In this paper, two-phase-flow oscillations at the natural-circulation CIRCUS test facility are investigated in a two-riser configuration. These oscillations are driven by flashing (and to some extent by geysering). For a given range of operating conditions of the facility, the oscillations exhibit erratic behavior. This study demonstrates that this behavior can be attributed to deterministic chaos. This is proven by performing a continuous wavelet transform of the measured time series. Any hidden self-similarity in the measurement is seen in the corresponding scale-space plane. The novelty of the present investigation lies with the multifractal approach used for characterizing the chaos. Both nonlinear time series analysis and wavelet-based analysis methods show that the dynamics of the flow oscillations has a multifractal structure. For the former, both Higuchi's method and detrended fluctuation analysis (DFA) were used, whereas for the latter, the wavelet-transform modulus-maxima method was used. The strange attractor corresponding to the dynamics of the system can thus be described as a set of interwoven monofractal objects. The global singular properties of the measured time series is then fully characterized by a spectrum of singularities f(), which is the Hausdorff dimension of the set of points where the multifractal object has singularities of strength (or Hölder exponents of) . Whereas Higuchi's method and DFA allow easily determining whether the deterministic chaos has a monofractal or multifractal hierarchy, the wavelet-transform modulus-maxima has the advantage of giving a quantitative estimation of the fractal spectrum. The time-modeling of such behavior of the facility is therefore difficult since there is sensitive dependence on initial conditions. From a regulatory point of view, such behavior of natural-circulation systems in a multiple-riser configuration has thus to be avoided.