This thesis examines the sound transmission loss (STL) through composite sandwich panel systems commonly used in the marine industry. Experimental, predictive and optimisation methods are used to evaluate the acoustic performance of these systems and to improve their acoustic performance with noise treatment. The complex nature of the material properties of composite sandwich panels was found to be dependent not only on the physical properties but also the frequency of incident noise. Young’s modulus was found to reduce with increasing frequency as has been predicted in the literature which is due to the shear stiffness dominating over the bending stiffness. Two methods for measuring these properties were investigated; ‘fixed-free’ and ‘free-free’ beam boundary condition modal analyses. The disagreement between these methods was identified as the clamping fixed nature that increased flexibility of the beam. Composite sandwich panels can be modelled as homogeneous isotopic materials when predicting their acoustic performance provided the dilatational resonance is above the frequency range of interest. Two such panels were modelled using this simple sound insulation prediction method, but the agreement between theory and experimental results was poor. A variable Young’s modulus was included in the model but agreement remained relatively poor especially in the coincidence frequency region due to variation of Young’s modulus with frequency. A statistical method of optimisation of the parameter settings by fractional factorial design proved successful at identifying the important parameters that affect the sound transmission class (STC) of a noise treatment material applied to a panel. The decouple foam layer and attachment method were the most significant factors. The same method, with higher resolution was then used to identify the important parameters that affected the noise reduction class (NRC) finding that the outer foam thickness without a face sheet were the most significant factors. The independent optimisation studies performed for each of the STC and NRC produced conflicting results meaning that both could not be achieved simultaneously.