Guided ultrasonic waves enable the structural health monitoring of thin-walled structures. For optimal sensitivity in damage detection, frequency and excitation signal selection is crucial. A broadband structural identification using frequency response functions allows subsequent evaluation of different virtual excitation signals and frequencies. This reduces the experimental effort, but linearity of the entire system is a prerequisite to obtain unique frequency response functions. This study presents a system linearity check of a typical laboratory structural health monitoring setup investigating the harmonic distortion in the respective components. These consist of a signal generator, high-voltage amplifiers, piezoceramic transducers adhesively bonded to an aluminium specimen and a digital oscilloscope. The distortion factors of the outputs at different positions in the effect chain are compared. The distortion factor values at the outputs of the signal generator, high-voltage amplifiers and the receiving piezoelectric transducers are all below 2 %. The frequency-dependence of the values is explained by digitization and sampling frequency, noise, structural dynamic behaviour of the specimen and piezoelectric transducers as well as excitability of guided ultrasonic waves with the piezoelectric actuator. In conclusion, this study verifies the assumption of linearity used in this exemplary structural health monitoring system based on frequency response functions. This assumption is made in a lot of publications, but rarely verified.