Bones have amazing mechanical characteristics making them a main contributor for structurally stabilizing our body and protecting inner organs. Unfortunately, bones may fail due to an overload or due to a degraded bone structure. Bone degradation varies between individuals and is typically a consequence of diseases such as osteoporosis. To reduce the number of osteoporotic fractures and to develop new pharmaceutical treatments accurate patient specific bone fracture risk predictions are required. In current clinical practice, fracture risk predictions are based on cost effective bone density measurements. In addition to that, recent developments of high resolution peripheral quantitative computed tomography (HRpQCT) devices allow 3D micro-structural imaging of bones. These images allow performing fracture risk predictions based on finite element (FE) simulations. However, straight forward linear FE models show only moderate improvements over densitometric measurements. Improvements are promised by the implementation of sophisticated non-linear FE simulations which are able to simulate detailed failure mechanisms. However, developments of non-linear FE model are bound by the limited knowledge of micro-structural failure mechanisms and the required large computational resources. For this reason, the present thesis was divided into three aims:​ (i) develop and perform systematic classification and analysis of experimental failure in human bone specimens, (ii) develop combined experimental and computational multiscale failure assessment to improve failure predictions and (iii) investigate local elastic properties of human trabecular bone by an inverse finite element algorithm. All three aims were based on experimental data gained through image guided failure assessment (IGFA) of human bone specimens.

For the first aim, a systematic classification approach was developed for 3D analysis of the images from the IGFA experiments conducted using synchrotron radiation. For every micro-structural element of trabecular bone (rods and plates) our classi- fication scheme differentiated between three local load states and ten local fracture morphologies. With this classification scheme over 1000 failure sites were identified and classified. Although our specimens were loaded in compression we discovered that failure was primarily caused by local bending. Furthermore, rods exhibited two different fracture morphologies depending on the load state, i.e. compression or tension. To our knowledge, this was the first study to systematically evaluate failure in IGFA experiments.

For the second aim, the experimental results were analyzed together with a multiscale computational model. We successfully showed the determination of local experimental failure based on local experimental and computational strains. Subsequently we developed a non-linear multiscale model containing local strain based failure criteria. We validated the local behavior of our non-linear multiscale model with our experimental results. The non-linear multiscale model performed better than the linear multiscale model. Furthermore the non-linear multiscale model used significantly less computational power than a non-linear finite element simulation.

For the third aim, we investigated the influence of local variations of elastic material properties in finite element simulations. We developed an algorithm to iteratively adapt the local elastic properties until the simulated and experimental strain matched. We showed that linear elastic isotropic FE could simulate artificially and experimentally generated local strains correctly, if local elastic properties were predicted a priori. We also investigated the possibility of predicting the local properties based on the local attenuation measurements. Nevertheless iteratively determined moduli did not correlate with local attenuation measurements.

In conclusion, this thesis revealed that experimental results, combined with computational modeling, allow multiscale failure assessment in human trabecular bone. It is expected that the insight gained from these findings may have far-reaching consequences for the better understanding of bone failure related to osteoporosis.