Glued laminated timber (glulam) is manufactured by gluing and stacking timber lamellas, which are sawn and finger-jointed parallel to the wood grain direction. This results in a sustainable and competitive construction material in terms of dimensional versatility and load-carrying capacity. With the proliferation of glued timber constructions, there is an increasing concern about safety problems related to adhesive bonding. Delaminations are caused by manufacturing errors and in service climate variations simultaneously combined with long-sustained loads (snow, wind and gravel filling on flat roofs). Several recent building collapses were related to bonding failure, which should be prevented in the future with a timely defect detection. Although ultrasound testing shows a high sensitivity to debonding, traditional diagnostic methods for wood generally require that the transducers are pressed with a well-controlled force onto the inspected sample. Air-coupled ultrasound (ACU) removes this limitation, which allows a high measurement reproducibility as well as automated transducer positioning and orientation in an arbitrary measurement grid. The main disadvantage is that the pressure level coupled into the sample is reduced by about 50 dB with respect to the one coupled with the transducers pressed onto the sample. As a consequence, until now only material thicknesses of up to 50mm could be tested without resonance, which has limited the applicability of this method to particleboards, fiberboards, plywoods and laminated veneered lumbers. The goal of the thesis was the development of novel non-destructive testing methodologies capable of imaging the position and geometry of delaminations within the bonding planes of glulam. An ACU system prototype capable of detecting an ultrasound beam transmitted through up to 500mm thick glulam was developed, consisting of off-the-shelf ACU transducers, high-power pulsed excitation electronics and a low-noise amplification chain. A five-axes computerized scanning system and a low-cost micro-electromechanic sensors (MEMS) linear array design allowed ultrasound imaging with fix or independent transmitter and receiver transducer units. The bonding assessment was fundamentally based on the evaluation of the attenuation of the ultrasound beam, which significantly increases when transmitted through a material discontinuity (delamination) with respect to a defect-free glue line. Wood is a highly structured anisotropic, heterogeneous and porous material, all of which have an impact on ultrasound wave propagation. For glulam laminates the long dimension of the lamellas is oriented parallel to the wood grain direction. The density distribution, the annual ring structure and the grain angle are the main material parameters for the simulation of ultrasound wave propagation. A full-wave finite-difference time-domain (FDTD) model was developed to simulate pulsed ACU wave propagation in heterogeneous timber laminates. The model allows local definition of viscoelastic anisotropic material properties at each pixel (gas, fluid or solid). The model was validated with analytical derivations for: a) plane wave velocity and attenuation in homogeneous anisotropic media, b) point source wavefronts in unbounded timber with straight and curved annual rings, c) energy flux shifts within anisotropic timber lamellas, d) wave interference as a function of the air gap separation between delaminated timber lamellas, and e) sound field radiation in air. The latter three were as well quantitatively validated with experimental data. Wave paths as a function of the insonification position in a 65mm thick laminate were described with 2mm accuracy. Additional tests were performed to evaluate the impact of wood growth heterogeneity (knots, earlywood/latewood transitions) on ACU wave propagation, based on expected or experimentally determined material property distributions. An experimental bonding quality assessment was first performed by perpendicularly insonifying the bonding planes (normal transmission - NT setup). The transmitter T and receiver R transducers were scanned as a single unit along the width and length of the laminate. For each measured pixel a global assessment (defective/defect-free) of the gluing state of all bonding planes was obtained. Paper I shows experimental results for thin laminates and Paper II for multilayered glulam. Dedicated signal processing (deconvolution imaging, overlapped averaging, amplitude tracking, difference imaging) profited from the highlights of the ACU method to improve the lateral resolution, the signal-to-noise ratio and the contrast of the defect maps. A second experimental setup (slanted lateral transmission - SLT setup) allowed to detect bonding defects in glulam beams of arbitrary height and length. Moreover, the defects could be associated to each of the individual bonding planes. In this case, the transmitter and receiver transducers were scanned as a single unit along the height and length of the laminate, so that a slanted ultrasound beam was coupled at a defined refraction angle within the sample and the fields transmitted or reflected through a specific bonding plane were detected. This method became the subject of a patent application. As an outlook, the feasibility of air-coupled ultrasound tomography was demonstrated with numerical tests and preliminary experiments on glulam. The FDTD wave propagation model was excited by the difference of the time-reversed sound fields transmitted through a test and a reference (defect-free) glulam cross-section. Both datasets were obtained with the same SLT setup. Wave convergences then provided a map of bonding defects along the height and width of the inspected glulam cross-sections. Further research is envisaged in this direction. A separate investigation was performed to evaluate the applicability of limited-angle X-ray computed tomography (LCT) to the detection of gluing defects in timber laminates (Paper III). Radiographies measure the integrated attenuation along the ray path, which is a function of the density and thickness of the wood sample. Computed tomography combines radiographies at all possible orientations between sample and detector and reconstructs three-dimensional information of the inner state of the sample. A limited-angle reconstruction was proposed, which only required radiographies in a narrow angular range parallel to the glue line. This method filters out undesired wood heterogeneity, is theoretically limited in only one sample dimension and achieves reduced measurement and reconstruction times. The feasibility of the inspection was experimentally demonstrated with a microfocus laboratory system. As a complementary investigation, the ACU NT setup was applied to material property characterization in particle composites (Paper IV). Acoustic parameters (sound velocity and frequency-dependent amplitude transmission) were correlated with X-ray radiographies and allowed prediction of the horizontal density distribution (HDD) and a classification according to particle geometry. The effects of grain scattering and multi-scale porosity in ACU propagation were discussed with the help of the FDTD wave propagation model.