]. After failure, the fracture surfaces of the specimens were investigated intensively by scanning electron microscopy (SEM). According to Krewerth et al. [5], the correlation of the results from thermographic and fractographic investigations can shed some light on the crack initiation point. Figure 1a shows the S-N curves of the fatigue tests on the aluminum specimens produced by sand casting (batch 1) and die-casting (batch 2) in hot isostatically pressed condition, respectively. Specimens of both batches failed due to casting porosity. However, the casting porosity in the sand cast material was significantly higher leading to a lower fatigue strength. Several specimens were investigated by thermographic measurements accompanied by simultaneous analysis of the resonant frequency as well as the nonlinearity parameter. Exemplarily, specimen 1 (sand casting, see Fig. 1a) is explained here in more detail. The development of the resonant frequency (fres), the nonlinearity parameter ( rel), the ambient temperature Tamb, the maximum temperature Tmax and the temperature variation T is shown in fig. 1b. This specimen failed due to an internal defect at Nf = 2.16 × 106 cycles. The dashed black