For over a century, model tests have supported in ship and offshore structures design. Experimental results can be used to (1) validate theory to apply numerical simulations in the design and (2) for design by testing, directly using the (scaled) results in the design. Recently, at the Ship Hydromechanics and Structures lab of the TU Delft, a series of experiments was carried out with a 1/20 scale 42 m fast patrol vessel. The model was segmented to enable internal loads measurement. Unfortunately the results were unsatisfactory, in all probability due to errors in the measurement set-up. As a consequence, a new routine of the numerical program fasthip could not be validated. The current research goal was to identify flaws of the prior measurement set-up, develop an improved set-up and determine its performance. First, a broad range of experiments was conducted to gain a deeper understanding of the load path of the previously used set-up. By systematically increasing the disturbance factors, the defects in the set-up could be identified. Due to its versatility and relative easy construction, the two segment rigid backbone construction was used. Two measurement options were developed and implemented, both with the aim to measure the vertical bending moment amidships. The bending moment was measured: 1. Indirectly using 9 force transducers. 2. Directly using 4 coupled fixed strain gauges. The set-up was tested before final assembly. This included an extensive calibration of the strain gauges placed directly on the backbone. Hereafter, experiments with increasing complexity were carried out. During calm water tests, the model was towed at a range of forward speeds. The equations forming the indirect method were verified using the measurement data and the free-free beam boundary conditions. To assess the performance in dynamic conditions, the model was towed at three velocities through a frequency range of regular head waves, forming 48 conditions. The signals were processed using a digital band pass filter and fitted against the excitation frequency. Inertia correction at the indirect method was conducted per time step by using the backbone mass properties and the vertical and rotational accelerations. The methods were compared regarding the bending moment values and the associated 95% confidence limits. Both options deliver equal and accurate results in calm water tests. In regular head waves the required processing to obtain the indirectly measured vertical bending moment proved specific and devious. This leaves the direct method the preferred choice for in particular the semi-planing regime velocities. By increased adaption of the backbone design to the expected excitation forces, the confidence bound limits will turn out more favorable. For this, it is key to incorporate a structural member with limited mass and obtain the required strain levels at the section cuts while maintaining sufficient global stiffness to avoid unwanted resonance. The improved measurement accuracy will further enlarge the practicability of measurement results for direct use in the design and numerical code validation.